DISS. ETH NO.

26895

Spatial and temporal modelling of insect pests

and their host plants under climate change

A thesis submitted to attain the degree of

DOCTOR OF SCIENCES of ETH ZURICH

(Dr. sc. ETH Zurich)

presented by

Marc Philip Grünig

MSc in Geosciences
University of Basel

Born on 27.10.1988
Citizen of Burgistein BE

accepted on the recommendation of

Prof. Dr. Loïc Pellissier

PD Dr. Dominique Mazzi
Dr. Pierluigi Calanca
Prof. Dr. Sven Bacher

2020
2
Table of Contents
SUMMARY 4

ZUSAMMENFASSUNG 6

GENERAL INTRODUCTION 9

CHAPTER I 31
Crop and forest pest metawebs shift towards increased linkage and suitability
overlap under climate change

CHAPTER II 88
Inflection point in climatic suitability of insect pest species in Europe suggests
non-linear responses to climate change

CHAPTER III 134

Applying deep neural networks to predict incidence and phenology of plant
pests and diseases

CONCLUSION AND PERSPECTIVES 168

ACKNOWLEDGEMENTS 186

3
SUMMARY

Agriculture is facing the major challenge of providing food security for a growing
human population without jeopardizing the environment. Meanwhile, climate change is putting
additional pressure on crop production, as current challenges are amplified and new ones
emerge. Besides the well-known direct impact of abiotic changes, such as increasing heat stress
or changing precipitation regimes, a key question remains in how far biotic factors in form of
insect pests will further undermine crop production under climate change. Insects, being
ectotherms, are highly dependent on thermal changes and therefore directly respond to raising
temperatures. Under warmer temperatures, local, already established insect pests are expected
to exhibit more generations, faster population growth, and increasing metabolism and newly
occurring, invasive insect pests may put additional pressure on agriculture. While globalisation
supports species to overcome natural barriers, they find increasingly suitable conditions as
climate warms. Therefore, the risk of insect pest invasion is increasing with climate change,
particularly in higher latitudes, where cold winter temperatures have previously limited insect
pest distributions. Preparing for increasing pest pressure and anticipating invasions of insect
pest species to protect agricultural food production, will require a better understanding of
potential shifts in insect pest distribution under climate change.

This thesis aimed at providing a better understanding of how the distribution of insect
pests will respond to climate change. To tackle this question, I used different spatial modelling
approaches to investigate the climatic niches of insect pest species and their host plants.
Mapping those niches spatially provided information on the potential distribution ranges of
insect pest species. The first chapter investigated potential range shifts of black-listed insect
pest species and their host plants using correlative models. I evaluated the spatial and temporal
dynamics of predicted future interactions between insect pests and their known host plants using
a metaweb approach. The second chapter compared two complementary modelling approaches
to examine differences between fundamental and realized climatic niches of insect pests and
their distribution along temperature gradients. The third chapter used deep neural network
algorithms to identify pest damages in the field and couples the occurrence of pest damages
with meteorological data. This chapter served as a proof of concept study for the integration of
new technologies into pest damage forecasting.

4
 SUMMARY

This thesis showed that climate change leads to shifts of potential distributions of insect
pest species (Chapter 1 and Chapter 2) and cultivated plants (Chapter 1) within Europe. In
particular, Chapter 1 used datasets of 89 black-listed insect pest species and 126 crops and
forest tree species to show that southern European regions are under high pest pressure posed
by non-native species already under current climate. Meanwhile, in northern European regions
agriculture may benefit from improved climatic suitability for cultivated plants. However, at
the same time, in these regions, yield may become more vulnerable to insect pests as the
changing climate allows the survival of new pest species. Chapter 2 compared physiological
and correlative species distribution models, yielding good model agreement despite conceptual
differences, therefore providing additional robustness to the inferences. With both modelling
approaches, I identified a threshold of temperature beyond which the number of pest species
with suitable climatic conditions will rapidly increase. Together, the results from Chapter 1 and
Chapter 2 deliver strong arguments that new invasive insect pest species will enter Europe in
the near future, underlining that there is a growing urge for development of new pest
management strategies to cope with such invasions. Chapter 3 showed how the combination of
big data and deep learning may provide new solutions for pest damage forecasting, uncovering
new opportunities for sustainable plant protection from exploiting innovative technologies.

Together, the three chapters of this thesis contribute to a better understanding of the
impact of climate change on the distribution of insect pests and their host plants. In particular,
this thesis combines comprehensive datasets of insect pests, agricultural crops and forest tree
species in order to identify regions under elevated pest pressure and introduces a new method
for pest forecasting. The chapters of this thesis provide a scientific basis for improvement of
plant protection and pest management.

5
ZUSAMMENFASSUNG

Die Landwirtschaft steht der grossen Herausforderung gegenüber, die

Ernährungssicherheit für eine wachsende Bevölkerung zu gewährleisten und gleichzeitig die
unerwünschten Umweltwirkungen zu minimieren. Der Klimawandel übt zusätzlichen Druck
auf den Pflanzenbau aus, indem gegenwärtige Herausforderungen verstärkt und gleichzeitig
neue hervorgerufen werden. Neben den direkten Einflüssen von abiotischen Veränderungen,
wie zum Beispiel erhöhtem Hitzestress und veränderten Niederschlagsmustern, die durch den
Kimawandel hervorgerufen werden, eröffnet sich eine weitere Schlüsselfrage bezüglich des
Einflusses von biotischen Faktoren, welche in Form von Schadinsekten den Pflanzenbau
gefährden. Insekten sind als ektotherme Organismen stark von Veränderungen der
Umgebungstemperatur abhängig und reagieren deshalb direkt auf den Klimawandel. Man muss
davon ausgehen, dass mit den steigenden Temperaturen die lokalen, etablierten Schadinsekten
eine erhöhte Anzahl Generationen aufweisen, die Populationen schneller wachsen, der
Stoffwechsel erhöht wird, und neu auftretende, gebietsfremde Arten zusätzlichen Druck auf die
Landwirtschaft ausüben. Die Globalisierung hilft Arten natürliche Grenzen zu überwinden und
dadurch neue Gebiete zu besiedeln, in denen sie zunehmend geeignete klimatische
Bedingungen vorfinden. Vor allem in höheren Breitengraden haben kalte Wintertemperaturen
bis anhin die Verbreitung eingeschränkt, weshalb sich dort das Risiko durch neue invasive
Arten erhöht. Eine bessere Kenntnis über die potentiellen Verschiebungen der
Verbreitungsgebiete von Schadinsekten im Klimawandel trägt zur Vorbereitung auf den
steigenden Schädlingsdruck bei und hilft somit die landwirtschaftliche Nahrungsproduktion
davor zu schützen.

Diese Dissertation hatte das Ziel ein besseres Verständnis für den Einfluss des
Klimawandels auf die Verbreitung von Schadinsekten zu erarbeiten. Für die Bearbeitung dieses
Themas nutzte ich verschiedene räumliche Modellierungsansätze um damit die klimatischen
Nischen von Schadinsekten und ihren Wirtspflanzen zu untersuchen. Die räumliche Darstellung
dieser Nischen verschaffte neue Erkenntnisse über die potentielle Verbreitung von
Schadinsekten. Das erste Kapitel untersuchte die potentielle Verschiebung der
Verbreitungsgebiete von ausgewählten, als besonders gefährlich eingestuften Schadinsekten
und deren bekannten Wirtspflanzen mit korrelativen Modellen. Dafür evaluierte ich die
räumliche und zeitliche Dynamik von möglichen Interaktionen zwischen Schadinsekten und

6
 ZUSAMMENFASSUNG

ihren Wirtspflanzen mit einem Metaweb-Ansatz. Das zweite Kapitel verglich zwei
komplementäre Modellierungsansätze um Unterschiede zwischen den fundamentalen und den
realisierten Nischen von Schadinsekten und deren Verteilung entlang eines
Temperaturgradienten zu untersuchen. Das dritte Kapitel verwendete Deep Neural Network
Algorithmen um durch Insekten verursachte Schadsymptome nachzuweisen und deren
Auftreten mit meteorologischen Daten zu koppeln. Dieses Kapitel brachte einen
konzeptionellen Nachweis für die Verwendung von neuen Technologien hervor um das
Auftreten von Schadinsekten vorauszusagen.

Diese Dissertation zeigte, dass der Klimawandel in Europa eine Veränderung der
potentiellen Verbreitungsgebiete von Schadinsekten (Kapitel 1 und Kapitel 2) und deren
Wirtspflanzen (Kapitel 1) herbeiführt. Kapitel 1 verwendete einen Datensatz von 89
Schadinsekten die als Quarantäneorganismen eingestuft sind sowie 126 Kulturpflanzen und
Waldbaumarten um den hohen Druck von invasiven Schadinsekten auf den produzierenden
Pflanzenbau in Südeuropa unter derzeitigen klimatischen Bedingungen aufzuzeigen.
Gleichzeitig zeigte dieses Kapitel, dass die Landwirtschaft im Norden Europas von den
verbesserten klimatischen Bedingungen für Kulturpflanzen profitieren könnte. In diesen
Regionen werden jedoch die Erträge vermehrt durch Schadinsekten gefährdet, da dort das sich
verändernde Klima das Überleben neuer Schadinsekten ermöglicht. Das zweite Kapitel verglich
physiologische und korrelative Artenverbreitungsmodelle, welche trotz konzeptionellen
Unterschieden gut übereinstimmten und daher für zusätzliche Robustheit der
Schlussfolgerungen sorgten. Mit beiden Modellierungsansätzen konnte ich einen
Temperaturgrenzwert bestimmen, welcher einen plötzlich rasch ansteigenden Zuwachs der
Anzahl Schadinsekten mit geeigneten klimatischen Bedingungen begrenzt. Zusammen lieferten
die Resultate von Kapitel 1 und Kapitel 2 starke Argumente, dass neue invasive Schadinsekten
in der näheren Zukunft in Europa auftreten werden. Dadurch wird die Dringlichkeit für die
Entwicklung von neuen Pflanzenschutzmassnahmen unterstrichen, um besser mit zukünftigen
Invasionen umgehen zu können. Das dritte Kapitel zeigte, dass Big Data und Deep Learning
zusammen neue Lösungen für die Prognose von Schadinsekten hervorbringen können. Als
Beispiel zeigte ich, wie innovative Technologien für die Voraussage des Auftretens von
Schadinsekten genutzt werden können und somit neue Möglichkeiten für einen nachhaltigeren
Pflanzenschutz entstehen.

7
 ZUSAMMENFASSUNG

Zusammengefasst, tragen die drei Kapitel zu einem besseren Verständnis zum Einfluss
des Klimawandels auf die potentielle Verbreitung von Schadinsekten und deren Wirtspflanzen
bei. Diese Dissertation kombiniert umfassende Datensätze für Schadinsekten, Kulturpflanzen
und Waldbaumarten um Regionen mit erhöhtem Schädlingsdruck zu identifizieren und
präsentiert einen neuen Ansatz für die Schädlingsprognose. Die drei Kapitel liefern eine
wissenschaftliche Grundlage für die Verbesserung des Pflanzenschutzes und insbesondere der
Regulierung von Schadinsekten.

8
GENERAL INTRODUCTION

Evolution of food production

Agriculture has been central to the development of humankind, providing people with
food and fibre since the dawns of civilisation (Federico, 2008). Since the 1960s, crop production
drastically increased, supporting an acceleration of human population growth (Khush, 2001).
This rapid development after the green revolution owes to advances in biotechnology and the
development of more productive crop varieties (Pingali, 2012, Navarro, 2006; Evenson &
Gollin, 2003). With advancing globalisation, large-scale mechanisation and motorisation
helped to further intensify agricultural production (Knudsen et al., 2006). The expansion of
arable land and irrigated surface, as well as the development and application of chemical
fertilizers and pesticides to counteract against pests and pathogens, promoted further
intensification of agriculture (Knudsen et al., 2006; Carvalho, 2006; Carvalho, 2017; Popp, Petõ
and Nagy, 2013). However, the intensification of cropping systems fostered environmental
challenges concerning land-use change, disruption of the global nitrogen cycle and pesticide
use (Knudsen et al., 2006; Pimentel & Pimentel, 1990).

Covering the growing demand for food production in the near future without
jeopardizing the environment requires reduced pesticide use. Chemical pesticides have been
centre to the agricultural intensification in the last decades and allowed crop production to
increase to the current levels (Oerke, 2006). However, excessive use of chemical pesticides has
detrimental consequences on the environment and biodiversity, as well as for human health
(Geiger et al., 2010; Tegtmeier & Duffy, 2004) and entails additional costs such as negative
effects on non-target species (Lamichhane et al. 2015), the occurrence of pesticide resistance
(Georghiou & Mellon, 1983; Carvalho, 2006) and the erosion of biodiversity (Hooper et al.
2005). To mitigate these consequences, progress towards more sustainable and less intense
production is crucial (Horrigan, Lawrence & Walker, 2002). At the same time, challenges
concerning global food security are pressing, since about 14% percent of the world’s population
still suffers from hunger and the demand for food production is expected to increase by up to
70% (FAO, 2009), while human population is predicted to reach 9 billion in 2050 (Godfray et
al., 2010). To tackle these challenges and to feed the world’s population without unnecessarily
harming the environment, closing the current yield gap and more efficient food production must
be attained (Godfray et al., 2010). Besides a systematic shift towards plant-based diets (Shepon

9
 GENERAL INTRODUCTION

et al. 2018), decreasing food and water waste as well as using nutrients and land more efficiently
(West et al. 2014), an additional major challenge is cutting yield losses owed to pests and
pathogens (Chakraborty and Newton, 2011; Riegler, 2018).

The concept of sustainable intensification has emerged to address the dual challenge of
increasing food demand and sustainable production. The concept of sustainable intensification
includes processes that aim at increasing productivity of cropping systems while simultaneously
decreasing environmental impacts, for example by lowering pesticide inputs (Tilman et al.,
2011; Garnett et al., 2013; Godfray & Garnett, 2014; Pretty and Bharucha, 2014). Novel
technologies relying on modern information technology may support the goals of sustainable
intensification (Tilman et al., 2011). Additionally, social trends have emerged to counteract
against the excessive pesticide use and to support sustainable and local food production, aiming
at reducing environmental impacts associated to agricultural production and downstream
processing and marketing (Lamichhane et al., 2016; Feldmann & Hamm, 2015). Particularly,
in western countries, consumer awareness for food production is growing and social trends are
helping to tackle current challenges of food production (Falguera, Aliguer & Falguera, 2012).
However, sustainable food systems will be increasingly challenged by climate change (Smith
& Gregory, 2013) and local food production may fail to meet demand in regions where climate
becomes unsuitable for crop production (Herrero & Thornton, 2013). While there is a growing
demand for sustainable agriculture by decreasing pesticide use and preserving the environment,
current challenges in agriculture will be amplified and new challenges will arise under climate
change, threatening crop production directly and indirectly.

Food production under climate change

Human-induced climate change is strongly affecting agriculture (Lobell, Schlenker &

Costa-Roberts, 2011; Bindi & Olesen, 2011) and challenging global food production and food
security on various levels (Rosenzweig & Parry, 1994; Wheeler & Von Braun, 2013; Nelson et
al., 2009; Parry et al., 2004). Cropping systems are expected to undergo worldwide changes,
including shifts in spatial distribution due to changing patterns in temperature and precipitation
(Walther et al., 2002; Lindner et al., 2010). On a global level, abiotic factors have negative
impacts on crop production under climate change, but show regionally differing signals (Nelson
et al., 2009; Bindi & Olesen, 2011). In temperate regions, agriculture might benefit from
increased CO2 levels and prolonged growth periods (Maracchi, Sirotenko & Bindi, 2005;

10
 GENERAL INTRODUCTION

Tubiello, Soussana & Howden, 2007), leading to an increase in crop diversity, which will in
turn support the resilience of cropping systems (Lin, 2011). Meanwhile, increasing frequency
of droughts and other extreme weather events will threaten crop production in many regions
across the globe (Rosenzweig & Parry, 1994; Wheeler & Von Braun, 2013). For example,
Mediterranean regions are predicted to suffer from increasing heat and drought stress (Olesen
et al., 2011) while tropical regions are at high risk due to a combination of factors, including
competition for land and water, soil degradation and changes in temperature and precipitation
patterns and variability (Brown & Funk, 2008; Challinor & Wheeler, 2008; Challinor et al.,
2014). In fact, regions where the people are already vulnerable to hunger are expected to face
drastic impacts of climate change on food production (Rosenzweig & Parry, 1994; Wheeler &
Von Braun, 2013). Besides these abiotic impacts, biotic factors in form of insect pests and
pathogens are threatening the agricultural sector under climate change (Deutsch et al., 2018;
Battisti & Larsson, 2015; Gregory et al., 2009). With abiotic components of climate change,
negative future impacts from insect pests and pathogens will likely intensify existing challenges
and raise new problems for food production globally.

Insect pests under climate change

Pests and pathogens cause enormous pre- and post-harvest yield losses, which can each
sum up to 10-16% of total annual crop production (Oerke, 2006; Bebber, Ramotowksi & Gurr,
2014; Bradshaw et al., 2016). In particular, the negative effects from insect pests are expected
to worsen under climate change, inflicting even greater yield losses and costs associated to plant
protection (Deutsch et al., 2018; Riegler, 2018). Insects are ectotherm organisms with biological
rates that are highly sensitive to temperature, making them particularly susceptible to climate
change (Paaijmans et al., 2013). Major problematic effects of climate change on pest species
include increased metabolism, faster population growth, higher voltinism (i.e. more generations
per year), and the extension of their distribution range (Porter, Parry and Carter, 1991; Bale et
al., 2002; Deutsch et al., 2018). Warmer temperatures increase the performance of insects up to
a species-specific thermal maximum (Deutsch et al., 2008). In tropical regions, crossing of this
threshold may lead to a decrease in insect activity as species are presently close to their thermal
optimum and would eventually exceed it with further warming (Deutsch et al., 2008).
Meanwhile, in temperate regions insect populations are expected to benefit from climate
warming, as species approach their optimal thermal conditions (Deutsch et al., 2008; Frazier,
Huey & Berrigan, 2006). Additionally, survival increases in milder winters, allowing a faster

11
 GENERAL INTRODUCTION

population growth in the following spring (Das, Singh and Vennila, 2011). Further, faster
development rates under higher temperatures and prolonged growth periods enable the
development of more generations per year (Porter, Parry and Carter, 1991; Altermatt, 2010;
Forrest, 2016). Together, higher metabolism and coherent increase in food consumption of
individuals, combined with faster population growth will lead to greater crop yield losses
(Deutsch et al., 2018). At the same time, raising pest pressure from the shift in species
distributions and the spread of invasive pest species is expected (Bebber et al., 2013; Paini et
al., 2016).

The warming climate supports the spread of pest species to new regions (Porter, Parry
and Carter, 1991), as species track their climatic niche by moving towards newly available
climates in higher latitudes and altitudes (Parmesan & Yohe, 2003; Chen et al., 2011).
Particularly in higher latitude regions, minimum temperatures are expected to increase more
than average temperatures (IPCC, 2007), enlarging and shifting the area with suitable
conditions of many species towards higher latitudes (Bebber et al., 2013; Yan et al., 2017),
because minimum temperature is the limiting factor for insect distribution in many regions
(Hill, 1987; Jarošík et al., 2015). Recent examples for range expansion of pests linked to climate
change have been attributed to enhanced winter survival (e.g. Thaumetopoea pitycampa in
Europe; Battisti et al., 2005). While on regional scales, neo-native species (i.e. range expanding
without human involvement; Essl et al., 2019) are expected to increasingly affect agricultural
production in the near future, on a global scale, human-induced introductions bear even higher
risks of invasions (Jeschke & Strayer, 2005; Wallingford et al., 2020).

Invasive species put additional pest pressure on agriculture under climate change
(Pimentel et al., 2000; Ziska et al., 2011; Paini et al., 2016). The term invasive describes non-
native species that have successfully overcome the four main stages of the invasion process:
transport, introduction, establishment and spread (Blackburn et al., 2011). The main drivers of
successful pest invasions are propagule pressure (i.e. the number of dispersing individuals:
Lockwood et al., 2005), climatic suitability (Walther et al., 2009) and host availability (Niemelä
& Matteson, 1996). Propagule pressure is closely linked to international trade and travel
(Hulme, 2009; Roques, 2010; Bacon et al., 2012, Liebhold et al., 2006), promoting transport
and introduction of pest species to new areas. Meanwhile, climate change plays an important
role in lifting abiotic barriers for establishment and spread (Walther et al., 2009; Hulme, 2009;
Levine & D’Antonio, 2003). Regions that were previously unsuitable are becoming more
12
 GENERAL INTRODUCTION

suitable with climate change, supporting establishment and spread of non-native species after
the introduction to a new region (Robinet & Roques, 2010; Walther et al., 2009; Roques, 2010).
For example, Drosophila suzukii, a polyphagous fruit pest native to Southeast Asia, was
introduced to southern France and rapidly spread across Europe (Cini et al., 2014). Moreover,
climate change favours establishment and spread, as ecosystems become more vulnerable to
invasions due to increased disturbance, for instance through extreme weather events, creating
empty niche space (Walther et al. 2009). The number of invasive insect species has increased
in the recent past (Roques et al., 2009), but the species pool for potential invasive insects seems
far from being saturated (Seebens et al. 2017). With higher activity and spread of insect pests,
increasing need for plant protection measures in cropping systems is inevitable (Delcour,
Spanoghe & Uyttendaele, 2015). In addition, contemporary measures to prevent pest species
from invading new regions, including capacities for border surveillance and border control are
insufficient (Poland & Rassati 2019; Cook et al., 2011; Bacon et al. 2012), because resources
for plant protection are lacking (Flood, 2010). Therefore, collecting information on potential
spread of insect pests is necessary and spatial modelling helps to gain knowledge on the current
and future climatic suitability of these species.

Spatial modelling of insect pests

Information on the potential distribution of pest species and anticipation of new invasive
species support proactive development of strategies for pest management. To investigate the
distribution of pest species, two main types of modelling approaches have been applied:
inductive and deductive models (Venette et al., 2010; Tonnang et al., 2017; Hill & Thomson,
2015). On the one hand, inductive modelling approaches, applied in correlative models couple
the geographical distribution records of a species with environmental conditions in these
locations (Guisan & Zimmermann, 2000; Guisan & Thuiller, 2005; Elith and Leathwick, 2009).
These, often termed species distribution models (SDMs), are powerful tools to model the
realized niche of species (Guisan & Zimmermann, 2000; Guisan & Thuiller, 2005), and have
been widely used to investigate the potential distribution of invasive species (Thuiller et al.,
2005; Elith and Leathwick, 2009; Jiménez-Valverde et al., 2011; Barbet-Massin et al., 2018).
By projecting the modelled realized niche spatially, regions that share the environmental
conditions preferred by a species are identified, informing on the potential distribution of the
species under current and future climate conditions (Thuiller et al., 2005). On the other hand,
deductive modelling approaches are implemented by physiological models (or process-based

13
 GENERAL INTRODUCTION

models), relying on physiological thresholds to identify the fundamental climate niche of a

species (Venette et al., 2010; Kearney & Porter, 2004; Kearney & Porter, 2009). Thresholds
marking the thermal tolerance of a species are often used for phenological models (e.g. Schaub
et al., 2017), but can also be linked to spatial climate data to identify areas where a species
encounters conditions within its thermal tolerance (Kearney & Porter, 2004; Kearney & Porter,
2009). Further, inductive and deductive approaches can be combined, such as in the CLIMEX
model (Sutherst & Maywald, 1985; Sutherst et al. 2003), providing another approach for species
distribution modelling that has been widely used to investigate potential pest distributions
(Tonnang et al., 2017). The type of approach used to model a species’ distribution depends of
the data availability (Hill & Thomson, 2015) and both, inductive and deductive modelling
approaches have been successfully applied to map areas at risk and to anticipate future invasions
of insect pest species (e.g. Biber-Freudenberger et al., 2016; Fand et al., 2014).

Pest risk analyses are assessments conducted to inform decision makers on the
likelihood that a pest species will enter, establish, spread and cause economic damage in a
certain region, supporting the selection of risk-mitigation measures and management strategies
(Venette et al., 2010; Robinet et al., 2012). Information on potential pest distributions is critical
for developing such pest risk analyses (Venette et al., 2010; Eyre et al., 2012), and spatial
models are applied for pest risk mapping to inform on the climatic suitability of pests in the
region of interest (Eyre et al., 2012). As climate change is expected to increase the invasion
success of pest species, investigating common patterns of climate suitability of invasive species,
informing on spatial and temporal trends of potential invasion risk becomes increasingly
important (Bebber, Gurr & Holmes, 2014). Previous studies used species distribution models
to show that pest species will generally move towards higher latitudes and that an increasing
number of pest species will encounter suitable conditions in temperate regions (Yan et al.,
2017).

Patterns of changes in distribution of insect pest species have rarely been the centre of
attention in species distribution modelling studies (but for Yan et al., 2017), and this study did
not restrict pest distribution with biotic interactions in any form. In particular, the potential
distribution of known host plants, reflecting host availability for pests, has never been included
for future distribution modelling of insect pests. Host availability is a crucial determinant of
invasion risk, and neglecting host availability will increase uncertainty of invasion risk
predictions based on climatic factors alone (Niemelä & Matteson, 1996; Bacon et al., 2014).
14
 GENERAL INTRODUCTION

Plant distribution will undergo changes in the future (Walther et al., 2002; Tubiello, Soussana
& Howden, 2007) as crops are also limited by climatic factors (Connor, Loomis & Cassmann,
2011). Therefore, investigating the overlap between future distribution of pests and their host
plants provides valuable insights and more accurate predictions of areas experiencing pest
pressure in the future. Finally, climate change has shown to stimulate non-linear reactions in
ecological systems (Lenton et al., 2011), resulting in abrupt shifts to potentially irreversible
ecosystem states (Dakos et al., 2019), for instance the sudden collapse of pollinator populations
(Lever et al. 2014). However, little is known about the temporal dynamics of the shifting
patterns of pest distribution. Previous studies focused on projections to one or two future
scenarios (e.g. Yan et al., 2017). With higher temporal resolution, more information can be
obtained on potential linearity or non-linearity of changes. Investigating climatic niches of
insect pest species will contribute to a better understanding of spatial and temporal trends of
potential invasion risk.

Smart farming: seasonal pest forecasting and decision support systems

The field of precision agriculture concentrates on using information technology to

manage production systems in a high temporal and spatial precision (i.e. within-field precision),
aiming at increasing productivity, environmental quality of cropping systems and time- and
cost-efficient farm management (Pierce & Nowak, 1999; Gebbers & Adamchuk, 2010; Zhang,
Wang and Wang, 2002). In the last decades, modern information technologies have reached the
agricultural sector, fostering the development of new tools to overcome challenges concerning
sustainable crop production (Cox, 2002). For instance, such tools rely on sensors, drones or
satellites to monitor plant growth, plant nutrients and plant pest outbreaks (Klerkx, Jakku &
Labarthe, 2019). Smart farming is going a step further, using data to produce real-time decisions
on farm management (Sundmaeker et al., 2016). Further, big data are expected to support the
development of applications that address challenges in agriculture, including the efficiency of
the full supply chain (Wolfert et al., 2017). The development of novel applications based on
information technology particularly brings new opportunities for sustainable intensification
(Lindblom et al., 2017; El Bilali & Allahyari, 2018). For instance, unmanned aerial vehicles
with autonomous flight control (i.e. drones; Floreano & Wood, 2015) can be used to detect dry
sections in a field using hyperspectral cameras, allowing efficient use of water resources (Bacco
et al., 2018). Additionally, smart farming may make agriculture economically more profitable
for farmers, as input resources drop and farm management becomes more time- and cost-

15
 GENERAL INTRODUCTION

efficient (Walter et al., 2017). Plant protection may also benefit from new technologies in the
near future. With continuous implementation of drones and hyperspectral cameras, valuable
information on the condition of crops is collected (Walter et al., 2017) and deep learning
applications have great potential to analyse the resulting data, for instance for plant protection
purposes (Zhu et al., 2018). Big data analyses have a yet unexploited potential to overcome
current limitations in pest forecasting (Orlandini et al., 2020).

Developing improved decision support systems becomes more important as the demand
for sustainable crop production will grow and pest pressure increases under climate change.
Numerous models have been developed in the last decades to describe the occurrence and the
seasonal dynamics of pest populations (e.g. Welch et al., 1978, Prues, 1983, Schaub et al., 2005;
Samietz et al., 2007). Generally, such phenological models are based on thermal requirements
(e.g. degree-days) and species-specific development thresholds (Damos and Savopoulou-
Soultani, 2010) to describe non-linear relationships between temperature and insect
development rates obtained from controlled experiments (Kemp, 2019). Although phenological
models are often not designed for growers, consultants or extension services, some have been
developed for pest management purposes and implemented in decision support systems
(Samietz et al., 2007). These decision support systems often rely on phenological models and
meteorological data to forecast insect population development under given weather conditions
and aim to optimize the timing of management interventions (Strand, 2000; Olatinwo &
Hoogenboom, 2014). The implementation of decision support systems has strongly benefited
from improvement of computer technology, accessibility to meteorological data, and
dissemination through media (e.g. internet) (Olatinwo & Hoogenboom, 2014). For example,
SOPRA (www.sopra.admin.ch) is a decision support system for ten common pests of pome and
stone fruits in Switzerland, relying on phenological data obtained from climate chamber
experiments to simulate initial emergence and development of the relevant life stages over time,
depending on input weather data (Samietz et al., 2007). While such models are extremely
helpful to optimize timing, management and control measures (Samietz et al., 2007), some
limitations have been reported. First, phenological data on pest species are rare and often only
available for species that have been subject to research for a while (Hill & Thomson, 2015). To
collect new data in climate chamber experiments is labour- and time-intensive and therefore,
once established, phenological models are rarely updated. Other models are based on adult
trapping and temperature measurements in the field, resulting in even greater workloads (up to
four years for one species; Damos & Savopoulou-Soultani, 2010). However, climate change is

16
 GENERAL INTRODUCTION

affecting the phenology of insects and the synchrony of pest-host interactions (Kingsolver et
al., 2011). Hence, measured relationships between temperature and development, on which
phenological models rely, may be offset or outdated under climate change, resulting in
unreliable predictions. Second, forecasting models often target specific pest-crop systems and
are location-dependent (Donatelli et al., 2017). Current limitations lead to underuse of decision
support systems by growers and other stakeholders (Mir & Quadri, 2009) and underline demand
for new approaches. Compared to other branches, pest modelling seems to be lagging behind
in the implementation of state-of-the-art artificial intelligence applications but is expected to
bring new opportunities to pest forecasting together with big data (Orlandini et al., 2020). While
deep neural networks have been applied to different issues in agriculture (Kamilaris &
Prenafeta-Boldú, 2018), studies on the implementation to pest forecasting are lacking. Modern
digital technologies can generate enormous quantities of data by continuous monitoring of a
system (Kamilaris et al., 2017). Drones can generate large datasets (Floreano & Wood, 2015)
and machine learning, particularly deep learning algorithms, are powerful tools to analyse such
big data and extract valuable information from them (LeCun et al., 2015), for example to
identify plant pathogens (Mohanty et al., 2016). Applying new technologies based on machine
learning for pest forecasting may enhance the robustness of decision support systems and the
implementation of models for new species.

Thesis outline

This thesis aims to provide a better understanding of the impact of climate change on
insect pest distribution, and provides a proof of concept for the development of a new smart
farming tool to better anticipate pest and pathogen damages in agricultural systems. This thesis
is structured as follows:

Chapter 1 investigates how climate change affects the spatial dynamics of potential
interactions between managed plants and black-listed insect pests under present and future
climate conditions. In particular, I examine the potential interactions between black-listed insect
pests and their host plants using a metaweb approach to ask how the linkage and exposure of
crops and forest trees to insect pests may change with climate warming. I hypothesize that
increasing climatic suitability for both, insect pests and their host plants across Europe will lead
to larger areas of potential interaction and hence higher pest pressure posed by potentially
invasive pest species. This chapter highlights a simultaneous increase of crop suitability and

17
 GENERAL INTRODUCTION

pest pressure across Europe, indicating that novel pest invasions may hamper the arising
opportunities for crop cultivation under climate change. To attain more detailed insights on the
temporal dynamics of increasing pest pressure, detailed evaluation of climatic niches of pest
species is needed.

Chapter 2 explores potential pest accumulation in Europe under climate change, by

comparing the expectations of an inductive and a deductive modelling approach. By
investigating the potential distribution of a comprehensive set of insect pests, I explore whether
model predictions of the two approaches show agreeing patterns of pest climatic suitability
despite conceptual differences. Further, I investigate whether the climatic niches of these pests
are gradually distributed along a temperature gradient or rather different groups can be
identified. The evaluation of pest climatic niches contributed to the understanding of future pest
pressure on European plant production and allowed the detection of a temperature threshold
beyond which pest pressure is predicted to drastically increase. Together with Chapter 1, the
evidence for increasing pest pressure in Europe underlines the growing demand for
development of novel tools to support sustainable plant protection.

Chapter 3 presents a new framework for pest forecasting based on deep learning and
big data approaches. In particular, I evaluate in a case study whether deep neural networks can
be applied for pest damage recognition, and whether the occurrence of damages predicted by
deep neural networks coupled with meteorological data can inform on the phenology of pest
damages. This study strongly emphasizes on exploiting state-of-the-art technologies to develop
new approaches for pest forecasting tools and to overcome current data scarcity of agricultural
datasets.

Together, these three chapters highlight that pest pressure in Europe will increase under
climate change, threatening current efforts towards sustainable intensification, and that modern
technologies provide powerful tools to support the development of applications aimed at
making agricultural production more sustainable.

18
 GENERAL INTRODUCTION

References

Altermatt, F. (2010). Climatic warming increases voltinism in European butterflies and moths.
Proceedings of the Royal Society B: Biological Sciences, 277(1685), 1281-1287.
Bacco, M., Berton, A., Ferro, E., Gennaro, C., Gotta, A., Matteoli, S., ... & Zanella, A. (2018).
Smart farming: Opportunities, challenges and technology enablers. In 2018 IoT Vertical
and Topical Summit on Agriculture-Tuscany (IOT Tuscany) (pp. 1-6). IEEE.
Bacon, S. J., Bacher, S., & Aebi, A. (2012). Gaps in border controls are related to quarantine
alien insect invasions in Europe. PLoS ONE, 7(10), e47689.
Bacon, S. J., Aebi, A., Calanca, P., & Bacher, S. (2014). Quarantine arthropod invasions in
Europe: the role of climate, hosts and propagule pressure. Diversity and Distributions,
20(1), 84-94.
Bale, J. S., Masters, G. J., Hodkinson, I. D., Awmack, C., Bezemer, T. M., Brown, V. K., ... &
Good, J. E. (2002). Herbivory in global climate change research: direct effects of rising
temperature on insect herbivores. Global Change Biology, 8(1), 1-16.
Barbet-Massin, M., Rome, Q., Villemant, C., & Courchamp, F. (2018). Can species
distribution models really predict the expansion of invasive species? PloS ONE, 13(3),
e0193085.
Battisti, A., Stastny, M., Netherer, S., Robinet, C., Schopf, A., Roques, A., & Larsson, S.
(2005). Expansion of geographic range in the pine processionary moth caused by
increased winter temperatures. Ecological Applications, 15(6), 2084-2096.
Battisti, A., & Larsson, S. (2015). Climate change and insect pest distribution range. In
Björkman, C., Niemelä, P., (Eds.), Climate Change and Insect Pests (pp. 1-15), CABI:
Wallingford, UK.
Bebber, D. P., Ramotowski, M. A., & Gurr, S. J. (2013). Crop pests and pathogens move
polewards in a warming world. Nature Climate Change, 3(11), 985.
Bebber, D. P., Holmes, T., & Gurr, S. J. (2014). The global spread of crop pests and
pathogens. Global Ecology and Biogeography, 23(12), 1398-1407.
Biber-Freudenberger, L., Ziemacki, J., Tonnang, H. E., & Borgemeister, C. (2016). Future
risks of pest species under changing climatic conditions. PloS One, 11(4), e0153237.
Bindi, M., & Olesen, J. E. (2011). The responses of agriculture in Europe to climate change.
Regional Environmental Change, 11(1), 151-158.
Blackburn, T. M., Pyšek, P., Bacher, S., Carlton, J. T., Duncan, R. P., Jarošík, V., ... &
Richardson, D. M. (2011). A proposed unified framework for biological invasions.
Trends in Ecology & Evolution, 26(7), 333-339.
19
 GENERAL INTRODUCTION

Bradshaw, C. J., Leroy, B., Bellard, C., Roiz, D., Albert, C., Fournier, A., ... & Courchamp, F.
(2016). Massive yet grossly underestimated global costs of invasive insects. Nature
Communications, 7(1), 1-8.
Brown, M. E., & Funk, C. C. (2008). Food security under climate change. Science, 319(5863),
580-581.
Carvalho, F. P. (2006). Agriculture, pesticides, food security and food safety. Environmental
Science & Policy, 9(7-8), 685-692.
Carvalho, F. P. (2017). Pesticides, environment, and food safety. Food and Energy Security,
6(2), 48-60.
Chakraborty, S., & Newton, A. C. (2011). Climate change, plant diseases and food security:
an overview. Plant Pathology, 60(1), 2-14.
Challinor, A. J., & Wheeler, T. R. (2008). Crop yield reduction in the tropics under climate
change: processes and uncertainties. Agricultural and Forest Meteorology, 148(3), 343-
356.
Challinor, A. J., Watson, J., Lobell, D. B., Howden, S. M., Smith, D. R., & Chhetri, N.
(2014). A meta-analysis of crop yield under climate change and adaptation. Nature
Climate Change, 4(4), 287-291.
Chen, I. C., Hill, J. K., Ohlemüller, R., Roy, D. B., & Thomas, C. D. (2011). Rapid range
shifts of species associated with high levels of climate warming. Science, 333(6045),
1024-1026.
Cini, A., Anfora, G., Escudero-Colomar, L. A., Grassi, A., Santosuosso, U., Seljak, G., &
Papini, A. (2014). Tracking the invasion of the alien fruit pest Drosophila suzukii in
Europe. Journal of Pest Science, 87(4), 559-566.
Connor, D. J., Loomis, R. S., & Cassman, K. G. (2011). Crop Ecology: Productivity and
Management in Agricultural Systems, Cambridge University Press, New York, USA.
Cook, D. C., Fraser, R. W., Paini, D. R., Warden, A. C., Lonsdale, W. M., & De Barro, P. J.
(2011). Biosecurity and yield improvement technologies are strategic complements in
the fight against food insecurity. PLoS One, 6(10), e26084.
Cox, S. (2002). Information technology: the global key to precision agriculture and
sustainability. Computers and Electronics in Agriculture, 36(2-3), 93-111.
Dakos, V., Matthews, B., Hendry, A. P., Levine, J., Loeuille, N., Norberg, J., ... & De
Meester, L. (2019). Ecosystem tipping points in an evolving world. Nature Ecology &
Evolution, 3(3), 355-362.

