Ecological Informatics 31 (2016) 49–58

Contents lists available at ScienceDirect

Ecological Informatics

journal homepage: www.elsevier.com/locate/ecolinf

Applying spatial analysis to the agroecology-led management of an

indigenous farm in New Zealand
Antoni Moore a,⁎, Marion Johnson b, Janice Lord c, Sam Coutts a, Mariana Pagan a,
Jeremiah Gbolagun a, G. Brent Hall a,d
a
School of Surveying, University of Otago, Dunedin, New Zealand
b
Future Farming Centre, Lincoln, New Zealand
c
Department of Botany, University of Otago, Dunedin, New Zealand
d
Esri Canada, 12 Concorde Place, 9th Floor, Toronto, ON M3C 3R8, Canada

a r t i c l e i n f o a b s t r a c t

Article history: In Aotearoa New Zealand Māori land is often owned by communities and managed by trusts. Under communal
Received 7 January 2014 ownership, trust managers are expected to provide for their communities in culturally responsive ways, using
Received in revised form 11 November 2015 alternative land-related paradigms. In the context of Māori trust rural land management, geographic information
Accepted 23 November 2015
systems (GIS) are seen as a beneficial resource to plan and support important decisions that have community-
Available online 30 November 2015
wide implications. For example, it is much easier to demonstrate visually the putative outcomes of spatial change
Keywords:
to large numbers of people with the help of a GIS than it is to verbalise the effects of change. This paper reports on
Indigenous farming one aspect of such an application to a Māori trust-owned South Island farm, where a GIS-based multi-criteria
Agroecology analysis of terrain and proximity was conducted using also local botanical knowledge input to indicate optimal
GIS areas of growth for traditional medicinal plants. The analysis yielded very useful results, with positive feedback
Medicinal plants received from the local community. As well as the need to verify the approach empirically, there are plans under-
Rongoā way to apply GIS in this way to other similarly-managed farms in Aotearoa New Zealand. However, in order to
build a more comprehensive decision support resource for farm management, further plant species need to be
modelled, and informed by a wider data set that includes detailed soil data.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction 1.1. The need for farming based on biological principles and indigenous
agroecology
There has recently been a resurgence of interest in low impact
agricultural practices, arising out of a need to minimise the effects of Agriculture is the backbone of the economy in Aotearoa New
farming on the environment. To maximise the productivity of such Zealand.1 Over one million hectares in the country is designated as
biologically-led methods, explore alternative income streams for ‘Māori land’, held by Māori communities, and often administered by
farms, and to take full advantage of the positive public attention these trusts. When managers and trustees are considering changes to the
methods attract, spatial analysis is of immense value. Use of a state-of- land, protocols require that community members be consulted. Such
the-art geographic information systems (GIS) application for this pur- initiatives often seem abstract and unclear when described verbally.
pose can optimise the selection and placement of plant species, promote However, through the use of a GIS they can become much clearer
the positive environmental and economic impact of sustainable agricul- when presented visually to community members and to the public at
ture, and allow highly visual outputs of spatial analysis to be used for large.
planning and communication purposes. Many farm trusts envisage the implementation of a sustainable agri-
cultural system utilising organic, biological or low chemical methods,
thus decreasing the impact of farming on the environment. Changing
from chemical-based agriculture yet maintaining healthy animals at
economical stocking rates can be a challenge without a significant alter-
ation to prevailing management practices. Hence, the trusts have to
⁎ Corresponding author. Tel.: +64 3 479 7589; fax: +64 3 479 7586.
E-mail addresses: [email protected] (A. Moore), [email protected]
1
(M. Johnson). This is the correct bilingual (Māori and English) name.