20
 GENERAL INTRODUCTION

Damos, P. T., & Savopoulou-Soultani, M. (2010). Development and statistical evaluation of

models in forecasting moth phenology of major lepidopterous peach pest complex for
Integrated Pest Management programs. Crop Protection, 29(10), 1190-1199.
Das, D. K., Singh, J., & Vennila, S. (2011). Emerging crop pest scenario under the impact of
climate change–a brief review. Journal of Agricultural Physics, 11, 13-20.
Delcour, I., Spanoghe, P., & Uyttendaele, M. (2015). Literature review: Impact of climate
change on pesticide use. Food Research International, 68, 7-15.
Deutsch, C. A., Tewksbury, J. J., Huey, R. B., Sheldon, K. S., Ghalambor, C. K., Haak, D. C.,
& Martin, P. R. (2008). Impacts of climate warming on terrestrial ectotherms across
latitude. Proceedings of the National Academy of Sciences of the United States of
America, 105(18), 6668-6672.
Deutsch, C. A., Tewksbury, J. J., Tigchelaar, M., Battisti, D. S., Merrill, S. C., Huey, R. B., &
Naylor, R. L. (2018). Increase in crop losses to insect pests in a warming climate.
Science, 361(6405), 916-919.
Donatelli, M., Magarey, R. D., Bregaglio, S., Willocquet, L., Whish, J. P., & Savary, S.
(2017). Modelling the impacts of pests and diseases on agricultural systems.
Agricultural Systems, 155, 213-224.
El Bilali, H., & Allahyari, M. S. (2018). Transition towards sustainability in agriculture and
food systems: Role of information and communication technologies. Information
Processing in Agriculture, 5(4), 456-464.
Elith, J., & Leathwick, J. R. (2009). Species distribution models: ecological explanation and
prediction across space and time. Annual Review of Ecology, Evolution, and
Systematics, 40, 677-697.
Essl, F., Dullinger, S., Genovesi, P., Hulme, P. E., Jeschke, J. M., Katsanevakis, S., ... &
Rabitsch, W. (2019). A conceptual framework for range-expanding species that track
human-induced environmental change. BioScience, 69(11), 908-919.
Evenson, R. E., & Gollin, D. (2003). Assessing the impact of the Green Revolution, 1960 to
2000. Science, 300(5620), 758-762.
Eyre, D., Baker, R. H., Brunel, S., Dupin, M., Jarošík, V., Kriticos, D. J., ... & Worner, S.
(2012). Rating and mapping the suitability of the climate for pest risk analysis. EPPO
Bulletin, 42(1), 48-55.
Falguera, V., Aliguer, N., & Falguera, M. (2012). An integrated approach to current trends in
food consumption: Moving toward functional and organic products? Food Control,
26(2), 274-281.

21
 GENERAL INTRODUCTION

Fand, B. B., Tonnang, H. E., Kumar, M., Bal, S. K., Singh, N. P., Rao, D. V. K. N., ... &
Minhas, P. S. (2014). Predicting the impact of climate change on regional and seasonal
abundance of the mealybug Phenacoccus solenopsis Tinsley (Hemiptera:
Pseudococcidae) using temperature-driven phenology model linked to GIS. Ecological
Modelling, 288, 62-78.
Food and Agriculture Organisation of the United Nations (FAO) (2009). The state of food
insecurity in the world: Economic crises-impacts and lessons learned.
Federico, G. (2008). Feeding the World: an Economic History of Agriculture, 1800-2000.
Princeton University Press, Princeton, USA.
Feldmann, C., & Hamm, U. (2015). Consumers’ perceptions and preferences for local food: A
review. Food Quality and Preference, 40, 152-164.
Flood, J. (2010). The importance of plant health to food security. Food Security, 2(3), 215-
231.
Floreano, D., & Wood, R. J. (2015). Science, technology and the future of small autonomous
drones. Nature, 521(7553), 460-466.
Forrest, J. R. (2016). Complex responses of insect phenology to climate change. Current
Opinion in Insect Science, 17, 49-54.
Frazier, M. R., Huey, R. B., & Berrigan, D. (2006). Thermodynamics constrains the evolution
of insect population growth rates: “warmer is better”. The American Naturalist, 168(4),
512-520.
Garnett, T., Appleby, M. C., Balmford, A., Bateman, I. J., Benton, T. G., Bloomer, P., ... &
Herrero, M. (2013). Sustainable intensification in agriculture: premises and policies.
Science, 341(6141), 33-34.
Gebbers, R., & Adamchuk, V. I. (2010). Precision agriculture and food security. Science,
327(5967), 828-831.
Geiger, F., Bengtsson, J., Berendse, F., Weisser, W. W., Emmerson, M., Morales, M. B., ... &
Eggers, S. (2010). Persistent negative effects of pesticides on biodiversity and biological
control potential on European farmland. Basic and Applied Ecology, 11(2), 97-105.
Georghiou, G. P., & Mellon, R. B. (1983). Pesticide resistance in time and space. In Pest
Resistance to Pesticides (pp. 1-46). Springer, Boston, MA, USA.
Godfray, H. C. J., Beddington, J. R., Crute, I. R., Haddad, L., Lawrence, D., Muir, J. F., ... &
Toulmin, C. (2010). Food security: the challenge of feeding 9 billion people. Science,
327(5967), 812-818.

22
 GENERAL INTRODUCTION

Godfray, H. C. J., & Garnett, T. (2014). Food security and sustainable intensification.
Philosophical Transactions of the Royal Society B: Biological Sciences, 369(1639),
20120273.
Gregory, P. J., Johnson, S. N., Newton, A. C., & Ingram, J. S. (2009). Integrating pests and
pathogens into the climate change/food security debate. Journal of Experimental
Botany, 60(10), 2827-2838.
Guisan, A., & Zimmermann, N. E. (2000). Predictive habitat distribution models in ecology.
Ecological Modelling, 135(2-3), 147-186.
Guisan, A., & Thuiller, W. (2005). Predicting species distribution: offering more than simple
habitat models. Ecology Letters, 8(9), 993-1009.
Herrero, M., & Thornton, P. K. (2013). Livestock and global change: emerging issues for
sustainable food systems. Proceedings of the National Academy of Sciences of the
United States of America, 110(52), 20878-20881.
Hill, D.S. (1987). Agricultural Insect Pests of Temperate Regions and Their Control.
Cambridge University Press, Cambridge, UK.
Hill M. P., Thomson L. J., (2015). Species distribution modelling in predicting response to
climate change. In Björkman, C. & Niemelä, P. (Eds), Climate Change and Insect Pests
(pp 16–37). CABI: Wallingford, UK.
Hooper, D. U., Chapin, F. S., Ewel, J. J., Hector, A., Inchausti, P., Lavorel, S., ... & Schmid,
B. (2005). Effects of biodiversity on ecosystem functioning: a consensus of current
knowledge. Ecological Monographs, 75(1), 3-35.
Horrigan, L., Lawrence, R. S., & Walker, P. (2002). How sustainable agriculture can address
the environmental and human health harms of industrial agriculture. Environmental
Health Perspectives, 110(5), 445-456.
Hulme, P. E. (2009). Trade, transport and trouble: managing invasive species pathways in an
era of globalization. Journal of Applied Ecology, 46(1), 10-18.
IPCC, Climate Change. (2007). The physical science basis.
Jarošík, V., Kenis, M., Honěk, A., Skuhrovec, J., & Pyšek, P. (2015). Invasive insects differ
from non-invasive in their thermal requirements. PloS One, 10(6), e0131072.
Jeschke, J. M., & Strayer, D. L. (2005). Invasion success of vertebrates in Europe and North
America. Proceedings of the National Academy of Sciences of the United States of
America, 102(20), 7198-7202.

23
 GENERAL INTRODUCTION

Jiménez-Valverde, A., Peterson, A. T., Soberón, J., Overton, J. M., Aragón, P., & Lobo, J. M.
(2011). Use of niche models in invasive species risk assessments. Biological Invasions,
13(12), 2785-2797.
Kamilaris, A., Kartakoullis, A., & Prenafeta-Boldú, F. X. (2017). A review on the practice of
big data analysis in agriculture. Computers and Electronics in Agriculture, 143, 23-37.
Kamilaris, A., & Prenafeta-Boldú, F. X. (2018). Deep learning in agriculture: A survey.
Computers and Electronics in Agriculture, 147, 70-90.
Kearney, M., & Porter, W. P. (2004). Mapping the fundamental niche: physiology, climate,
and the distribution of a nocturnal lizard. Ecology, 85(11), 3119-3131.
Kearney, M., & Porter, W. (2009). Mechanistic niche modelling: combining physiological
and spatial data to predict species’ ranges. Ecology Letters, 12(4), 334-350.
Kemp, W. P. (2019). 24. Predictive Phenology Modeling in Rangeland Pest Management. In
Capinera, J. L. (Eds.), Integrated Pest Management On Rangeland: A Shortgrass
Prairie Perspective. CRC Press.
Khush, G. S. (2001). Green revolution: the way forward. Nature Reviews Genetics, 2(10),
815-822.
Kingsolver, J. G., Woods, H. A., Buckley, L. B., Potter, K. A., MacLean, H. & Higgins, J. K.
(2011) Complex life cycles and the responses of insects to climate change. Integrative
and Comparative Biology, 51, 719–732.
Klerkx, L., Jakku, E., & Labarthe, P. (2019). A review of social science on digital agriculture,
smart farming and agriculture 4.0: New contributions and a future research agenda.
NJAS-Wageningen Journal of Life Sciences, 90-91, 100315.
Knudsen, M. T., Halberg, N., Olesen, J. E., Byrne, J., Iyer, V., & Toly, N. (2006). Global
trends in agriculture and food systems. In Halberg, N. (Eds.), Global Development of
Organic Agriculture: Challenges and Prospects (pp 1-48), CABI Publishing.
Lamichhane, J. R., Barzman, M., Booij, K., Boonekamp, P., Desneux, N., Huber, L., ... &
Sarah, J. L. (2015). Robust cropping systems to tackle pests under climate change. A
review. Agronomy for Sustainable Development, 35(2), 443-459.
Lamichhane, J. R., Dachbrodt-Saaydeh, S., Kudsk, P., & Messéan, A. (2016). Toward a
reduced reliance on conventional pesticides in European agriculture. Plant Disease,
100(1), 10-24.
LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444.
Lenton, T. M. (2011). Early warning of climate tipping points. Nature Climate Change, 1(4),
201-209.

24
 GENERAL INTRODUCTION

Lever, J. J., van Nes, E. H., Scheffer, M., & Bascompte, J. (2014). The sudden collapse of
pollinator communities. Ecology Letters, 17(3), 350-359.
Levine, J. M., & D'Antonio, C. M. (2003). Forecasting biological invasions with increasing
international trade. Conservation Biology, 17(1), 322-326.
Liebhold, A. M., Work, T. T., McCullough, D. G., & Cavey, J. F. (2006). Airline baggage as
a pathway for alien insect species invading the United States. American Entomologist,
52(1), 48-54.
Lin, B. B. (2011). Resilience in agriculture through crop diversification: adaptive
management for environmental change. BioScience, 61(3), 183-193.
Lindblom, J., Lundström, C., Ljung, M., & Jonsson, A. (2017). Promoting sustainable
intensification in precision agriculture: review of decision support systems development
and strategies. Precision Agriculture, 18(3), 309-331.
Lobell, D. B., Schlenker, W., & Costa-Roberts, J. (2011). Climate trends and global crop
production since 1980. Science, 333(6042), 616-620.
Lockwood, J. L., Cassey, P., & Blackburn, T. (2005). The role of propagule pressure in
explaining species invasions. Trends in Ecology & Evolution, 20(5), 223-228.
Maracchi, G., Sirotenko, O., & Bindi, M. (2005). Impacts of present and future climate
variability on agriculture and forestry in the temperate regions: Europe. Climatic
Change, 70(1-2), 117-135.
Mir, S. A., & Quadri, S. M. K. (2009). Decision support systems: concepts, progress and
issues–A review. In Lichtfouse E. (Eds.), Climate Change, Intercropping, Pest Control
and Beneficial Microorganisms (pp. 373-399). Springer, Dordrecht, Netherlands.
Mohanty, S. P., Hughes, D. P., & Salathé, M. (2016). Using deep learning for image-based
plant disease detection. Frontiers in Plant Science, 7, 1419.
Navarro, M. (2006). How can agricultural and extension educators contribute to a successful
New Green revolution? Journal of Agricultural Education and Extension, 12(2), 83-96.
Nelson, G. C., Rosegrant, M. W., Koo, J., Robertson, R., Sulser, T., Zhu, T., ... & Magalhaes,
M. (2009). Climate change: Impact on agriculture and costs of adaptation. IFPRI Food
Policy Report No. 21.
Niemelä, P., & Mattson, W. J. (1996). Invasion of North American forests by European
phytophagous insects. BioScience, 46(10), 741-753.
Oerke, E. C. (2006). Crop losses to pests. The Journal of Agricultural Science, 144(1), 31-43.

25
 GENERAL INTRODUCTION

Olatinwo, R., & Hoogenboom, G. (2014). Weather-based pest forecasting for efficient crop
protection. In Abrol, D. P., (Eds.), Integrated Pest Management (pp. 59-78). Academic
Press, San Diego, CA, USA.
Olesen, J. E., Trnka, M., Kersebaum, K. C., Skjelvåg, A. O., Seguin, B., Peltonen-Sainio, P.,
... & Micale, F. (2011). Impacts and adaptation of European crop production systems to
climate change. European Journal of Agronomy, 34(2), 96-112.
Orlandini, S., Magarey, R. D., Park, E. W., Sporleder, M., & Kroschel, J. (2020). Methods of
agroclimatology: modeling approaches for pests and diseases. Agroclimatology: Linking
Agriculture to Climate, 60, 453-488.
Paaijmans, K. P., Heinig, R. L., Seliga, R. A., Blanford, J. I., Blanford, S., Murdock, C. C., &
Thomas, M. B. (2013). Temperature variation makes ectotherms more sensitive to
climate change. Global Change Biology, 19(8), 2373-2380.
Paini, D. R., Sheppard, A. W., Cook, D. C., De Barro, P. J., Worner, S. P., & Thomas, M. B.
(2016). Global threat to agriculture from invasive species. Proceedings of the National
Academy of Sciences of the United States of America, 113(27), 7575-7579.
Parmesan, C., & Yohe, G. (2003). A globally coherent fingerprint of climate change impacts
across natural systems. Nature, 421(6918), 37-42.
Pierce, F. J., & Nowak, P. (1999). Aspects of precision agriculture. In Advances in Agronomy
(Vol. 67, pp. 1-85). Academic Press, San Diego, CA, USA.
Pimentel, D., & Pimentel, M. (1990). Comment: Adverse environmental consequences of the
Green Revolution. Population and Development Review, 16, 329-332.
Pimentel, D., Lach, L., Zuniga, R., & Morrison, D. (2000). Environmental and economic costs
of nonindigenous species in the United States. BioScience, 50(1), 53-65.
Pingali, P. L. (2012). Green revolution: impacts, limits, and the path ahead. Proceedings of
the National Academy of Sciences of the United States of America, 109(31), 12302-
12308.
Poland, T. M., & Rassati, D. (2019). Improved biosecurity surveillance of non-native forest
insects: a review of current methods. Journal of Pest Science, 92(1), 37-49.
Popp, J., Pető, K., & Nagy, J. (2013). Pesticide productivity and food security. A review.
Agronomy for Sustainable Development, 33(1), 243-255.
Porter, J. H., Parry, M. L., & Carter, T. R. (1991). The potential effects of climatic change on
agricultural insect pests. Agricultural and Forest Meteorology, 57(1-3), 221-240.
Pretty, J., & Bharucha, Z. P. (2014). Sustainable intensification in agricultural systems.
Annals of Botany, 114(8), 1571-1596.

26
 GENERAL INTRODUCTION

Pruess, K. P. (1983). Day-degree methods for pest management. Environmental Entomology,

12(3), 613-619.
Riegler, M. (2018). Insect threats to food security. Science, 361(6405), 846.
Robinet, C., & Roques, A. (2010). Direct impacts of recent climate warming on insect
populations. Integrative Zoology, 5(2), 132-142.
Robinet, C., Kehlenbeck, H., Kriticos, D. J., Baker, R. H., Battisti, A., Brunel, S., ... & Kenis,
M. (2012). A suite of models to support the quantitative assessment of spread in pest
risk analysis. PLoS One, 7(10), e43366.
Rosenzweig, C., & Parry, M. L. (1994). Potential impact of climate change on world food
supply. Nature, 367(6459), 133-138.
Roques, A., Rabitsch, W., Rasplus, J. Y., Lopez-Vaamonde, C., Nentwig, W., & Kenis, M.
(2009). Alien terrestrial invertebrates of Europe. In Handbook of Alien Species in
Europe (pp. 63-79), vol. 3. Springer, Dordrecht, Netherlands.
Roques, A. (2010). Alien forest insects in a warmer world and a globalised economy: impacts
of changes in trade, tourism and climate on forest biosecurity. New Zealand Journal of
Forestry Science, 40, S77-S94.
Samietz, J., Graf, B., Höhn, H., Schaub, L., & Höpli, H. U. (2007). Phenology modelling of
major insect pests in fruit orchards from biological basics to decision support: the
forecasting tool SOPRA. EPPO Bulletin, 37(2), 255-260.
Schaub, L., Graf, B., & Butturini, A. (2005). Phenological model of pear psylla Cacopsylla
pyri. Entomologia Experimentalis et Applicata, 117(2), 105-111.
Schaub, L., Breitenmoser, S., Derron, J., & Graf, B. (2017). Development and validation of a
phenological model for the univoltine European corn borer. Journal of Applied
Entomology, 141(6), 421-430.
Seebens, H., Blackburn, T. M., Dyer, E. E., Genovesi, P., Hulme, P. E., Jeschke, J. M., ... &
Bacher, S. (2017). No saturation in the accumulation of alien species worldwide. Nature
Communications, 8(1), 1-9.
Shepon, A., Eshel, G., Noor, E., & Milo, R. (2018). The opportunity cost of animal based
diets exceeds all food losses. Proceedings of the National Academy of Sciences of the
United States of America, 115(15), 3804-3809.
Smith, P., & Gregory, P. J. (2013). Climate change and sustainable food production.
Proceedings of the Nutrition Society, 72(1), 21-28.
Strand, J. F. (2000). Some agrometeorological aspects of pest and disease management for the
21st century. Agricultural and Forest Meteorology, 103(1-2), 73-82.

27
 GENERAL INTRODUCTION

Sundmaeker, H., Verdouw, C., Wolfert, S., Pérez Freire, L. (2016). Internet of food and farm
2020. In Vermesan, O., Friess, P. (Eds.), Digitising the Industry - Internet of Things
Connecting Physical, Digital and Virtual Worlds (pp. 129-151), River Publishers,
Gistrup, Denmark.
Sutherst, R. W., & Maywald, G. F. (1985). A computerised system for matching climates in
ecology. Agriculture, Ecosystems & Environment, 13(3-4), 281-299.
Sutherst, R. W. (2003). Prediction of species geographical ranges. Journal of Biogeography,
30(6), 805-816.
Tegtmeier, E. M., & Duffy, M. D. (2004). External costs of agricultural production in the
United States. International Journal of Agricultural Sustainability, 2(1), 1-20.
Thuiller, W., Richardson, D. M., Pyšek, P., Midgley, G. F., Hughes, G. O., & Rouget, M.
(2005). Niche‐based modelling as a tool for predicting the risk of alien plant invasions
at a global scale. Global Change Biology, 11(12), 2234-2250.
Tilman, D., Balzer, C., Hill, J., & Befort, B. L. (2011). Global food demand and the
sustainable intensification of agriculture. Proceedings of the National Academy of
Sciences of the United States of America, 108(50), 20260-20264.
Tonnang, H. E., Hervé, B. D., Biber-Freudenberger, L., Salifu, D., Subramanian, S., Ngowi,
V. B., ... & Niassy, S. (2017). Advances in crop insect modelling methods-Towards a
whole system approach. Ecological Modelling, 354, 88-103.
Tubiello, F. N., Soussana, J. F., & Howden, S. M. (2007). Crop and pasture response to
climate change. Proceedings of the National Academy of Sciences of the United States
of America, 104(50), 19686-19690.
Venette, R. C., Kriticos, D. J., Magarey, R. D., Koch, F. H., Baker, R. H., Worner, S. P., ... &
De Barro, P. J. (2010). Pest risk maps for invasive alien species: a roadmap for
improvement. BioScience, 60(5), 349-362.
Wallingford, P. D., Morelli, T. L., Allen, J. M., Beaury, E. M., Blumenthal, D. M., Bradley,
B. A., ... & Ibáñez, I. (2020). Adjusting the lens of invasion biology to focus on the
impacts of climate-driven range shifts. Nature Climate Change, 10, 398-405.
Walter, A., Finger, R., Huber, R., & Buchmann, N. (2017). Opinion: Smart farming is key to
developing sustainable agriculture. Proceedings of the National Academy of Sciences of
the United States of America, 114(24), 6148-6150.
Walther, G. R., Post, E., Convey, P., Menzel, A., Parmesan, C., Beebee, T. J., ... & Bairlein,
F. (2002). Ecological responses to recent climate change. Nature, 416(6879), 389.

28
 GENERAL INTRODUCTION

Walther, G. R., Roques, A., Hulme, P. E., Sykes, M. T., Pyšek, P., Kühn, I., ... & Czucz, B.
(2009). Alien species in a warmer world: risks and opportunities. Trends in Ecology &
Evolution, 24(12), 686-693.
Welch, S. M., Croft, B. A., Brunner, J. F., & Michels, M. F. (1978). PETE: an extension
phenology modeling system for management of multi-species pest complex.
Environmental Entomology, 7(4), 487-494.
West, P. C., Gerber, J. S., Engstrom, P. M., Mueller, N. D., Brauman, K. A., Carlson, K. M.,
... & Siebert, S. (2014). Leverage points for improving global food security and the
environment. Science, 345(6194), 325-328.
Wheeler, T., & Von Braun, J. (2013). Climate change impacts on global food security.
Science, 341(6145), 508-513.
Wolfert, S., Ge, L., Verdouw, C., & Bogaardt, M. J. (2017). Big data in smart farming–a
review. Agricultural Systems, 153, 69-80.
Yan, Y., Wang, Y. C., Feng, C. C., Wan, P. H. M., & Chang, K. T. T. (2017). Potential
distributional changes of invasive crop pest species associated with global climate
change. Applied Geography, 82, 83-92.
Zhang, N., Wang, M., & Wang, N. (2002). Precision agriculture-a worldwide overview.
Computers and Electronics in Agriculture, 36(2-3), 113-132.
Zhu, N., Liu, X., Liu, Z., Hu, K., Wang, Y., Tan, J., ... & Guo, Y. (2018). Deep learning for
smart agriculture: Concepts, tools, applications, and opportunities. International Journal
of Agricultural and Biological Engineering, 11(4), 32-44.
Ziska, L. H., Blumenthal, D. M., Runion, G. B., Hunt, E. R., & Diaz-Soltero, H. (2011).
Invasive species and climate change: an agronomic perspective. Climatic Change,
105(1-2), 13-42.

29
GENERAL INTRODUCTION

30
CHAPTER I

Crop and forest pest metawebs shift towards increased linkage and
suitability overlap under climate change

Marc Grünig*1,2,3, Dominique Mazzi1, Pierluigi Calanca2, Dirk Nikolaus Karger4 and Loïc
Pellissier3,4

1
Agroscope, RD Plant Protection, Wädenswil, Switzerland
2
Agroscope, RD Agroecology and Environment, Zurich, Switzerland
3
ETH, Landscape Ecology, Zurich, Switzerland
4
Swiss Federal Research Institute WSL, Birmensdorf, Switzerland

Published in Communications Biology (2020), 3, 233

Doi: https://doi.org/10.1038/s42003-020-0962-9
Post-print version

31
 CHAPTER I

Abstract

Global changes pose both risks and opportunities to agriculture and forestry and
biological forecasts can inform future management strategies. Here, we investigate potential
land-use opportunities arising from climate change for these sectors in Europe, and risks
associated with the introduction and establishment of novel insect pests. Adopting a metaweb
approach including all interaction links between 126 crops and forest tree species and 89 black-
listed insect pest species, we show that the metawebs shift toward increased numbers of links
and overlap of suitable area under climate change. Decomposing the metaweb across regions
shows large saturation in southern Europe, while many novel interactions are expected for
northern Europe. In light of the rising consumer awareness about human health and
environmental impacts of food and wood production, the challenge will be to effectively exploit
new opportunities to create diverse local agriculture and forestry while controlling pest species
and reducing risks from pesticide use.

32
 CHAPTER I

Introduction

Global changes, including biological invasions and climate change, have already
affected human-managed ecosystems1 and are expected to continue to shape the productivity
and diversity of agricultural and forest landscapes 2,3,4,5. Agricultural and forest systems provide
a variety of food and manufacturing resources, which are central to the functioning of
societies6,7,8. European agriculture currently strives towards more sustainable management
practices, including enhanced local food production and reduced use of pesticides9,10. Climate
change might oppose these trends, and the design of innovative management practices will
require adaptations to new environmental conditions11. Agriculture and forestry are particularly
sensitive to abiotic changes12. Climate change may increase the productivity of crops and forest
trees, e.g. via positive responses to higher CO2 concentrations13, but also increase yield losses
from pests and pathogens14,15,16. Investigating future opportunities for crop cultivation and
forest management under impending new threats from pest species is therefore crucial for
addressing risks and opportunities in the agricultural and forestry sectors associated with future
climate change.

At the global scale, climate change is expected to decrease crop production and hamper
food security17,18. However, in some areas of Europe climate change may enhance productivity
and provide opportunities for diversifying agriculture and forestry19,20. Like their natural
counterparts, agricultural crop species and managed forest trees display an ecological niche of
climatic preference21, and the suitable area for growth is expected to shift with increasing
temperatures22. Particularly in northern regions, new opportunities for intensifying agricultural
and forest resource utilization are predicted for the future12,23. Cropping area is expected to
expand towards higher latitudes, raising productivity in Northern Europe24. Moreover, positive
in-situ effects, such as a prolonged growing season and increased CO2 fertilization, could boost
the productivity of agricultural systems24 and forests alike25,26. For instance, the distribution
range, production, quantity and quality of grapevines have been projected to benefit from
climate change, thanks to higher CO2 concentrations27. Overall, the beneficial effects related to
climate change are expected to provide new opportunities for crop and forest tree species in
some European regions, but these gains might be counteracted by greater risks from climate
extreme events26 and pest pressure28.

33
 CHAPTER I

Insect pests already inflict major costs to the agricultural and forestry sectors, and their
impact is predicted to increase under climate change14,15,29. Native and recently introduced alien
insect pest species cause major costs to agricultural and forest production annually30,31. Pre- and
post-harvest yield losses can each sum up to 10–16% of total annual crop production30. The
extra pressure from invasive pests associated with the globalization of trades is expected to
increase these costs further32. Whereas in the past the movement of species through commercial
networks and their establishment in new regions was hampered by climatic barriers33, future
climate change might lift abiotic barriers and enable the proliferation and spread of species22.
Additionally, milder winters will enable increased survival of more insect species at higher
latitudes29,34. Following recent warming and globalization, the number of newly established
alien species, including insect pests, has been rising in Europe35,36,37. For instance, the
polyphagous fruit pest Drosophila suzukii has successfully colonized Europe, and is already
causing large financial losses to growers38. In contrast to many native pests, for which effective
management practices are in place, invasive pests require the deployment of new, still largely
underdeveloped control measures. Anticipating the arrival of new pest species and
understanding their interactions with crops and managed forests is crucial for designing
management strategies for different invasion scenarios.

Here, we adopt a metaweb approach39,40 to study the present and future links and
exposure of managed plants with their novel pests under climate change. We expect that climate
change will promote: i) new opportunities for cropping and forest systems owing to an increase
of areas with suitable climate for growing more diverse crops; ii) higher pest pressure caused
by increasing feeding interactions from novel invasive pests on managed plants (increasing
number of links), iii) greater risks caused by larger overlaps of climatically suitable areas for
host plants and their pests (increasing exposure). We forecast future climatic suitability for 96
economically relevant crops and 30 forest tree species from Europe and 89 insect pest species
included in lists of the European Plant Protection Organisation (EPPO). The considered pests
are either recommended by experts to be regulated as quarantine species or have been recently
identified as posing a risk to the EPPO region (www.eppo.int). We use species distribution
modelling (SDM) and future climate scenarios in high spatial and temporal resolution to
forecast climatically suitable areas for all species. We investigate the potential for plant growing
under climate change within five categories (“fruit crops”, “vegetable crops”, “arable crops”,
“other crops” and “forest trees”). Coupling the metaweb with forecasted climatically suitable
areas, we predict how the linkage properties between host plants and pests, and the plant species

34
 CHAPTER I

exposure are affected by climate change. We further quantify pest pressure, as the number of
pests with suitable climatic habitat, for five categories of pests (“fruit pests”, “vegetable pests”,
“arable crop pests”, “polyphagous pests” and “forest pests”).

Results

Changing area of suitable climate for crops and forest trees

We predict that the area of suitable climate will increase for most crops and forest tree
species within Europe between 2020 and 2100. We estimate a median increase in the area with
suitable climate for crops from 1,925,265 km2 in 2020 to 2,790,484 km2 (+47%) under the
representative concentration pathway (RCP) 8.5 and 2,487,919 km2 (+27%) under the RCP4.5
scenario in 2100. For forest tree species, the median area of suitable climate increases from
4,225,050 km2 in 2020 to 4,366,851 km2 (+3%) under the RCP8.5 scenario, less than forecasted
under the RCP4.5 scenario (4,561,816 km2) (+8%) until the end of the century, because of the
smaller loss in southern Europe.

35
 CHAPTER I

Figure 1: Predicted shifts of climatic suitability for exemplary host plant and insect pest species. Grapes (Vitis
vinifera), maize (Zea mays) and European beech (Fagus sylvatica) show shifting climatic suitability towards higher
latitudes. Aleurocanthus spiniferus, a fruit pest, Helicoverpa zea, an arable crop pest and Ips pini, a forest pest,
also show northwards shifting climatic suitability under future climate conditions. Europe’s current climate already
provides suitable conditions for these pests. Red areas show climatic suitability loss from 2020 to 2100. Together
the red and the dark blue area show the modelled distribution in 2020. Projections under the RCP8.5 scenario are
shown here.
Images are licensed under the Creative Commons Attribution 3.0 Unported (https://creativecommons.org/license
s/by/3.0/deed.en/: Vitis vinifera: https://commons.wikimedia.org/wiki/File:Grapes_during_pigmentation.jpg; Ips
pini: https://commons.wikimedia.org/wiki/File:Ips_pini.jpg), or the Creative Commons Attribution-Share Alike
2.0
Generic (https://creativecommons.org/licenses/by- sa/2.0/deed.en; Zea mays: https://commons.wikimedia.org/wi
ki/File:Mahane_Yehuda_Market_(9629714152).jpg; Fagus sylvatica: https://commons.wikimedia.org/wiki/File:
Beech_(Fagus_sylvatica)_(19185865168).jpg; Helicoverpa zea: https://commons.wikimedia.org/wiki/File:Helic
overpa_zea_larva.jpg.) The image of Aleurocanthus spiniferus was offered as copyright free on
http://www.ces.csiro.au/aicn/name_s/b_164.htm. All images were cropped to the fitting extent but remained
otherwise unchanged.

Assuming a relationship between economic profit and climatically suitable areas, we

predict increased cultivation opportunities for 82 (RCP8.5), respectively 91 (RCP4.5) out of

36
 CHAPTER I

126 crop and forest tree species. Some of these species will have largely increased suitable
climatic area and thus scope for growing economic significance. For example, in Europe
soybeans currently have a gross production value of roughly 3.5 billion dollars (FAOSTAT41)
and their suitable climate area will increase by 190% under the RCP8.5 scenario (95% RCP4.5)
by the end of the century. We also predict an increase in the suitable area for many specialty
crops with high market values (e.g. RCP8.5: apple +29%, grapefruit +756%, lemon lime
+105%, melon +87%, tomato +42%; RCP4.5: apple +47%, grapefruit +225%, lemon lime 70%,
melon +50%, tomato +23%). Meanwhile, for other economically relevant crops, the suitable
climatic area within Europe is forecasted to decline substantially by 2100 under the RCP8.5
scenario (wheat -9%, maize -14% (Fig. 1), oats -44%, rye -76%, potatoes -20%), while more
limited under RCP4.5 (wheat +4%, maize +7%, oats -9%, rye -28%, potatoes +1%). Forest tree
species are predicted to lose suitable climatic area under steeper temperature increase (RCP8.5
Abies alba -73%, Fagus sylvatica -12% (Fig. 1), Picea abies -77%), but slower decline under
the RCP4.5 (Abies alba -36%, Fagus sylvatica +8%, Picea abies -39%). We investigated the
geographic differences in the change of climatically suitable areas. Our results for the five
European regions highlight that new opportunities for the exploitation of crops and forest trees
will open up, particularly in Northern Europe (RCP8.5: 48 species in 2020, +33 in 2100;
RCP4.5: 48 +16) and the British Isles (RCP8.5: 53 +28; RCP4.5: 53 +10). In Western (RCP8.5:
85 +9; RCP4.5: 83 +9) and Eastern Europe (RCP8.5: 90 +6; RCP4.5: 90 +7), slightly more
species are expected to encounter suitable climate in the future, while in Southern Europe
(RCP8.5: 101 -7; RCP4.5: 100 +0) the number of species with suitable climate is predicted to
decrease.

37
 CHAPTER I

Figure 2: Predicted dynamics of realized interactions between insect pests and their host plants for Southern (left),
and Northern (right) Europe under the RCP8.5 scenario. The interaction network for Southern Europe shows that
most interactions are already possible under current climate conditions. Icons show different categories of pests
(from bottom to top: “fruit pests”, “arable crop pests”, “vegetable pests”, “polyphagous pests” and “forest pests”)
and host plants (from left to right: “fruit crops”, “arable crops”, “vegetable crops”, “other crops” and “forest trees”).
“Fruit pests”, “polyphagous pests” and “forest pests” face the highest risk. Coloured points show the time step of
first potential overlap between each pair of host plant and pest. In contrast, the interaction network for Northern
Europe shows that many interactions become realizable only in the second half of the current century or not until
2100 (grey links). Interaction networks for other regions and RCP4.5 scenarios are shown in Supplementary Figure
1 – 5. Icons are pictures licensed under the Creative Commons CC0 1.0 Universal Public Domain Dedication
https://creativecommons.org/publicdomain/zero/1.0/deed.en), the Creative Commons Attribution-Share Alike 2.0
Generic licence (https://creativecommons.org/licenses/by-sa/2.0/deed.en), Creative Commons Attribution 3.0
Unported (https://creativecommons.org/licenses/by/3.0/deed.en) or Creative Commons Attribution-Share Alike
4.0 International (https://creativecommons.org/licenses/by-sa/4.0/deed.en). Images are available at the following
URLs: Forest pests: https://commons.wikimedia.org/wiki/File:Lymantor_coryli_(Perris,_1855)_Syn.- Triotemnu
s_coryli_(Perris,_1855)_(15286593562).png; Polyphagus pests: https://commons.wikimedia.org/wiki/File:Halyo
morpha_halys_s2a.jpg; Vegetable pests: https://commons.wikimedia.org/wiki/File:CSIRO_ScienceImage_7410_
A_larva_of_Helicoverpa_armigera_the_worlds_worst_insect_pest.jpg Arable crop pests: https://commons.wiki
media.org/wiki/File:Diabrotica_virgifera_LeConte,_1868.jpg; Apple: https://commons.wikimedia.org/wiki/File:
Manzana.svg; Pear: https://commons.wikimedia.org/wiki/File:Pear_icon.png; Arable crop: https://commons.wiki
media.org/wiki/File:Agriculture_- The_Noun_Project.svg; Tomato : https://commons.wikimedia.org/wiki/File:T
wemoji_1f345.svg; Eggplant: https://commons.wikimedia.org/wiki/File:Twemoji_1f346.svg; Coffee : https://co
mmons.wikimedia.org/wiki/File:Coffee_beans_by_gnokii.svg; Forest trees: https://commons.wikimedia.org/wiki
/File:Noun_883674_cc_Symbolon_tree_icon.svg

38
 CHAPTER I

Increasing linkage between plants and pests

We built a metaweb recording all known interactions between host plants and insect
pests for Europe. We constrained the metaweb with potential range suitability overlap in order
to quantify general changes in the incidence of pests on crops, under current and future climate
(Fig. 2). The measured overlap of modelled climatic suitability of host plants and pests indicates
increasing number of links and exposure (as mean overlap area per link) for Europe. The
metaweb filtered by suitability overlap under climate change indicates that by 2100, up to 80%
(RCP4.5: 79%) of links are predicted to be possible, notwithstanding large variation among
regions (i.e. Southern Europe, Western Europe, Northern Europe, Eastern Europe, British Isles;
Fig. 3) and time periods (2020 – 2100). In Southern Europe, 64% (RCP4.5: 63%) of the links
can already be realized under the current climate. In contrast, in Northern Europe currently only
7% (RCP4.5: 7%) of the links can currently be realized. This leaves a large potential for increase
by 2100, when up to 25% (RCP4.5: 15%) of all links become possible. Most interactions in
Europe affect “fruit crops” (RCP8.5: 251 +11; RCP4.5: 244 +31) and “forest trees” (RCP8.5:
176 -4; RCP4.5: 171 +8). Economically relevant crops will be affected by many more potential
pest species in Northern Europe towards the end of the century (maize: +7 links under RCP8.5/
+3 links under RCP4.5; wheat +4/+3, potato +6/+3, grapevine +5/+4). We find a stronger
increase in links per pest species in Northern regions. The numerous links illustrate that
Southern (RCP8.5: 595; RCP4.5: 586), Western (RCP8.5: 287; RCP4.5: 297) and Eastern
Europe (RCP8.5: 318; RCP4.5 297) are already potentially threatened by pest invasions under
the current climate. In contrast, Northern Europe (RCP8.5: +166 links; RCP4.5: +77) and the
British Isles (RCP8.5: +190; RCP4.5: +78) show a strong increase in network links under
climate change, indicating that crop diversification will come at the cost of higher pest pressure.
While in Southern Europe, the number of links per species decreases on average from 6.7 to
6.1, it strongly increases in Northern Europe (0.8 to 2.6) and the British Isles (1.1 to 3.2). The
predicted greater occurrence of generalist pests, such as the polyphagous Spodoptera frugiperda
and Helicoverpa zea contributes to the rising number of links and links per species under
climate change. While in Southern Europe the links with most of their host plants are already
possible under current climate, in Northern Europe, the number of links will drastically increase
for both of them (Spodoptera frugiperda: +19 links from 2020 to 2100 under RCP8.5, +10
RCP4.5; Helicoverpa zea: +21 RCP8.5, +14 RCP4.5). The increase in the incidence of
generalist species further causes a decrease in network specialization under climate change.
Additionally, the observed modular structure is predicted to be disrupted over time, where more
generalist pests will attack different categories of host plants. These trends are supported by
39
 CHAPTER I

other network metrics such as increasing partner diversity and increasing number of shared
partners for host plants and pests (Supplementary Figure 6).

Figure 3: Network properties for European regions. European regions (A: British Isles; B: Northern Europe; C:
Eastern Europe; D: Western Europe; E: Southern Europe; upper left). The graphs show properties derived from
metawebs of different regions. The number of links (red), exposure, as the mean overlap area per link in percent
increase (orange), host plants (green) and pests (blue) with suitable climate conditions in 5000 grid cells or more
are shown per time step (2020 – 2100) for all regions (A to E). Scales for the links are shown on the left side, for
host plants and pests on the right side on the upper panels, for exposure on the right side on the lower panels of
the figure. Thin lines show the results of the single GCM predictions and the thick lines the medians of the 4 GCMs
per scenario. Additional metrics (specialization, modularity, increasing partner diversity and increasing number of
shared partners) for host plants and pests are shown in the supplementary material (Supplementary Figure 6).

Area of suitability overlap increases with climate change

Coupling the metaweb with projected climatic suitability indicates shifts in the exposure
of managed plants to pests over time (Fig. 3). The mean area of overlap per link of the modelled

40
 CHAPTER I

climatic suitability of host plants and pests is predicted to increase by 51% (RCP4.5: 38%)
between 2020 and 2100 in Europe. The exposure increases most in Northern Europe (RCP8.5:
173%; RCP4.5: 75%) and the British Isles (RCP8.5: 165%; RCP4.5: 57%). Also for Western
(RCP8.5: 90%; RCP4.5: 43%) and Eastern Europe (RCP8.5: 60%; RCP4.5: 49%), we observe
a marked increase in contrast to Southern Europe, where we predict little changes (RCP8.5:
0%; RCP4.5: is 19%). For instance, the increasing climatic suitability for some pest species
promotes a marked increase of potential overlapping area with their host plants (Spodoptera
frugiperda: +176% RCP8.5; +70% RCP4.5; Helicoverpa zea +88% RCP8.5; +42% RCP4.5).
For different categories of host plants, we find similar patterns of slightly increasing area of
overlap in all regions but Southern Europe, where we predict decreasing area of overlap for
forest trees and arable crops (Supplementary Figure 7 and 8). For pest categories, we predict
that forest pests will overlap less with their host plants in the future, while all other categories
will have larger overlap of suitable areas towards the end of the century (Supplementary Figure
9 and 10). Although the climatic suitability drops for some crops, the area of overlap of these
crops with their pests is nevertheless predicted to increase, as found for maize (RCP8.5: +110%;
RCP4.5: +39%), wheat (RCP8.5: +135%; RCP4.5: +40%) and potatoes (RCP8.5: +80%;
RCP4.5: +44%).

41
 CHAPTER I

Figure 4: Climatic suitability shift for pests and host plants. Arrows show for each grid cell the average direction
of climatic suitability shift over all species. For each species, we calculated the direction from where each newly
suitable grid cell can be reached from its closest suitable grid cell in the previous time step. The length of the
arrows is proportional to the number of new colonisations of each grid cell. The coloured maps show the change
in total number of pests (top) and host plants (bottom) with suitable conditions during the time steps of 2020 –
2060 (left) and 2060 – 2100 (right). Red shadings indicate an increase of the number of species with suitable
climate; blue shadings indicate decreasing numbers. Climatic suitability shift and change in number of species are
shown for the RCP8.5 scenario (see Supplementary Figure 11 for RCP4.5).

Changing area of climate suitability for pests

Overall, the median area with suitable climate for insect pests (2,491,321 km2 in 2020)
will increase under climate change associated with a northward expansion of pest species. We
forecast an average increase in the suitable area for pest species of 294,176 km2 (+12%) under
the RCP8.5 scenario and 229,981 km2 (+9%) under the RCP4.5 scenario. Most of the

42
 CHAPTER I

considered pests already have suitable climatic conditions in Europe. In particular, Southern
Europe (RCP8.5: 71; RCP4.5: 71) is already threatened by many of the pests on the EPPO Lists.
In Northern Europe (RCP8.5: 26 +14; RCP4.5: 27 +5), Western Europe (RCP8.5: 43 +13;
RCP4.5: 44 +8), Eastern Europe (RCP8.5: 47 +11; RCP4.5: 45 +6) and the British Isles
(RCP8.5: 25 +17; RCP4.5: 26 +9), pest pressure will increase until 2100 (Fig. 3). Under the
RCP8.5 we predict increasing suitable climate area for 60 of 89 pest species (71 under RCP4.5),
and hence important potential expansions such as for the fall armyworm (Spodoptera
frugiperda), which will increase by 81% (3,341,038 km2 under current conditions; 51% under
RCP4.5), corresponding to additional 2,758,535 km2 (1,676,514 km2 under RCP4.5). We
quantified the dynamic shifts of climatic suitability for host plant and pest species from and to
each colonized grid cell in Europe between 2020 – 2060 and 2060 – 2100 (Fig. 4). The shift of
climatic suitability for host plants shows a gradient towards higher latitudes, underlining the
opportunities arising in northern regions in the second half of the century. Meanwhile, the shift
of climatic suitability for insect pest shows no clear south-north gradient, possibly because of
more complex and diverse climatic niche shapes of pest species. Consequently, the dynamic of
the shift of insect pest species is expected to be more idiosyncratic than that of their host plants
under climate change. We observe a slight decrease of the number of pest species in central and
northeastern regions, caused by the gap in climatic niches between cold-adapted pests and more
warm-adapted pests (Supplementary Figure 12). While cold-adapted species will move further
north with increasing temperatures, warm-adapted species are lacking behind. Finally, we
analysed the shift in centroid position of all modelled ranges of host plants and pest species by
measuring the direction and distance of the movement between 2020 and 2100 (Supplementary
Figure 13 and 14). The centroid analysis shows a median distance of 519 km, and speed of 6.5
km/year for pests under RCP8.5 (240 km; 3.0 km/year for RCP4.5), values that are consistent
with published estimates of dispersal capacity42. The analysis indicated a median distance of
588 km for host plants, resulting in a speed of 7.3 km/year (269 km; 3.4 km/year for RCP4.5).

Discussion

The metaweb approach adopted in our study indicates a general increase in susceptibility
of managed plants to pests under climate change owing to 1) an increase in the number of links
between crops, forest trees and their pests, and 2) an increase in the area of climatic suitability
overlap between pests and plants, which will challenge the benefit of climate change on
agricultural diversification. In contrast to the increase in potential distribution for single crops
43
 CHAPTER I

(e.g. for maize43), here we show a general pattern of increasing climatic suitability for a wide
variety of crops and forest tree species, indicating that climate change will favour diversification
of European plant production across different subsectors. While in Southern Europe future
climate will become increasingly unsuitable for staple crops like wheat, maize and potatoes,
future suitable areas are identified in northern countries, partly offsetting the loss in the South.
In northern European regions, the potential for growing more valuable crops and trees provides
scope for enhanced economic profit. For example, grapes are currently harvested on 3,429,137
ha in Europe, resulting in a gross production value of roughly 30 billion dollars (FAOSTAT41).
We forecast that the suitable climate for growing grapes will increase by 136% under the
RCP8.5 (71% under RCP4.5) by 2100. If the production area increases in proportion, gross
production values could increase by roughly 22 to 40 billion Euro. While the agronomic and
economic implications are far more complex, we highlight a wider range of opportunities for
growing crops across Europe under climate change. Since multiple crops cannot spread
simultaneously into new areas, the decision of realizing new crop potential will depend on the
market prices, consumer demands, regulatory frameworks and cultivation decisions at the farm
level.

Increasing climatic suitability and associated positive effects for host plants might be
offset by simultaneously increasing number of links between managed plants and pests at their
degree of exposure. By quantifying changes in interactions between pests and their host plants
under climate change within a metaweb, we demonstrated that increasing plant climatic
suitability is accompanied by increasing pest pressure across Europe. Using a metaweb
including 89 pests and 126 host plants allows to investigate the change in the system as a whole,
including a variety of climatic niches (Supplementary Figure 12) and all European regions. We
show how the web of agricultural pests and cultivated plants is forecasted to change, rather than
e.g. 43,44
predicting the suitable climate for single pests and crops . We found that polyphagous
pests will most expand the interaction area with their host plants, while for forest trees exposure
to pests decreases on average (Supplementary Figure 9). Generalist species affecting crops will
benefit most from warmer temperatures with a larger potential distribution and increasing
number of possible links. This is underlined by several metrics of the interaction network
(decreasing modularity and specialisation, and increasing partner diversity and number of
shared partners), showing that the average number of interactions per pest and the links of pests
with other categories of hosts will increase in most regions, while specialisation will decrease.
The invasion success, and thus the nature and extent of pest threats, and the damage caused

44
 CHAPTER I

depend on host availability45,46, while larger areas of overlap between plants and pests have a
larger potential for interactions and therefore larger population sizes and higher invasion risk47.
We observe a sharp rise of exposure to pests, especially in Northern Europe and the British
Isles. Although northern regions are predicted to benefit most from increasing climatic
suitability for crops and forest tree species under climate change19, they will also become more
targeted by pest invasions. Global changes in climatic suitability of pest species has been shown
with correlative models, indicating higher pest pressure in high latitude regions48, and rising
pest pressure has been associated with increasing metabolic rates of pests and therefore
increasing crop losses in warmer climates14. Our metaweb approach corroborates this trend, and
adds a thorough quantification of the nature of the risks in terms of link distribution and their
strength. Seizing new opportunities will require weighting the benefits of new exploitation
opportunities against the costs of co-occurrence of the novel crop or tree species and their
associated novel pest species, whereby the latter may also collaterally affect other host plants.

The general pattern of pest range shift to higher latitudes will likely be associated with
increasing yield losses and pest management costs48. Our approach to model the change in
climatic suitability with a high temporal resolution of climate change illustrates the potential
direction and speed at which species can be expected to spread. We show that under current
climate conditions, most species could invade parts of Southern and Western Europe, and from
there spread north-east with the changing climate. The predicted median speed of 6.5 km/year
for the RCP8.5 emission scenario (RCP4.5: 3.0 km/year) is well in line with previous estimates
for invasive insect species. Roques et al. (2016) estimated spreading rates of accidentally
introduced species of 3.5 km/year, but results varied widely among insect orders (e.g. 7 km/year
for Coleoptera)42. Assuming that the EPPO lists are proportionately representative for the
categories of pests in the pool of pest species, we predict that most interactions will occur for
pests of “fruit crops”, “vegetable crops” and “forest trees”, indicating that these are the crop,
respectively tree categories most jeopardized by pest invasions. The number of interceptions at
European borders between 1995 and 2004 show that Hemiptera (sub-order Sternorrhyncha),
Diptera and Coleoptera are the orders intercepted most often of all insect pests49. In our dataset,
60 species belong to these orders and 47 of these species are either “fruit pests”, “vegetable
pests” or “forest pests” (Supplementary Data 1). This indicates that our selection reliably
reflects current propagule pressure and that the number of included species allows representing
these different categories adequately. Further, we point out the difference in predicted pest
pressure between the two RCPs. The median area of the modelled distribution of pests and the

45
 CHAPTER I

median overlapping area under the RCP4.5 scenario increase much less than under the RCP8.5
scenario (Supplementary Figure 12). Northern regions might therefore suffer from fewer pest
invasions under the RCP4.5 than under the RCP8.5 scenario (e.g. Northern Europe: +14
RCP8.5, +5 RCP4.5; British Isles: +17 RCP8.5, +9 RCP4.5), thus corroborating the urgency of
policies aimed at restricting CO2 emissions in the near future.

An alarming implication of our results is that in large parts of Europe (i.e. mainly
Southern and Western Europe) many of the invasive pests included in our analysis can survive
under current climate conditions. In these regions, many host plants of these invasive pests can
already be grown, and most network links are thus feasible, highlighting that invasion risks are
an impending reality with the potential to severely disrupt the ecology and economics of
managed ecosystems. This finding underscores the urgency of rapidly deploying support to
phytosanitary services in Mediterranean countries. Pathways of insect pest invasions are often
associated with accidental introductions by international trade, cargo movement and individual
travel35,50,51. Once introduced and established in a new region, a pest might spread further to
other regions with suitable conditions. Interception statistics from cargo control show that live
plant imports bear an especially high risk of transporting insects52. Phytosanitary services have
strict regulations for the inspection and control of live vegetal goods45,53, which will become
even stricter in the European Union under the Regulation EU 2016/2031. However,
international trade and travel have reached such a large volume, that screening and inspecting
all potential routes of invasions is no longer feasible51. Finally, we emphasize the importance
of preparing for scenarios where pests overcome natural barriers by human-aided transport. As
noticed before, Southern Europe is already an entry gate for many subtropical pest species such
as Aleurocanthus spiniferus (Fig. 1).

In conclusion, we showed that the structure of the plant-pest metaweb will be altered
under climate change, favouring greater diversity of managed plants and incidence of pests,
especially of generalist ones. In Europe, climate change could overall have beneficial effects
on the diversity of crop production. However, to exploit this potential, it is crucial to monitor
and prepare for potential collateral risks of pest pressure. Pest pressure presents a severe threat
to European agriculture and forestry already under the current climate and will keep rising in
the future. Reaping the benefits from the newly arising opportunities while minimizing the costs
associated with the risks of climate change requires strong efforts and collaborations among all
stakeholders in the food and wood production chains.
46
 CHAPTER I

Methods

Data collection

We considered all crops for which distribution ranges are available from Earthstat54,55
(www.earthstat.org) and economically important forest tree species of Europe. We downloaded
distribution ranges for crops as raster from Earthstat54,55 and for forest trees as shapefiles from
EUFORGEN (www.euforgen.org) whenever available and presence records from GBIF
(www.gbif.org) for all other species. We only included crops and forest trees listed as host
plants for at least one of the pest species included in the EPPO plant quarantine lists (Alert, A1,
A2; www.eppo.int). Vectors of plant pathogens were not considered. Occurrence records for
pests were collected from various databases and from the published literature (see complete list
of host plants and pests in Supplementary Data 1 and 2). We considered only species reported
as present in fewer than five European countries in order to abstract from species already
established on the continent, and strengthen the focus on pests to be expected in the future. We
did not consider occurrences from stepping stones such as greenhouses and other structures that
provide protection from unsuitable climate and thus promote the proliferation and spread of
invasive species by enabling them to bridge unsuitable conditions and build up early
generations in spring (e.g. Tuta absoluta56). Coordinates of occurrence records were mapped
for each individual species and checked for unreasonable records by comparing with EPPO
PQR database (https://gd.eppo.int) distribution maps, which show for each country if a species
is present or absent. To prevent from multiple records per cell and reduce sampling bias, we
filtered the data with a minimum distance between each pair of records. With the remaining
occurrence records the geographic extent of the species range is represented as reliably as
possible (i.e. records in native and invasive range). To secure adequate SDM performance, we
excluded species with fewer than 24 occurrence records (8 records per explanatory variable).
In total, 128 host plant species and 94 insect pest species met these criteria.

Species distribution modelling

SDMs were calibrated using ensembles (unweighted average) of four widely used
modelling techniques (Generalized linear models (GLM), Generalized additive models (GAM),
Gradient boosting machine (GBM), RandomForest (RF)) or a subset. We used a pseudo-
absence approach, which is widely recognized as a solution for overcoming the lack of species

47
 CHAPTER I

absence data57. For each species, we randomly sampled 5,000 pseudo-absences from biomes in
which the species’ occurrence records lay. We down weighted the pseudo-absences to reach a
prevalence of 0.5. In a first step, we projected the models globally to check the potential
distribution under current climate visually. For future projections, we projected our models only
to Europe. To evaluate model performance, we used the area under the ROC-plot curve
(AUC)58,59 and true skill statistics (TSS)60. We used a split sample approach (70% calibration
data and 30% evaluation data) with 20 repetitions. Models were considered to have a reliable
performance with AUC scores > 0.761 and TSS values > 0.462. Models with AUC < 0.7 were
not included in the ensemble. Five pest species and two crop species with unsatisfying
evaluation metrics were excluded from the analysis (see Supplementary Data 3 – 6 for model
performances).

Using SDMs to model the climatic suitability of pests and plants is a common and
widely accepted approach. However, we are well aware of shortcomings when applying SDMs
to invasive species. Invasive species tend to occur in a broader climatic niche in their invasive
range than in their native range, for example because of the lack of natural enemies. This may
lead to an underestimation of the region of climatic suitability for pests when only the native
climatic niche is modelled. To overcome this caveat, we covered the native and invasive range
of pests by including distribution records from their entire known range whenever possible63
(see Supplementary Data 7 for more detailed description). Further, we omitted biotic factors
and dispersal limitations in our models. While we can assess the climatic niche of the species,
in reality their distribution may be constrained by these factors. For host plants, the soil
properties are also a major restricting factor. Therefore, we expect to overestimate potential
distributions and the changes in the network. However, this limitation should mostly impact
forest species because the movement of crops and pest species is affected by agronomic
decisions, and thus only partly dependent on dispersal abilities. Finally, irrigation has a large
impact on the distribution of many crop species. As precipitation patterns will differ from
current conditions under climate change, water scarcity may limit crop irrigation in many parts
of Europe during growth periods, restricting the distribution of crops. However, these changes
are difficult to predict and beyond the scope of this study. Here, we addressed the issue by
comparing models including both temperature and precipitation variables with models based
on temperature alone. We found good agreement between the two approaches and hence applied
SDMs based on only temperature variables to crops.

48
 CHAPTER I

Climate data and climate change scenarios

For historical climate data, we used the CHELSA V1.2 dataset64 (www.chelsa-
climate.org) with a 2.5 arcmin (~ 5 km) resolution. For future scenarios, we used model output
statistics in combination with mechanistic downscaling (the CHELSA algorithm) to calculate
mean monthly maximum and minimum temperatures, as well as monthly precipitation sums at
a ~ 5 km spatial resolution globally for the years 2006-2100. Projected future climate variables
were taken from four global circulation models (GCMs) driven by two scenarios of
representative concentration pathways (RCP4.5 and RCP8.5) in a factorial manner. The four
selected models originate from the CMIP5 collection of model runs used in IPCC's 5th
Assessment Report65. Different GCMs are, however, often based on similar code, and hence
generate similar output66,67. We therefore chose models characterized by only a small amount
of interdependence to allow for a good representation of uncertainty in climate projections.
Model selection was based on model interdependence in ensembles67. Data were taken from the
following four models: CESM1-BGC, run by the National Center for Atmospheric Research
(NCAR); CMCC-CM, run by the Centro Euro-Mediterraneo per i Cambiamenti Climatici
(CMCC); MIROC5, run by the University of Tokyo; and ACCESS1-3, run by the
Commonwealth Scientific and Industrial Research Organization (CSIRO) and the Bureau of
Meteorology (BOM), Australia.

We aggregated current climate data (1979-2020) in five-year time intervals, from which
we extracted climate data for all presence and absence records of pests, considering their
sampling year if available. Records older than 1979 were coupled with the first time step.
Records with no sampling year were coupled with an average of the historical data. We coupled
host plant records with a baseline of future climate (2006 – 2020) for each GCM and RCP.
Additionally, we considered the resolution of the presence records. For low-resolution records
(lower than 2.5 arc min), we extracted climate data from aggregated variable layers (5 arc min).
Presence records with a precision of less than 5 arc min were excluded. For model projections,
we aggregated time series of future climate (2011 – 2100) into 10-year time steps. For host
plant SDMs, we used subsets of the following five explanatory variables: mean annual
temperature, temperature seasonality, growing degree-days above 5°C, annual precipitation and
precipitation seasonality. In parallel, we ran models for all crop species without precipitation
variables. Due to crop irrigation, precipitation might be an unreliable predictor of the
distribution of crop species. We tested both approaches and found very similar results. For the

49
 CHAPTER I

final analysis, we thus used models based on temperature alone. For pest SDMs, we chose the
following variables: minimum temperature of the coldest month, growing degree-days above
5°C, annual precipitation and precipitation seasonality. For 15 pest species, we added
temperature seasonality to the explanatory variables and used a subset of the five variables to
reach better model performance. All explanatory variables were chosen based on ecological
significance68,69. We chose the variables based on ecological importance rather than statistical
information criteria70 (see Supplementary Data 1 and 2 for the variables included for each
species). Overall, we followed the recommendations to meet sufficient best practice standards
of species distribution modelling70.

For further analyses, we applied a binary classification of the climatic suitability to each
model output. We used the sensitivity-specificity sum maximization approach to define the
threshold that separates suitable from unsuitable climate71 (R package presenceAbsence
1.1.972). To apply binary classification to the ensembles, we used the average of the thresholds
of all individual models included in the ensemble. Further, in all models we restricted the area
of crop distribution with a cropland mask derived from Earthstat54,55. We analysed the number
of species per grid cell within Europe by overlaying binary model outputs of all species. We
measured the area of suitable climate for each species as the sum of the specific area of all cells
classified as suitable. For each European region (i.e. Southern, Western, Eastern and Northern
Europe and the British Isles), we calculated the number of species with suitable climate
conditions in at least 5,000 grid cells (representing about 10% of the median agricultural surface
of the different regions) per time step.

Metawebs

We generated a metaweb39,40 recording all known interactions between pests and their
host plants, which defines all possible network links. For each pest, we used a list of host plants
given in the EPPO database on quarantine pests (http://www.eppo.int). We investigated all
potential links for each time step by testing whether modelled distributions of host plants and
pests overlap in at least 5,000 grid cells. We did not consider host shifts that can occur in the
invaded range, climate-driven evolutionary change, or phenotypic plasticity in either host plants
or pests, their interactions and their interactions with other species73. We measured the area of
overlap in the same way as the area of suitable climate (see above). For each European region,
we counted the interactions that occur in at least one grid cell. To summarize the numbers from

50
 CHAPTER I

the different GCMs we calculated the median of each metric. We calculated exposure as the
mean of the overlap area of all links for each timestep. Modularity, specialization, partner
diversity and number of shared partners were calculated with the R package bipartite 2.1374.

Climatic niches

Climatic niches of host plants and pests were computed based on the worldwide
modelled distribution under current temperature. For each species, we isolated all grid cells
predicted as suitable by our models and obtained the annual minimum temperature for these
grid cells for current conditions. The range of occupied minimum temperatures was then used
as the climatic niche over minimum temperature. For the borders of minimum temperature in
Europe, we obtained the minimum and maximum value of the annual minimum temperature
raster of Europe for 2020, 2060 and 2100 as a mean of the four GCMs.

Spatial analysis of shift in climate suitability

To calculate the direction of shifts of climatic suitability, we extracted newly suitable

grid cells for each species and all 10-year time steps. For each newly colonized grid cell, we
identified the closest already occupied grid cell in the previous time step and measured the
direction from there to the focal cell in a 25 km resolution. We averaged the direction for each
grid cell over all species and counted the number of species colonizing each grid cell. We
displayed the direction in a 250 km resolution to highlight general regional patterns. Further,
we averaged the direction and number of colonization of the single time steps for 2020 - 2060
and 2060 – 2100. We did not apply a crop mask to the modelled distribution of crops for this
analysis because excluding grid cells led to unreliable averages during the interpolation of the
direction. The analysis of the shift of climatic suitability is a qualitative representation to
investigate general patterns and, therefore, small-scale inaccuracies might occur. Centroids
were calculated as the latitudinal mean and the longitudinal mean (weighted with the cosine of
latitude) of all occupied grid cells. All analyses were done in R version 3.5.175.

Statistics and Reproducibility

For producing random numbers, we applied the set.seed function of the R package base
(version 3.5.1)75 to enable reproducibility.

51
 CHAPTER I

Data availability

Climate data is available from www.chelsa-climate.org. Crops distribution maps

available from www.earthstat.org, forest distribution maps from www.euforgen.org. Pest
distribution records were gathered from published literature and databases and are available
from the corresponding author upon request. All source data underlying the Figures 1-4 are
available in Zenodo with the identifier https://doi.org/10.5281/zenodo.374610376

Code availability

The R codes for species distribution modelling, calculations of the results and analysis
are available from https://doi.org/10.5281/zenodo.374610376.

References

1. Lobell, D. B., Schlenker, W., & Costa-Roberts, J. Climate trends and global crop
production since 1980. Science, 1204531 (2011).
2. Bindi, M., & Olesen, J. E. The responses of agriculture in Europe to climate change.
Regional Environmental Change, 11(1), 151-158 (2011).
3. Seidl, R., Schelhaas, M. J., Rammer, W., & Verkerk, P. J. Increasing forest disturbances
in Europe and their impact on carbon storage. Nature Climate Change, 4(9), 806 (2014).
4. Rosenzweig, C., Iglesias, A., Yang, X.B., Epstein, P.R. and Chivian, E. Climate change
and extreme weather events; implications for food production, plant diseases, and pests.
Global Change and Human Health, 2(2), 90-104 (2001).
5. Challinor, A.J., et al. A meta-analysis of crop yield under climate change and adaptation.
Nature Climate Change, 4(4), 287 (2014)
6. Easterling, W. E., et al. Food, fibre and forest products. Climate Change, 273-313
(2007).
7. Tilman, D., Cassman, K. G., Matson, P. A., Naylor, R., & Polasky, S. Agricultural
sustainability and intensive production practices. Nature, 418(6898), 671 (2002).

52
 CHAPTER I

8. Endreny, T. A. Strategically growing the urban forest will improve our world. Nature
Communications, 9(1), 1160 (2018).
9. Lamichhane, J. R., Dachbrodt-Saaydeh, S., Kudsk, P., & Messéan, A. Toward a reduced
reliance on conventional pesticides in European agriculture. Plant Disease, 100(1), 10-
24 (2016).
10. Feldmann, C., & Hamm, U. Consumers’ perceptions and preferences for local food: A
review. Food Quality and Preference, 40, 152-164 (2015).
11. Howden, S.M. Adapting agriculture to climate change. Proceedings of the National
Academy of Sciences of the U.S.A., 104(50), 19691-19696 (2007).
12. Lindner, M., et al. Climate change impacts, adaptive capacity, and vulnerability of
European forest ecosystems. Forest Ecology and Management, 259(4), 698-709 (2010).
13. Cao, M., & Woodward, F. I. Net primary and ecosystem production and carbon stocks
of terrestrial ecosystems and their responses to climate change. Global Change Biology,
4(2), 185-198 (1998).
14. Deutsch, C. A., et al. Increase in crop losses to insect pests in a warming climate.
Science, 361(6405), 916-919 (2018).
15. Seidl, R., et al. Invasive alien pests threaten the carbon stored in Europe’s forests.
Nature Communications, 9(1), 1626 (2018).
16. Paini, D. R., et al. Global threat to agriculture from invasive species. Proceedings of the
National Academy of Sciences of the U.S.A., 113(27), 7575-7579 (2016)
17. Rosenzweig, C., & Parry, M. L. Potential impact of climate change on world food
supply. Nature, 367(6459), 133-138 (1994).
18. Wheeler, T., & Von Braun, J. Climate change impacts on global food security. Science,
341(6145), 508-513 (2013).
19. Maracchi, G., Sirotenko, O., & Bindi, M. Impacts of present and future climate
variability on agriculture and forestry in the temperate regions: Europe. Climatic
Change, 70(1-2), 117-135 (2005).
20. Schröter, D., et al. Ecosystem service supply and vulnerability to global change in
Europe. Science (2005).
21. Connor, D. J., Loomis, R. S., & Cassman, K. G. Crop ecology: productivity and
management in agricultural systems. (Cambridge University Press, New York, 2011).
22. Walther, G. R., et al. Ecological responses to recent climate change. Nature, 416(6879),
389 (2002).

53
 CHAPTER I

23. Olesen, J. E., & Bindi, M. Consequences of climate change for European agricultural
productivity, land use and policy. European Journal of Agronomy, 16(4), 239-262
(2002).
24. Tubiello, F. N., Soussana, J. F., & Howden, S. M. Crop and pasture response to climate
change. Proceedings of the National Academy of Sciences, 104(50), 19686-19690
(2007).
25. Bussotti, F., Pollastrini, M., Holland, V., & Brueggemann, W. Functional traits and
adaptive capacity of European forests to climate change. Environmental and
Experimental Botany, 111, 91-113 (2015).
26. Pretzsch, H., Biber, P., Schütze, G., Uhl, E., & Rötzer, T. Forest stand growth dynamics
in Central Europe have accelerated since 1870. Nature Communications, 5, 4967
(2014).
27. Bindi, M., Fibbi, L., & Miglietta, F. Free Air CO2 Enrichment (FACE) of grapevine
(Vitis vinifera L.): II. Growth and quality of grape and wine in response to elevated CO2
concentrations. European Journal of Agronomy, 14(2), 145-155 (2001).
28. Lamichhane, J. R., et al. Robust cropping systems to tackle pests under climate change.
A review. Agronomy for Sustainable Development, 35(2), 443-459 (2015).
29. Bebber, D. P., Ramotowski, M. A., & Gurr, S. J. Crop pests and pathogens move
polewards in a warming world. Nature Climate Change, 3(11), 985 (2013).
30. Oerke, E. C. Crop losses to pests. The Journal of Agricultural Science, 144(1), 31-43
(2006).
31. Pimentel, D., Zuniga, R., & Morrison, D. Update on the environmental and economic
costs associated with alien-invasive species in the United States. Ecological Economics,
52(3), 273-288 (2005).
32. Bradshaw, C. J., et al. Massive yet grossly underestimated global costs of invasive
insects. Nature Communications, 7, 12986 (2016).
33. Robinet, C., & Roques, A. Direct impacts of recent climate warming on insect
populations. Integrative Zoology, 5(2), 132-142 (2010).
34. Bale, Jeffery S., et al. Herbivory in global climate change research: direct effects of
rising temperature on insect herbivores. Global Change Biology, 8.1, 1-16 (2002).
35. Hulme, P. E. Trade, transport and trouble: managing invasive species pathways in an
era of globalization. Journal of Applied Ecology, 46(1), 10-18 (2009).

54
 CHAPTER I

36. Seebens, H., et al. No saturation in the accumulation of alien species worldwide. Nature
Communications, 8, 14435 (2017).
37. Bebber, D.P., Holmes, T. and Gurr, S.J. The global spread of crop pests and pathogens.
Global Ecology and Biogeography, 23(12), 1398-1407 (2014).
38. Mazzi, D., Bravin, E., Meraner, M., Finger, R., & Kuske, S. Economic impact of the
introduction and establishment of Drosophila suzukii on sweet cherry production in
Switzerland. Insects, 8(1), 18 (2017).
39. Dunne, J.A., 2006. The network structure of food webs. Ecological networks: linking
structure to dynamics in food webs, pp.27-86 (2006).

40.Albouy, C., et al. The marine fish food web is globally connected. Nature ecology &
evolution, 3(8), p.1153 (2019).
41. Food and Agriculture Organization of the United Nations. FAOSTAT Database
http://www.fao.org/faostat/en/ (2018).
42. Roques, A., et al. Temporal and interspecific variation in rates of spread for insect
species invading Europe during the last 200 years. Biological Invasions, 18(4), 907-920
(2016).
43. Ramirez-Cabral, N. Y., Kumar, L., & Shabani, F. Global alterations in areas of
suitability for maize production from climate change and using a mechanistic species
distribution model (CLIMEX). Scientific Reports, 7(1), 5910 (2017).
44. Arthur, F. H., Morrison III, W. R., & Morey, A.C. Modeling the potential range
expansion of larger grain borer, Prostephanus truncatus (Coleoptera: Bostrichidae).
Scientific Reports, 9, 6862 (2019).
45. Bacon, S. J., Aebi, A., Calanca, P., & Bacher, S. Quarantine arthropod invasions in
Europe: the role of climate, hosts and propagule pressure. Diversity and Distributions,
20(1), 84-94 (2014).
46. Hill, M. P., et al. Predicted decrease in global climate suitability masks regional
complexity of invasive fruit fly species response to climate change. Biological
Invasions, 18(4), 1105-1119 (2016).
47. Niemelä, P. & Mattson, W.J. Invasion of North American forests by European
phytophagous insects. BioScience, 46(10), pp.741-753 (1996).
48. Yan, Y., Wang, Y. C., Feng, C. C., Wan, P. H. M., & Chang, K. T. T. Potential
distributional changes of invasive crop pest species associated with global climate
change. Applied Geography, 82, 83-92 (2017).

55
 CHAPTER I

49. Kenis, M., Rabitsch, W., Auger-Rozenberg, M. A., & Roques, A. How can alien species
inventories and interception data help us prevent insect invasions? Bulletin of
Entomological Research, 97(5), 489-502 (2007).
50. Liebhold, A. M., Work, T. T., McCullough, D. G., & Cavey, J. F. Airline baggage as a
pathway for alien insect species invading the United States. American Entomologist,
52(1), 48-54 (2006).
51. Roques, A. Alien forest insects in a warmer world and a globalised economy: impacts
of changes in trade, tourism and climate on forest biosecurity. New Zealand Journal of
Forestry Science, 40(Suppl), S77-S94 (2010).
52. Liebhold, A. M., Brockerhoff, E. G., Garrett, L. J., Parke, J. L., & Britton, K. O. Live
plant imports: the major pathway for forest insect and pathogen invasions of the US.
Frontiers in Ecology and the Environment, 10(3), 135-143 (2012).
53. Bacon, S. J., Bacher, S., & Aebi, A. Gaps in border controls are related to quarantine
alien insect invasions in Europe. PLoS ONE, 7(10), e47689 (2012).
54. Ramankutty, N., Evan, A. T., Monfreda, C., & Foley, J. A. Farming the planet: 1.
Geographic distribution of global agricultural lands in the year 2000. Global
Biogeochemical Cycles, 22(1) (2008).
55. Monfreda, C., Ramankutty, N., & Foley, J. A. Farming the planet: 2. Geographic
distribution of crop areas, yields, physiological types, and net primary production in the
year 2000. Global Biogeochemical Cycles, 22(1) (2008).
56. Van Damme, V., et al. Overwintering potential of the invasive leafminer Tuta absoluta
(Meyrick) (Lepidoptera: Gelechiidae) as a pest in greenhouse tomato production in
Western Europe. Journal of Pest Science, 88(3), 533-541 (2015).
57. Stockwell, D. The GARP modelling system: problems and solutions to automated
spatial prediction. International Journal of Geographical Information Science, 13(2),
143-158 (1999).
58. Hanley, J. A., & McNeil, B. J. The meaning and use of the area under a receiver
operating characteristic (ROC) curve. Radiology, 143(1), 29-36 (1982).
59. Fielding, A. H., & Bell, J. F. A review of methods for the assessment of prediction errors
in conservation presence/absence models. Environmental Conservation, 24(1), 38-49
(1997).

56
 CHAPTER I

60. Allouche, O., Tsoar, A., & Kadmon, R. Assessing the accuracy of species distribution
models: prevalence, kappa and the true skill statistic (TSS). Journal of Applied Ecology,
43(6), 1223-1232 (2006).
61. Hosmer Jr, D. W., Lemeshow, S., & Sturdivant, R. X. Applied logistic regression (Vol.
398). (John Wiley & Sons, New Jersey, 2013).
62. Descombes, P., et al. Forecasted coral reef decline in marine biodiversity hotspots under
climate change. Global Change Biology, 21(7), 2479-2487 (2015).
63. Broennimann, O., et al. Measuring ecological niche overlap from occurrence and spatial
environmental data. Global ecology and biogeography, 21(4), pp.481-497 (2012).
64. Karger, D. N., et al. Climatologies at high resolution for the earth’s land surface areas.
Scientific Data, 4, 170122 (2017).
65. Stocker, T. F., et al. IPCC, 2013: Climate Change 2013: The Physical Science Basis.
Contribution of Working Group I to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change, 1535 pp (2013)
66. Knutti, R., & Sedláček, J. Robustness and uncertainties in the new CMIP5 climate
model projections. Nature Climate Change, 3(4), 369 (2013).
67. Sanderson, B. M., Knutti, R., & Caldwell, P. A representative democracy to reduce
interdependency in a multimodel ensemble. Journal of Climate, 28(13), 5171-5194
(2015).
68. Neuvonen S. and Virtanen T. 2015. Abiotic factors, climatic variability and forest insect
pests. In: Climate Change and Insect Pests. Eds: C. Björklund, P. Niemelä. Chapter 9
CAB International (CABI, Wallingford , 2015).
69. Olatinwo, R., & Hoogenboom, G. Weather-based pest forecasting for efficient crop
protection. Integrated Pest Management, 59-78 (2014).
70. Araújo, M. B., et al. Standards for distribution models in biodiversity assessments.
Science Advances, 5(1), p.eaat4858 (2019).
71. Liu, C., Berry, P. M., Dawson, T. P., & Pearson, R. G. Selecting thresholds of
occurrence in the prediction of species distributions. Ecography, 28(3), 385-393 (2005).
72. Freeman, E. A., & Moisen, G. PresenceAbsence: An R package for presence absence
analysis. Journal of Statistical Software. 23 (11), 31 (2008).
73. Bush, A., et al. Incorporating evolutionary adaptation in species distribution modelling
reduces projected vulnerability to climate change. Ecology Letters, 19(12), 1468-1478
(2016).

57
 CHAPTER I

74. Dormann, C.F., et al. Indices, graphs and null models: analyzing bipartite ecological
networks. The Open Ecology Journal, 2, 7-24 (2009).
75. R Core Team. R: A language and environment for statistical computing. R Foundation
for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/. (2018).
76. Grünig M., et al. Code for pest and crops species distribution modelling (Version 1).
Zenodo. https://doi.org/10.5281/zenodo.3746103 (2020)

Acknowledgements

This work contributes to the newly established Swiss National Centre for Climate
Services (www.nccs.ch) and received financial support by the Swiss Federal Office of
Agriculture. We would like to thank Olivier Benz (b-data-GmbH) for maintaining the
Agroscope RStudio-Server.

Author contributions

M.G. contributed to conceiving ideas, gathering distribution data, establishing and

running models, performing analysis and writing the manuscript; D.M. contributed to
conceiving ideas and writing; P.C contributed to conceiving ideas and writing; D.N.K. provided
climate change scenario data; L.P. contributed to conceiving ideas, establishing models and
writing. All authors gave final approval for publication.

58
 CHAPTER I

Supplementary material

Supplementary figures

Supplementary Figure 1: Evolution 2020-2100 of the interaction network for Southern (left) and Northern
Europe (right) under the RCP4.5 scenario. The geographic extent of the regions is defined in Fig. 3.

Supplementary Figure 2: Evolution 2020-2100 of the interaction network for the UK (left) and Western Europe
(right) under the RCP8.5 scenario.

59
 CHAPTER I

Supplementary Figure 3: Same as supplementary Fig. 2 but for the RCP4.5 scenario.

Supplementary Figure 4: Evolution 2020-2100 of the interaction network for Eastern Europe (left) and total
Europe (right) under the RCP8.5 scenario.

60
 CHAPTER I

Supplementary Figure 5: Same as Fig. S4 but for the RCP4.5 scenario

61
 CHAPTER I

Supplementary Figure 6: Same as Fig. 3 but for modularity and specialisation. Modularity was calculated with
the calculateModule function (R package bipartite 2.11). For specialization, we calculated the sum of all links for
each pest species and used the mean of all sums as a metric.

62
 CHAPTER I

Supplementary Figure 7: Change in the area of overlap over time for the five different host categories under the
RCP8.5 scenario. Panels from left top to right bottom: British Isles, Northern Europe, Eastern Europe, Western
Europe, Southern Europe.

Supplementary Figure 8: Same as fig. S7 but under RCP4.5 scenario.

63
 CHAPTER I

Supplementary Figure 9: Change in the area of overlap over time for the five different pest categories under the
RCP8.5 scenario. Panels from left top to right bottom: British Isles, Northern Europe, Eastern Europe, Western
Europe, Southern Europe.

64
 CHAPTER I

Supplementary Figure 10: Same as fig. S9 but under RCP4.5 scenario.

Supplementary Figure 11: Movement patterns of pests and host plants, same as figure 4 but for RCP4.5.

65
 CHAPTER I

Supplementary Figure 12: Niche distribution of pests (left) and host plants (right) over Min Temp. Blue boxes
show the niche of each pest. Vertical lines show the border of Min Temp conditions in Europe (red for 2020,
orange for 2060 and green for 2100). Dashed horizontal lines are borders of pest categories. From bottom to top:
fruit pests, arable crop pests, vegetable pests, polyphagous pests, forest pests.

66
 CHAPTER I

Supplementary Figure 13: Centroid shift between 2020 and 2100 in four time steps (2020 – 2040, 2040 – 2060,
2060 – 2080, 2080 – 2100) for the five different categories of host plants (left) and pests (right). Centroids were
calculated on the individual distribution maps under the RCP8.5 scenario.

Supplementary Figure 14: Same as figure S12 but for RCP4.5.

67
 CHAPTER I

Supplementary tables

Supplementary Table 1: List of pest species. Categories: 1 = Fruit pests, 2 = Arabale crop pests, 3 = Vegetable
pests, 4 = Polyphagous pests, 5 = Forest pests.