http://dx.doi.org/10.1016/j.ecoinf.2015.11.009
1574-9541/© 2015 Elsevier B.V. All rights reserved.
50 A. Moore et al. / Ecological Informatics 31 (2016) 49–58

balance economic output with culturally and socially appropriate man- agroecology twinned with GIS and can help to establish and consolidate
agement systems. In addition to a move toward low input farming, farm land-based links with tipuna (ancestors) and turangawaewae (a place of
managers are also facing a requirement to fence waterways to preserve belonging) (Pacey, 2005). Combining indigenous knowledge and GIS in
water quality, which has become seriously degraded in many areas. agroecology, if achieved sensitively, and mindful of culture, security and
It is possible to combine the natural support of animal health with im- property rights, gives the advantage of capturing and storing cultural
provements to land and water quality by the judicious planting of multi- and traditional values and concepts, along with geographic data
purpose species to address welfare, management, biodiversity, and (Landcare Research, 2013). Used in this way and at this level, manage-
cultural concerns. Associated with this, the concept of providing a range ment of the GIS requires community participation, according to the now
of plant species with which animals can self-medicate when given the op- well established principles of public participation (Weiner et al., 2002).
portunity is gaining credence (Clayton and Wolfe, 1993; Huffman, 2003; An example of such an implementation in Motueka, Aotearoa New
Huffman and Seifu, 1989). The literature surrounding the effectiveness Zealand was reported on by Harmsworth et al. (2005), who described
of traditional plant remedies and the ability of farmed animals to treat an iwi (tribe)-led GIS project founded on local participation principles.
themselves is growing, for example in Africa (Githiori et al., 2004),
South America (Lans and Brown, 1998), India (Sharma and Singh, 1.3. Aim
1989), Europe (Pieroni et al., 2004) and the United States (Villalba et al.,
2006). This paper seeks to apply GIS-based multi-criteria analysis (MCA) or
Māori have an excellent knowledge of native plants and their prop- multi-criteria decision analysis (MCDA) to the indigenous agroecology-
erties (Riley, 1994) and used them extensively as medicine. Rongoā, or based management of a Māori trust-owned farm to discover the most
Māori traditional medicine, is holistic. Plants are seen as integral to suitable growing areas, given a set of land-based criteria, for native
healing but the land also plays an important role in the maintenance plants with traditional medicinal properties. MCA/MCDA is used in
of health. Accordingly, for the true personal health of the tangata decision-making contexts where there is normally a mixture of certain
whenua (people of the land), the land where one belongs, should also data/information and available, but less tangible, expert evaluations to
be healthy (Durie, 2001; Mark and Lyons, 2010; Pere, 1995). Hence, ap- factor in, but also where there is uncertainty present and conflicting
plying the principles of rongoā on a Māori trust farm assures the health criteria affecting the decision outcome (Malczewski, 1999). For this rea-
of the land, water, stock and overall environmental health. Fig. 1 illus- son, it is a powerful technique to apply to indigenous farm manage-
trates a number of commonly used rongoā species that can be grown ment, where there is a mix of decision makers with differing opinions,
on-farm in Aotearoa New Zealand. as well as variable criteria to assess the growth potential of different
In Aotearoa New Zealand the term indigenous agroecology, or the ap- plant species. Since expert knowledge and criteria may conflict, MCA al-
plication of ecological principles to agricultural systems, has been coined lows solutions to be discovered through variable criteria values and dif-
to describe the contribution and inclusion of mātauranga (Māori tradi- ferential weighting of criteria importance.
tional knowledge) to agroecological practices (Johnson et al., 2013). By
extension, rongoā rākau, or traditional medicine using plants, is consid- 2 . Materials and methods
ered to be an integral part of the indigenous agroecology paradigm.
2.1. Te Kaio farm, Te Pātaka o Rākaihautū, Te Waipounamu, Aotearoa New
1.2. The proposed role of GIS as a decision support tool for indigenous Zealand
farming practices
Te Pātaka o Rākaihautū or Banks Peninsula covers an area of approx-
The Māori notion of kaitiaki or guardianship of land and resources can imately 1000 km2 on the east coast of Te Waipounamu South Island of
be incorporated into resource management through the use of indigenous Aotearoa New Zealand. The land is of volcanic origin, and rises steadily