Species Order Cat Variable 1 Variable 2 Variable 3 Variable 4

Acleris variana Lepidoptera 5 Min Temp GDD An precip Precip season
Agrilus anxius Coleoptera 5 Min Temp GDD An precip Precip season
Agrilus planipennis Coleoptera 5 Min Temp GDD An precip Precip season
Aleurocanthus spiniferus Hemiptera 1 Min Temp GDD An precip Precip season
Aleurocanthus woglumi Hemiptera 1 Min Temp GDD An precip Precip season
Anastrepha fraterculus Diptera 1 GDD Temp season NA NA
Anastrepha ludens Diptera 1 Min Temp GDD An precip Precip season
Anastrepha obliqua Diptera 1 GDD Temp season NA NA
Anoplophora glabripennis Coleoptera 5 Min Temp GDD An precip Precip season
Anthonomus bisignifer Coleoptera 1 Min Temp GDD An precip Precip season
Anthonomus grandis Coleoptera 2 Min Temp GDD An precip Precip season
Anthonomus quadrigibbus Coleoptera 1 Min Temp GDD An precip Precip season
Anthonomus signatus Coleoptera 1 Min Temp GDD An precip Precip season
Bactericera cockerelli Hemiptera 3 Min Temp GDD An precip Precip season
Bactrocera cucumis Diptera 3 Min Temp GDD An precip Precip season
Bactrocera cucurbitae Diptera 3 GDD An precip NA NA
Bactrocera invadens Diptera 1 GDD An precip NA NA
Bactrocera latifrons Diptera 3 Min Temp GDD An precip Precip season
Bactrocera tryoni Diptera 1 Min Temp GDD An precip Precip season
Bactrocera zonata Diptera 1 Min Temp GDD An precip Precip season
Ceratitis rosa Diptera 1 GDD An precip NA NA
Choristoneura conflictana Lepidoptera 5 Min Temp GDD An precip Precip season
Choristoneura occidentalis Lepidoptera 5 Min Temp GDD An precip Precip season
Choristoneura rosaceana Lepidoptera 4 Min Temp GDD An precip Precip season
Conotrachelus nenuphar Coleoptera 1 Min Temp GDD An precip Precip season
Dacus ciliatus Diptera 3 GDD An precip NA NA
Dendroctonus adjunctus Coleoptera 5 GDD An precip NA NA
Dendroctonus brevicomis Coleoptera 5 Min Temp GDD An precip Precip season
Dendroctonus frontalis Coleoptera 5 Min Temp GDD An precip Precip season
Dendroctonus ponderosae Coleoptera 5 Min Temp GDD An precip Precip season
Dendroctonus pseudotsugae Coleoptera 5 Min Temp GDD An precip Precip season
Dendroctonus rufipennis Coleoptera 5 Min Temp GDD An precip Precip season
Dendrolimus superans Lepidoptera 5 Min Temp GDD An precip Precip season
Diabrotica barberi Coleoptera 2 Min Temp GDD An precip Precip season
Temp
Diabrotica speciosa Coleoptera 2 GDD An precip season NA
Diabrotica undecimpunctata
howardi Coleoptera 4 Min Temp GDD An precip Precip season
Diabrotica undecimpunctata
undecimpunctata Coleoptera 4 Min Temp GDD An precip Precip season
Diaphorina citri Hemiptera 1 Min Temp GDD An precip Precip season
Epitrix cucumeris Coleoptera 3 Min Temp GDD An precip Precip season
Epitrix subcrinita Coleoptera 3 Min Temp GDD An precip Precip season
Euwallacea fornicatus Coleoptera 5 Min Temp GDD An precip Precip season
Gnathotrichus sulcatus Coleoptera 5 GDD An precip NA NA

68
 CHAPTER I

Gonipterus scutellatus Coleoptera 5 Min Temp GDD An precip Precip season

Grapholita packardi Lepidoptera 1 Min Temp GDD An precip Precip season
Helicoverpa zea Lepidoptera 4 Min Temp GDD An precip Precip season
Heteronychus arator Coleoptera 3 GDD An precip NA NA
Homalodisca vitripennis Hemiptera 1 Min Temp GDD An precip Precip season
Ips calligraphus Coleoptera 5 Min Temp GDD An precip Precip season
Ips grandicollis Coleoptera 5 Min Temp GDD An precip Precip season
Ips pini Coleoptera 5 Min Temp GDD An precip Precip season
Leucinodes orbonalis Lepidoptera 3 Min Temp GDD An precip Precip season
Limonius californicus Coleoptera 3 Min Temp GDD An precip Precip season
Liriomyza sativae Diptera 4 Min Temp GDD An precip Precip season
Listronotus bonariensis Coleoptera 2 Min Temp GDD An precip Precip season
Lycorma delicatula Hemiptera 1 Min Temp GDD An precip Precip season
Lymantria mathura Lepidoptera 4 Min Temp GDD An precip Precip season
Maconellicoccus hirsutus Hemiptera 4 Min Temp GDD An precip Precip season
Malacosoma americanum Lepidoptera 1 Min Temp GDD An precip Precip season
Malacosoma disstria Lepidoptera 5 Min Temp GDD An precip Precip season
Massicus raddei Coleoptera 1 Min Temp GDD An precip Precip season
Megacopta cribraria Hemiptera 2 Min Temp GDD An precip Precip season
Melanotus communis Coleoptera 3 Min Temp GDD An precip Precip season
Metamasius hemipterus Coleoptera 4 Min Temp GDD An precip Precip season
Naupactus leucoloma Coleoptera 4 Min Temp GDD An precip Precip season
Naupactus xanthographus Coleoptera 1 Min Temp GDD An precip Precip season
Neodiprion abietis Hymenoptera 5 Min Temp GDD An precip Precip season
Oemona hirta Coleoptera 4 GDD An precip NA NA
Orgyia pseudotsugata Lepidoptera 5 Min Temp GDD An precip Precip season
Pissodes strobi Coleoptera 5 Min Temp GDD An precip Precip season
Platynota stultana Lepidoptera 2 Min Temp GDD An precip Precip season
Polygraphus proximus Coleoptera 5 Min Temp GDD An precip Precip season
Popilia japonica Coleoptera 4 Min Temp GDD An precip Precip season
Pseudacysta perseae Hemiptera 1 Min Temp GDD An precip Precip season
Rhagoletis mendax Diptera 1 Min Temp GDD An precip Precip season
Rhagoletis pomonella Diptera 1 Min Temp GDD An precip Precip season
Rhynchophorus palmarum Coleoptera 5 Min Temp GDD An precip Precip season
Saperda candida Coleoptera 1 Min Temp GDD An precip Precip season
Scirtothrips dorsalis Thysanoptera 4 Min Temp GDD An precip Precip season
Spodoptera eridania Lepidoptera 4 GDD An precip NA NA
Spodoptera frugiperda Lepidoptera 4 Min Temp GDD An precip Precip season
Spodoptera litura Lepidoptera 4 Min Temp GDD An precip Precip season
Strauzia longipennis Diptera 2 Min Temp GDD An precip Precip season
Temp
Tecia solanivora Lepidoptera 3 GDD An precip season NA
Thaumatotibia leucotreta Lepidoptera 1 Min Temp GDD An precip Precip season
Trioza erytreae Hemiptera 1 Min Temp GDD An precip Precip season
Xylosandrus compactus Coleoptera 1 Min Temp GDD An precip Precip season
Xylosandrus crassiusculus Coleoptera 4 Min Temp GDD An precip Precip season
Zaprionus indianus Diptera 1 GDD An precip NA NA
Zaprionus tuberculatus Diptera 1 GDD An precip NA NA

69
 CHAPTER I

Supplementary Table 2: List of host plant species. Categories: 1 = Fruit crops, 2 = Arabale crops, 3 = Vegetables,
4 = Other crops, 5 = Forest trees

Species Category Variable 1 Variable 2

almond 1 Min Temp GDD
apple 1 Min Temp GDD
apricot 1 Min Temp GDD
blueberry 1 Min Temp GDD
carob 1 Min Temp GDD
cherry 1 Min Temp GDD
cranberry 1 Min Temp GDD
gooseberry 1 Min Temp GDD
grape 1 Min Temp GDD
kiwi 1 Min Temp GDD
peachetc 1 Min Temp GDD
pear 1 Min Temp GDD
plum 1 Min Temp GDD
rasberry 1 Min Temp GDD
sourcherry 1 Min Temp GDD
strawberry 1 Min Temp GDD
tangetc 1 Min Temp GDD
fig 1 Min Temp GDD
avocado 1 Min Temp GDD
cashew 1 Min Temp GDD
citrusnes 1 Min Temp GDD
date 1 Min Temp GDD
grapefruitetc 1 Min Temp GDD
groundnut 1 Min Temp GDD
karite 1 Min Temp GDD
lemonlime 1 Min Temp GDD
mango 1 Min Temp GDD
olive 1 Min Temp GDD
orange 1 Min Temp GDD
papaya 1 Min Temp GDD
persimmon 1 Min Temp GDD
pineapple 1 Min Temp GDD
pistachio 1 Min Temp GDD
plantain 1 Min Temp GDD
quince 1 Min Temp GDD
areca 2 Min Temp GDD
barley 2 Min Temp GDD
bean 2 Min Temp GDD
broadbean 2 Min Temp GDD
buckwheat 2 Min Temp GDD
castor 2 Min Temp GDD
chickpea 2 Min Temp GDD
cowpea 2 Min Temp GDD
maize 2 Min Temp GDD
oats 2 Min Temp GDD
pea 2 Min Temp GDD

70
 CHAPTER I

pigeonpea 2 Min Temp GDD

rapeseed 2 Min Temp GDD
rice 2 Min Temp GDD
rye 2 Min Temp GDD
sorghum 2 Min Temp GDD
soybean 2 Min Temp GDD
triticale 2 Min Temp GDD
wheat 2 Min Temp GDD
alfalfa 3 Min Temp GDD
beetfor 3 Min Temp GDD
cabbage 3 Min Temp GDD
carrot 3 Min Temp GDD
cassava 3 Min Temp GDD
cauliflower 3 Min Temp GDD
chilleetc 3 Min Temp GDD
clover 3 Min Temp GDD
cucumberetc 3 Min Temp GDD
eggplant 3 Min Temp GDD
grassnes 3 Min Temp GDD
greenbean 2 Min Temp GDD
greencorn 2 Min Temp GDD
greenpea 2 Min Temp GDD
lettuce 3 Min Temp GDD
melonetc 3 Min Temp GDD
okra 3 Min Temp GDD
onion 3 Min Temp GDD
potato 3 Min Temp GDD
pumpkinetc 3 Min Temp GDD
spinach 3 Min Temp GDD
swedefor 3 Min Temp GDD
sweetpotato 3 Min Temp GDD
taro 3 Min Temp GDD
tomato 3 Min Temp GDD
turnipfor 3 Min Temp GDD
watermelon 3 Min Temp GDD
cinnamon 4 Min Temp GDD
cocoa 4 Min Temp GDD
coconut 4 Min Temp GDD
coffee 4 Min Temp GDD
cotton 4 Min Temp GDD
flax 4 Min Temp GDD
hemp 4 Min Temp GDD
kapokseed 4 Min Temp GDD
oilpalm 4 Min Temp GDD
pimento 4 Min Temp GDD
ramie 4 Min Temp GDD
rubber 4 Min Temp GDD
sugarcane 4 Min Temp GDD
sunflower 4 Min Temp GDD
tobacco 4 Min Temp GDD
Abies alba 5 An precip GDD

71
 CHAPTER I

Acer campestre 5 An precip GDD

Acer platanoides 5 An precip GDD
Acer pseudoplatanus 5 An precip GDD
Alnus glutinosa 5 An precip GDD
Betula pendula 5 An precip GDD
Castanea sativa 5 An precip GDD
Corylus avellana 5 An precip GDD
Eucalyptus globulus 5 An precip GDD
Fagus sylvatica 5 An precip GDD
Fraxinus excelsior 5 An precip GDD
Juglans regia 5 An precip GDD
Larix decidua 5 An precip GDD
Picea abies 5 An precip GDD
Pinus nigra 5 An precip GDD
Pinus pinaster 5 An precip GDD
Pinus pinea 5 An precip GDD
Pinus sylvestris 5 An precip GDD
Populus alba 5 An precip GDD
Populus nigra 5 An precip GDD
Populus tremula 5 An precip GDD
Prunus avium 5 An precip GDD
Pseudotsuga menziesii 5 An precip GDD
Quercus petraea 5 An precip GDD
Quercus robur 5 An precip GDD
Quercus suber 5 An precip GDD
Robinia pseudoacacia 5 An precip GDD
Salix alba 5 An precip GDD
Tilia cordata 5 An precip GDD
Ulmus laevis 5 An precip GDD

72
 CHAPTER I

Supplementary Table 3: Pest evaluation metrics: AUC scores and TSS values of the ensemble of each species
and average of all GCMs and RCPs.
Species name AUC TSS tss.threshold
Acleris variana 0.8009 0.5605 0.3307
Agrilus anxius 0.8535 0.6377 0.3962
Agrilus planipennis 0.959 0.8118 0.5347
Aleurocanthus spiniferus 0.8887 0.7315 0.386
Aleurocanthus woglumi 0.9124 0.7883 0.3656
Anastrepha ludens 0.8677 0.6808 0.4299
Anoplophora glabripennis 0.8783 0.6562 0.4054
Anthonomus bisignifer 0.9737 0.9129 0.377
Anthonomus grandis 0.8904 0.6899 0.402
Anthonomus quadrigibbus 0.7682 0.4983 0.3979
Anthonomus signatus 0.7912 0.6034 0.4516
Bactericera cockerelli 0.9255 0.7254 0.478
Bactrocera cucumis 0.9249 0.7757 0.4612
Bactrocera latifrons 0.8174 0.5482 0.436
Bactrocera tryoni 0.9302 0.7687 0.5479
Bactrocera zonata 0.8493 0.5571 0.4596
Choristoneura conflictana 0.8087 0.5833 0.3609
Choristoneura occidentalis 0.8832 0.6679 0.463
Choristoneura rosaceana 0.8879 0.6393 0.4176
Conotrachelus nenuphar 0.859 0.6467 0.3586
Dendroctonus brevicomis 0.8823 0.6885 0.4099
Dendroctonus frontalis 0.89 0.7075 0.3694
Dendroctonus ponderosae 0.7966 0.499 0.4323
Dendroctonus pseudotsugae 0.8363 0.5779 0.4242
Dendroctonus rufipennis 0.8442 0.5657 0.4356
Dendrolimus superans 0.8973 0.7921 0.4996
Diabrotica barberi 0.9824 0.9121 0.3881
Diabrotica undecimpunctata howardi 0.8505 0.599 0.3981
Diabrotica undecimpunctata undecimpunctata 0.9766 0.8907 0.4419
Diaphorina citri 0.8584 0.6093 0.4361
Epitrix cucumeris 0.9052 0.7346 0.4392
Epitrix subcrinita 0.9608 0.848 0.428
Euwallacea fornicatus 0.8042 0.6321 0.4829
Gonipterus scutellatus 0.8985 0.7028 0.4125
Grapholita packardi 0.8936 0.7149 0.4467
Helicoverpa zea 0.8702 0.5821 0.4816
Homalodisca vitripennis 0.9713 0.8453 0.4206
Ips calligraphus 0.7984 0.5481 0.4888
Ips grandicollis 0.9177 0.7105 0.4496
Ips pini 0.8989 0.6819 0.5206
Leucinodes orbonalis 0.9097 0.7505 0.5007
Limonius californicus 0.8326 0.5638 0.4392
Liriomyza sativae 0.8988 0.6747 0.4077

73
 CHAPTER I

Listronotus bonariensis 0.8877 0.7082 0.3901

Lycorma delicatula 0.9666 0.8468 0.3582
Lymantria mathura 0.9381 0.7945 0.44
Maconellicoccus hirsutus 0.9388 0.7754 0.3846
Malacosoma americanum 0.9725 0.8425 0.6019
Malacosoma disstria 0.9462 0.75 0.5034
Massicus raddei 0.9281 0.8768 0.5304
Megacopta cribraria 0.9843 0.9221 0.4146
Melanotus communis 0.8533 0.6871 0.4449
Metamasius hemipterus 0.9937 0.9545 0.4083
Naupactus leucoloma 0.9263 0.7265 0.4179
Naupactus xanthographus 0.9322 0.7718 0.4723
Neodiprion abietis 0.9229 0.7877 0.3857
Orgyia pseudotsugata 0.7699 0.5643 0.3768
Pissodes strobi 0.8304 0.6493 0.3558
Platynota stultana 0.9607 0.8794 0.5311
Polygraphus proximus 0.9809 0.9236 0.4528
Popilia japonica 0.9773 0.8585 0.5781
Pseudacysta perseae 0.9263 0.7942 0.4091
Rhagoletis mendax 0.8488 0.6429 0.4137
Rhagoletis pomonella 0.9195 0.7233 0.4141
Rhynchophorus palmarum 0.9717 0.8825 0.3985
Saperda candida 0.8577 0.6631 0.4414
Scirtothrips dorsalis 0.786 0.5174 0.3705
Spodoptera frugiperda 0.8917 0.6621 0.4795
Spodoptera litura 0.9135 0.7054 0.5139
Strauzia longipennis 0.866 0.6499 0.4057
Thaumatotibia leucotreta 0.905 0.7377 0.3684
Trioza erytreae 0.9284 0.784 0.416
Xylosandrus compactus 0.9607 0.8687 0.498
Xylosandrus crassiusculus 0.886 0.6758 0.4881
Anastrepha fraterculus 0.7829 0.4472 0.4493
Anastrepha obliqua 0.7263 0.3435 0.4555
Bactrocera invadens 0.7962 0.4672 0.4793
Ceratitis rosa 0.8458 0.54 0.4835
Dacus ciliatus 0.8287 0.5295 0.4757
Diabrotica speciosa 0.8051 0.5333 0.5333
Gnathotrichus sulcatus 0.7851 0.4825 0.4061
Heteronychus arator 0.9425 0.8274 0.418
Oemona hirta 0.9361 0.8192 0.496
Spodoptera eridania 0.8251 0.6246 0.3625
Tecia solanivora 0.9694 0.9087 0.415
Bactrocera cucurbitae 0.8335 0.5694 0.419
Dendroctonus adjunctus 0.7264 0.4053 0.3989
Zaprionus tuberculatus 0.7311 0.4425 0.3991
Zaprionus indianus 0.8133 0.5389 0.4392

74
 CHAPTER I

Supplementary Table 4: Host evaluation metrics: AUC scores and TSS values of the ensemble of each species
and average of all GCMs and RCPs.

Species name AUC TSS tss.threshold

almond 0.9291 0.7611 0.5451
apple 0.8828 0.5979 0.5503
apricot 0.8483 0.5454 0.5076
blueberry 0.8881 0.6556 0.5673
carob 0.9227 0.7268 0.5467
cherry 0.8437 0.5492 0.5266
cranberry 0.9733 0.8911 0.6226
gooseberry 0.9392 0.7915 0.4983
grape 0.8938 0.6564 0.5425
kiwi 0.8742 0.6504 0.553
peachetc 0.8639 0.5998 0.5306
pear 0.8836 0.6094 0.5779
plum 0.8563 0.5903 0.4539
rasberry 0.9197 0.7147 0.553
sourcherry 0.9333 0.7456 0.5799
strawberry 0.8822 0.6561 0.5653
tangetc 0.8033 0.4769 0.5587
fig 0.8556 0.5845 0.5772
avocado 0.8431 0.5828 0.4794
cashew 0.7784 0.449 0.4969
citrusnes 0.815 0.536 0.5589
date 0.8724 0.603 0.5114
grapefruitetc 0.8653 0.5636 0.5493
groundnut 0.8212 0.5143 0.5188
karite 0.8295 0.5666 0.5074
lemonlime 0.8047 0.4811 0.5496
mango 0.8171 0.4927 0.5663
olive 0.9352 0.7516 0.5485
orange 0.8673 0.6053 0.5353
papaya 0.8127 0.4782 0.5343
persimmon 0.8869 0.6183 0.4864
pineapple 0.7534 0.3707 0.5132
pistachio 0.8864 0.6568 0.5697
plantain 0.8018 0.4828 0.4836
quince 0.8674 0.6204 0.5843
areca 0.7822 0.4615 0.4727
barley 0.891 0.6567 0.5082
bean 0.7407 0.3617 0.4765
broadbean 0.8056 0.5021 0.5563
buckwheat 0.8378 0.5625 0.5036
castor 0.9282 0.7145 0.4904

75
 CHAPTER I

chickpea 0.8225 0.5076 0.447

cowpea 0.8659 0.5851 0.5869
maize 0.8489 0.5802 0.5042
oats 0.8709 0.6434 0.5126
pea 0.8697 0.6166 0.4521
pigeonpea 0.8617 0.5791 0.5707
rapeseed 0.8556 0.6071 0.4733
rice 0.7901 0.4548 0.5469
rye 0.9281 0.7368 0.5923
sorghum 0.8143 0.4683 0.5284
soybean 0.8417 0.5255 0.5229
triticale 0.909 0.7022 0.5739
wheat 0.8113 0.4874 0.4897
alfalfa 0.8373 0.5539 0.5094
beetfor 0.8822 0.6521 0.4569
cabbage 0.8804 0.6097 0.5397
carrot 0.8431 0.5678 0.5116
cassava 0.7212 0.3409 0.5315
cauliflower 0.8816 0.6613 0.5976
chilleetc 0.8549 0.5796 0.5141
clover 0.9293 0.7441 0.5424
cucumberetc 0.8344 0.5212 0.4986
eggplant 0.8141 0.5183 0.5403
grassnes 0.8993 0.6626 0.5429
greenbean 0.795 0.4711 0.4922
greencorn 0.8092 0.4823 0.5734
greenpea 0.7931 0.4325 0.4921
lettuce 0.8836 0.6554 0.5099
melonetc 0.8221 0.5239 0.5182
okra 0.8392 0.5488 0.5078
onion 0.8017 0.4604 0.5116
potato 0.831 0.5112 0.491
pumpkinetc 0.77 0.4017 0.5407
spinach 0.868 0.6001 0.4903
swedefor 0.952 0.8535 0.5777
sweetpotato 0.8524 0.5524 0.5134
taro 0.8543 0.5921 0.5196
tomato 0.8254 0.5264 0.5317
turnipfor 0.9278 0.8009 0.628
watermelon 0.8661 0.5903 0.4994
cinnamon 0.8796 0.6302 0.5176
cocoa 0.8311 0.5108 0.5666
coconut 0.8417 0.5267 0.5204
coffee 0.8323 0.5424 0.4703
cotton 0.8323 0.5036 0.5068
flax 0.8139 0.5198 0.5019

76
 CHAPTER I

hemp 0.8232 0.5288 0.5454

kapokseed 0.8192 0.5386 0.3957
oilpalm 0.8787 0.5976 0.5056
pimento 0.7671 0.3985 0.4918
ramie 0.8116 0.5341 0.5228
rubber 0.8263 0.5111 0.5192
sugarcane 0.8036 0.4483 0.535
sunflower 0.8731 0.6087 0.561
tobacco 0.8415 0.532 0.5034
Abies alba 0.8266 0.5003 0.5136
Acer campestre 0.8665 0.6407 0.5361
Acer platanoides 0.7991 0.5265 0.4749
Acer pseudoplatanus 0.7878 0.474 0.4721
Alnus glutinosa 0.8347 0.5812 0.4631
Betula pendula 0.7237 0.3835 0.4264
Castanea sativa 0.9307 0.7639 0.5176
Corylus avellana 0.8795 0.6388 0.446
Eucalyptus globulus 0.9531 0.812 0.5011
Fagus sylvatica 0.8575 0.6018 0.4299
Fraxinus excelsior 0.8648 0.6664 0.4527
Juglans regia 0.9413 0.7682 0.5156
Larix decidua 0.8811 0.6008 0.5109
Picea abies 0.7593 0.4346 0.489
Pinus nigra 0.7466 0.3877 0.5004
Pinus pinaster 0.8722 0.5906 0.47
Pinus pinea 0.9209 0.7314 0.4678
Pinus sylvestris 0.786 0.4906 0.4551
Populus alba 0.7556 0.4332 0.4823
Populus nigra 0.7877 0.5002 0.4889
Populus tremula 0.8113 0.5243 0.4902
Prunus avium 0.9123 0.7319 0.4486
Pseudotsuga menziesii 0.8801 0.6455 0.5606
Quercus petraea 0.9161 0.7508 0.4724
Quercus robur 0.8853 0.6989 0.4148
Quercus suber 0.9167 0.7267 0.4797
Robinia pseudoacacia 0.9053 0.6709 0.5956
Salix alba 0.9097 0.6868 0.5417
Tilia cordata 0.8494 0.6235 0.5022
Ulmus laevis 0.8526 0.6265 0.5073

77
 CHAPTER I

Supplementary Table 5: Area with suitable climatic habitat of all host plants under RCP8.5.

Species name 2020 2100 change change in %

almond 1032408.07 4000516.28 2968108.21 293.751729
apple 3254431.37 4257849.09 1003417.72 31.0394461
apricot 2342347.98 5299097.64 2956749.66 126.823638
blueberry 3525838.97 480630.156 -3045208.8 -87.175005
carob 1742656.4 4667254.17 2924597.77 167.759557
cherry 3814216.46 3771293.94 -42922.523 0.45703138
cranberry 627159.497 230315.215 -396844.28 -41.989644
gooseberry 4399173.16 2027781.7 -2371391.5 -53.721071
grape 2117999.18 5072976.89 2954977.71 140.847296
kiwi 2785289.85 4776420.56 1991130.71 109.915061
peachetc 2360522.47 4243017.65 1882495.18 82.9495409
pear 1404409.18 2907795.85 1503386.67 104.573722
plum 3164038.43 4374252.36 1210213.93 38.034578
rasberry 3840614.97 2425054.57 -1415560.4 -36.935591
sourcherry 3759434.8 2691685.84 -1067749 -28.166283
strawberry 2935115.16 3063824.64 128709.475 8.6546808
tangetc 2899457.14 4746209.19 1846752.05 65.5299122
fig 2083935.04 5154998.35 3071063.31 148.981288
avocado 465484.892 2561016.82 2095531.93 472.406533
cashew 0 3196.95732 3196.95732 NA
citrusnes 0 1644.59491 1644.59491 NA
date 102378.955 152628.116 50249.161 48.734971
grapefruitetc 92174.3665 842567.555 750393.188 1074.16263
groundnut 481038.689 2351871.31 1870832.63 436.534163
karite 0 0 0 NA
lemonlime 358290.294 745684.187 387393.894 126.315463
mango 0 7121.55421 7121.55421 Inf
olive 883921.454 3704732.46 2820811.01 319.609949
orange 209649.8 1265377 1055727.2 545.758296
papaya 49028.8219 133.786999 -48895.035 -99.705848
persimmon 2043955.47 4327522.15 2283566.68 110.568605
pineapple 756.998354 829.242054 72.2436991 NA
pistachio 443665.803 2026242.57 1582576.76 353.336576
plantain 5613399.03 5615324.84 1925.80465 0.03434248
quince 2194102.08 3608384.59 1414282.52 64.6968248
areca 1395866.63 2001903.03 606036.398 934.654187
barley 5487322.96 3856374.86 -1630948.1 -29.638863
bean 3790633.82 4527342.68 736708.856 20.1068616
broadbean 5136263.22 4559991.76 -576271.47 -10.779358
buckwheat 4164889.72 4324321.79 159432.071 3.95250488
castor 0 0 0 NA
chickpea 406918.253 88460.6274 -318457.63 Inf
cowpea 0 0 0 NA
maize 4992465.1 4318892.54 -673572.56 -13.23313
oats 5468277.8 3066210.59 -2402067.2 -43.975787
pea 5362412.42 3739453.07 -1622959.3 -30.123327
pigeonpea 0 18791.8831 18791.8831 Inf

78
 CHAPTER I

rapeseed 2991882.92 3200704.7 208821.778 6.13920335

rice 631263.607 2496404.62 1865141.01 283.078963
rye 3774543.31 912503.706 -2862039.6 -75.998295
sorghum 518294.2 2757769.2 2239475 479.840802
soybean 1446633.93 4301807.96 2855174.03 197.487267
triticale 3969335.04 1963572.03 -2005763 -50.62388
wheat 5338827.86 4881246.25 -457581.61 -8.5182259
greenbean 3163911.19 4282194.66 1118283.47 41.1522969
greencorn 3395222.38 2897705.87 -497516.51 -12.362191
greenpea 2904564.25 3547577.09 643012.842 22.944007
alfalfa 4139848.27 4491963.13 352114.854 9.84173378
beetfor 1624857.26 717533.172 -907324.08 -55.75436
cabbage 1478942.46 3781500.83 2302558.37 157.767757
carrot 4209533.81 4460270.12 250736.314 6.1328124
cassava 3839.80272 2677.12849 -1162.6742 135.819435
cauliflower 511607.139 2362897.09 1851289.96 353.62449
chilleetc 2158286.46 4047356.09 1889069.62 89.2725041
clover 2439721.96 1499349.44 -940372.52 -39.794485
cucumberetc 2050501.39 4259275.31 2208773.92 107.012327
eggplant 1059460.06 3105879.59 2046419.53 195.879232
grassnes 2838740.6 2660485.53 -178255.06 -3.4614892
lettuce 1295692.67 2312812.67 1017120 82.9838183
melonetc 2110056.91 4011397.53 1901340.63 90.2884075
okra 0 18503.0539 18503.0539 Inf
onion 2912561.02 4228476.04 1315915.02 48.0266479
potato 5294226.9 4229209.98 -1065016.9 -19.889347
pumpkinetc 2053599.62 3907137.24 1853537.62 86.8545631
spinach 1638084.79 4202618.74 2564533.96 155.548539
swedefor 481069.331 48163.3112 -432906.02 -91.01766
sweetpotato 1834246.74 5161476.46 3327229.72 181.977798
taro 1799.19487 9430.87683 7631.68197 417.208163
tomato 2806672.79 4028381.72 1221708.93 44.3833547
turnipfor 430496.655 142689.2 -287807.45 -65.244387
watermelon 1491868.28 4660723.92 3168855.64 211.668417
cinnamon 29.822491 127.215108 97.3926175 NA
cocoa 0 1017.30682 1017.30682 NA
coconut 0 598.77051 598.77051 NA
coffee 51992.97 155256.22 103263.25 247.172385
cotton 359196.026 2742743.65 2383547.62 697.475698
flax 5163731.71 4533971.14 -629760.57 -12.243186
hemp 2513349.45 3790367.59 1277018.14 52.795279
kapokseed 977.971089 0 -977.97109 -100
oilpalm 0 189.693927 189.693927 NA
pimento 183152.264 1445073.81 1261921.54 734.411259
ramie 1792442.6 5272141.66 3479699.05 195.33548
rubber 0 0 0 NA
sugarcane 27787.8788 627015.968 599228.09 3747.26585
sunflower 3470541.57 3746681.87 276140.292 10.613495
tobacco 2612044.49 4007455.38 1395410.9 54.8536923
Abies alba 1271989.58 326329.85 -945659.73 -72.947618
Acer campestre 7017142.64 7425246.26 408103.614 5.9161451

79
 CHAPTER I

Acer platanoides 5614096.14 5108623.84 -505472.3 -8.9475111

Acer pseudoplatanus 3688017.19 3193738.24 -494278.94 -12.806784
Alnus glutinosa 5169549.65 5691394.38 521844.73 10.0752586
Betula pendula 5259409.84 2750962.97 -2508446.9 -47.607361
Castanea sativa 3300893.93 5394369.75 2093475.82 63.6746686
Corylus avellana 4598058.85 4101699.33 -496359.52 -10.783591
Eucalyptus globulus 621525.039 2451903.37 1830378.33 298.758654
Fagus sylvatica 4053818.14 3552331.2 -501486.94 -12.610831
Fraxinus excelsior 6153831.79 6259737.03 105905.24 1.69774433
Juglans regia 5067588.26 5946602.33 879014.077 17.375463
Larix decidua 755532.775 167251.248 -588281.53 -77.80547
Picea abies 3431528.58 790049.61 -2641479 -77.156949
Pinus nigra 3594491.46 5155629.51 1561138.05 44.3395886
Pinus pinaster 1898928.16 4613178.84 2714250.68 142.911838
Pinus pinea 686961.674 2945673.25 2258711.58 328.534287
Pinus sylvestris 4650755.21 1418925.29 -3231829.9 -69.520825
Populus alba 4283613.4 7441974.08 3158360.68 73.7436181
Populus nigra 6817675.51 8124045.84 1306370.33 19.1983868
Populus tremula 7009383.7 4883827.97 -2125555.7 -30.382759
Prunus avium 5472761.08 6013285.53 540524.456 9.86227798
Pseudotsuga menziesii 3971050.13 3143780.2 -827269.93 -20.984379
Quercus petraea 5102275.26 6018105.57 915830.308 17.9370936
Quercus robur 5858345.59 5511073.17 -347272.42 -5.9151801
Quercus suber 1037210.77 4378159.8 3340949.04 320.424422
Robinia pseudoacacia 3736400.64 4334947.19 598546.546 16.3041782
Salix alba 4415357.67 4870576.64 455218.972 10.2764455
Tilia cordata 6166294.63 5029658.92 -1136635.7 -18.292572
Ulmus laevis 5812747.51 4305493.79 -1507253.7 -25.74863

80
 CHAPTER I

Supplementary Table 6: Same as table S5 but for RCP4.5

Species name 2020 2100 change change in %

almond 995735.491 1974709.03 978973.54 99.4702786
apple 3062689.63 4537806.9 1475117.27 48.810419
apricot 2354265.87 3922718.88 1568453.01 66.0455366
blueberry 3622185.36 2199023.04 -1423162.3 -40.24475
carob 1720935.76 3058316.89 1337381.13 76.9471877
cherry 4128295.36 4293391.76 165096.405 3.96469352
cranberry 546601.86 645084.389 98482.5294 46.2373844
gooseberry 4491385.19 4002593.34 -488791.85 -10.873489
grape 2153349.41 3704148.69 1550799.28 72.4024069
kiwi 2459537.65 3717285.48 1257747.83 58.638474
peachetc 2423274.19 3782216.87 1358942.68 55.7807432
pear 1084403.53 2129646.58 1045243.05 96.3052426
plum 3456835.95 4129752.97 672917.023 19.8894947
rasberry 3947170.44 4180453.07 233282.632 6.11115364
sourcherry 3760152.95 4141158.58 381005.63 10.1011873
strawberry 3020040.72 3727927.12 707886.396 24.8113889
tangetc 2641468.41 4805519.25 2164050.84 82.105503
fig 2098212.97 3598939.13 1500726.16 71.1844379
avocado 442700.232 924957.268 482257.035 108.592634
cashew 0 0 0 NA
citrusnes 0 0 0 NA
date 95848.7401 65524.5744 -30324.166 -32.3668
grapefruitetc 88096.5409 301692.139 213595.598 264.376279
groundnut 360223.515 1280887.14 920663.623 271.596388
karite 0 0 0 NA
lemonlime 301834.608 520810.362 218975.754 78.6876135
mango 0 0 0 NA
olive 861743.972 1748886.34 887142.363 102.755451
orange 200688.452 442227.607 241539.156 123.208597
papaya 49311.7466 0 -49311.747 -100
persimmon 1987636.22 3501629.54 1513993.33 76.0373234
pineapple 430.807814 921.490115 490.682301 NA
pistachio 456365.956 1291113.07 834747.119 187.788686
plantain 5615103.38 5615336.41 233.03306 0.00415061
quince 2440740.84 3640325.98 1199585.14 48.9565925
areca 1225608.59 1146203.51 -79405.076 12221.1478
barley 5483908.27 5319558.95 -164349.31 -2.9930095
bean 3986138.84 5019217.21 1033078.38 26.162475
broadbean 4523421.53 5459579.41 936157.883 22.4770532
buckwheat 4132220.52 4985504.77 853284.251 20.8368679
castor 17.2217734 0 -17.221773 NA
chickpea 359196.18 246317.458 -112878.72 NA
cowpea 0 0 0 NA
maize 5016768.07 5361862.26 345094.182 6.88229005
oats 5484365.6 4980559.93 -503805.67 -9.1800933
pea 5340716.25 5208528.7 -132187.54 -2.4687154
pigeonpea 0 146.126991 146.126991 NA

81
 CHAPTER I

rapeseed 3368748.57 3896716.35 527967.778 16.4127488

rice 578870.708 1389540.6 810669.897 141.057513
rye 3760677.78 2711360.28 -1049317.5 -28.097845
sorghum 479847.972 1601665.7 1121817.73 230.581283
soybean 1344377.17 2657540.3 1313163.12 100.645407
triticale 4333674.81 3094276.15 -1239398.7 -28.494682
wheat 5341128.65 5546053.11 204924.457 3.84789031
greenbean 3349525.25 4216034.02 866508.773 26.3584221
greencorn 3272023.3 4076237.15 804213.847 25.4116333
greenpea 3095990.66 4423240.78 1327250.12 44.3873504
alfalfa 3839838.41 5233728.09 1393889.68 37.9436615
beetfor 1896115.79 1312058.07 -584057.72 -30.667721
cabbage 1491567.19 2731042.87 1239475.68 83.0245331
carrot 4516350.95 5061983.94 545632.99 12.0553394
cassava 3853.84541 29.7554671 -3824.0899 -99.336511
cauliflower 478410.545 1203873.79 725463.248 151.451298
chilleetc 2190062.04 3255059.26 1064997.22 48.6295682
clover 2728040.96 2160340.95 -567700.01 -20.834906
cucumberetc 1943609.88 3224528.21 1280918.32 64.4833253
eggplant 1017945.04 1911807.82 893862.774 88.3361704
grassnes 3031526.96 3511872.56 480345.599 16.464193
lettuce 1261279.08 1607526.58 346247.508 26.6766838
melonetc 2118733.94 3208257.38 1089523.44 50.3885228
okra 0 233.010643 233.010643 NA
onion 2977529.51 3695368.81 717839.306 24.3864615
potato 5349118.52 5398754.72 49636.2036 0.96707098
pumpkinetc 2109917.57 2775523.21 665605.634 31.1538486
spinach 1492262.62 2583424.14 1091161.52 73.1150613
swedefor 604391.042 119701.621 -484689.42 -79.639466
sweetpotato 1493585.46 3516516.26 2022930.8 135.217303
taro 1527.14398 9465.86866 7938.72468 298.653776
tomato 2870242.17 3644148.53 773906.361 26.7002892
turnipfor 490470.504 296650.766 -193819.74 -37.884101
watermelon 1330132.79 2743961.14 1413828.35 105.755389
cinnamon 0 405.547184 405.547184 NA
cocoa 0 0 0 NA
coconut 0 0 0 NA
coffee 50060.1774 95797.9368 45737.7594 104.511194
cotton 305475.106 1417421.39 1111946.28 364.758762
flax 5204417.01 5512945.49 308528.482 5.93108332
hemp 2721088.4 3967131.57 1246043.17 46.5180414
kapokseed 189.093776 0 -189.09378 NA
oilpalm 0 0 0 NA
pimento 92344.6063 289946.602 197601.996 385.122641
ramie 1477414.75 4100123.22 2622708.47 181.05325
rubber 0 0 0 NA
sugarcane 28340.9642 176610.8 148269.835 1139.11216
sunflower 3568303.75 4547492.76 979189.007 27.9383825
tobacco 2541010.4 3610495.4 1069484.99 41.5006953
Abies alba 1492479.69 919435.226 -573044.47 -37.607238
Acer campestre 6953244.25 7541805.13 588560.878 8.48154223

82
 CHAPTER I

Acer platanoides 5608296.41 5798058.95 189762.534 3.38642812

Acer pseudoplatanus 3868408.35 4018327.01 149918.658 4.10225379
Alnus glutinosa 5198083.05 5829599.8 631516.745 12.2188155
Betula pendula 5286766.76 4540247.06 -746519.69 -14.11471
Castanea sativa 3309709.78 4645142.75 1335432.97 40.2826023
Corylus avellana 4614749.69 5046613.31 431863.622 9.40347802
Eucalyptus globulus 625881.757 1194147.72 568265.963 91.10437
Fagus sylvatica 4013382.14 4347870.75 334488.605 8.3182686
Fraxinus excelsior 6104329.79 6634397.86 530068.073 8.72316745
Juglans regia 4980098.15 5888650.23 908552.073 18.8525898
Larix decidua 902888.649 400496.99 -502391.66 -54.984924
Picea abies 3326374.67 2014187.22 -1312187.4 -39.468556
Pinus nigra 3772055.49 5094948.59 1322893.1 35.5377123
Pinus pinaster 1921573.24 2990659.69 1069086.45 55.6045355
Pinus pinea 647679.921 1318965.37 671285.448 101.480365
Pinus sylvestris 4595767.54 3172867.33 -1422900.2 -31.024946
Populus alba 4365130.89 5828611 1463480.11 33.5333944
Populus nigra 6894700.78 7646445.37 751744.585 10.9343351
Populus tremula 6988862.72 6304708.9 -684153.82 -9.7952038
Prunus avium 5412977.54 6206020.49 793042.958 14.6701209
Pseudotsuga menziesii 3986621.83 4225875.59 239253.762 6.05085763
Quercus petraea 4994663.78 5828703.96 834040.175 16.7908093
Quercus robur 5892440.7 6235103.27 342662.572 5.84095851
Quercus suber 1057426.45 2020201.65 962775.196 90.7019568
Robinia pseudoacacia 3739009.05 4504309.99 765300.937 21.1411671
Salix alba 4347579.68 5033856.72 686277.04 15.8552175
Tilia cordata 6234680.16 6160419.61 -74260.545 -1.1992308
Ulmus laevis 5892354.77 5636339.97 -256014.8 -4.3236439

83
 CHAPTER I

Supplementary Table 7: Pest species, description of their native and invasive and the geographical origin of their
occurrence records.

Species name Native range Invasive range Records

Acleris variana North America Full
Agrilus anxius North Amercia Southern Africa Full
Agrilus planipennis East Asia North America Full
No records in Africa
Aleurocanthus spiniferus South Asia Africa, Australia and Australia
Africa, North America, Central
America, South America, Only 2 records from
Aleurocanthus woglumi South Asia Oceania native range
South America,
Anastrepha fraterculus Central America Full
Anastrepha ludens Central America Full
South America,
Anastrepha obliqua Central America Full
Only 3 records in North
America, 2 records in
Anoplophora glabripennisChina North America Europe
Anthonomus bisignifer Japan Russia Native range only
Anthonomus grandis Central America North America, South America Full
Anthonomus quadrigibbus North America Full
Anthonomus signatus North America Full
Bactericera cockerelli North America Central America; New Zealand Full
Bactrocera cucumis Australia Full
Bactrocera cucurbitae South Asia Africa Full
Bactrocera invadens Sri lanka Africa Full
Bactrocera latifrons Asia Africa, Hawaii Full
Bactrocera tryoni Austalia Full
Bactrocera zonata Asia West Asia, Africa Full
Ceratitis rosa East and South Africa Full
Choristoneura conflictana
North America Full
Choristoneura freemani North America Full
Choristoneura rosaceana North America Full
Conotrachelus nenuphar North America Full
Dacus ciliatus Africa, Asia Full
Dendroctonus adjunctus North America Full
Dendroctonus brevicomis North America Full
North America,
Dendroctonus frontalis Central America Full
Dendroctonus ponderosae North America Full
Dendroctonus
pseudotsugae North America Full
Dendroctonus rufipennis North America Full
Dendrolimus superans Japan, Russia Only 4 records in Russia
Diabrotica barberi North America Full
Diabrotica speciosa South America Full
Diabrotica
undecimpunctata howardi North America Full
Diabrotica
undecimpunctata
undecimpunctata North America Full

84
 CHAPTER I

South America, Central

America, North America, East
Diaphorina citri South Asia Africa No records from Africa
Epitrix cucumeris North America Central America Full
Epitrix subcrinita North America South America Native range only
North America, Central No records from Central
Euwallacea fornicatus Asia America, Oceania America
Gnathotrichus sulcatus North America Central America Full
South America, Africa, No records from South
Gonipterus scutellatus Australia Southern Europe America
Grapholita packardi North America Full
North America,
Central America, Only 1 record in South
Helicoverpa zea South America America
Heteronychus arator Africa, Oceania Full
Homalodisca vitripennis North America Full
North America,
Ips calligraphus Central America Philippines Native range only
North America, Only 3 records in
Ips grandicollis Central America Oceania Oceania
Ips pini North America Full
Only 3 records in
Leucinodes orbonalis Asia, Oceania Oceania
Limonius californicus North America Full
Africa, North America, Central
America, South America,
Liriomyza sativae China Oceania Invasive range only
Listronotus bonariensis South America Oceania Full
Only 4 records in North
Lycorma delicatula Asia North America America
Lymantria mathura Russia, East Asia Full
Africa, Oceania, North
America, Central America, No records in Asia and
Maconellicoccus hirsutus Southern Asia South America only 1 record in Africa
Malacosoma americanum North America Full
Malacosoma disstria North America Full
Only two records not
Massicus raddei China, Japan Russia from Japan
Megacopta cribraria Asia North America Invasive range only
Melanotus communis North America Melanotus communis Full
Central America,
Metamasius hemipterus South America Africa Native range only
North America, Australia, Only 1 records from
Naupactus leucoloma South America South Africa Africa
Naupactus xanthographus South America Full
Neodiprion abietis North America Full
Oemona hirta New Zealand Full
Orgyia pseudotsugata North America Full
Pissodes strobi North America Full
Platynota stultana North America Full
Polygraphus proximus Russia, East Asia Full
North America, Russia,
Popilia japonica Japan Southern Europe No records in Russia
North America,
Central America, No records in Soth
Pseudacysta perseae South America Portugal America

85
 CHAPTER I

Rhagoletis mendax North America Full

Rhagoletis pomonella Western USA Eastern USA Full
Central America,
Rhynchophorus palmarum South America Full
Saperda candida North America German Island Full
South Africa, Hawaii, No records from South
Scirtothrips dorsalis South Asia, Oceania Carribean, Israel Africa and Hawaii
North America, Central Only 4 records in South
Spodoptera eridania South America America America
Spodoptera frugiperda South America Africa, Asia No records in Asia
Spodoptera litura South Asia East Asia, Oceania Full
Strauzia longipennis North America Germany (1 record) Full
Central America,
Tecia solanivora South America Tenerifa Full
Thaumatotibia leucotreta Africa Full
Trioza erytreae Africa Saudi-Arabia, Portugal Full
Africa, North America, Central
America, South America,
Xylosandrus compactus Asia Oceania Invasive range only
No records in Oceania
Africa, North America, Central and little records in Asia
Xylosandrus crassiusculus Asia America, Oceania and Africa
Asia, North America, Central
America, South America,
Zaprionus indianus Africa Southern Europe No records in Europe
Zaprionus tuberculatus Africa Mediterranean Only 1 record in Italy

86
87
CHAPTER II

Inflection point in climatic suitability of insect pest species in Europe

suggests non-linear responses to climate change

Marc Grünig*1,2,3, Pierluigi Calanca2, Dominique Mazzi1, and Loïc Pellissier3,4

1
Agroscope, RD Plant Protection, Wädenswil, Switzerland
2
Agroscope, RD Agroecology and Environment, Zurich, Switzerland
3
ETH, Landscape Ecology, Zurich, Switzerland
4
Swiss Federal Research Institute WSL, Birmensdorf, Switzerland

Published in Global Change Biology (2020)

Doi: https://doi.org/10.1111/gcb.15313
Pre-print version

88
 CHAPTER II

Abstract

Climate change and globalization affect the suitable conditions for agricultural crops
and insect pests, threatening future food security. It remains unknown whether shifts in species’
climatic suitability will be linear or rather non-linear, with crop exposure to pests suddenly
increasing when a critical temperature threshold is crossed. Moreover, uncertainty of forecasts
can arise because of the modelling approach based either on species distribution data or on
physiological measurements. Here, we compared the predictions of two modelling approaches
(physiological models and species distribution models) for forecasting the potential distribution
of agricultural insect pests in Europe. Despite conceptual differences, we found good agreement
overall between the two approaches. We further identified a potential regime change in pest
pressure along a temperature gradient. With both modelling approaches, we found an inflection
point in the number of pest species with suitable climatic conditions around a minimum
temperature of the coldest month of -3°C. Our results could help decision-makers anticipate the
onset of rising pest pressure and provide support for intensifying surveillance measures,
particularly in regions where temperatures are already beyond the inflection point.

Keywords: agricultural crop, climate change, insect pest, physiological model, species
distribution model, temperature threshold

89
 CHAPTER II

Introduction

Climate change is predicted to shift the distribution of agricultural crops and of the insect
pest species feeding on them (Tubiello et al., 2007; Bebber, Holmes & Gurr, 2014; Sloat et al.,
2020). While changing climatic conditions might increase climatic stress factors and reduce
resources for water irrigation of crops (Fader et al., 2016), higher temperatures favour the
expansion of cultivation and crop diversification at higher latitudes (Tuck et al., 2006; Walther
et al., 2002; Grünig et al., 2020). However, climate change is also associated with greater pest
pressure (Deutsch et al., 2018). Further, trade flows and human travel help insect pests to
overcome natural barriers, linking climate change and globalization to pest invasions (Robinet
& Roques, 2010; Hulme, 2009; Paini et al., 2016). Distribution ranges of pest species are
expected to shift, to the detriment of cropping systems (Bebber et al., 2013), and thus threaten
food production (Schmidhuber & Tubiello, 2007) and undermine increasing consumer demands
for local and more sustainably produced food (Lamichhane et al., 2016; Feldmann & Hamm,
2015). Investigating the distribution of niches of pest species along climatic gradients can point
to future opportunities and risks under climate change (Grünig et al., 2020) and help develop
effective crop-protection strategies.

The development of insects is a function of temperature over time (Jarosik et al., 2011).
Under climate change, temperatures in Europe, particularly at higher latitudes, are expected to
increase more than the global average warming (MacDonald, 2010; IPCC, 2007). Ongoing
warming allows the cultivation of more crops (Maracchi, Sirotenko, & Bindi, 2005) but also
favours insect pest survival in these regions (Bale & Hayward, 2010). Higher winter
temperatures are crucial for the survival of insect pest species at higher latitudes (Bebber et al.,
2013; Jarosik et al., 2015), although in practice their realized distribution is restricted by
additional factors, particularly biotic interactions (Hutchinson, 1957). To account for the
multitude of factors that potentially define the climate suitability for insect pests, different
modelling approaches have been developed, which should imply differences in predictions
(Roberston et al., 2003; Kearny, Wintle & Porter, 2010; Newman, 2005).

To model the climatic niche of pest species, deductive and inductive approaches have
been developed (Venette et al., 2010; Tonnang et al., 2017). For the deductive approach adopted
in physiological models, data from controlled experiments provide a basis to project the
potential distribution of a species (Tonnang et al., 2017). Studies on the life history of species
90
 CHAPTER II

under different environmental conditions help define physiological thresholds for species
survival, development and performance, which are often used to develop phenological models
(e.g. Schaub et al., 2017). By relating these thresholds to climatic variables in a spatial context
(e.g. spatial raster layers of temperature), areas where species meet conditions that allow their
development can be identified (Kearney & Porter, 2004; Kearney & Porter, 2009). Inductive
models use the occurrence records of a species and link them to climatic data to infer the
conditions that are suitable for the species (Guisan & Zimmermann, 2000; Elith & Leathwick,
2009). Species distribution models (SDMs) implementing the inductive approach have been
widely used to model climatic suitability of insect pests (e.g. Arthur, Morrison & Morey, 2019).
While deductive and inductive modelling approaches target similar goals of mapping suitable
climatic conditions for a species in space, they differ in terms of the data used for model
development. A better understanding of how systems can be expected to shift under climate
change could potentially be achieved by comparing forecasts between those approaches.

The response of a biological system to climate change can be linear or non-linear. When
the response is not linear, the ecosystem state typically shifts faster than what would be expected
for a linear response (Dakos et al., 2019), eventually exhibiting discontinuities when critical
thresholds in the driving variables are crossed (Lenton, 2011). Shifts in pest distributions under
climate change may display discontinuous behaviour, implying a sudden change in the potential
for colonization by pest species, including the invasion of non-native species, as a result of
global warming (Paini et al., 2016). Invasive insects are likely to be introduced from geographic
areas sharing similar climatic conditions (Brockerhoff & Liebhold, 2017; Walther et al., 2009).
With climate change, temperate regions will become warmer and therefore climatically more
similar to regions where insect diversity is presently disproportionally higher. The study of pest
species’ climatic niches coupled with climate change scenarios help identify potential inflection
points for pest suitability.

Here, we investigated the potential pest accumulation in Europe under climate change
by comparing predictions obtained from deductive physiological models and inductive SDMs.
We considered a comprehensive set of insect pests that includes a large number of quarantine
insects. We evaluated whether climatic niches shift along a smooth gradient, or whether an
inflection point exists, beyond which the increase in the number of potential niches accelerates.
Our working hypotheses were as follows:

91
 CHAPTER II

i) We expected congruent forecasts of suitable climate with the two modelling

approaches. Under climate change, we expected a general pattern of increasing
climatic suitability for insect pests across Europe, irrespective of the modelling
approach applied.
ii) We expected non-linear responses to minimum temperatures for insect pests,
involving thresholds beyond which the number of species with a suitable climate
increases considerably.
iii) We expected to observe that temperatures in Southern Europe already exceed
the threshold temperature identified under (ii). Hence, we expected to find that
these regions already provide suitable climatic conditions for many insect pest
species because temperatures rarely dip below freezing.

Materials and Methods

Physiological data collection

For physiological data, we assembled insect developmental thresholds from the

PRATIQUE database (Jarosik et al., 2011) and published literature. From the PRATIQUE
database, we selected insect species classified as pests in the EPPO Global Database
(www.eppo.org), which comprises information generated or assembled by the European and
Mediterranean Plant Protection Organization. Further, we searched on GoogleScholar for
studies applying CLIMEX models for pest species (keyword combinations of “CLIMEX” +
“insect” + “pest”). We obtained physiological parameters on the lower development threshold
(LDT; minimum development threshold in CLIMEX studies) and the sum of effective
temperatures (SET), that is, the number of growing degree days above the LDT required for the
completion of a generation. LDTs and SETs reflect a linear relationship between developmental
rate and temperature, and are calculated from the proportion of development occurring per unit
of time (Jarosik et al., 2011). Where multiple entries per species were available, we used the
average of the values. Further, we searched for lethal temperatures of all species. Whenever we
did not find a lower lethal temperature ( , ), we set it to 0°C for species not known to
survive winter in a dormant stage and to -20°C for species with winter diapause. This implies
that warm-adapted species are predicted to not occur in regions where the minimum
temperature of the coldest month is below 0°C. We used 40°C as the upper lethal temperature

92
 CHAPTER II

( , ) for all species, as insects are known to have little variability in this regard, with
values ranging from 40°C to 50°C (Heinrich, 1981). We collected data on physiological
parameters for 75 species (Table S1 in the Supplementary Material).

Occurrence data collection

We compiled the distribution records of pest species in the published literature and the
Global Biodiversity Information Facility (GBIF, www.gbif.org) for all pest species in the EPPO
Global Database. We searched Google Scholar for the following keywords in various
combinations: “Pest name”, “distribution”, “records”, “occurrence”, “sampling”, “spatial” and
“data”. GBIF data were carefully checked for unreliable records. We classified the pest species
into two categories: all species on the EPPO quarantine lists (A1, A2, Alert) were considered
quarantine pests, and all others established pests. Species for which we could not obtain more
than 20 occurrence records were dismissed (Wisz et al., 2008). In total, we gathered occurrence
data for 173 species (Table S2).

Climate data and future scenarios

We acquired climate data from the CHELSA database (www.chelsa-climate.org). We

used CHELSA V1.2 data for monthly minimum, maximum and mean temperatures, as well as
bioclimatic variables, with a 2.5 arcmin (5 km) grid size resolution (Karger et al., 2017) to
represent current climatic conditions. Further, we used climate change scenarios from the
CMIP5 family representing two different scenarios (RCP 4.5, RCP 8.5) and four global
circulation models (GCMs). We selected the following GCMs based on model interdependence
to achieve a good representation of uncertainty in climate projections (Sanderson et al., 2015):
CESM1-BGC (US National Center for Atmospheric Research, NCAR); CMCC-CM (Centro
Euro-Mediterraneo per i Cambiamenti Climatici, CMCC); MIROC5 (University of Tokyo);
and ACCESS1-3 (Commonwealth Scientific and Industrial Research Organization, CSIRO,
and Bureau of Meteorology, BOM, Australia).

Physiological models

We developed physiological models by coupling physiological data with monthly mean,

minimum and maximum temperatures following FAO-ECOCROP (Hijmans et al., 2001;
Ramirez-Villegas et al., 2013) and CLIMEX (Sutherst & Maywald, 1985) approaches. The

93
 CHAPTER II

output of the models is a suitability index (Si) calculated on cell basis of a spatial raster (5 km).
As a global equation, we calculated Si as the product of a temperature index (Ti), a growing
index (Gi) and a killing index (Ki) (equation 1).

= ∗ ∗ (1)

Ti was calculated as the fraction of months in which the monthly mean temperature (Tk) exceeds
LDT (equation 2).

=
(2)

with

= >
!"

We evaluated Gi based on the growing degree days needed for completion of development
(GDDreq) (equation 3).

$ − &<
=#
(3)
(

!"

For Ki, we checked whether the monthly minimum temperature (Tmin,k) fell below the lethal
minimum temperature ()*+,,,-+. ) and whether the monthly maximum temperature (Tmax,k)
surpassed the lethal maximum temperature ()*+,,,-/0 ) (equation 4).

=1 1
(4)

94
 CHAPTER II

with

>
= , ,
!"

<
= , ,
!"

We evaluated the physiological models using a maximum sensitivity approach based on

occurrence data. To calculate the sensitivity of the model predictions, we projected the models
to a global extent in order to include all available occurrence records for the evaluations. Models
with a sensitivity score < 0.5 were excluded from the analyses. The outputs were classified into
binary predictions using the maximum sensitivity threshold calculated with the
optimal.thresholds function in the ‘PresenceAbsence’ package (version 1.1.9; Freeman &
Moisen, 2008).

Species distribution models

SDMs were calibrated using ensembles (unweighted averages) of generalized linear

models (GLMs) and generalized additive models (GAMs; Wood, 2006). For each species, we
randomly sampled 5,000 pseudo-absences from the species biomes. We weighted presence
records in order to balance their weights with the large number of pseudo-absence records. We
assumed a binomial error distribution for both modelling techniques and used fourth-order
polynomials to adjust the flexibility of the response curve. As predictor variables, we used
growing degree days above 5°C and minimum temperature of the coldest month to reflect the
variables used in the physiological models. We used a variable selection procedure to reach an
acceptable model performance and projection for as many species as possible (see Note S1).
Further, we followed the standards and guidelines for distribution modelling (Araujo et al.,
2019). GLMs were fitted with the ‘base’ R-package, whereas GAMs were fitted with functions
in the R-package ‘gam’ (version 1.16.1; Hastie, 2019). For the evaluation of model
performance, we used a split sample approach (70% calibration data and 30% evaluation data)
with 20 repetitions to calculate the area under the ROC-plot curve (AUC) and true skill statistics
(TSS). We considered models to be reliable at AUC > 0.7 (Hosmer, Lemeshow & Sturdivant,
2013) and TSS > 0.4 (Descombes et al., 2015) and discarded all others. Additionally, we

95
 CHAPTER II

inspected the quality of all model projections visually. For binary classifications of the model
outputs, we used the optimal TSS threshold (Allouche et al., 2006).

Comparison of physiological and species distribution models

For each species with reliable projections in both modelling approaches, we calculated
the Pearson correlation coefficient between the predicted suitability indices of the two model
projections and the percentage of agreeing grid cells of the binary projections. Further, we
calculated the Sørensen index for the community similarity between the modelling approaches.
For this, we stacked together the predicted binary projections of all species for each timestep,
resulting in species richness raster stacks for the different modelling approaches for each
timestep. For these raster stacks we calculated the Sorensen dissimilarity index with the
beta.pair function of the ‘betapart’ R-package (version 1.5.1; Baselga et al., 2018), which we
then subtracted from 1 to obtain a similarity index ranging from 0 to 1, where 1 indicates
complete agreement for the exact same set of species and 0 no overlap between the predicted
communities. To quantify differences between projections of the two modelling approaches
(i.e. absolute model agreement), we subtracted the model projection of the SDM from the
physiological model projection for each species and summed the resulting differences across
all species. Positive values indicate overrepresentation by the physiological models, and
negative values signify overrepresentation by the SDMs.

Analysis of inflection points under climate change

Based on the predicted distribution in Europe for each species under current climatic
conditions, we calculated the relationship between the predicted occurrence and the minimum
temperature of the coldest month by extracting values from all grid cells in the study area.
Minimum temperature represents the most limiting factor for insect distributions (Jarosik et al.,
2015). We calculated the average probability of occurrence for each temperature class (step of
0.1°C). We identified the turning-point temperature, marking the border of the climatic niche,
for each species by fitting GAMs, using the R-package ‘mgcv’ (version 1.8-31; Wood, 2011),
for the relationship between minimum temperature of the coldest month and average probability
for the temperature class. For this, we extracted minimum temperatures, corresponding to
occurrence probabilities in the range 0.4 to 0.6 (in steps of 0.05) from the fitted values of the
GAM (Figure S1). We ordered species according to the turning-point temperature and then
investigated the shape of the pattern of the number of species along the temperature gradient.
96
 CHAPTER II

We used the ‘segmented’ R-package (version 1.1.0; Muggeo 2003) to identify

breakpoints, which mark a change in slope in the ordered sequence of turning-point
temperatures. To find a suitable initial estimate of the number and location of the breakpoints,
we used the final estimates of the breakpoints and the slopes of the linear regression models for
the intervals between the breakpoints. We isolated the slopes of the two distinct clusters of
niches and scaled them to the relative number of species used for each modelling approach.
Finally, we mapped the breakpoint temperatures marking the end of the first cluster, the start
of the second cluster and the end of the second cluster to highlight the areas that are likely to
become climatically suitable for warm-adapted pests and potentially see a rapid accumulation
of newly occurring pests. We used representative values of -6°C, 0°C and 6°C as limits for
these domains, taking into account the good agreement between the breakpoints identified for
the two different modelling approaches (Figure S2).

To determine the inflection point marking the transition between the first and the second
cluster of pest species, we fitted a GAM to the ordered turning-points for each cut-off threshold
(0.4–0.6) and identified the inflection point temperature with the function ese in the ‘inflection’
R-package (version 1.3.5; Christopoulos, 2016). We compared the AIC values of the GAMs
with linear functions to check whether the data follows a non-linear distribution (Table S3). We
calculated the mean and the confidence interval (0.95%) for the inflection zone from the
inflection points of the different cut-off thresholds. All analyses were done in R version 3.6.3
(R Core Team, 2020).

Results

Model performance and niche limits

The development thresholds varied widely over the study species (Figure S3). LDTs
were between 1.6°C and 15°C. Degree days required to complete development ranged from
148 to 1800 and minimum killing temperature ranged from -40°C to 4.5°C. We obtained
acceptable model sensitivity (> 0.5) for all 75 species, with a median of 0.97. We found lower
niche limits between -19.3°C and 10.8°C, with a median over all species of -7.7°C. The upper
niche limit was the same for all species (12.4°C) and the median niche breadth was 20.3°C. For
SDMs, we gathered occurrence records for 173 pest species. We obtained satisfying model

97
 CHAPTER II

performance for 159 pest species, with a median model performance of 0.83 in terms of AUC
score and 0.59 for TSS. We found lower niche limits between -22.8°C and 11.9°C, and upper
limits between -4.4°C and 12.4°C. The median lower niche limit was 0.1, whereas the median
upper niche limit was 12.4°C. We estimated a median niche breadth of 9.2°C.

Agreement between modelling approaches

Generally, we observed good agreement between the two modelling approaches

regarding pest species distributions, despite large variation among species and regions.
Comparing the pair of model projections for all species, we found a median Pearson correlation
coefficient between the occurrence probability predictions of the two modelling approaches of
0.68 under current climatic conditions, ranging from -0.8 to 0.88 and with a standard deviation
of 0.31. For binary projections, this corresponded to agreement of predictions in 85% of the
grid cells. For future climate change scenarios, the pairwise model agreement decreased until
2100, where we found a median Pearson correlation of 0.46 with higher uncertainty (standard
deviation 0.48), corresponding to agreement in 70% of the grid cells (Figures S4 & S5).
Concerning the spatial model agreement, we observed differences in the community similarity
predicted by the two modelling approaches. Disagreement arose mainly in Northern and
Northeastern Europe, as well as in mountain ranges (Figure 1). Under future projections, the
community similarity increased towards Northeastern Europe, but decreased in southern
regions. The cumulated model disagreement showed that SDMs were responsible for the
discrepancies in southern regions, because they were more restricted at southern range borders
(Figures S6 & S7). When comparing the niche breadth predicted by the two modelling
approaches, we observed much broader niches predicted by physiological models than by
SDMs. The median of the pairwise niche breadth difference was 3.8°C.

98
 CHAPTER II

Figure 1: Maps of model agreement (Sørensen similarity index) for 2010 (a) and 2100 (b). Blue/red colours
indicate lower/higher community similarity of predicted species. Results in (b) are based on the RCP8.5 climate
change scenario (for results based on the RCP4.5 scenario, see Figure S8).

Shift in modelled pest distribution

Regarding changes in pest species richness over time, agreement between the modelling
approaches was good over Central Europe and the UK (Figure 2), while in marginal areas
(Eastern Europe, in particular Belarus and the Ukraine) changes were predicted differently by
the two modelling approaches. We observed an increase in Northeastern Europe and
mountainous regions of up to 30 species with physiological model projections and up to around
70 species with SDM projections. This corresponds to about 50% of species for physiological
models and 40% for SDMs. When comparing the results of the different RCP scenarios, we
found a greater increase in the number of species with a suitable climate under RCP8.5 than
under RCP4.5. In particular, physiological models showed a much greater increase in northern
regions by 2100. For SDMs, the increase across Europe was very similar under the two
scenarios, but differences arose in southern regions (e.g. Iberian peninsula), where fewer
species with suitable climatic conditions were predicted for the year 2100 under RCP8.5.

99
 CHAPTER II

Figure 2: Change in the number of species between 2010 and 2100, as predicted by (a) physiological models and
(b) SDMs. The number of modelled pest species was higher for the SDMs (159) than for the physiological models
(75), and we therefore scaled the changes to 100 species to enable model comparison despite this difference. The
white area in central Spain marks the region where maximum monthly temperatures above the assumed upper
lethal temperature of 40°C in 2100 prevented species occurrence. Results shown here are based on the RCP8.5
climate change scenario (for results based on the RCP4.5 scenario, see Figure S9).

Relationship between predicted niche distribution and temperature

We observed a non-linear relationship between species richness and temperature with

the physiological models and SDMs (Figure 3). Temperature niches showed two distinct groups
with both modelling approaches. The first group included cold-adapted species, predicted to
occur in regions that experience minimum temperatures of the coldest month below the
breakpoint temperature of -6°C. The second group consisted of warm-adapted species, existing
only in regions where minimum temperatures of the coldest month never fall below the
breakpoint temperature of 0°C. For both modelling approaches, there was only a partial overlap
between cold-adapted and established species, and between warm-adapted and quarantine
pests. In the warm-adapted group, quarantine species were overrepresented (Figure S10). The

100
 CHAPTER II

inflection point, marking the switch between the two groups, i.e. the inflection point,
corresponded to a minimum temperature of the coldest month of -3.0°C for physiological model
predictions and -3.1°C for SDM predictions. Additionally, we calculated inflection points for
the subset of species for which we generated physiological models and SDMs (53 species).
Including only these species, we observed inflection points at -3.0°C (physiological models;
±1.5°C) and -1.7°C (SDMs; ±1.5°C).

Figure 3: Climatic niches of pest species along a gradient of the minimum temperature of the coldest month. Light
blue areas show the range in the climatic niche predicted by the physiological models (a) and SDMs (b) for each
species, sorted by the turning-point temperature obtained with a cut-off threshold of 0.5. Denser blue for the
temperature range of a species corresponds to a higher proportion of grid cells with this temperature predicted to
be suitable for the species (prob). The dark blue lines show the distribution of turning-point temperatures.
Inflection points were identified at -3.0°C for physiological model predictions and -3.1°C for SDMs (indicated by
the red dashed lines). Yellow boxes indicate the group of cold-adapted species (up to the breaking point of -6°C)
and red boxes the group of warm adapted species (between the breaking points of 0°C and 6°C). For the subset
with 53 species, see Figure S11.

The spatial dynamics of the inflection point indicated clear shifts towards the northeast
(Figure 4). The breakpoints (-6°C, 0°C, 6°C) are predicted to be dislocated at different rates.
While the area above the -6°C isoline, marking the border of suitability for all cold-adapted
species, almost covered the entire European continent, the other two isolines, marking the start
and the end of the cluster of warm-adapted pests, were more stable. The area between 0°C and
6°C in Figure 4 corresponds to where warm-adapted species were predicted to encounter
suitable climatic conditions. This area increased mainly in Western and Central Europe, but
was predicted to reach the Atlantic coast of Norway in the North, as well as Poland and the

101
 CHAPTER II

coast of the Baltic Sea in the East, by 2100. Slopes of the increasing number of pests with
suitable climatic conditions varied between the two groups. With physiological models we
found an increase of about five species per degree for cold-adapted species and seven species
per degree for warm-adapted species, and SDMs predicted four (cold-adapted) and seven
(warm-adapted) species per degree.

Figure 4: Figure 4: Spatial analysis of the inflection point across Europe. The maps show the minimum
temperature of the coldest month across Europe in 2010 (a), 2040 (b), 2070 (c) and 2100 (d) under the RCP8.5
scenario. The yellow shaded area marks the region where minimum temperatures of the coldest month is below
the -6°C breaking point. The red shaded area marks the region between the 0°C and the 6°C breaking points, where
a strong acceleration of pest climatic suitability is expected. Figure S12 shows the results for the RCP4.5 scenario.

102
 CHAPTER II

Discussion

In this study, we compared predictions of species distributions by physiological models

and SDMs to investigate climatic niches of a comprehensive set of insect pest species of
agricultural crops. Employing both physiological (deductive) and statistical (inductive) models
entails robustness in cases of agreement and prompts interesting hypotheses when differences
arise (Hijmans & Graham, 2006). We show general agreement in the predicted species
distribution between the two modelling approaches. Moreover, with both modelling
approaches, we found inflection points around -3°C in the minimum temperature of the coldest
month, indicating increasing pest pressure after this threshold temperature is crossed.

Physiological and statistical models provide coherent results despite conceptual

differences. Model projections for current climatic conditions showed good agreement for
pairwise suitability index predictions 0.68), corresponding to 85% of the grid cells for binary
projections. We found the highest model agreement in areas where both approaches predicted
similar pest community composition to encounter suitable climatic conditions. We found lower
similarity in Northern Europe (i.e. Scandinavia and the European part of Russia), as well as in
high mountain ranges (Pyrenees, Alps and Carpathian Mountains), mainly because few insect
pest species were predicted to occur in those regions. Furthermore, the niche breadth obtained
from models reflects differences between the fundamental and the realized niche. We expected
niches modelled based on physiological limits to be broader than realized niches (Venette et
al., 2010; Soberón & Arroyo-Peña, 2017). The results confirmed this expectation, with realized
pest niches predicted by SDMs (median niche breadth 9.2°C) much lower than fundamental
climatic niches predicted by physiological models (17.1°C). Pairwise comparisons likewise
suggested narrower climatic niches predicted with SDMs than with physiological models. The
broader climatic niches predicted with physiological models indicates that SDMs may
underestimate the climatically suitable area, implying that the regions of potential invasions
could be larger than estimated based on SDMs alone. Yet, physiological models have also a
methodological limitation regarding upper development thresholds, because data are not
available for most species. Available data for 31 species indicated an average of 34.4°C, with
only two species having their upper development thresholds below 30°C. Monthly mean
temperatures above 34°C do not occur in Europe under the current climate and are unlikely to
occur during this century (Figure S13) and thus it is unlikely that including upper development
thresholds would have altered our predictions for pest species.

103
 CHAPTER II

Forecasts of species distributions under climate change have been reported to be

congruent for deductive and inductive models (Kearney, Wintle & Porter, 2010), although
inductive models have also been reported to produce more pessimistic predictions than
deductive models (Lobell & Asseng, 2017). Here, we found a decline in model agreement for
future projections, with a decrease in median correlation from 0.68 in 2010 to 0.46 in 2100,
although the latter value still represents relatively good agreement for binary projections (70%).
We observed that the southern range borders modelled with physiological models remained
mostly in the same locations, while SDM borders moved towards higher latitudes, resulting in
lower community similarity in Southern Europe and decreasing model agreement. SDMs for
future projections are more prone to extrapolation errors in areas where new climatic conditions
will occur, which is not the case for physiological models (Kearney & Porter, 2009). This
problem is particularly important for invasive species because the distribution is often projected
to new environmental conditions (Elith, 2017), which could explain the decreasing model
agreement and increasing differences in pairwise comparisons as time progresses.

On the individual species level, we observed differences between modelling approaches

in the predictions of the area with suitable climatic conditions for a few species (e.g.
Dendroctonus ponderosae; Pearson correlation of -0.8). For these species, the occurrence
records used for the SDMs may not reflect the climatic niche represented by the physiological
thresholds obtained from laboratory studies. For other species, we observed a good spatial
match between the two approaches regarding distribution ranges (e.g. Spodoptera litura;
Pearson correlation of 0.88), suggesting that the occurrence records for these species well
represent their fundamental climatic niche, giving additional robustness to the prediction
(Hijmans & Graham, 2006). For future investigations, we suggest using ensemble approaches
of physiological models and SDMs for a proper understanding of the modelled system, as
proposed in previous studies (Overmars, de Groot & Huigen, 2007). One approach could be to
restrict the predicted distribution from correlative SDMs with the fundamental niche produced
by physiological models (Kearney & Porter 2009). Physiological models alone are too general,
predicting only the fundamental climatic niche, while SDMs may be too restricting, capturing
non-accountable factors (e.g. restriction through competition), which distort model projections
to future conditions (Sinclair, White and Newell, 2010). However, physiological models give a
more direct biological understanding than statistical models, which can be more valuable than
exact prediction (Lobell & Asseng, 2017).

104
 CHAPTER II

Our results suggest that climate change will lead to an accelerated increase in pest
pressure over large parts of Europe in the near future. We observed a non-linear relationship
between niche limit positions and the prevalence of insect pest species. Crossing the inflection
point implies a transition from cold- to warm-adapted species, and crossing the breakpoint
temperature of 0°C of minimum temperature of the coldest month implies an abrupt increase in
the number of pests with suitable climatic conditions and therefore in the number of pest species
that may threaten European crop production. Previous studies have shown that climate change
will entail northward movement of insect pest species in the near future (Bebber et al., 2013;
Grünig et al., 2020). Here, we identified an inflection point around -3°C in the minimum
temperature of the coldest month, beyond which the number of pest species with suitable
climatic conditions in Europe accelerates.

Finding an inflection point for the minimum temperature of the coldest month below
0°C would be reasonable, given the physiology of insects and their reaction to freezing. Cold
tolerance limits the distribution of many insects that do not go through winter diapause (Bale
& Hayward, 2010). As we considered only monthly mean temperatures, the temperatures
actually experienced by insects are likely to be more extreme than those reported here. For a
monthly minimum temperature around -3°C, freezing events are very likely to occur, even in
microhabitats that potentially buffer air temperature fluctuations and provide shelter to insects
for overwintering (Danks, 1978). Minimum temperature and freezing events are known to act
as limiting factors for insect spreading and protect many regions from invasions (Maxmen,
2013; Jarosik et al., 2015), underlining the importance of the inflection point.

Based on the non-linear response of pest niche distributions along a temperature

gradient, we investigated the temporal and spatial dynamics of inflection points under climate
change. We showed that the presence of two groups of species, warm-adapted and cold-adapted,
could lead to two waves of increasing pest pressure. While northern regions are currently facing
the first wave of insect pest invasion, with the cluster of cold-adapted species, Southern and
Central European regions have already passed the inflection point temperatures, and are
therefore expected to face the second wave, with the cluster of warm-adapted species, in the
near future. The expected timing of the arrival of the second wave in these regions depends on
the RCP scenario. Our results support the findings of previous studies, showing that abating

105
 CHAPTER II

CO2 emissions could be crucial in preventing the impact of pests on crop yields (Deutsch et al.,
2018), as we find a greater number of species with suitable climatic conditions under the
RCP8.5 scenario than under RCP4.5. Moreover, our results show that some areas are already
confronted with increased pest pressure, including most parts of Southern Europe, the British
Isles and Western Europe. Indeed, these regions serve regularly as entry gates to Europe for
invasive crop pests, such as Drosophila suzukii (Calabria et al., 2012) and Tuta absoluta
(Desneux et al., 2010). France and Italy have the highest recorded numbers of established alien
invertebrates in Europe (Roques et al., 2009). While these countries are part of the major
pathways of global trade and entry gates to Europe for international shipping traffic, the
Mediterranean climate has mild winters, supporting the establishment of more species than in
northern regions. These findings support the importance of border control and improved
inspection capacity with increasing trading volume (Poland & Rassati 2019).

The analyses applied in our study have limitations arising from data availability and
differences in the total number of species for which suitable models could be developed
depending on the modelling approach. Including data on more detailed requirements for insect
development (e.g. on diapause initiation and termination) would improve the individual
physiological models, however such data is scarce and not available for a broad range of
species. Further, the size of the sample influences the outcome of the general response of species
prevalence with respect to temperature. We tried to circumvent this problem by including data
for a wide range of pests, including both cold- and warm-adapted species, but our list of pests
is not complete. In addition, the global species pool for invasive insects shows no sign of
saturation (Seebens et al., 2017). Indeed, quarantine (i.e. potentially invasive) species were
underrepresented in this study, suggesting an even stronger acceleration of pest pressure after
the inflection point is crossed. We observed that the inferred inflection zones for pests are very
similar, irrespective of the chosen modelling approach and despite the fact that different sets of
species were included in the analysis (75 species with physiological models and 159 with
SDMs). We checked whether the same pattern occurs with only a subset of the species by only
using the 53 species for which we could develop models for both the deductive and the inductive
approach. The results confirmed the existence of an inflection point around -3°C. Within the
warm-adapted cluster, the majority are quarantine species, implying a high invasion risk in new
areas if minimum temperatures exceed the inflection point in these regions. Established pest
species, for which we also predicted range shifts towards higher latitudes and expanding areas
with suitable climatic conditions within Europe, heighten pest pressure in these regions.

106
 CHAPTER II

Increasing temperatures will not only promote more invasions of quarantine pests, but also
increase the spread of established pests, threatening agricultural cropping areas that are
expanding to more northern latitudes in response to global warming.

Conclusions

We used deductive and inductive models to highlight trends of increasing climate

suitability for insect pest species across Europe and a non-linear distribution of their climatic
niches along a minimum temperature gradient. We found good model agreement between
physiological models and SDMs under current climatic conditions. Forecasts under climate
change showed diverging model agreement for pest species over time, indicating increasing
uncertainty. Further, investigating the non-linear relationship between pest prevalence and
minimum temperature made it possible to identify an inflection point beyond which the number
of pest species with suitable climatic conditions increases rapidly. By mapping the inflection
temperature spatially, we showed temporal and spatial dynamics of potential pest pressure
under future climate change. Such information can inform policy-makers and stakeholders on
where and when climatic conditions approach the transition point for the onset of accelerated
pest invasions. New insights are necessary for planning crop protection strategies that can
effectively help control the new threats. Ultimately, with expanding areas in Europe becoming
susceptible to pest pressure in the near future, we advise a strengthening of surveillance
measures in general and border control in particular, and for an improvement in inspection
capacity.