Fig. 1. Examples of the medicinal plant species used for healing in Aotearoa New Zealand: Kawakawa (Piper excelsum), Mapou (Myrsine australis), Harakeke (Phormium tenax), Kahikatea
(Dacrycarpus dacrydioides), Ti Kouka (Cordyline australis), Totara (Podocarpus totara), and Manuka (Leptospermum scoparium).
Photocredit NZ Plant Conservation Network.
 A. Moore et al. / Ecological Informatics 31 (2016) 49–58 51

from sea level to a height of 919 m, the summit of Te Ahu Patiki Mt like to attract kereru or native pigeon (Hemiphaga novaeseelandiae), tui
Herbert. (Prosthemadera novaeseelandiae), Pīwakawaka or fantail (Rhipidura
Te Kaio (Fig. 2) is a 449 ha sheep and beef farm located in the south- fuliginosa) and weka (Gallirallus australis). Planned plantings must there-
ern facing bays of the area. The farm is lightly stocked, currently carry- fore include forbs, shrubs and trees that provide the required habitat for
ing 800 cross-bred ewes and 60 cattle. During the last 20 years Te Kaio these bird species.
has had few chemical inputs and is progressing towards organic There should be as many benefits as possible accruing to each spe-
certification. cies chosen and a forb, shrub or tree must contribute to the farm in mul-
Te Kaio was gifted to the people of Wairewa and is managed by a tiple ways. For example, snowberry (Gaultheria antipoda) is excellent
trust. In line with the aspirations of many Māori land owners, the trust for wounds, promotes lactation and is a favourite food of weka. It also
wishes that Te Kaio should become a more biodiverse and eco- provides good ground cover. Hebe (Hebe strictissima) is renowned as a
friendly farm, while providing an economic return. Tourism may pro- remedy for diarrhoea, a tonic and encourages pollinators. Mapou
vide an option for income beyond farming outputs, particularly if the (Myrsine australis) is a tonic and good for wounds and provides
farm is managed according to the principles of indigenous agroecology. favoured habitat. The cabbage tree (Ti Kouka; Cordyline australis) is
iconic in Aotearoa New Zealand (Simpson, 2000), with many uses
2.1.1. Planting on Te Kaio from food and weaving to medicine. Totara (Podocarpus totara and
It is a basic tenet of rongoā that plants are part of the landscape Podocarpus hallii) forest previously clothed Te Pātaka o Rākaihautū
and their medicinal properties relate to the environment in which Banks Peninsula (Norton and Fuller, 1994), and there are many old
they grow. For example, research into the chemistry of Manuka Totara fence posts to be found on Te Kaio. Māori had multiple uses for
(Leptospermum scoparium) across New Zealand bears this view out the wood and bark of the Totara. Specifically, the berries are good to
(Maddocks-Jennings et al., 2005; Perry et al., 1997; Porter and Wil- eat and the tree has medicinal properties. Kawakawa (Macropiper
kins, 1998). In order to adhere to this principle and to maintain a excelsum) has an important role in Māori culture (Riley, 1994). It is an
good conservation ethic, planning for plantings on Te Kaio was re- excellent tonic, and is antiparasitic, good for wounds and toothache
stricted to species that are found growing naturally on Te Pātaka o (Porritt, 1967).
Rākaihautū/Banks Peninsula. In addition, all selected plants must
be non-toxic, withstand browsing by stock, and regenerate within a 2.2. GIS, spatial data used, and analysis
reasonable time frame.
It is neither possible nor sensible simply to plant large numbers of 2.2.1. Application of GIS to agriculture
species. Hence, a multifaceted approach is required so that plantings ad- The application of GIS (specifically analysis and mapmaking) in
dress conservation and protection, the historical existence of species, and farming activity provides stakeholders with valuable decision-support
the aspirations of the local community. Local people often wish to see, or capabilities. GIS use is well established in the agriculture production
have access to, particular plants and have a desire to encourage certain process throughout the world. It helps farmers to increase production,
fauna. For example, on Te Kaio, Wairewa community members would manage their land more efficiently, reduce costs (Pierce and Clay,

Fig. 2. Location map of Te Kaio farm.