Data Sharing and Accessibility statement

The data and code that support the findings of this study are available from the
corresponding author upon reasonable request.

107
 CHAPTER II

References

Allouche, O., Tsoar, A., & Kadmon, R. (2006). Assessing the accuracy of species distribution
models: prevalence, kappa and the true skill statistic (TSS). Journal of Applied Ecology,
43(6), 1223-1232.
Araújo, M., et al. (2019). Standards for distribution models in biodiversity assessments. Science
Advances, 5(1), eaat4858.
Arthur, F. H., Morrison, W. R., & Morey, A. C. (2019). Modeling the potential range expansion
of larger grain borer, Prostephanus truncatus (Coleoptera: Bostrichidae). Scientific
Reports, 9(1), 1-10.
Bale, J., & Hayward, S. (2010). Insect overwintering in a changing climate. Journal of
Experimental Biology, 213(6), 980-994.
Baselga, A., Orme, D., Villeger, S., De Bortoli, J., & Leprieur, F. (2017). Partitioning beta
diversity into turnover and nestedness components. R package betapart, version 1-4.
Bebber, D. P., Ramotowski, M. A., & Gurr, S. J. (2013). Crop pests and pathogens move
polewards in a warming world. Nature Climate Change, 3(11), 985.
Bebber, D. P., Holmes, T., & Gurr, S. J. (2014). The global spread of crop pests and pathogens.
Global Ecology and Biogeography, 23(12), 1398-1407.
Brockerhoff, E., & Liebhold, A. (2017). Ecology of forest insect invasions. Biological
Invasions, 19(11), 3141-3159.
Calabria, G., Máca, J., Bächli, G., Serra, L., & Pascual, M. (2012). First records of the potential
pest species Drosophila suzukii (Diptera: Drosophilidae) in Europe. Journal of Applied
Entomology, 136(1‐2), 139-147.
Christopoulos, D. T. (2016). Inflection: Finds the inflection point of a curve. R package version
1.
R Core Team. (2020). R: a language and environment for statistical computing. R Foundation
for statistical computing, Vienna, Austria.
Dakos, V., Matthews, B., Hendry, A., Levine, J., Loeuille, N., Norberg, J., & De Meester, L.
(2019). Ecosystem tipping points in an evolving world. Nature Ecology and Evolution,
3(3), 355-362.
Danks, H. V. (1978). Modes of seasonal adaptation in the insects: I. Winter survival. The
Canadian Entomologist, 110(11), 1167-1205.
Descombes, P., et al. (2015). Forecasted coral reef decline in marine biodiversity hotspots under
climate change. Global Change Biology, 21(7), 2479-2487.

108
 CHAPTER II

Desneux, N., Wajnberg, E., Wyckhuys, K.A., Burgio, G., Arpaia, S., Narváez-Vasquez, C.A.,
González-Cabrera, J., Ruescas, D.C., Tabone, E., Frandon, J. & Pizzol, J. (2010).
Biological invasion of European tomato crops by Tuta absoluta: ecology, geographic
expansion and prospects for biological control. Journal of Pest Science, 83(3), 197-215.
Deutsch, C. A., Tewksbury, J. J., Tigchelaar, M., Battisti, D. S., Merrill, S. C., Huey, R. B., &
Naylor, R. L. (2018). Increase in crop losses to insect pests in a warming climate. Science,
361(6405), 916-919.
Elith, J. (2017). Predicting distributions of invasive species. Invasive species: Risk Assessment
and Management, pp. 93-129.
Elith, J., & Leathwick, J. (2009). Species distribution models: Ecological explanation and
prediction across space and time. Annual Review of Ecology, Evolution, and Systematics,
40(1), 677-697.
Fader, M., Shi, S., Von Bloh, W., Bondeau, A., & Cramer, W. (2016). Mediterranean irrigation
under climate change: More efficient irrigation needed to compensate for increases in
irrigation water requirements. Hydrology and Earth System Sciences, 20(2), 953-973.
Feldmann, C., & Hamm, U. (2015). Consumers’ perceptions and preferences for local food: A
review. Food Quality and Preference, 40, 152-164.
Freeman, E. A., & Moisen, G. (2008). PresenceAbsence: An R package for presence absence
analysis. Journal of Statistical Software, 23 (11), 31.
Grünig, M., Mazzi, D., Calanca, P., Karger, D. N. & Pellissier, L. (2020). Crop and forest pest
metawebs shift towards increased linkage and suitability overlap under climate change.
Communications Biology 3, 233.
Guisan, A., & Zimmermann, N. E. (2000). Predictive habitat distribution models in ecology.
Ecological modelling, 135 (2-3), 147-186.
Hastie, T. (2019). gam: Generalized Additive Models. R package version 1.16.1.
https://CRAN.R-project.org/package=gam
Heinrich, B. (1981). Ecological and evolutionary perspectives. Insect Thermoregulation, 235-
302.
Hijmans, R. J., Guarino, L., Cruz, M., & Rojas, E. (2001). Computer tools for spatial analysis
of plant genetic resources data: 1. DIVA-GIS. Plant Genetic Resources Newsletter, 15-
19.
Hijmans, R. J., & Graham, C. H. (2006). The ability of climate envelope models to predict the
effect of climate change on species distributions. Global Change Biology, 12(12), 2272-
2281.

109
 CHAPTER II

Hosmer Jr, D. W., Lemeshow, S., & Sturdivant, R. X. (2013). Applied logistic regression (Vol.
398). (John Wiley & Sons, New Jersey).
Hulme, P. (2009). Trade, transport and trouble: Managing invasive species pathways in an era
of globalization. Journal of Applied Ecology, 46(1), 10-18.
Hutchinson, G. E. (1957). A treatise on limnology. Vol. I, Part 1: Geography and physics of
lakes. XIV, 540 S., 154 Abb., 63 Tab., 11 Tafeln. Vol. I, Part 2: Chemistry of lakes. XIV,
474 S., 74 Abb., 69 Tab. (Verbesserter Neudruck der Ausgabe von 1957). New York–
London–Sydney–Toronto: John Wiley & Sons.
IPCC (2007). Climate change 2007: the physical science basis: summary for policymakers.
Geneva: IPCC. Jarošík, V., Honěk, A., Magarey, R., & Skuhrovec, J. (2011).
Developmental Database for Phenology Models: Related Insect and Mite Species Have
Similar Thermal Requirements. Journal of Economic Entomology, 104(6), 1870-1876.
Jarošík, V., Kenis, M., Honěk, A., Skuhrovec, J., & Pyšek, P. (2015). Invasive insects differ
from non-invasive in their thermal requirements. PLoS ONE, 10(6).
Karger, D., Conrad, O., Böhner, J., Kawohl, T., Kreft, H., Soria-Auza, R., & Kessler, M. (2017).
Climatologies at high resolution for the earth's land surface areas. Scientific Data, 4.
Kearney, M., & Porter, W. P. (2004). Mapping the fundamental niche: physiology, climate, and
the distribution of a nocturnal lizard. Ecology, 85(11), 3119-3131.
Kearney, M., & Porter, W. (2009). Mechanistic niche modelling: combining physiological and
spatial data to predict species’ ranges. Ecology Letters, 12(4), 334-350.
Kearney, M., Wintle, B., & Porter, W. (2010). Correlative and mechanistic models of species
distribution provide congruent forecasts under climate change. Conservation Letters,
3(3), 203-213.
Lamichhane, J. R., Dachbrodt-Saaydeh, S., Kudsk, P., & Messéan, A. (2016). Toward a reduced
reliance on conventional pesticides in European agriculture. Plant Disease, 100(1), 10-
24.
Lenton, T. M. (2011). Early warning of climate tipping points. Nature Climate Change, 1(4),
201-209.
Lobell, D. B., & Asseng, S. (2017). Comparing estimates of climate change impacts from
process-based and statistical crop models. Environmental Research Letters, 12(1),
015001.
MacDonald, G. (2010). Global warming and the Arctic: A new world beyond the reach of the
Grinnellian niche? Journal of Experimental Biology, 213(6), 855-861.

110
 CHAPTER II

Maracchi, G., Sirotenko, O., & Bindi, M. (2005). Impacts of present and future climate
variability on agriculture and forestry in the temperate regions: Europe. Climatic change,
70(1-2), 117-135.
Maxmen, A. (2013). Crop pests: under attack. Nature, 501(7468), S15-S17.
Muggeo, V.M.R. (2003) Estimating regression models with unknown break-points. Statistics
in Medicine 22, 3055–3071.
Newman, J. A. (2005). Climate change and the fate of cereal aphids in Southern Britain. Global
Change Biology, 11(6), 940-944.
Overmars, K., de Groot, W., & Huigen, M. (8 2007). Comparing inductive and deductive
modeling of land use decisions: Principles, a model and an illustration from the
Philippines. Human Ecology, 35(4), 439-452.
Paini, D., Sheppard, A., Cook, D., De Barro, P., Worner, S., & Thomas, M. (2016). Global
threat to agriculture from invasive species. Proceedings of the National Academy of
Sciences of the United States of America, 113(27), 7575-7579.
Poland, T., & Rassati, D. (2019). Improved biosecurity surveillance of non-native forest insects:
a review of current methods. Journal of Pest Science, 92(1), 37-49.
Ramirez-Villegas, J., Jarvis, A., & Läderach, P. (2013). Empirical approaches for assessing
impacts of climate change on agriculture: The EcoCrop model and a case study with grain
sorghum. Agricultural and Forest Meteorology, 170, 67-78.
Robertson, M. P., Peter, C. I., Villet, M. H., & Ripley, B. S. (2003). Comparing models for
predicting species’ potential distributions: a case study using correlative and mechanistic
predictive modelling techniques. Ecological Modelling, 164(2-3), 153-167.
Robinet, C., & Roques, A. (2010). Direct impacts of recent climate warming on insect
populations. Integrative Zoology, 5(2), 132-142.
Roques, A., Rabitsch, W., Rasplus, J. Y., Lopez-Vaamonde, C., Nentwig, W., & Kenis, M.
(2009). Alien terrestrial invertebrates of Europe. In Handbook of alien species in Europe
(pp. 63-79). Springer, Dordrecht.
Sanderson, B., Knutti, R., & Caldwell, P. (2015). Addressing interdependency in a multimodel
ensemble by interpolation of model properties. Journal of Climate, 28(13), 5150-5170.
Schaub, L., Breitenmoser, S., Derron, J., & Graf, B. (2017). Development and validation of a
phenological model for the univoltine European corn borer. Journal of Applied
Entomology, 141(6), 421-430.
Schmidhuber, J., & Tubiello, F. N. (2007). Global food security under climate change.
Proceedings of the National Academy of Sciences, 104(50), 19703-19708.

111
 CHAPTER II

Seebens, H., Blackburn, T. M., Dyer, E. E., Genovesi, P., Hulme, P. E., Jeschke, J. M., ... &
Bacher, S. (2017). No saturation in the accumulation of alien species worldwide. Nature
Communications, 8(1), 1-9.
Sinclair, S. J., White, M. D., & Newell, G. R. (2010). How useful are species distribution
models for managing biodiversity under future climates? Ecology and Society, 15(1), 8.
Sloat, L. L., Davis, S. J., Gerber, J. S., Moore, F. C., Ray, D. K., West, P. C., & Mueller, N. D.
(2020). Climate adaptation by crop migration. Nature Communications, 11(1), 1-9.
Sutherst, R. W. & Maywald, G. F. (1985). A computerised system for matching climates in
ecology. Agriculture, Ecosystems & Environment, 13(3-4), 281-299.
Tonnang, H. E., Hervé, B. D., Biber-Freudenberger, L., Salifu, D., Subramanian, S., Ngowi, V.
B., ... & Niassy, S. (2017). Advances in crop insect modelling methods—Towards a
whole system approach. Ecological Modelling, 354, 88-103.
Tubiello, F. N., Soussana, J. F., & Howden, S. M. (2007). Crop and pasture response to climate
change. Proceedings of the National Academy of Sciences, 104(50), 19686-19690.
Tuck, G., Glendining, M. J., Smith, P., House, J. I., & Wattenbach, M. (2006). The potential
distribution of bioenergy crops in Europe under present and future climate. Biomass and
Bioenergy, 30(3), 183-197.
Venette, R., Kriticos, D., Magarey, R., Koch, F., Baker, R., Worner, S., … & Pedlar, J. (2010).
Pest Risk Maps for Invasive Alien Species: A Roadmap for Improvement. BioScience,
60(5), 349-362.
Walther, G. R., Post, E., Convey, P., Menzel, A., Parmesan, C., Beebee, T. J., ... & Bairlein, F.
(2002). Ecological responses to recent climate change. Nature, 416(6879), 389.
Walther, G. R., Roques, A., Hulme, P. E., Sykes, M. T., Pyšek, P., Kühn, I., ... & Czucz, B.
(2009). Alien species in a warmer world: risks and opportunities. Trends in Ecology &
Evolution, 24(12), 686-693.
Wisz, M. S., Hijmans, R. J., Li, J., Peterson, A. T., Graham, C. H., Guisan, A. NCEAS
Predicting Species Distributions Working Group. (2008). Effects of sample size on the
performance of species distribution models. Diversity and Distributions, 14(5), 763-773.
Wood, S. (2006). Low-rank scale-invariant tensor product smooths for generalized additive
mixed models. Biometrics, 62(4), 1025-1036.
Wood, S. N. (2011) Fast stable restricted maximum likelihood and marginal likelihood
estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society
(B), 73(1),3-36.

112
 CHAPTER II

Acknowledgements

This work contributes to the newly established Swiss National Centre for Climate
Services (www.nccs.ch) and received financial support by the Swiss Federal Office of
Agriculture.

Author contributions

M.G. contributed to conceiving ideas, gathering distribution data, establishing and

running models, performing analysis and writing the manuscript; D.M. contributed to
conceiving ideas and writing; P.C contributed to conceiving ideas, performing analysis and
writing; L.P. contributed to conceiving ideas, establishing models and writing. All authors gave
final approval for publication.

113
 CHAPTER II

Supplementary Material

Note S1: Variable selection procedure

We modelled all species using two variables: minimum temperature of the coldest
month and growing degree days above 5°C. We investigated model performance with area
under the ROC-plot curve (AUC) and true skill statistics (TSS) values, as well as visual
inspection of the projections. For species not reaching acceptable performance, we tested the
models using second-order polynomials and later also using one of the two variables alone or
in combination with monthly maximum temperature. As a result, for 23 crops and 24 pest
species we used growing degree days as a single predictor variable, for two crop species we
used maximum temperature as a single predictor variable, and for two pest species we used
maximum temperature in combination with growing degree days. For eight pest species we
used minimum temperature and maximum temperature as predictor variables. All species with
the variables used for their models are listed in Table S1.

114
 CHAPTER II

Supplementary Figures

Figure S1: Schematic display of the method used to find the turning point for each species. The blue line represents
the fitted GAM and the red lines indicate the temperatures of the turning points for all thresholds between 0.4 and
0.6 in steps of 0.05.

115
 CHAPTER II

Figure S2: Analysis using the ‘segmented’ package in R to identify breakpoints in the distribution of turning
points. The upper panel shows results for physiological models, and the lower panel shows results for SDMs.

116
 CHAPTER II

Figure S3: Histograms of species development thresholds used in the study. We could not obtain data on minimum
killing temperature (Tkill,min) for all species, and this variable therefore contains a smaller number of data points
than lower development threshold (LDT) and sum of effective temperatures (SET), i.e. the number of growing
degree days (GDD) above the LDT required for the completion of a generation.

117
 CHAPTER II

Figure S5: Median of model agreement as Pearson correlation of suitability indices over time, with RCP8.5 in the
upper panel and RCP4.5 in the lower panel.

118
 CHAPTER II

Figure S6: Median of model agreement as percentage of agreeing grid cells over time, with RCP8.5 in the upper
panel andRCP4.5 in the lower panel.

119
 CHAPTER II

Figure S6: Absolute model disagreement between physiological models and SDMs under the RCP8.5 climate
change scenario. We subtracted the raster layer of the SDM projection from the raster layer of the physiological
model projection, and summed up the resulting layers of all species. Red areas show where SDMs predict more
species occurrence than physiological models, and blue areas vice versa. The panels show disagreement (a) for
pests in 2010, (b) for pests in 2100, (c) for crops in 2010, and (d) for crops in 2100.

Figure S7: Same as Figure S6 but for climate change scenario RCP4.5.

120
 CHAPTER II

Figure S8: Maps of model agreement (Sorensen similarity index) for 2010 (a) and 2100 (b). Blue/red colours
indicate lower/higher community similarity of predicted species. Results in (b) are based on the RCP4.5 climate
change scenario (for results based on the RCP8.5 scenario, see Figure 1).

121
 CHAPTER II

Figure S9: Same as Figure 3 but for RCP4.5.

122
 CHAPTER II

Figure S10: Turning-point lines for the five different cut-off thresholds (red lines). The blue polygon indicates the
inflection zone, including the median (black lines). Red and black marks on the left side of the graphs show the
category of pests: invasive (red) or established (black).

123
 CHAPTER II

Figure S11: Same as Figure 3 but for the subset of 53 species for each modelling approach.

124
 CHAPTER II

Figure S12: Same as Figure 4 but for physiological models.

125
 CHAPTER II

Figure S13: Areas predicted to reach a monthly mean temperature of more than 34°C in 2100 in one or more
months (red). Values were calculated as the mean of the four GCMs in the RCP8.5 scenario.

126
 CHAPTER II

Supplementary tables

Table S1: List of pest species modelled with physiological models. Quarantine species are indicated with the value
1, established species with the value 0.

Species Quarantine/Established
Acrolepiopsis assectella 0
Adoxophyes orana 0
Aleurocanthus woglumi 1
Anarsia lineatella 0
Anastrepha fraterculus 1
Anastrepha ludens 1
Anoplophora glabripennis 1
Anthonomus grandis 1
Anthonomus signatus 1
Bactrocera cucurbitae 1
Bactrocera dorsalis 1
Bactrocera latifrons 1
Bactrocera tryoni 1
Bactrocera zonata 1
Baris lepidii 1
Bemisia tabaci 0
Cacoecimorpha pronubana 0
Cacopsylla pyricola 0
Carposina sasakii 1
Ceratitis capitata 1
Ceratitis rosa 1
Choristoneura occidentalis 1
Choristoneura rosaceana 1
Thaumatotibia leucotreta 1
Cydia pomonella 0
Delia antiqua 0
Delia platura 0
Dendroctonus ponderosae 1
Diabrotica virgifera virgifera 1
Diaphorina citri 1
Diuraphis noxia 0
Euwallacea fornicatus 1
Frankliniella occidentalis 1
Grapholitha molesta 0
Halyomorpha halys 0
Helicoverpa armigera 1
Helicoverpa zea 1
Hyphantria cunea 0
Ips calligraphus 1
Ips cembrae 0
Ips typographus 0
Leptinotarsa decemlineata 1
Leucoptera malifoliella 0
Liriomyza bryoniae 0

127
 CHAPTER II

Liriomyza huidobrensis 1
Liriomyza sativae 1
Liriomyza trifolii 1
Lobesia botrana 0
Lycorma delicatula 1
Lymantria dispar 0
Mayetiola destructor 0
Neoleucinodes elegantalis 1
Orgyia pseudotsugata 1
Ostrinia nubilalis 0
Oulema melanopus 0
Pectinophora gossypiella 0
Phthorimaea operculella 0
Phyllotreta cruciferae 0
Pieris brassicae 0
Popilia japonica 1
Psacothea hilaris 0
Pseudaulacaspis pentagona 0
Rhagoletis cerasi 0
Rhagoletis indifferens 1
Rhagoletis pomonella 1
Scirtothrips dorsalis 1
Spodoptera eridania 1
Spodoptera frugiperda 1
Spodoptera littoralis 1
Spodoptera litura 1
Stephanitis takeyai 0
Tecia solanivora 1
Thrips palmi 1
Toxoptera citricida 0
Tuta absoluta 1

128
 CHAPTER II

Table S2: List of pest species modelled with species distribution models (SDMs). Variables used for the SDMs
for each species are shown. Quarantine species are indicated with the value 1, established species with the value
0.

Species Quarantine/Established Variable 1 Variable 2

Acrolepiopsis assectella 0 GDD MinTemp
Aleurocanthus woglumi 1 GDD MinTemp
Aleurodicus dispersus 1 GDD MinTemp
Amyelois transitella 1 GDD MinTemp
Anastrepha fraterculus 1 GDD MaxTemp
Anastrepha ludens 1 GDD MinTemp
Anastrepha obliqua 1 GDD MinTemp
Anoplophora chinensis 1 GDD MinTemp
Anoplophora glabripennis 1 GDD MinTemp
Anthonomus bisignifer 1 GDD MinTemp
Anthonomus grandis 1 GDD MinTemp
Anthonomus signatus 1 GDD NA
Argyrotaenia velutinana 1 GDD MinTemp
Aroga trialbamaculella 1 GDD MinTemp
Arvelius albopunctatus 1 GDD MinTemp
Bactericera cockerelli 1 GDD MinTemp
Bactrocera carambolae 1 GDD NA
Bactrocera cucurbitae 1 GDD MinTemp
Bactrocera latifrons 1 GDD MaxTemp
Bactrocera occipitalis 1 GDD NA
Bactrocera tryoni 1 GDD MinTemp
Bemisia tabaci 0 GDD MinTemp
Cacoecimorpha pronubana 0 GDD MinTemp
Cactoblastis cactorum 1 GDD MinTemp
Cameraria ohridella 0 GDD MinTemp
Carposina sasakii 1 GDD MinTemp
Ceratitis capitata 1 GDD NA
Ceratitis cosyra 1 GDD MinTemp
Ceratitis quinaria 1 MinTemp MaxTemp
Ceratitis rosa 1 GDD NA
Ceroplastes sinensis 0 MinTemp MaxTemp
Chinavia marginata 1 MinTemp MaxTemp
Chloridea virescens 1 GDD MinTemp
Chlorochroa sayi 1 GDD MinTemp
Choristoneura occidentalis 1 GDD MinTemp
Choristoneura rosaceana 1 GDD MinTemp
Chrysodeixis eriosoma 1 GDD MinTemp
Cingilia catenaria 1 GDD MinTemp
Conotrachelus nenuphar 1 GDD NA
Coscinoptycha improbana 1 GDD MinTemp
Cotinis nitida 1 GDD MinTemp
Cydia pomonella 0 GDD MinTemp
Dacus ciliatus 1 GDD MinTemp
Daktulosphaira vitifoliae 0 GDD NA
Delia antiqua 0 GDD MinTemp
129
 CHAPTER II

Dendroctonus ponderosae 1 GDD MinTemp

Diabrotica barberi 1 GDD NA
Diabrotica speciosa 1 GDD MinTemp
Diabrotica undecimpunctata howardi 1 GDD NA
Diabrotica undecimpunctata
undecimpunctata 1 GDD MinTemp
Diabrotica virgifera virgifera 0 GDD MinTemp
Diabrotica virgifera zeae 0 GDD MinTemp
Diaphorina citri 1 GDD MinTemp
Diaprepes abbreviatus 1 GDD MinTemp
Diuraphis noxia 0 GDD MinTemp
Drosophila suzukii 0 GDD MinTemp
Dryocosmus kuriphilus 0 GDD MinTemp
Egira curialis 1 MinTemp MaxTemp
Epicauta abadona 1 GDD MinTemp
Epicauta immaculata 1 GDD MinTemp
Epicauta occidentalis 1 GDD MinTemp
Epicauta vittata 1 GDD MinTemp
Epiglaea apiata 1 GDD MinTemp
Epilachna vigintioctopunctata 1 GDD MinTemp
Epitrix cucumeris 1 GDD MinTemp
Erthesina fullo 1 GDD MinTemp
Eudocima fullonia 1 GDD MinTemp
Euschistus conspersus 1 GDD MinTemp
Euschistus servus 1 GDD MinTemp
Euwallacea fornicatus 1 GDD MinTemp
Frankliniella occidentalis 0 GDD NA
Grapholita packardi 1 GDD MinTemp
Grapholitha molesta 0 GDD MinTemp
Halyomorpha halys 0 GDD MinTemp
Helicoverpa armigera 0 GDD MinTemp
Helicoverpa assulta 1 MinTemp MaxTemp
Helicoverpa punctigera 1 GDD NA
Helicoverpa zea 1 GDD MinTemp
Heteronychus arator 1 GDD MinTemp
Homalodisca vitripennis 1 GDD MinTemp
Ips typographus 0 GDD MinTemp
Lacanobia subjuncta 1 GDD MinTemp
Lamprolonchaea brouniana 1 GDD MinTemp
Leptinotarsa decemlineata 1 GDD MinTemp
Leptoglossus zonatus 1 GDD MinTemp
Leucinodes cordalis 1 GDD MinTemp
Leucinodes orbonalis 1 GDD MinTemp
Limonius californicus 1 GDD MinTemp
Lineodes integra 1 GDD MinTemp
Liriomyza bryoniae 0 GDD NA
Liriomyza huidobrensis 0 GDD NA
Liriomyza sativae 1 GDD MinTemp
Liriomyza trifolii 0 GDD MinTemp
Lissorhoptrus oryzophilus 1 GDD MinTemp
Listronotus bonariensis 1 GDD NA

130
 CHAPTER II

Lobesia botrana 0 GDD MinTemp

Lobiopa insularis 1 GDD MinTemp
Lycorma delicatula 1 GDD MinTemp
Lygus elisus 1 GDD MinTemp
Lygus hesperus 1 GDD NA
Lygus lineolaris 1 GDD MinTemp
Lygus shulli 1 GDD MinTemp
Lymantria dispar 0 GDD MinTemp
Maconellicoccus hirsutus 1 GDD MinTemp
Malacosoma americana 1 GDD MinTemp
Manduca sexta 1 GDD MinTemp
Megacopta cribraria 1 GDD MinTemp
Melanotus communis 1 GDD MinTemp
Metamasius hemipterus 1 GDD MinTemp
Naupactus leucoloma 1 GDD MinTemp
Neoceratitis cyanescens 1 MinTemp MaxTemp
Nipaecoccus viridis 1 GDD MinTemp
Nysius huttoni 0 GDD MinTemp
Ochropleura implecta 1 GDD MinTemp
Oemona hirta 1 GDD NA
Opogona sacchari 1 GDD NA
Orgyia pseudotsugata 1 GDD MinTemp
Orthosia hibisci 1 GDD MinTemp
Paralobesia viteana 1 GDD NA
Phlyctinus callosus 1 GDD NA
Phthia picta 1 GDD MinTemp
Phyllotreta cruciferae 0 GDD MinTemp
Phyrdenus divergens 1 GDD MinTemp
Platynota flavedana 1 GDD MinTemp
Platynota idaeusalis 1 GDD MinTemp
Platynota stultana 1 GDD MinTemp
Popilia japonica 1 GDD MinTemp
Popillia japonica 1 GDD MinTemp
Psacothea hilaris 0 GDD MinTemp
Rhagoletis cerasi 0 GDD MinTemp
Rhagoletis cingulata 0 GDD MinTemp
Rhagoletis completa 0 GDD MinTemp
Rhagoletis fausta 1 GDD MinTemp
Rhagoletis indifferens 1 GDD MinTemp
Rhagoletis pomonella 1 GDD MinTemp
Rhagoletis suavis 1 GDD MinTemp
Saperda candida 1 GDD MinTemp
Scirtothrips dorsalis 1 GDD MinTemp
Sparganothis sulfureana 1 GDD MinTemp
Spodoptera albula 1 GDD NA
Spodoptera eridania 1 GDD NA
Spodoptera frugiperda 1 GDD MinTemp
Spodoptera latifascia 1 GDD NA
Spodoptera littoralis 1 GDD NA
Spodoptera litura 1 GDD MinTemp
Spodoptera ornithogalli 1 GDD MinTemp

131
 CHAPTER II

Spodoptera praefica 1 GDD MinTemp

Strauzia longipennis 1 GDD MinTemp
Systena frontalis 1 GDD MinTemp
Thaumatotibia leucotreta 1 GDD MinTemp
Thrips imaginis 1 GDD MinTemp
Thrips palmi 1 GDD MinTemp
Trichoferus campestris 1 GDD MinTemp
Trioza erytreae 1 GDD NA
Tuta absoluta 1 GDD MinTemp
Unaspis citri 1 GDD MinTemp
Xylena nupera 1 GDD MinTemp
Zaprionus indianus 1 GDD MinTemp
Zonosemata electa 1 GDD MinTemp

132
CHAPTER II

133
CHAPTER III

Applying deep neural networks to predict incidence and phenology of plant

pests and diseases

Marc Grünig*1,4, Elisabeth Razavi1, Pierluigi Calanca2, Dominique Mazzi1, Jan Dirk Wegner3
and Loïc Pellissier4,5

1
Agroscope, RD Plant Protection, Wädenswil, Switzerland
2
Agroscope, RD Agroecology and Environment, Zurich, Switzerland
3
ETH Zürich, EcoVision Lab, Zurich, Switzerland
4
ETH Zürich, Landscape Ecology, Zurich, Switzerland
5
Swiss Federal Research Institute WSL, Birmensdorf, Switzerland

Manuscript status
Submitted to Ecosphere

134
 CHAPTER III

Abstract

A major challenge of agriculture is to improve the sustainability of food production

systems in order to provide enough food for a growing human population. Pests and pathogens
cause vast yield losses, while crop protection practices raise environmental and human health
concerns. Decision support systems provide detailed information on optimal timing and
necessity of crop protection interventions, but are often based on phenology models that are
time-, cost- and labour-intensive in development. Here, we aim to develop a data-driven
approach for pest damage forecasting, relying on big data and deep learning algorithms. We
present a framework for the development of deep neural networks for pest and pathogen damage
classification and show their potential for predicting the phenology of damages. As a case study,
we investigate the phenology of the pear leaf blister moth (Leucoptera malifoliella, Costa). We
employ a set of 52,322 pictures taken during a period of 19 weeks and establish deep neural
networks to categorize the images into six main damage classes. Classification tools achieved
good performance scores overall, with differences between the classes indicating that the
performance of deep neural networks depends on the similarity to other damages and the
number of training images. The reconstructed damage phenology of the pear leaf blister moth
matches mine counts in the field. We further develop statistical models to reconstruct the
phenology of damages with meteorological data and find good agreement with degree-day
models. Hence, our study indicates a yet underexploited potential for data-driven approaches to
enhance the versatility and cost-efficiency of plant pest and disease forecasting.

Keywords: Decision support system; deep neural network; image classification; insect pest;
phenological modelling.

135
 CHAPTER III

Introduction

Agriculture is facing major challenges concerning food security and food production for
a global human population predicted to grow to nine billion by 2050 (Godfray et al., 2010).
Besides shifting towards more plant-based diets and decreasing food waste (Shepon et al. 2018;
West et al. 2014), sustainable intensification is necessary for food and environmental security
(Godfray et al., 2010; Garnett and Godfray, 2012). The rapidly growing field of precision
agriculture uses modern information technology, including computer vision and artificial
intelligence, provides enormous potential to contribute to the goals of more sustainable
agriculture (Bongiovanni and Lowenberg-deBoer, 2004; Lindblom et al. 2017; Patricio and
Rieder, 2018). Referred to as “smart farming”, technological development is foreseen to support
a more efficient use of natural resources and better target plant protection from pests and
pathogens while minimizing hazards to environmental and human health (Chakraborty and
Newton, 2011; Tilman et al. 2011; Garnett et al. 2013; Walter et al., 2017). Pests and pathogens
are responsible for large yield losses (Oerke, 2006), and are often counteracted with pesticides
(Lamichhane et al. 2015). Raising concerns regarding the negative externalities of pesticides
on human health and environmental safety (Geiger et al. 2010; Tegtmeier and Duffy 2004,
Pimentel and Burgess, 2014) foster strategies to reduce the risks from their use without
compromising productivity and profitability (Lechenet et al., 2017). Reliable pest and pathogen
detection and prediction support more timely and precise interventions, and thus reduced
pesticide use.

Machine learning has great potential to assist the development of innovative methods
for pest and pathogen management, supporting more sustainable plant protection (Behmann et
al., 2015). The identification of pest and pathogens and the detection of damages on crops are
challenging for farmers, yet crucial for the decision on appropriate control measures (Martinelli
et al., 2015; Lamichhane et al., 2016). Deep neural networks (DNNs) and algorithms for image
classification (Goodfellow et al., 2016) can serve the detection of pests and pathogens for plant
protection (e.g. Mohanty et al. 2016; Sladojevic et al. 2016; Ferentinos 2018). For example,
convolutional neural networks (CNNs; Krizhevsky et al., 2012; Szegedy et al., 2015) have been
used to identify different pathogens on apple leaves, based on image data (Liu et al. 2018;
Zhong and Zhao 2020), and for the classification of insect pest species occurring on crops
(Cheng et al., 2017; Thenmozhi & Reddy, 2019). The goal of these technological
implementations is to help growers to recognize and detect pests and pathogens in the field,

136
 CHAPTER III

fostering faster and more self-reliant evaluation of the pest situation in situ and support
decision-making processes to optimize yields in a sustainable way (e.g. Sladojevic et al. 2016).
Appropriate timing of pesticide applications is not trivial (Tang et al., 2010) and anticipating
the damage can improve application precision. Temporal precision of pesticide application can
increase the efficacy and reduce the number of required applications, and therefore lower the
total amount (Möhring et al., 2020). Therefore, advancing novel technologies may lead to faster
and more efficient recognition processes and eventually contribute to decrease risks associated
to pesticide use, as well as yield losses.

Decision support systems (DSS) assist crop producers with the surveillance, decision on
optimal timing and anticipation for the need of pesticide applications (Samietz et al. 2007) and
big data in combination with deep learning can increase the precision of DSS. In general, DSS
for pests and pathogens rely on phenological models, where the timing of crucial events in the
life-cycle of those damaging organisms, measured under controlled temperature conditions, are
coupled with meteorological data to predict their seasonal occurrence. Phenological models
have reached good performances for the prediction of pest and pathogen occurrence, offering
an effective complement to field observations (e.g. Schaub et al. 2017), but are still time- and
cost- intensive in development. Resulting models often target specific pest-crop or pathogen-
crop systems, are location-dependent (Donatelli et al., 2017) and rarely updated once
established. In the meantime, the life cycles of many insects are altered by changing climate
conditions (Kingsolver et al. 2011). With warming climate, phenological shifts and disruption
of synchrony between host plants and pests are widespread reactions (Forrest, 2016). This leads
to increasing discrepancies between model predictions and actual observations, because
forecast models at the core of DSS are seldom reparametrized to account for altered insect
biology. Further, climate change promotes the introduction and spread of invasive species into
newly suitable, so far uncolonized regions (Bebber et al. 2013; Grünig et al., 2020), requiring
fast development of new DSS. Novel technologies to analyze big data based on deep learning
(LeCun, 2015) can support monitoring and deliver the baseline for developing phenological
models needed, in combination with weather data and forecasts, to anticipate pest damages.

Using 52,322 photographs taken under field and standard conditions during the spring-
summer of 2019, here we develop a framework toward pest phenology forecasting based on big
data and deep learning algorithms. We focus on a proof of concept for the development of
damage classification tools, which, in combination with meteorological data, are used to
137
 CHAPTER III

produce phenological models (Figure 1). We use DNNs to classify damages on apple tree leaves
and investigate the phenology of six main classes of damages predicted by the DNNs. We
couple the predicted occurrence of damages with meteorological data to model damage
phenology. Our case study targets the apple crop because it is the most important fruit crop in
Switzerland, with a production varying between 250,000 and 450,000 tons per year depending
on weather (SBV, 2019). Concerning damages, the focus lays on the mines of the pear leaf
blister moth (Leucoptera malifoliella Costa, Lepidoptera: Lyonetiidae; from here on blister
moth), a pest that has recolonized orchards in central Switzerland since 2013 (Zwahlen et al.,
2017). The blister moth prefers apple trees as host plants and in case of heavy infestation can
affect the photosynthesis and cause premature leaf drop (Ivanov, 1976; www.cabi.org). Larvae
of the blister moth are solitary miners producing characteristic brown, round mines that are
distinguishable from physical damages. Our working hypotheses are as follows:

1. We expect that the development of DNNs to categorize different classes of damages

on apple leaves is feasible with a subset of the 52,332 images collected during one
season.
2. We expect that by applying the resulting classification tools to the full dataset, we
can reconstruct the phenology of blister moth mines, which should match
conventional monitoring methods.
3. We expect to find a meteorological signature in damage phenology of the blister
moth, providing the basis for the development of statistical prediction models.

138
 CHAPTER III

Figure 1: Conceptual figure showing the overall goal of the framework for developing pest damage forecasting
tools. Data collection can be implemented with drones or citizen science approaches (a). The collected data can be
used to train deep neural networks (DNN) for image classification to recognize pest and pathogen damages (b).
Once the classification tool is established, predictions using deep neural networks can be used for reconstructing
the phenology of pest damages (c). Coupling damage phenology with meteorological data enables establishing
phenology models (d). Eventually, these phenology models can be used for predictions and in DSS, aiming at
informing on pest occurrence to support growers and experts, for instance by implementing the tool in a
smartphone app (e). Image classification tools can then reinforce the data collection, by making it available to
citizen scientists (f). In this study, we focus on the deep learning and phenology modelling aspects within this
framework.

139
 CHAPTER III

Methods

Data collection

We sampled leaves and collected images weekly between April 15th and August 28th
2019 in three apple orchards in central Switzerland, in Kleinwangen (47°11'49.3"N;
8°17'22.8"E; 536 m a.s.l.), Gelfingen (47°13'12.9"N; 8°16'10.0"E ; 557 m a.s.l.) and
Waedenswil (47°13'18.4"N; 8°40'38.6"E; 483 m a.s.l.; see Figure S1 for map with the
locations). Using smartphone cameras, we collected pictures of leaves in the field (from here
on referred to as field pictures) and sampled leaves to take pictures under standardized
conditions in the lab (from here on standardized pictures). Taking pictures using smartphones
results in pictures of a similar quality expected if they were taken by growers, untrained citizen
scientists or from automated devices such as drones. We conducted a structural sampling in the
three orchards. In each location, we sampled at least 400 leaves per week. With this structural
sampling, we aimed to capture a representative set of pest symptoms. As the orchards have
different number of trees planted in a different number of rows, we conducted three different
sampling strategies. In Kleinwangen and Gelfingen, we took two pictures from every third,
respectively fourth tree per row. In Waedenswil we took four pictures from every tree. Field
pictures and leaves were taken from the lower part and upper part of the trees. Collected leaves
were kept in a 3°C storage room before we took pictures under standardized conditions. We
used scotch tape to stick the leaves on a white paper in order to have a uniform background.
The pictures were taken with one of two different mobile phones, an iPhone 6 (8 megapixel
camera) and a Sony Xperia X (23 megapixel camera). At the time of sampling, in Kleinwangen
and Gelfingen extension services were conducting tests on the efficacy of pesticides targeted
against blister moth. We distinguished between treated and untreated sections of the orchards
for the image data collection, irrespective of the management of the treated trees.

In parallel to the leaf and picture sampling, population density of blister moth was
monitored with pheromone trapping and mine counting in Gelfingen. The orchard was exposed
to testing of eight different control methods including a control section. We placed one trap
(Delta-trap pheromone traps, Andermatt Biocontrol AG, Grossdietwil, Switzerland) in the
control section to document the occurrence of adults on a weekly basis. Weekly, 50 randomly
selected leaves per treatment section were inspected visually for the presence of mines, resulting

140
 CHAPTER III

in 400 leaves per week. Each leaf was carefully checked by eye and the occurrence of mines
was noted. We registered the total number of mines per 50 leaves.

Meteorological data

Meteorological data were extracted from the gridded dataset (2 km x 2 km) obtained
from the Swiss Federal Office of Meteorology and Climatology (MeteoSwiss;
meteoswiss.admin.ch). Daily data were extracted for the year 2019 and aggregated to the
weekly resolution in order to match the weekly sampling rate of the damage monitoring. The
statistical models employed to create the gridded data are described in Ceppi et al. (2010) and
Frei (2014) (daily minimum, maximum and mean temperature), Frei and Isotta (2019)
precipitation and Dürr and Zelenka. (2009) (solar radiation). To track phenology, we further
calculated the accumulated temperature sum (i.e. degree-days) as the cumulative sum of the
mean temperature over 5°C on daily basis.

Data preparation

In total, we gathered 52,322 pictures of apple leaves. We manually classified 8,735

randomly sampled standardized and field pictures into damage classes. We found 42 classes,
including classes containing combined damages and classes with a low number of pictures
(less than 100). We focused on seven classes with at least 100 pictures for further processing
(Figure 2):

- Undamaged: no damages detected on the leaf.

- PLBM: mines of blister moth detected;

- Physical damages: holes, cracks, fissures or deformations;

- Brown spots: brownish spots distinguishable from blister moth mines;

- Lepidoptera: rolled in leaf edges indicating pupae of Lepidoptera species;

- Mildew: powdery mildew (Podosphaera leucotricha) detected;

- Feeding: feeding damage from herbivore insects

We cropped all images to an extent of 2840 x 1560 pixels in the centre of the image to focus on
the leaf rather than the background (e.g. Figure 2 left-most pictures of class Undamaged).

141
 CHAPTER III

Deep neural networks

We used DNNs to apply image classification of the entire image in order to allocate it
to one out of several classes. We implemented our deep neural network approach in R (version
3.5.3; R Core Team, 2019), using the R-package ‘reticulate’ (version 1.13.0-9003; Ushey,
Allaire & Tang, 2019) to open an interface to python. We used ‘Keras’ (version 2.2.5.0; Allaire
& Chollet, 2019) and ‘Tensorflow’ (version 1.9; Allaire & Tang, 2019) R-packages as DNN
frameworks. We loaded image data with the image_data_generator function of the ‘Keras’ R-
package. Images were imported and resized to 256 x 256 pixels with three channels (i.e. RGB
colour channels) and rescaled to values between 0 and 1. Further, we applied data augmentation
to the training dataset during the image import. Data augmentation is a common strategy to
increase the number of images in the training dataset. We applied the following specifications
in the image_data_generator function to augment the data: zooming (range = 0.4), rotations
(range = 90), width and height shifts (range = 0.2), shearing (range = 0.2), horizontal and
vertical flips. As network structure we used the ResNet50 (He et al. 2016) model, loaded with
weights pre-trained on ImageNet (Deng et al. 2009) as base for our model architecture. We
fine-tuned all layers of the ResNet50 network and added one dense layer with 256 nodes and a
ReLU activation function, as well as an output layer with a softmax activation function on top
of the ResNet50 to adapt to our dataset. Moreover, we added dropout (0.5) after the ResNet50
network and as well after the densely connected layer to prevent model overfitting. Further, we
used an RMSprop optimizer with a base learning rate of 0.0001 and a decay of 0.00001 for
gradient descent. We set the mini-batch size, which defines how many images the DNN takes
into account per step for calculating the model error and updating the model coefficients, to 32.
All networks were trained for 100 epochs (i.e. iterations over the full training dataset).

We trained DNNs on different combinations of classes (i.e. different classification

tasks). First, we trained DNNs for each damage class to distinguish images of this class, from
all other images (i.e. all other classes as one summary class), resulting in six classification tasks.
Second, we trained full model DNNs to classify the six main classes (PLBM, Undamaged,
Physical damages, Brown spots, Lepidoptera and Mildew) simultaneously in one DNN. In a
preliminary analysis step, we used Feeding and Physical damages as independent classes, but
compounded them for the final analysis because DNNs struggled with differentiating these two

142
 CHAPTER III

classes. We trained DNNs for the same classification tasks with field pictures. To establish
DNNs, we split the categorized images of each class into five subsets for a 5-fold cross-
validation. DNNs were trained on four subsets (80% of the data) and tested on the left out subset
(20%). From the 80% training data, 20% were used for validation to tune hyperparameters (i.e.
settings to control the learning process of a deep neural network). This procedure was repeated
five times resulting in five different DNNs per classification task. We measured the
performance of DNNs for each classification task on the test set with F1-score averaged over
the five DNNs per classification task. F1-scores (equation 1) were calculated with the scores
for true positives (TP), false positives (FP) and false negatives (FN), resulting in values ranging
from 0 to 1, 1 being perfect classification:

2 =
3
$ 34 .6∗$23427&

(1)

Additionally, we measured performance with classification accuracy as the percentage

of correctly classified images. DNNs were evaluated with the performance on the test data set,
which was not included in the construction of the network.

Coupling pest damage with meteorological data

We used DNNs to classify all images from Kleinwangen and Gelfingen, because in these
two locations we found blister moths. We used the predict_class and the predict_proba function
of the ‘Keras’ R-package (version 2.2.5.0; Allaire & Chollet, 2019) to obtain predictions on the
class and the pseudo-probabilities per class for all images using the full model DNNs. The
pseudo-probability prediction are output scores from DNNs showing how confident the DNN
is in predicting a class for an image. We grouped the predictions into locations and dates.
Further, we calculated the percentages of damaged leaves per sampling event, scaling the
number of damaged leaves of each class with the total number of collected leaves to correct for
uneven sampling, as the sampling events did not result in the exact same number of images.
We used GLMs to model the percentage of damaged leaves with weekly meteorological data
as predictor variables. Climatic variables included growing degree-days (i.e. cumulative sum
of mean temperature over base temperature of 5°C), mean temperature, precipitation, solar

143
 CHAPTER III

radiation and diurnal temperature range. We ran GLMs for each meteorological predictor, and
one multivariate GLM with all predictors, allowing second-degree polynomials and assuming
binomial error distribution. We used the ecospat.adj.D2.glm function of the ‘ecospat’ R-
package (version 3.0; Broennimann et al., 2018) to obtain model deviance as adjusted D2 values
of all models. Additionally, we use the glarma function of the ‘glarma’ R-package (version 1.6-
0, Dunsmuir & Scott, 2015) to run GLARMA (Generalized Linear Autoregressive Moving
Average) models for the same predictor variables, to check whether accounting for temporal
autocorrelation would change the model estimation of parameters.

Results

Data collection

We collected 52,322 pictures of apple tree leaves in total over 19 weeks of sampling.
35,903 pictures were taken in the field and 21,087 under standardized conditions. For the two
locations, where we used the image data to reconstruct the blister moth phenology, we gathered
14,466 images in Gelfingen and 18,384 in Kleinwangen. We did not find different signals from
the different treatments and therefore only present the results for the collection of treated and
control section. From the 8,735 categorized pictures, we found that class PLBM contained 1,390
images, Undamaged 1,415, Physical damages 1,139, Brown spots 2,025, Lepidoptera 103 and
Mildew 134 images.

Deep neural networks

We established DNNs for 14 different classification tasks of apple tree leaves. F1-scores
for classification tasks of standardized pictures ranged from 0.69 to 0.93 with the exception of
the class Lepidoptera (0.32), where the number of training images was very low (with a total
of 103 manually classified images, for model training we used between 26 and 44 images
depending on classification task (standardized or field) and the cross-validation chunk).
Classification accuracy ranged from 91.3% to 99.5%. The full model including all six classes
reached a F1-score of 0.89 (standard deviation across the five cross-validation runs: ± 0.035)
and a classification accuracy of 95.4% (± 1.5%). DNNs performed generally more poorly on
images taken in the field (Figure 2). F1-scores for field pictures ranged from 0.52 to 0.90 with
the exception of Physical damage where no F1-score could be determined because none of the
test images was assigned to Physical damages, meaning that this class was not recognized by
144
 CHAPTER III

the DNN based on field pictures. Classification accuracy for all classes ranged from 87.5% to
99.4%. The full model reached a F1-score of 0.85 (± 0.02) and a classification accuracy of 87.7
% (± 1.6 %) for field images.

Figure 2: DNN performance for the 14 classification tasks measured as classification accuracy. Boxes show the
classification accuracy variation over the 5-fold cross validation. Red boxes show the results for standardized
pictures, blue boxes for field pictures. For the results of F1-scores see Figure S2. Grey dashed lines mark the 85%,
90% and 95% lines. Images show the different classes considered in the study. F.l.t.r: Undamaged, PLBM, Physical
damage, Lepidoptera, Brown spots, Mildew. Right-most boxes show the performance of the full model.

We used the trained DNNs to classify the images of the dataset that were not categorized
a priori. From the five DNNs trained for the cross-validation for the full model we selected the
one with the best performance. With the full model for standardized pictures, we found 7,627
images of class PLBM, 5,598 Undamaged, 3,063 Physical damages, 156 Brown spots, 122
Lepidoptera and 850 Mildew. The full model for field pictures resulted in a prediction of 7,350
PLBM, 10,638 Undamaged, 899 Physical damages, 131 Brown spots, 282 Lepidoptera and 693
Mildew. The full model for field images was unable to detect the class Lepidoptera (see Figure
S2 & Figure S3 for reconstructed phenologies of all damage classes). We used the DNNs to
145
 CHAPTER III

reconstruct the phenology of the blister moth (Figure 3). We found very similar patterns for the
two locations, with an increase of blister moth mines in mid-June and a first peak in early July.
Standardized and field pictures show very similar results, although field pictures indicated the
peak one week later than the standardized pictures.

Although we found different DNN performance for the standardized pictures and the
field pictures, predictions to the full dataset resulted in very similar patterns of the phenology
of the different damage classes. Further, we compared the blister moth phenology predictions
of the DNNs with count data of blister moth adults in traps and mines obtained from surveys in
the same orchards (Fig. 4). The count data supports our findings for the blister moth phenology
based on the predictions of DNNs. The phenology of the trapped adults explained patterns of
the mines, which start to emerge 3 to 4 weeks after the peak of a generation of adults. These
results matched well with literature descriptions of development times for one generation (e.g.
36 days at 18°C; Sáringer et al., 1985). We also observed the same pattern of decreasing
numbers of mines in early to mid-July, which is explained by the simultaneous emergence of
new leaves and the gap between the first and the second generation of blister moth larvae. Mine
counts and reconstructed mine phenology showed a Pearson correlation coefficient between
0.938 and 0.978 for both locations with standardized and field pictures.

146
 CHAPTER III

Figure 3: Comparison between blister moth mines phenology reconstructed with DNNs (red line) and count data
of mines from the field (blue line). The black line marks the count data of adults caught in traps. The hatched areas
in the background show a 4 week timespan between the peaks of the blister moth generations (adults) and the local
peak of mines. Panels show the data for a) Gelfingen standardized pictures, b) Gelfingen field pictures, c)
Kleinwangen standardized pictures, d) Kleinwangen field pictures.

Coupling pest damages with meteorological data

We quantified the relationship between the phenology of the blister moth and climate
using GLMs (Figure 4). For both locations and for standardized and field pictures we found
that degree-days is the most important variable, with adjusted D2 values between 0.950 and
0.967 (Table 1). The full model explained only slightly more of the deviance ranging from
0.956 to 0.969. We found similar results with GLARMA models, confirming that degree-days

147
 CHAPTER III

is the best predictor for blister moth phenology from Gelfingen, but only the second best
predictor for Kleinwangen after mean temperature (Table 2).

Table 1: Model deviance (as adjusted D2 values) for GLM on blister moth phenology with different
meteorological predictor variables. Degree-days above 5°C (GDD), mean temperature (Tmean), precipitation
(Precip), radiation (SRad), diurnal temperature (Diur) and the model with all predictors (Full).

Gel. stand. Kle, stand. Gel. field Kle. field

GDD 0.966 0.967 0.959 0.950
Tmean 0.841 0.767 0.773 0.804
Precip 0.006 0.014 0.032 0.018
Srad 0.332 0.287 0.286 0.317
Diur 0.202 0.176 0.211 0.194
Full 0.969 0.965 0.961 0.956

Table 2: Akaike’s information criterion (AIC) for GLARMA models on blister moth phenology with different
meteorological predictor variables. Degree-days above 5°C (GDD), mean temperature (Tmean), precipitation
(Precip), radiation (SRad), diurnal temperature (Diur) and the model with all predictors (Full).

Gel. stand. Kle, stand. Gel. field Kle. field

GDD 344.6685 569.2 443.3927 534.1602
Tmean 346.8743 452.2735 504.6914 355.1499
Precip 2107.805 1949.889 2323.896 1555.169
Srad 448.142 592.8213 465.8917 396.9868
Diur 841.1444 795.4927 928.275 551.5733
Full 185.3556 233.9368 268.5195 227.7823

148
 CHAPTER III

Figure 4: Comparison of the seasonal evolution of the inferred damage phenology (DNN) and the phenology
modelled with GLMs. Panels show the data for a) Gelfingen standardized pictures, b) Gelfingen field pictures, c)
Kleinwangen standardized pictures, d) Kleinwangen field pictures.

Discussion

In this study, we tested the feasibility of different segments of a framework for

developing pest damage forecasting models, relying on big data, DNNs and meteorological data
(Figure 1). Our case study on blister moth mines suggests that DNNs coupled with
meteorological data are suitable tools for the proposed method. We use DNNs to build image
classification tools to classify the damages of a large dataset of apple leaf images and to
reconstruct the phenology of different damage classes (Figure S3). Using blister moth as a case
study species, we show that the blister moth phenology predicted with DNNs matched count
data of mines and adults in the field. Finally, we quantify the phenology of blister moth mines

149
 CHAPTER III

with meteorological variables and show that phenological models based on degree-days are
well fitting the blister moth phenology. While we show here that, the proposed framework is
feasible in principle, for the full implementation of the framework, data collection processes
need to be optimized and the phenological models need to be validated with independent data.

Our results show that DNNs are suitable tools for pest damage classification based on
image data, similarly as it has earlier been shown for other classification tasks (e.g. Mohanty et
al., 2016; Cheng et al., 2017). Trained DNNs reached good model performance for categorizing
blister moth damages with a classification accuracy of 93.8% (F1-score of 93.2); all other single
class DNNs were trained successfully with F1-scores above 0.86 except Physical damages
(0.69) and Lepidoptera (0.32) (see Figure S4 and S5 for some examples of misclassifications).
In general, we observe better and more robust results for classes where more data for training
were available. DNNs classifying multiple classes were struggling with distinguishing some of
the classes, but performed well overall (F1-score of 0.89 for six classes). Compared to studies
predicting several classes of pathogens, the accuracies of our DNNs are slightly lower (e.g.
Oppenheim and Shani, 2017 reached 96% accuracy with five classes; see Barbedo, 2018 for an
overview). A reason for this could be that the manually classified part of our dataset is rather
small and applying a cross-validation lowers the number of pictures available to train the DNNs.
In general, we observe that DNNs struggle with distinguishing between classes with similar
symptoms, for example between Physical damages and Feeding, which we discriminated in
preliminary analyses. Another reason for the decreasing model performance with more classes
may lay in the co-occurrence of damages on one leaf, but general solutions to properly identify
such simultaneous damages are still lacking (Barbedo, 2018). Further, we show that DNNs are
also capable of classifying pest damages with images taken under field conditions (see results,
section deep neural networks), which is crucial to develop useful pest or disease recognition
tools, as the goal should be the application in the field (Sladojevic et al., 2016; Picon et al.,
2019). Overall, DNNs established with field pictures show good results, but as expected, reach
slightly lower performance than DNNs for standardized pictures, because external influences
such as shading effects, multiple leaves, other plant parts or irrelevant objects in the background
can be disturbing (Ferentinos, 2018). Particularly, DNNs struggled with the class of Physical
damages. Still, the DNN for the class PLBM with field pictures reached a F1-score of 0.90
(classification accuracy of 93.1%). The full model DNN for field pictures was successful with
an F1-score of 0.85 and classification accuracy of 87.7%. Our case study shows that the
development of pest damage classification tools using DNNs is realizable, allowing to use those

150
 CHAPTER III

tools to obtain the phenology of classes by analyzing big datasets, given that sufficient data is
available.

We highlight that reconstructing damage phenology with DNNs, coupled with

meteorological data opens up new possibilities to produce phenological models for pest
forecasting. Big data science has recently been proposed to help to overcome current limitations
in pest forecasting (Orlandini et al. 2018). In our case study, we find that degree-days is the
most important variable to model blister moth phenology. Degree-days have been shown to be
a reliable predictor of insect development (e.g. Cayton et al. 2015) and are important
components in phenology models for insects (Nietschke et al., 2007). This is important for the
development of operational systems, as in many regions of the world temperature data are
available at high spatial and temporal resolution. We validate the reconstructed phenology of
the blister moth, with count data on adults and mines obtained from the same study sites,
showing that classification tools are able to reconstruct the real phenology. Further, we show
that our approach is suitable to reconstruct the phenology of other classes and therefore, could
be used to investigate not only the phenology of insects, but also the phenology of other types
of damages (e.g. Mildew). With sufficient image data, the prediction approach may also be
implemented for pests and pathogens to find meteorological signals behind their seasonal
occurrence or the occurrence of the entailed damages. Successful recognition tools for
pathogens have already been developed (e.g. Fuentes et al., 2017; Liu et al. 2018) and seasonal
occurrence of pathogens is often limited by abiotic factors (Rossi, Giosuè and Caffi, 2010). In
addition, we emphasize that the framework for the establishment of this approach would also
be suitable for invasive species phenology modelling, due to the potential of fast
implementation. However, expert recommendations on management interventions need to be
based on solid testing of control strategies by plant protection experts. Particularly for invasive
species, this is crucial to implement sustainable control. We find promising results in this case
study, underlining that the proposed framework could bring new opportunities for pest
forecasting, given that new methods will help to overcome the lack of data availability.

While this case study highlights new possibilities for pest damage forecasting, we came
across some limitations that need to be addressed in future studies. One of the main limitations
is the low number of images of some classes we used to train the DNNs. Here, we use only a
subset of our dataset where the frequency of the different classes is not equally distributed. This
means that for some classes only few images are available for training and evaluation of a DNN.
151
 CHAPTER III

Meanwhile, previous studies establish DNNs using the full data of large datasets (e.g.
PlantVillage) for model training and evaluation (e.g. Mohanty et al., 2016). However, such
image data is limited and previous studies in the field of plant disease classification often rely
on the same dataset and similar tools, yielding in a low variation between the results of these
studies (Barbedo, 2018; Arsenovic et al., 2019). Increasing the input data for DNNs could
therefore promote higher robustness and performance of the classification tools of some classes
(Sladojevic et al., 2016). Further, within our sampling period, the variation in meteorological
conditions was rather small. Longer-term surveys are required to capture a broader scope of
meteorological settings, leading to more robust phenological models. Similarly, data from long-
term monitoring programs are needed to validate phenological models. With the
implementation of new data collection strategies, these limitations may be overcome.

To address the limitations of the current work, as well as provide the base of the
proposed framework for pest forecasting, innovative data collection strategies must be
established. We present perspectives and potential approaches for acquisition of data for the
proposed framework for pest damage forecasting. The main disadvantage of deep learning is
the amount of data needed (Kamilaris & Prenafeta-Boldu, 2018) and not many agricultural
image datasets are publicly available (Kamilaris et al. 2017; Arsenovic et al., 2019). To
overcome this data scarcity, we propose two approaches for data collection. First, with the
increasing number of smartphones used worldwide, allowing to record images, sound and
location, there is a wide scope for gathering large datasets, in particular in the context of citizen
science (Teacher et al. 2013). In agriculture, particularly in relation to pest and pathogen
monitoring, there is pressing interest for this approach and farmers are traditionally interested
in participating in research projects (Ryan et al., 2018). While citizen scientists benefit from the
classification tool and the pest forecasting model, the collected image data can be used to create
a feedback loop where new images can be used for updating the classification tool (see Figure
1). Additionally, an advantage of a citizen science approach could come from detection of new
invasive species (Hulbert et al. 2017; Johnson et al., 2020), as famers might want to inform
themselves and alert the responsible experts when they encounter a yet unknown damage.
Finally, the acceptance for DSS is expected to be higher if users are involved in their
development (Lynch, Gregor and Midmore, 2000).

The second approach we propose here is the implementation of drones (Floreano &
Wood, 2015). Drones are expected to revolutionise precision agriculture by delivering big data
152
 CHAPTER III

that can be used for various purposes (Tripicchio et al., 2015; Finn and Donovan, 2016). For
example, drones have been used for weed detection, irrigation equipment monitoring, or crop
health monitoring (Veroustraete, 2015). Drones programmed with GIS inputs and equipped
with high-resolution cameras (e.g. 15 megapixel, Shankar et al., 2018) are suitable tools to
collect data in a structured way, which can be analysed with deep learning algorithms (Shankar
et al., 2018). Together, these approaches highlight opportunities to overcome the lack of image
datasets on pest and pathogens, allowing to advance with the proposed framework for pest
forecasting and providing groundwork for other novel technologies supporting sustainable
agriculture.

Conclusions

In conclusion, we present a framework for developing pest monitoring and forecasting

tools that rely on big data and deep learning. A non-representative survey suggested that farmers
are generally interested in the development of new forecasting tools, and that there is a demand
for new technologies with broad applicability in plant protection. In this study, we focus on the
segments of this framework connected to building DNNs and coupling the phenology
reconstructed with those DNNs with meteorological variables to produce phenology models.
The case study on blister moth phenology highlights that this approach is feasible. DNNs
showed good performance on categorizing different classes of damages with pictures of leaves
taken under standardized conditions and in the field. Further, the phenology of the blister moth
obtained from DNNs matched the phenology observed with count data in the field well. While
damage classification tools are valuable instruments for pest and pathogen monitoring, using
those classification tools to reconstruct the damage phenology and coupling them with
meteorological data, will promote new opportunities for early warning. Together, this study
highlights that big data and modern technologies provide new opportunities to advance
sustainable plant protection. To overcome the scarcity in data availability, which presents the
main limiting factor for such data-driven approaches, here we suggest to address this issue with
data collection based on citizen science or drones. Increasing data availability would not only
support this framework for pest damage forecasting, but also foster further development
towards applying modern information technology to tackle current agricultural challenges.

153
 CHAPTER III

Code availability statement

The R codes for the models and calculations of the results and analysis are available
from the corresponding author. The code and scripts will be published on a public repository
upon manuscript acceptance.

154
 CHAPTER III

References

Allaire, J.J. & Chollet, F. (2019). keras: R Interface to 'Keras'. R package version 2.2.5.0.
https://CRAN.R-project.org/package=keras.
Allaire, J.J. & Tang, Y. (2019). tensorflow: R Interface to 'TensorFlow'. R package version
2.0.0. https://CRAN.R-project.org/package=tensorflow.
Arsenovic, M., Karanovic, M., Sladojevic, S., Anderla, A., & Stefanovic, D. (2019). Solving
current limitations of deep learning based approaches for plant disease detection.
Symmetry, 11(7), 939.
Barbedo, J. G. (2018). Factors influencing the use of deep learning for plant disease recognition.
Biosystems Engineering, 172, 84-91.
Bebber, D., Ramotowski, M., & Gurr, S. (2013). Crop pests and pathogens move polewards in
a warming world. Nature Climate Change, 3(11), 985.
Behmann, J., Mahlein, A. K., Rumpf, T., Römer, C., & Plümer, L. (2015). A review of advanced
machine learning methods for the detection of biotic stress in precision crop protection.
Precision Agriculture, 16(3), 239-260.
Bongiovanni, R., & Lowenberg-DeBoer, J. (2004). Precision agriculture and sustainability.
Precision Agriculture, 5(4), 359-387.
Broennimann, O., Di Cola, V. & Guisan, A. (2018). ecospat: Spatial Ecology Miscellaneous
Methods. R package version 3.0. https://CRAN.R-project.org/package=ecospat
Cayton, H. L., Haddad, N. M., Gross, K., Diamond, S. E., & Ries, L. (2015). Do growing degree
days predict phenology across butterfly species? Ecology, 96(6), 1473-1479.
Ceppi, P., Scherrer, S. C., Fischer, A. M., & Appenzeller, C. (2012). Revisiting Swiss
temperature trends 1959–2008. International Journal of Climatology, 32(2), 203-213.
Chakraborty, S., & Newton, A. C. (2011). Climate change, plant diseases and food security: an
overview. Plant Pathology, 60(1), 2-14.
Cheng, X., Zhang, Y., Chen, Y., Wu, Y., & Yue, Y. (2017). Pest identification via deep residual
learning in complex background. Computers and Electronics in Agriculture, 141, 351-
356.
Dehnen-Schmutz, K., Foster, G. L., Owen, L., & Persello, S. (2016). Exploring the role of
smartphone technology for citizen science in agriculture. Agronomy for Sustainable
Development, 36(2), 25.

155
 CHAPTER III

Deng, J., Dong, W., Socher, R., Li, L. J., Li, K., & Fei-Fei, L. (2009, June). Imagenet: A large-
scale hierarchical image database. In 2009 IEEE conference on computer vision and
pattern recognition (pp. 248-255). IEEE.
Donatelli, M., Magarey, R. D., Bregaglio, S., Willocquet, L., Whish, J. P., & Savary, S. (2017).
Modelling the impacts of pests and diseases on agricultural systems. Agricultural
Systems, 155, 213-224.
Dunsmuir, W. T. M., Scott, D. J. (2015). The 'glarma' Package for Observation-Driven Time
Series Regression of Counts. Journal of Statistical Software, 67(7), 1-36.
Dürr, B., & Zelenka, A. (2009). Deriving surface global irradiance over the Alpine region from
METEOSAT Second Generation data by supplementing the HELIOSAT method.
International Journal of Remote Sensing, 30(22), 5821-5841.
Ferentinos, K. P. (2018). Deep learning models for plant disease detection and diagnosis.
Computers and Electronics in Agriculture, 145, 311-318.
Finn R., Donovan A. (2016) Big Data, Drone Data: Privacy and Ethical Impacts of the
Intersection Between Big Data and Civil Drone Deployments. In Custers B. (Eds.) The
Future of Drone Use. Information Technology and Law Series, vol 27. T.M.C. Asser
Press, The Hague, the Netherlands.
Floreano, D., & Wood, R. J. (2015). Science, technology and the future of small autonomous
drones. Nature, 521(7553), 460-466.
Forrest, J. R. (2016). Complex responses of insect phenology to climate change. Current
Opinion in Insect Science, 17, 49-54.
Frei, C. (2014). Interpolation of temperature in a mountainous region using nonlinear profiles
and non‐Euclidean distances. International Journal of Climatology, 34(5), 1585-1605.
Frei, C., & Isotta, F. A. (2019). Ensemble spatial precipitation analysis from rain gauge data:
methodology and application in the European Alps. Journal of Geophysical Research:
Atmospheres, 124(11), 5757-5778.
Fuentes, A., Yoon, S., Kim, S. C., & Park, D. S. (2017). A robust deep-learning-based detector
for real-time tomato plant diseases and pests recognition. Sensors, 17(9), 2022.
Garnett, T. & Godfray, H.C.J. (2012). Sustainable intensification in agriculture. Navigating a
course through competing food system priorities. A report on a workshop. Food Climate
Research Network and the Oxford Martin Programme on the Future of Food, University
of Oxford, UK.

156
 CHAPTER III

Garnett, T., Appleby, M. C., Balmford, A., Bateman, I. J., Benton, T. G., Bloomer, P., ... &
Herrero, M. (2013). Sustainable intensification in agriculture: premises and policies.
Science, 341(6141), 33-34.
Geiger, F., Bengtsson, J., Berendse, F., Weisser, W. W., Emmerson, M., Morales, M. B., ... &
Eggers, S. (2010). Persistent negative effects of pesticides on biodiversity and biological
control potential on European farmland. Basic and Applied Ecology, 11(2), 97-105.
Grünig, M., Mazzi, D., Calanca, P., Karger, D. N. & Pellissier, L. (2020). Crop and forest pest
metawebs shift towards increased linkage and suitability overlap under climate change.
Communications Biology 3, 233.
Godfray, H. C. J., Beddington, J. R., Crute, I. R., Haddad, L., Lawrence, D., Muir, J. F., ... &
Toulmin, C. (2010). Food security: the challenge of feeding 9 billion people. Science,
327(5967), 812-818.
Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. MIT press.
He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In
Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (770-
778).
Hulbert, J. M., Agne, M. C., Burgess, T. I., Roets, F., & Wingfield, M. J. (2017). Urban
environments provide opportunities for early detections of Phytophthora invasions.
Biological Invasions, 19(12), 3629-3644.
Ivanov S., 1976. Leaf-miner moths on fruit-trees in Bulgaria. Dissertation. Fruit-Growing
Institute, Plovdiv, Bulgaria.
Johnson, B. A., Mader, A. D., Dasgupta, R., & Kumar, P. (2020). Citizen science and invasive
alien species: An analysis of citizen science initiatives using information and
communications technology (ICT) to collect invasive alien species observations. Global
Ecology and Conservation, 21, e00812.
Kamilaris, A., Kartakoullis, A., & Prenafeta-Boldú, F. X. (2017). A review on the practice of
big data analysis in agriculture. Computers and Electronics in Agriculture, 143, 23-37.
Kamilaris, A., & Prenafeta-Boldú, F. X. (2018). Deep learning in agriculture: A survey.
Computers and Electronics in Agriculture, 147, 70-90.
Kingsolver, J., Arthur Woods, H., Buckley, L., Potter, K., MacLean, H., & Higgins, J. (2011).
Complex life cycles and the responses of insects to climate change. Integrative and
Comparative Biology, 51(5), 719-732.

157
 CHAPTER III

Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep
convolutional neural networks. In Advances in Neural Information Processing Systems,
1097-1105.
Lamichhane, J. R., Arendse, W., Dachbrodt-Saaydeh, S., Kudsk, P., Roman, J. C., van
Bijsterveldt-Gels, J. E., ... & Messéan, A. (2015). Challenges and opportunities for
integrated pest management in Europe: A telling example of minor uses. Crop Protection,
74, 42-47.
Lamichhane, J. R., Aubertot, J. N., Begg, G., Birch, A. N. E., Boonekamp, P., Dachbrodt-
Saaydeh, S., ... & Kiss, J. (2016). Networking of integrated pest management: A powerful
approach to address common challenges in agriculture. Crop Protection, 89, 139-151.
Lechenet, M., Dessaint, F., Py, G., Makowski, D., & Munier-Jolain, N. (2017). Reducing
pesticide use while preserving crop productivity and profitability on arable farms. Nature
Plants, 3(3), 1-6.
LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444.
Lindblom, J., Lundström, C., Ljung, M., & Jonsson, A. (2017). Promoting sustainable
intensification in precision agriculture: review of decision support systems development
and strategies. Precision Agriculture, 18(3), 309-331.
Liu, B., Zhang, Y., He, D., & Li, Y. (2018). Identification of apple leaf diseases based on deep
convolutional neural networks. Symmetry, 10(1), 11.
Lynch, T., Gregor, S., & Midmore, D. (2000). Intelligent support systems in agriculture: how
can we do better?. Australian Journal of Experimental Agriculture, 40(4), 609-620.
Mohanty, S. P., Hughes, D. P., & Salathé, M. (2016). Using deep learning for image-based
plant disease detection. Frontiers in Plant Science, 7, 1419.
Martinelli, F., Scalenghe, R., Davino, S., Panno, S., Scuderi, G., Ruisi, P., ... & Davis, C. E.
(2015). Advanced methods of plant disease detection. A review. Agronomy for
Sustainable Development, 35(1), 1-25.
Möhring, N., Wuepper, D., Musa, T., & Finger, R. (2020). Why farmers deviate from
recommended pesticide timing: The role of uncertainty and information. Pest
Management Science.
Nietschke, B. S., Magarey, R. D., Borchert, D. M., Calvin, D. D., & Jones, E. (2007). A
developmental database to support insect phenology models. Crop Protection, 26(9),
1444-1448.
Oerke, E. C. (2006). Crop losses to pests. The Journal of Agricultural Science, 144(1), 31-43.

158
 CHAPTER III

Oppenheim, D., & Shani, G. (2017). Potato disease classification using convolution neural
networks. Advances in Animal Biosciences, 8(2), 244-249.
Orlandini S., Magarey R.D., Park E.W., Sporleder M. and Kroschel J., 2018: Methods of
Agroclimatology: Modeling Approaches for Pests and Diseases. In: Hatfield J.L:,
Sivakumar M.V:K. and Prueger J.H. (eds), Agroclimatology: Linking Agriculture to
Climate, Volume 60, American Society of Agronomy & Crop Science Society of America
& Soil Science Society of America, 453-488.
Patrício, D. I., & Rieder, R. (2018). Computer vision and artificial intelligence in precision
agriculture for grain crops: A systematic review. Computers and Electronics in
Agriculture, 153, 69-81.
Picon, A., Alvarez-Gila, A., Seitz, M., Ortiz-Barredo, A., Echazarra, J., & Johannes, A. (2019).
Deep convolutional neural networks for mobile capture device-based crop disease
classification in the wild. Computers and Electronics in Agriculture, 161, 280-290.
Pimentel, D., & Burgess, M. (2014). Environmental and economic costs of the application of
pesticides primarily in the United States. In Integrated Pest Management (47-71).
Springer, Dordrecht, the Netherlands.
R Core Team (2019). R: A language and environment for statistical computing. R Foundation
for Statistical Computing, Vienna, Austria.
Rossi, V., Giosuè, S., & Caffi, T. (2010). Modelling plant diseases for decision making in crop
protection. In Precision Crop Protection - the Challenge and Use of Heterogeneity (241-
258). Springer, Dordrecht, the Netherlands.
Ryan, S. F., Adamson, N. L., Aktipis, A., Andersen, L. K., Austin, R., Barnes, L., ... & Cooper,
C. B. (2018). The role of citizen science in addressing grand challenges in food and
agriculture research. Proceedings of the Royal Society B, 285(1891), 20181977.
Samietz, J., Graf, B., Höhn, H., Schaub, L., & Höpli, H. U. (2007). Phenology modelling of
major insect pests in fruit orchards from biological basics to decision support: the
forecasting tool SOPRA. EPPO Bulletin, 37(2), 255-260.
Sáringer, G., Seprős, I., & Ángyán, F. (1985). Effect of temperature and photoperiod on the
course of development of Leucoptera scitella Zell. (Lep.: Leucopteridae).
Növényvédelem, 21(5), 213-214.
SBV (Schweizer Bauernverband - Swiss Farmers’ Union), 2019: Statistische Erhebungen und
Schätzungen über Landwirtschaft und Ernährung. Schweizer Bauernverband, Brugg.

159
 CHAPTER III

Schaub, L., Breitenmoser, S., Derron, J., & Graf, B. (2017). Development and validation of a
phenological model for the univoltine European corn borer. Journal of Applied
Entomology, 141(6), 421-430.
Shankar, R. H., Veeraraghavan, A. K., Sivaraman, K., & Ramachandran, S. S. (2018).
Application of UAV for Pest, Weeds and Disease Detection using Open Computer
Vision. In 2018 International Conference on Smart Systems and Inventive Technology
(ICSSIT) (287-292). IEEE.
Shepon, A., Eshel, G., Noor, E., & Milo, R. (2018). The opportunity cost of animal based diets
exceeds all food losses. Proceedings of the National Academy of Sciences of the United
States of America, 115(15), 3804-3809.
Sladojevic, S., Arsenovic, M., Anderla, A., Culibrk, D., & Stefanovic, D. (2016). Deep neural
networks based recognition of plant diseases by leaf image classification. Computational
Intelligence and Neuroscience.
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., ... & Rabinovich, A. (2015).
Going deeper with convolutions. In Proceedings of the IEEE Conference on Computer
Vision and Pattern Recognition (1-9). IEEE.
Tang, S., Tang, G., & Cheke, R. A. (2010). Optimum timing for integrated pest management:
modelling rates of pesticide application and natural enemy releases. Journal of
Theoretical Biology, 264(2), 623-638.
Teacher, A. G., Griffiths, D. J., Hodgson, D. J., & Inger, R. (2013). Smartphones in ecology
and evolution: a guide for the app‐rehensive. Ecology and Evolution, 3(16), 5268-5278.
Tegtmeier, E., & Duffy, M. (2004). External costs of agricultural production in the United
States. International Journal of Agricultural Sustainability, 2(1), 1-20.
Thenmozhi, K., & Reddy, U. S. (2019). Crop pest classification based on deep convolutional
neural network and transfer learning. Computers and Electronics in Agriculture, 164,
104906.
Tilman, D., Balzer, C., Hill, J., & Befort, B. (2011). Global food demand and the sustainable
intensification of agriculture. Proceedings of the National Academy of Sciences of the
United States of America, 108(50), 20260-20264.
Tripicchio, P., Satler, M., Dabisias, G., Ruffaldi, E., & Avizzano, C. A. (2015). Towards smart
farming and sustainable agriculture with drones. In 2015 International Conference on
Intelligent Environments (140-143). IEEE.
Ushey, K., Allaire, J.J. & Tang, Y. (2019). reticulate: Interface to Python. R package version
1.13.0-9003. https://github.com/rstudio/reticulate.

160
 CHAPTER III

Veroustraete, F. (2015). The rise of the drones in agriculture. EC Agriculture, 2(2), 325-327.
Walter, A., Finger, R., Huber, R., & Buchmann, N. (2017). Opinion: Smart farming is key to
developing sustainable agriculture. Proceedings of the National Academy of Sciences,
114(24), 6148-6150.
West, P. C., Gerber, J. S., Engstrom, P. M., Mueller, N. D., Brauman, K. A., Carlson, K. M., ...
& Siebert, S. (2014). Leverage points for improving global food security and the
environment. Science, 345(6194), 325-328.
Zhong, Y., & Zhao, M. (2020). Research on deep learning in apple leaf disease recognition.
Computers and Electronics in Agriculture, 168, 105146.
Zwahlen, D. & Hunkeler, M. (2017). Fleckenminiermotte in der Zentralschweiz – ein fast
vergessener Schädling. SZOW, 12/17.

161
 CHAPTER III

Acknowledgements

This work contributes to the newly established Swiss National Centre for Climate
Services (www.nccs.ch) and received financial support by the Swiss Federal Office for
Agriculture. Access to the daily gridded weather data was kindly provided by the Swiss Federal
Office of Meteorology and Climatology (MeteoSwiss). We thank Markus Thali and Dani Vogel
for allowing us to collect data in their orchards and Barbara Egger, Diana Zwahlen and Theresa
Steiner for sharing the mine count and trapping data.

Author contributions

M.G. contributed to conceiving ideas, collecting image data and monitoring data,
constructing and running deep neural networks, performing analysis and writing the
manuscript; E. R. contributed to collecting image data and monitoring data; P. C. contributed
to conceiving ideas, performing analysis and writing; D.M. contributed to conceiving ideas and
writing; J. D. W. contributed to constructing deep neural networks and writing; L.P. contributed
to conceiving ideas and writing. All authors gave final approval for publication.

162
 CHAPTER III

Supplementary material

Supplementary Figures

Figure S1: Location of the three study orchards (yellow stars) in Switzerland: Kleinwangen (lower left), Gelfingen
(upper left), Waedenswil (right).

163
 CHAPTER III

Figure S2: DNN performance for the 14 classification tasks measured as F1-scores. Boxes show the classification
accuracy variation over the 5-fold cross validation. Red boxes show the results for standardized pictures, blue
boxes for field pictures.

164
 CHAPTER III

Figure S3: Reconstruction of the phenology for all damage classes for Gelfingen standardized pictures (a),
Gelfingen field pictures (b), Kleinwangen standardized (c) and Kleinwangen field (d).

165
 CHAPTER III

Figure S4: Examples of wrongly classified standardized pictures. F.l.t.r.: Physical damage instead of Brownspots;
PLBM instead of Brownspots; PLBM instead of Brownspots; PLBM instead of Physical damage; Undamaged
instead of PLBM.

Figure S5: Examples of wrongly classified field pictures. F.l.t.r.: PLBM instead of Undamaged; PLBM instead of
Undamaged; Undamaged instead of Brownspots; Undamaged instead of Physical damage; Undamaged instead of
PLBM.

166
CHAPTER III

167
CONCLUSION AND PERSPECTIVES

The motivation for this thesis was to provide better understanding on how the
distribution of insect pest species will be affected by future climate change. In particular, using
spatial modelling approaches, this thesis contributes to a better understanding of: (i) the impact
of climate change on insect pest pressure, (ii) the consistency between inductive and deductive
modelling approaches for pest distribution modelling, (iii) how the structure of pest – host plant
metawebs will change in the near future, and (iv) implications of future insect pest occurrence
for plant protection services. Furthermore, using deep neural networks, this thesis also
investigates (v) how novel technologies may bring new opportunities to pest damage
forecasting.

The impact of climate change on insect pest pressure

This thesis contributes to a better understanding of spatial and temporal dynamics of

climatic suitability for insect pest species in Europe under climate change (Chapter 1 and
Chapter 2). As defined in Chapter 1, pest pressure is quantified here as the number of species
with suitable climatic conditions in a region. In Chapter 1, I show that the majority of 89
quarantine (i.e. black-listed) pest species, which have not yet established in Europe, encounter
suitable conditions in southern Europe under current climate, underlining high current-day pest
pressure in these regions. With progressing climate change, the suitable climate conditions of
insect pest species shift towards higher latitudes (Bebber et al., 2013), thus imposing increasing
pest pressure on European agriculture and forestry. Chapter 2 confirms this trend with two
comprehensive sets of insect pest species, including established pests and quarantine pests that
were previously considered mainly in isolation. The combination of two methodological
approaches (inductive and deductive), to compare fundamental and realized climatic niches,
further supports the findings of higher pest climatic suitability in temperate regions (Yan et al.,
2017) and increasing pest pressure on plant production systems under climate change (Deutsch
et al., 2018). Further, I corroborate the results of earlier studies that showed expanding potential
distribution for individual insect pest species across Europe (e.g. for Popilia japonica; Kistner-
Thomas, 2019) by investigating their climatic suitability as component of the comprehensive
lists of insect pest species included in Chapter 1 and Chapter 2. Finally, I find that western and
central Europe will be threatened by increasing pest pressure in the near future, and also

168
 CONCLUSION AND PERSPECTIVES

northern and eastern Europe will face higher pest pressure towards the end of the current
century.

Determining inflection points in pest niche accumulation is valuable to anticipate non-

linear changes in pest pressure under climate change. However, to the best of my knowledge,
studies investigating non-linearity of pest pressure are missing so far. Chapter 2 identifies an
inflection point around -3°C minimum temperature of the coldest month, beyond which the
number of pest species encountering suitable climatic conditions will drastically increase,
promoting a rapid enhancement of pest pressure. This inflection point marks a transition
between the groups of cold-adapted and warm-adapted pest species. Once this threshold is
crossed, warm-adapted pests that are not going through a winter diapause will find increasingly
suitable conditions. Therefore, under climate change, regions that cross this inflection point
temperature are threatened by an acceleration in pest invasions. The choice to use a minimum
temperature gradient to investigate niche distribution in climatic space is well justified, because
minimum temperature is the main limiting factor for species distribution in colder regions,
therefore also preventing from new pest invasions (Maxmen, 2013; Jarošik et al. 2015).

Comparison of inductive and deductive distribution models

Hutchinson’s niche concept distinguishes between the fundamental and the realized
niche of species (Hutchinson, 1957). While the fundamental niche defines the distribution of a
species based on abiotic environmental limits, the realized niche is additionally restricted by
biotic effects, such as competition or predation (Hutchinson, 1957; Wiens et al., 2009; Soberón
& Arroyo-Peña, 2017). Chapter 2 uses species distribution models (SDMs) to map the realized
niche (Guisan & Zimmermann, 2000; Guisan & Thuiller, 2005; Elith & Leathwick, 2009) and
physiological models to map the fundamental niche (Kearney & Porter, 2009; Venette et al.,
2010), giving two alternative perspectives on the potential distribution of established and
quarantine pest species. Although conceptual differences of the two modelling approaches
(Venette et al., 2010; Tonnang et al., 2017) are observable from species level comparisons,
where projections differ widely in some cases, at a broader assemblage scale results from these
different approaches tend to agree and predict an increasing pest pressure across Europe. The
benefit of combining approaches is that if models agree, their predictions are expected to be
more robust, while model disagreement can foster new insights and hypothesis (Hijmans &
Graham 2006). Specifically, we find model disagreements whereby the realized niche is

169
 CONCLUSION AND PERSPECTIVES

narrower than the fundamental niche (Chapter 2; Soberón & Arroyo-Peña, 2017). As a result,
there is less restriction from spatial predictions of physiological models at southern range
boarders, leading to differences in predicted species communities in southern Europe. However,
the broad scale increasing climate suitability for pest species towards higher latitudes is
congruent across models, adding additional robustness to the results from Chapter 1.

Looking at the similarities and discrepancies between fundamental and realized niches
is particularly interesting for species shifting their distribution range or invading new regions
(Tingley et al., 2014). This can foster insights on the invasion process, such as realized niche
shifts (Tingley et al., 2014), because those species are not at equilibrium with the environment
(Hill & Thomson, 2015). The realized niche in the invasive range of a species may differ from
the realized niche in the native range, because biotic constraints, such as competition or
predation, are absent (Broennimann et al., 2007). This allows the species to express another part
of its fundamental niche (Broennimann et al, 2007; Alexander and Edwards, 2010). Shifts in
the realized niche can also be caused by intraspecific variation in niche requirements, meaning
that a population in the invasive range may tolerate a different set of climatic conditions than
the population in the native range because the invasive population was drawn from a specific
area of the native range (Peterson & Holt, 2003; Strubbe et al., 2013). Meanwhile, the
fundamental niche does not change unless there is evolutionary or epigenetic change (Müller-
Schärer et al., 2004, Broennimann et al., 2007). Together, investigating the fundamental niches
using process-based models can give valuable insights on the potential species distribution in
the invasive range (Kearney et al., 2008), while investigating the realized niches using SDMs
is useful to anticipate the area of introduction and establishment of non-native species
(Broennimann et al., 2007). Pest pressure on European agriculture is increasing under climate
change, however, host availability (Niemelä & Matteson, 1996) and propagule pressure
(Lockwood et al., 2005) are also crucial drivers of invasion success (Bacon et al., 2014).

The impact of climate change on pest – host linkage and suitability overlap

Host availability is crucial for pest establishment and invasion (Niemelä & Matteson,
1996; Bacon et al., 2014). The changes in the linkage and the spatial overlap between insect
pests and their host plants under climate change inform on potential host availability for pest
species and therefore on regions at elevated invasion risk. Metawebs, or interaction network
approaches have been applied to ecological problems in natural ecosystems to investigate

170
 CONCLUSION AND PERSPECTIVES

network structures and impacts of climate change on species interactions and food webs
(Dunne, 2006; Pellissier et al., 2018). For instance, metaweb approaches based on species
distribution models have informed on changes in food web structures of fish species under
climate change (Albouy et al., 2014). While ecological network approaches can inform on
interactions in agricultural systems (e.g. Saunders & Rader, 2019) previous work has, to the
best of my knowledge, not quantified changing pest-host interactions under climate change.
Chapter 1 adopts such an ecological network approach to investigate the impact of climate
change on linkage and suitability overlap between insect pests and their host plants, delivering
new insights on general trends of pest-host interactions under climate change. Potential
interactions based on published host lists for each pest species are used to investigate the
realized linkage as modelled distribution overlap in Europe. Although the structure of the
metaweb is relatively simple, it creates valuable insights on the effects of climate change on
pest-host interactions and allows a more integrative understanding of the network system
(Pellissier et al., 2018), and therefore on pest pressure and invasion risk in different regions
under climate change.

Host availability for insect pests is dynamic and expected to change under climate
change, as the distribution of managed plants depends on climatic factors similarly as
unmanaged plants (Connor, Loomis & Cassman, 2011). To adequately inform on host
availability under climate change, I therefore use SDMs to investigate how crop and forest tree
distribution will change in the future (Chapter 1). This creates new insights on large-scale
patterns of crop and forest suitability across Europe. I investigate the climatic suitability of 96
crops and 30 forest tree species, therefore allowing inference on trends of host plant suitability
change rather than how the distribution of single plant species is affected by climate change
(e.g. maize, Ramirez-Cabral, Kumar & Shabani, 2017). While in southern Europe the climatic
suitability for crops and forest trees decreases until the end of the century, those plants will
generally find increasing climatic suitability in central, eastern and northern Europe. This will
open up new opportunities for cultivation (Maracchi, Sirotenko, & Bindi, 2005) without
increasing cropping areas. These results indicate that growers in large parts of Europe may
benefit from the cultivation of more valuable crops as they find increasingly suitable climatic
conditions, for instance wine production in northern Europe (Bindi et al., 2001). In addition,
diversification is often assumed to boost the resilience of cropping systems against disturbances
(Lin, 2011). However, the capitalization of these opportunities may be hampered by increasing
pest pressure associated to the new crop species. With larger scopes of different crops, host

171
 CONCLUSION AND PERSPECTIVES

availability increases for pests, promoting establishment and spread, and therefore putting more
pest pressure on the presently grown crops.

The analysis of the area of overlap between pests and their host plants is particularly
relevant because invasion success of pest species depends on host availability (Niemelä &
Matteson, 1996; Bacon et al., 2014). The metaweb approach combines two main drivers of
insect pest invasions, host availability and climatic suitability (Ward & Masters, 2007; Bacon
et al. 2014). The results of the metaweb approach show highest linkage and overlap in southern
Europe, which will slightly decrease by the end of the century, as climate conditions for both,
pests and plants will become increasingly unsuitable in southern Europe (Bindi & Olesen,
2011). Northern Europe is currently less threatened by pest invasions, but with increasing crop
suitability, more pest species could not only encounter suitable climate conditions but also
available host plants. Investigating the metaweb structure further shows that specialisation
decreases and more links per pest species become realizable. Particularly polyphagous pests
will benefit from climate change because their potential distribution overlaps with many host
plants. This confirms that generalist species pose the greatest risk of invasion, as those species
are most likely to encounter host plants in newly occupied areas (Ward & Masters, 2007).
Increasing pest pressure because of warming climate in Europe will thus be intensified by a
growing area of overlap between pests and their host plants because climate suitability of crops
and forest trees will also be affected by climate change. Chapter 1 shows that the increasing
linkage and climatic suitability overlap between pests and their host plants additionally favours
pest invasions across Europe, posing new challenges to plant protection in the near future.

Implications from increasing pest pressure for plant protection

Border control and surveillance measures are crucial to prevent introduction of insect
pests (Bacon et al., 2012). Strong efforts by plant protection services will be required to
minimize the risks associated to invasive insect pests, as climate change increases the invasion
risk (i.e., Chapter 1 and Chapter 2) and global trade and travel support the transport of insect
pests (Hulme et al., 2009; Liebhold et al., 2006; Roques, 2010). While I demonstrate that
climatic suitability and host availability for insect pest species are strongly increasing in the
future due to warming climate, propagule pressure largely depends on global trade and travel
(Liebhold et al., 2006; Hulme, 2009; Roques et al., 2010; Bacon et al., 2012). Global traffic is
expected to further increase in the near future (Levine & D’Antonio, 2003; Sardain, Sardain &

172
 CONCLUSION AND PERSPECTIVES

Leung, 2019). Meanwhile, the pool of invasive species seems to be insatiable (Seebens et al.,
2017), and continuously more regions are connected to the global trade market, serving as
newly accessible pools for potentially invasive species (Seebens et al., 2018). Currently, large
regions of the Mediterranean already provide suitable climate conditions for the majority of
quarantine pests and their host plants (Chapter 1 and Chapter 2). This thesis identifies these
regions as potential entry gates for pest invasions, as non-native pests establish and spread from
there as climate warms, as for example Tuta absoluta (Desneux et al. 2010). Indeed, those
regions exhibit the highest numbers of established non-native invertebrate species (Roques et
al., 2009). Furthermore, I identify a threshold releasing pests from abiotic constraints allowing
spread from the initial entry gates, for example, when rising temperatures in central Europe
cross a threshold in minimum temperatures of the coldest month around -3°C, obtained from
the results of Chapter 2 (i.e. inflection point temperature). Crossing this inflection point
temperature enables the spread of non-native pests from the entry gates to new areas. Therefore,
the results of Chapter 1 and Chapter 2 urge for strong efforts from plant protection services, as
decreasing propagule pressure is crucial to prevent invasions and will become even more
important under climate change (Bacon et al., 2012; Bacon et al., 2014). Although different
sampling strategies have been proposed to optimize inspections of plant material, capacities are
limited (Bacon et al., 2012; Chen et al, 2018) and only a small fraction of the trade goods are
inspected (McCullough et al., 2006; Poland & Rassati, 2019). To support the plant protection
services to more efficiently anticipate invasion of new pest species and prepare management
strategies, providing data on their current and potential future distribution serves as valuable
contribution.

A way to support plant protection services in conducting pest risk assessments is to

generate forecasts of the potential distribution of pest species by providing pest risk maps
(Venette et al., 2010). The findings of Chapter 1 and Chapter 2 underline the importance of
plant protection services for effective inspection of trade goods and prevention from pest
invasions. Hence, the results emphasize strongly that efforts from all stakeholders are needed
and that European plant health depends on the collaboration of all member states (MacLeod et
al., 2010) to minimize the costs associated to insect pest invasions. Information on potential
distribution contributes to pest risk assessments (Venette et al., 2010) for individual species,
helping to categorise and decide whether pests ought to be prioritized as quarantine pests (Baker
et al., 2009; EFSA Panel on Plant Health, 2018). Once potential locations for establishment of
a species are identified, targeted monitoring can be conducted, for instance with DNA based

173
 CONCLUSION AND PERSPECTIVES

tools (Darling & Blum, 2007), remote sensing (Rocchini et al., 2015), pheromone or kairomone
trapping (Poland & Rassati, 2019) or sentinel host plants (Kenis et al., 2018). The results of this
thesis can therefore be used for pest risk mapping and planning of monitoring actions. In light
of the increasing globalisation, even with stronger plant health regulations, more efficient
border surveillance and stronger efforts for pest monitoring, new species must be expected to
invade parts of Europe (Bacon et al., 2014). Therefore, strategies have to be developed to deal
with higher pest pressure and new tools for seasonal pest forecasting may help to absorb some
of the impacts of those invasive species.

Novel opportunities for pest forecasting

With increasing pest pressure challenging sustainable plant protection under climate
change, the demand for new tools for plant protection increases and big data and deep learning
open new opportunities to meet those demands and support pest management. Big data brings
many new opportunities to agriculture (Kamilaris et al., 2017; Coble et al., 2018), including
pest forecasting, as data-driven approaches can be used to create robust relationships between
pest occurrence and meteorological data (Orlandini et al. 2020). For instance, such data-driven
approaches may help to overcome current limitations of phenology models, such as labour- and
cost-intensive development and increasing discrepancies between model predictions and field
observations due to disruption of pest and host phenology under climate change (Kingsolver et
al., 2011). Although deep neural networks and big data have been used for pathogen and pest
classification (e.g. Mohanty et al., 2016, Cheng et al., 2017), to my knowledge these
technologies have never been applied in a pest forecasting context. Chapter 3 presents a novel
approach for a forecasting model as part of a proposed framework for the development of
decision support systems. In particular, the proposed framework connects data collection to
create big data sets with deep neural networks to produce phenological models for pest
damages. The results of Chapter 3 show that deep neural networks are capable of classifying
pest damages recorded with smartphone pictures and that the phenology of pest damages can
be obtained from a big dataset using a classification model established with deep neural
networks. Coupled with meteorological variables, the phenology can then be used to construct
a phenological model. More precisely, the case study presented in Chapter 3 highlights that this
approach is applicable to model the phenology of mines produced by larvae of the pear leaf
blister moth (Leucoptera malifoliella, Costa) on apple leaves. The established deep neural
networks perform well for different classes of damages individually and in multiclass

174
 CONCLUSION AND PERSPECTIVES

approaches. Performance is better on classes with many images available for training, and
models trained with pictures taken under standardized conditions perform better than those
trained with pictures taken in the field. Compared to studies using deep neural networks to
classify plant pathogens from leaf pictures, the performance of the networks presented in
Chapter 3 are slightly lower (Barbedo, 2018). However, earlier studies concentrated on the
classification tools alone and used full datasets for training and evaluation (e.g. Mohanty et al.,
2016). Meanwhile, Chapter 3 used only a small proportion of the images for model training.
The remaining images were used to test whether phenologies can be reconstructed. Still, the
results of this case study underline that pest forecasting could benefit greatly from exploiting
modern technologies for data collection (e.g. with drones; Finn and Donovan, 2016) and data
analysis (i.e. deep neural networks; Goodfellow et al., 2016), provided adequate data
availability. In the context of the framework for developing novel decision support systems, big
data and deep neural networks provide powerful tools, although new methods for data collection
must be established first.

Collectively, the three chapters presented show that the pest pressure from established
and quarantine insect pest species is increasing in Europe under climate change. I use two
complementary modelling approaches to provide additional robustness to the predictions of pest
climatic suitability and obtain an inflection point temperature beyond which the number of
species encountering suitable climate will drastically increase. In addition, I address host
availability as crucial factor for insect pest invasion using a metaweb approach and show that
linkage and suitability overlap between insect pests and their host plants must be expected to
increase in Europe under climate change. Finally, I show that big data and deep neural networks
could support developing data-driven pest forecasting models.

175
 CONCLUSION AND PERSPECTIVES

Perspectives

Pest risk analysis

The impact of invasive insect pests on global crop yields is increasingly considered as
one of the main challenges for food security (Oerke, 2006; Bradshaw et al., 2016; Deutsch et
al., 2018). This thesis provides a better understanding of the spatial and temporal patterns of
pest pressure for Europe under climate change and shows that increasing host availability
elevates the risk of invasion by insect pests. These findings are particularly valuable for
decision-makers in the plant protection sector and suggest action, such as dissemination of the
current knowledge and raising the awareness of this challenge. Insights on which quarantine
species pose risks to certain areas may help to better prepare for pest invasions and gain time
to implement management strategies, such as prevention, early detection and eradication,
containment and various forms of mitigation (Pyšek and Richardson, 2010). However, Chapter
1 and Chapter 2 would have benefited from the availability of more occurrence data and data
on physiological thresholds of pest species. For many quarantine species, no or little data were
available. With improving data availability, the set of pest species may become even more
representative as distribution models for additional species could be established. Additionally,
better computational power and climate data quality will help to produce predictions with
higher temporal and spatial resolution, allowing to extract even more valuable information from
such models.

Developing new tools that meet specific needs of pest risk analysts would improve the
support to plant protection services. Although, species distribution models and physiological
models can be directly used for basic pest risk mapping (Venette et al., 2010), other factors,
such as spread (Robinet et al., 2012) or economic impact (Soliman et al., 2010) need to be
addressed in pest risk analyses. Coupling those factors directly to pest distribution models could
be of great value for plant protection services. For instance, combining pest distribution models
with more sophisticated process-based crop models instead of crop distribution model as in
Chapter 1. Yield and productivity forecasts under climate change depend on complex
interactions between several mechanisms and process-based crop models can for instance
include beneficial effects from CO2 fertilization (Challinor et al., 2014) or detailed information
on irrigation regimes that might change in the future depending on water availability (Elliot et
al., 2014). Therefore, combining sophisticated crop models with pest distribution models would

176
 CONCLUSION AND PERSPECTIVES

allow to focus on the economically most relevant pest species threatening crop productivity
under climate change.

Pest damage forecasting

This thesis further shows that development of new pest forecasting models relying on
big data and deep neural networks is feasible, although future work needs to address the current
lack of data availability. Open image datasets in the agricultural domain are scarce, particularly
on pest and pathogen damages, therefore slowing down advances in pest and pathogen
classification (Barbado, 2018; Kamilaris & Prenafeta-Bolú, 2018). To develop decision support
systems based on big data and deep neural networks for broader application, new approaches
for data collection must be established first. Chapter 3 suggests two different approaches to
tackle this challenge. On the one hand, the implementation of drones is a matter of financial
resources, as high-tech cameras and drones are expensive, although this would be a one-time
investment that can be shared by grower communities. On the other hand, the citizen science
approach requires the initial investment for network establishment, which may pay off in the
longer term as additional beneficial effects may arise from including farmers in the
development (Lynch, Gregor and Midmore, 2000). Indeed, one of the main challenges for the
implementation of a decision support system is convincing growers to actually use it.
Nevertheless, in an unrepresentative survey I conducted during Chapter 3, growers showed
interest in such new tools. In addition, citizen science could help to detect invasive species
(Hulbert et al. 2017), as famers will be specifically interested in categorizing unknown
damages.

The development of robust and efficient pest forecasting tools could imply opportunities
for the management of invasive pest species. As pest pressure and invasion risk are expected to
increase under climate change, the implementation of novel technologies to support sustainable
intensification gains in relevance. The major advantage of the effective implementation strategy
of pest forecasting shown in Chapter 3 is that such tools can be quickly adapted to new pest
species. However, applying phenological models for decision support needs additional efforts.
To provide recommendations on management interventions and their optimal timing, solid
testing of plant protection measures is crucial. Nevertheless, the proposed approach for pest
forecasting presents a fast and efficient strategy to model pest damage phenology, which is
particularly useful in the situation where an increasing number of newly occurring pests is

177
 CONCLUSION AND PERSPECTIVES

expected. Finally, my hope is that the proposed framework approach can be used location
independent, meaning that countries lacking in resources and knowledge on pest and disease
management could benefit greatly from such tools.

This thesis shows that different modelling approaches are useful tools for several
applications in plant protection and pest management. Spatial modelling can provide valuable
information on the potential distribution of insect pests and could become even more useful for
pest risk analysts by including additional important factors that inform on the potentially most
relevant pest species. Although the need for development of new approaches for pest damage
forecasting is growing independently from increasing pest pressure under climate change, the
significance of perspectives arising from modern technologies to support sustainable food
production cannot be overemphasized.

178
 CONCLUSION AND PERSPECTIVES

References

Albouy, C., Velez, L., Coll, M., Colloca, F., Le Loc'h, F., Mouillot, D., & Gravel, D. (2014).
From projected species distribution to food‐web structure under climate change. Global
Change Biology, 20(3), 730-741.
Alexander, J. M., & Edwards, P. J. (2010). Limits to the niche and range margins of alien
species. Oikos, 119(9), 1377-1386.
Bacon, S. J., Aebi, A., Calanca, P., & Bacher, S. (2014). Quarantine arthropod invasions in
Europe: the role of climate, hosts and propagule pressure. Diversity and Distributions,
20(1), 84-94.
Baker, R. H. A., Battisti, A., Bremmer, J., Kenis, M., Mumford, J., Petter, F., ... & Karadjova,
O. (2009). PRATIQUE: a research project to enhance pest risk analysis techniques in
the European Union. EPPO Bulletin, 39(1), 87-93.
Barbedo, J. G. (2018). Factors influencing the use of deep learning for plant disease
recognition. Biosystems Engineering, 172, 84-91.
Bebber, D. P., Ramotowski, M. A., & Gurr, S. J. (2013). Crop pests and pathogens move
polewards in a warming world. Nature Climate Change, 3(11), 985.
Bindi, M., Fibbi, L., Lanini, M., & Miglietta, F. (2001). Free air CO2 enrichment (FACE) of
grapevine (Vitis vinifera L.): I. Development and testing of the system for CO2
enrichment. European Journal of Agronomy, 14(2), 135-143.
Bindi, M., & Olesen, J. E. (2011). The responses of agriculture in Europe to climate change.
Regional Environmental Change, 11(1), 151-158.
Bradshaw, C. J., Leroy, B., Bellard, C., Roiz, D., Albert, C., Fournier, A., ... & Courchamp, F.
(2016). Massive yet grossly underestimated global costs of invasive insects. Nature
Communications, 7(1), 1-8.
Broennimann, O., Treier, U. A., Müller‐Schärer, H., Thuiller, W., Peterson, A. T., & Guisan,
A. (2007). Evidence of climatic niche shift during biological invasion. Ecology Letters,
10(8), 701-709.
Challinor, A. J., Watson, J., Lobell, D. B., Howden, S. M., Smith, D. R., & Chhetri, N.
(2014). A meta-analysis of crop yield under climate change and adaptation. Nature
Climate Change, 4(4), 287-291.
Chen, C., Epanchin‐Niell, R. S., & Haight, R. G. (2018). Optimal inspection of imports to
prevent invasive pest introduction. Risk Analysis, 38(3), 603-619.

179
 CONCLUSION AND PERSPECTIVES

Cheng, X., Zhang, Y., Chen, Y., Wu, Y., & Yue, Y. (2017). Pest identification via deep
residual learning in complex background. Computers and Electronics in Agriculture,
141, 351-356.
Coble, K. H., Mishra, A. K., Ferrell, S., & Griffin, T. (2018). Big data in agriculture: A
challenge for the future. Applied Economic Perspectives and Policy, 40(1), 79-96.
Connor, D. J., Loomis, R. S., & Cassman, K. G. (2011). Crop Ecology: Productivity and
Management in Agricultural Systems, Cambridge University Press, New York, NY,
USA.
Darling, J. A., & Blum, M. J. (2007). DNA-based methods for monitoring invasive species: a
review and prospectus. Biological Invasions, 9(7), 751-765.
Desneux, N., Wajnberg, E., Wyckhuys, K. A., Burgio, G., Arpaia, S., Narváez-Vasquez, C.
A., ... & Pizzol, J. (2010). Biological invasion of European tomato crops by Tuta
absoluta: ecology, geographic expansion and prospects for biological control. Journal
of Pest Science, 83(3), 197-215.
Deutsch, C. A., Tewksbury, J. J., Tigchelaar, M., Battisti, D. S., Merrill, S. C., Huey, R. B., &
Naylor, R. L. (2018). Increase in crop losses to insect pests in a warming climate.
Science, 361(6405), 916-919.
Dunne, J.A. (2006). The network structure of food webs. In Pascual, M. & Dunne, J. A. (Eds.)
Ecological Networks: Linking Structure to Dynamics in Food Webs (pp. 27-86). Oxford
University Press, Oxford, UK.
EFSA Panel on Plant Health (PLH), Jeger, M., Bragard, C., Caffier, D., Candresse, T.,
Chatzivassiliou, E., ... & Navajas Navarro, M. (2018). Guidance on quantitative pest
risk assessment. EFSA Journal, 16(8), e05350.
Elith, J., & Leathwick, J. R. (2009). Species distribution models: ecological explanation and
prediction across space and time. Annual Review of Ecology, Evolution, and
Systematics, 40, 677-697.
Elliott, J., Deryng, D., Müller, C., Frieler, K., Konzmann, M., Gerten, D., ... & Eisner, S.
(2014). Constraints and potentials of future irrigation water availability on agricultural
production under climate change. Proceedings of the National Academy of Sciences of
the United States of America, 111(9), 3239-3244.
Finn R., Donovan A. (2016) Big data, drone data: Privacy and ethical impacts of the
intersection between big data and civil drone deployments. In Custers B. (Eds.) The
Future of Drone Use. Information Technology and Law Series, vol 27. T.M.C. Asser
Press, The Hague, The Netherlands.

180
 CONCLUSION AND PERSPECTIVES

Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. MIT Press, Cambridge,
MA, USA.
Guisan, A., & Thuiller, W. (2005). Predicting species distribution: offering more than simple
habitat models. Ecology Letters, 8(9), 993-1009.
Guisan, A., & Zimmermann, N. E. (2000). Predictive habitat distribution models in ecology.
Ecological Modelling, 135(2-3), 147-186.
Hijmans, R. J., & Graham, C. H. (2006). The ability of climate envelope models to predict the
effect of climate change on species distributions. Global Change Biology, 12(12), 2272-
2281.
Hulbert, J. M., Agne, M. C., Burgess, T. I., Roets, F., & Wingfield, M. J. (2017). Urban
environments provide opportunities for early detections of Phytophthora invasions.
Biological Invasions, 19(12), 3629-3644.
Hulme, P. E. (2009). Trade, transport and trouble: managing invasive species pathways in an
era of globalization. Journal of Applied Ecology, 46(1), 10-18.
Hutchinson, G. E. (1957). A Treatise on Limnology, vol. 1. Wiley, New York, USA.
Jarošík, V., Kenis, M., Honěk, A., Skuhrovec, J., & Pyšek, P. (2015). Invasive insects differ
from non-invasive in their thermal requirements. PloS One, 10(6), e0131072.
Kamilaris, A., Kartakoullis, A., & Prenafeta-Boldú, F. X. (2017). A review on the practice of
big data analysis in agriculture. Computers and Electronics in Agriculture, 143, 23-37.
Kamilaris, A., & Prenafeta-Boldú, F. X. (2018). Deep learning in agriculture: A survey.
Computers and Electronics in Agriculture, 147, 70-90.
Kearney, M. R., Wintle, B. A., & Porter, W. P. (2010). Correlative and mechanistic models of
species distribution provide congruent forecasts under climate change. Conservation
Letters, 3(3), 203-213.
Kearney, M., & Porter, W. (2009). Mechanistic niche modelling: combining physiological
and spatial data to predict species’ ranges. Ecology Letters, 12(4), 334-350.
Kearney, M., Phillips, B. L., Tracy, C. R., Christian, K. A., Betts, G., & Porter, W. P. (2008).
Modelling species distributions without using species distributions: the cane toad in
Australia under current and future climates. Ecography, 31(4), 423-434.
Kenis, M., Li, H., Fan, J. T., Courtial, B., Auger-Rozenberg, M. A., Yart, A., ... & Roques, A.
(2018). Sentinel nurseries to assess the phytosanitary risks from insect pests on
importations of live plants. Scientific Reports, 8(1), 1-8.

181
 CONCLUSION AND PERSPECTIVES

Kingsolver, J.G., Woods, H.A., Buckley, L.B., Potter, K.A., MacLean, H. & Higgins, J.K.
(2011) Complex life cycles and the responses of insects to climate change. Integrative
and Comparative Biology, 51, 719-732.
Kistner-Thomas, E. J. (2019). The potential global distribution and voltinism of the Japanese
beetle (Coleoptera: Scarabaeidae) under current and future climates. Journal of Insect
Science, 19(2), 16.
Levine, J. M., & D'Antonio, C. M. (2003). Forecasting biological invasions with increasing
international trade. Conservation Biology, 17(1), 322-326.
Liebhold, A. M., Work, T. T., McCullough, D. G., & Cavey, J. F. (2006). Airline baggage as
a pathway for alien insect species invading the United States. American Entomologist,
52(1), 48-54.
Lin, B. B. (2011). Resilience in agriculture through crop diversification: adaptive
management for environmental change. BioScience, 61(3), 183-193.
Lockwood, J. L., Cassey, P., & Blackburn, T. (2005). The role of propagule pressure in
explaining species invasions. Trends in Ecology & Evolution, 20(5), 223-228.
Lynch, T., Gregor, S., & Midmore, D. (2000). Intelligent support systems in agriculture: how
can we do better? Australian Journal of Experimental Agriculture, 40(4), 609-620.
MacLeod, A., Pautasso, M., Jeger, M. J., & Haines-Young, R. (2010). Evolution of the
international regulation of plant pests and challenges for future plant health. Food
Security, 2(1), 49-70.
Maracchi, G., Sirotenko, O., & Bindi, M. (2005). Impacts of present and future climate
variability on agriculture and forestry in the temperate regions: Europe. Climatic
Change, 70(1-2), 117-135.
Maxmen, A. (2013). Crop pests: under attack. Nature, 501(7468), S15-S17.
McCullough, D. G., Work, T. T., Cavey, J. F., Liebhold, A. M., & Marshall, D. (2006).
Interceptions of nonindigenous plant pests at US ports of entry and border crossings
over a 17-year period. Biological Invasions, 8(4), 611.
Mohanty, S. P., Hughes, D. P., & Salathé, M. (2016). Using deep learning for image-based
plant disease detection. Frontiers in Plant Science, 7, 1419.
Müller-Schärer, H., Schaffner, U., & Steinger, T. (2004). Evolution in invasive plants:
implications for biological control. Trends in Ecology & Evolution, 19(8), 417-422.
Niemelä, P., & Mattson, W. J. (1996). Invasion of North American forests by European
phytophagous insects. BioScience, 46(10), 741-753.
Oerke, E. C. (2006). Crop losses to pests. The Journal of Agricultural Science, 144(1), 31-43.

182
 CONCLUSION AND PERSPECTIVES

Orlandini, S., Magarey, R. D., Park, E. W., Sporleder, M., & Kroschel, J. (2020). Methods of
agroclimatology: modeling approaches for pests and diseases. Agroclimatology: Linking
Agriculture to Climate, 60, 453-488.
Pellissier, L., Albouy, C., Bascompte, J., Farwig, N., Graham, C., Loreau, M., ... & Rohr, R.
(2018). Comparing species interaction networks along environmental gradients.
Biological Reviews, 93(2), 785-800.
Peterson, A. T., & Holt, R. D. (2003). Niche differentiation in Mexican birds: using point
occurrences to detect ecological innovation. Ecology Letters, 6(8), 774-782.
Poland, T. M., & Rassati, D. (2019). Improved biosecurity surveillance of non-native forest
insects: a review of current methods. Journal of Pest Science, 92(1), 37-49.
Pyšek, P., & Richardson, D. M. (2010). Invasive species, environmental change and
management, and health. Annual Review of Environment and Resources, 35, 25-55.
Ramirez-Cabral, N. Y., Kumar, L., & Shabani, F. (2017). Global alterations in areas of
suitability for maize production from climate change and using a mechanistic species
distribution model (CLIMEX). Scientific Reports, 7(1), 1-13.
Robinet, C., Kehlenbeck, H., Kriticos, D. J., Baker, R. H., Battisti, A., Brunel, S., ... & Kenis,
M. (2012). A suite of models to support the quantitative assessment of spread in pest
risk analysis. PLoS One, 7(10), e43366.
Rocchini, D., Andreo, V., Förster, M., Garzon-Lopez, C. X., Gutierrez, A. P., Gillespie, T.
W., ... & Marcantonio, M. (2015). Potential of remote sensing to predict species
invasions: A modelling perspective. Progress in Physical Geography, 39(3), 283-309.
Roques, A., Rabitsch, W., Rasplus, J. Y., Lopez-Vaamonde, C., Nentwig, W., & Kenis, M.
(2009). Alien terrestrial invertebrates of Europe. In Handbook of Alien Species in
Europe (pp. 63-79), vol. 3. Springer, Dordrecht, Netherlands.
Roques, A. (2010). Alien forest insects in a warmer world and a globalised economy: impacts
of changes in trade, tourism and climate on forest biosecurity. New Zealand Journal of
Forestry Science, 40, S77-S94.
Sardain, A., Sardain, E., & Leung, B. (2019). Global forecasts of shipping traffic and
biological invasions to 2050. Nature Sustainability, 2(4), 274-282.
Saunders, M. E., & Rader, R. (2019). Network modularity influences plant reproduction in a
mosaic tropical agroecosystem. Proceedings of the Royal Society B: Biological
Sciences, 286(1899), 20190296.

183
 CONCLUSION AND PERSPECTIVES

Seebens, H., Blackburn, T. M., Dyer, E. E., Genovesi, P., Hulme, P. E., Jeschke, J. M., ... &
Bacher, S. (2017). No saturation in the accumulation of alien species worldwide. Nature
Communications, 8(1), 1-9.
Seebens, H., Blackburn, T. M., Dyer, E. E., Genovesi, P., Hulme, P. E., Jeschke, J. M., ... &
Ansong, M. (2018). Global rise in emerging alien species results from increased
accessibility of new source pools. Proceedings of the National Academy of Sciences of
the United States of America, 115(10), E2264-E2273.
Soliman, T., Mourits, M. C. M., Lansink, A. O., & Van der Werf, W. (2010). Economic
impact assessment in pest risk analysis. Crop Protection, 29(6), 517-524.
Soberón, J., & Arroyo-Peña, B. (2017). Are fundamental niches larger than the realized?
Testing a 50-year-old prediction by Hutchinson. PloS One, 12(4), e0175138.
Strubbe, D., Broennimann, O., Chiron, F., & Matthysen, E. (2013). Niche conservatism in
non‐native birds in Europe: niche unfilling rather than niche expansion. Global Ecology
and Biogeography, 22(8), 962-970.
Tingley, R., Vallinoto, M., Sequeira, F., & Kearney, M. R. (2014). Realized niche shift during
a global biological invasion. Proceedings of the National Academy of Sciences of the
United States of America, 111(28), 10233-10238.
Tonnang, H. E., Hervé, B. D., Biber-Freudenberger, L., Salifu, D., Subramanian, S., Ngowi,
V. B., ... & Niassy, S. (2017). Advances in crop insect modelling methods—Towards a
whole system approach. Ecological Modelling, 354, 88-103.
Venette, R. C., Kriticos, D. J., Magarey, R. D., Koch, F. H., Baker, R. H., Worner, S. P., ... &
De Barro, P. J. (2010). Pest risk maps for invasive alien species: a roadmap for
improvement. BioScience, 60(5), 349-362.
Ward, N. L., & Masters, G. J. (2007). Linking climate change and species invasion: an
illustration using insect herbivores. Global Change Biology, 13(8), 1605-1615.
Wiens, J. A., Stralberg, D., Jongsomjit, D., Howell, C. A., & Snyder, M. A. (2009). Niches,
models, and climate change: assessing the assumptions and uncertainties. Proceedings
of the National Academy of Sciences of the United States of America, 106(Supplement
2), 19729-19736.
Yan, Y., Wang, Y. C., Feng, C. C., Wan, P. H. M., & Chang, K. T. T. (2017). Potential
distributional changes of invasive crop pest species associated with global climate
change. Applied Geography, 82, 83-92.

184
CONCLUSION AND PERSPECTIVES

185
ACKNOWLEDGEMENTS

First, my gratitude goes to all members of my thesis committee, Dominique Mazzi, Pierluigi
Calanca and Loïc Pellissier who gave me the opportunity to start this journey and guided me
during this PhD. I am very thankful for the trust and freedom they granted me during this time
and for all their inputs and contributions to my work. I would also like to thank Sven Bacher
for accepting to be my external examiner.

I would like to thank Dirk Karger for his contributions to my first chapter and Jan Wegner for
his contributions to my third chapter and for helping me to learn valuable skills in deep learning.

I am also very grateful to Nicola Stäheli, Elisabeth Razavi and Simon Blaser from Agroscope
and Fanny Gutsche-Jones from the Citizen Science Center Zürich for all their contributions and
inputs.

A special thanks goes to all the members of the Landscape Ecology group from ETH Zürich, I
was very happy and proud to be part of this group! Thank you Patrice Descombes, Camille
Pitteloud, Giulia Donati, Oskar Hagen and Elena Haeler for your friendship and for all the
advice you gave me during this PhD and particularly in the final phase. I would also like to
thank Conor Waldock and Paul Dahlin for their helpful comments and suggestions on the thesis
manuscript.

A very special thanks goes to my partner Elena who was always there for me and supported me
during this time. I also want to thank my parents Maya and Hans-Martin to whom I owe a great
deal in all that I have accomplished so far. Finally, I would like to thank my late uncle Lukas
Schaub who inspired me to go in this direction and who was always a role model for me in
many aspects.

186