52 A. Moore et al. / Ecological Informatics 31 (2016) 49–58

2007), increase profit margins, and mitigate the effect of farm-related therefore a significant addition to the toolset addressing the challenge
environmental disasters (Wilson, 1999). Cassel (2007) reports on the of spatial modelling for agriculture, indigenous or not. In a similar com-
use of GIS in precision agriculture, modelling soil properties in cases bination of data and domain knowledge, Reshmidevi et al. (2009) de-
where crop productivity is compromised by excessive salinity. scribe the application of a fuzzy rule-based inference system used in
conjunction with a GIS for land suitability assessment for agriculture
2.2.2. Spatial analysis in West Bengal, India.
Spatial analysis adds value to geographic data, turning data into in- The process flow used in this paper for the integration of underlying
formation to support decision-making through, for example, transfor- spatial data layers is shown in Fig. 4. Input data such as a Digital Eleva-
mative processes as simple as combining two or more datasets tion Model, coastline and on-farm infrastructure were used to derive
through overlay (Longley et al., 2011). In this research, a GIS-based spa- surrogate criteria (e.g. slope, aspect, distance from coast) for conditions
tial analysis was undertaken to aid the planning and planting processes that directly affect plant growth, such as frost, drought, waterlogging
of traditional medicinal species on Te Kaio farm. Fig. 3 shows a high- and wind. Botanical expert knowledge was used in the definition of
level schematic of the data and methods used. how different criteria affect species growth, as well as upper and
lower indicator thresholds for optimal growth conditions (implement-
2.2.3. Data ed at the reclassification and normalisation stage). The third role for bo-
The analysis methods used were determined by the nature of the tanical knowledge was in the definition of pairwise importance ratios
available spatial data. In this instance, a file geodatabase was initially for factors along with the reclassed and normalised indicator data to
created using ArcGIS software for all spatial data relating to Te Kaio the MCA process. Finally, a suitability map for each plant was derived
farm. This included a collection of feature datasets, feature classes, ta- as output from the MCA before their combination into a single suitabil-
bles and raster grids that would serve as the foundation for all subse- ity map for all species on the farm. Details of the components and the
quent analysis and mapmaking. The base spatial data for the project steps in this analysis process are now discussed.
were field surveyed in 2011 and subsequently partitioned into various
thematic layers, including roads, buildings, and elevations (also tracks, 2.2.4.1. Preparation of input data. Out of the available datasets shown at
fences, powerlines, poles and tree shelter belts) in the form of points, the top of Fig. 4, the Digital Elevation Model (DEM) of the farm site pro-
polylines and polygons. These survey data were subsequently converted vided several of the initial criteria for the assessment of traditional me-
into Esri feature datasets and classes using ArcGIS software. dicinal plant growth requirements. The specific derived criteria used
include elevation and its morphometric derivatives slope and aspect
2.2.4. Multi-criteria analysis for medicinal plant siting (a straightforward operation in a GIS), as well as distance from the
Multi-criteria analysis (MCA; Eastman, 1999; Malczewski, 1999) is coast and position on slope. As suggested in Fig. 4, data for the position
an advanced form of overlay, with criteria (equivalent to attributes, on the slope criterion were derived through a more involved fuzzy
each represented by a raster data layer) weighted for importance before modelling process to segment the DEM into areas of specific topograph-
combination according to domain knowledge. The process is run with ic identity, such as “ridges”, “valleys”, “flat”, “pits”, and “planar”. The
an overall goal in mind, and optionally the criteria can be grouped by method used by Fisher et al. (2004) was applied using Wood's LandSerf
the objectives they are related to if there are multiple competing objec- software (LandSerf, 2009).
tives to model (these objectives are also weighted, informed by domain With vague geographic concepts, fuzzy modelling allows areas to be
knowledge prior to combination). This process represents an imple- assigned continuous values in the range 0 to 1, depending on how well
mentation of the analytical hierarchical process (MCA-AHP) method the location matches the core concept being modelled (Elaalem et al.,
of decision support developed by Saaty (1980). 2011). An example in the context of this paper is a pixel that has a
The objectives and related criteria are ranked based on their suitabil- fuzzy membership of 0.8 relating to “ridges”. After sharpening the ter-
ity when compared to the weights and evaluations revealed in the pref- rain class membership values, that pixel would be coded as a ridge
erences of the decision maker(s). The degree of suitability of a criterion due to its relatively high fuzzy membership score in that category
is indicated using an index that is mapped by raster grids. Any grid cells (assuming that no other class of topographic feature has a higher mem-
that contain a value above a given suitability threshold represent loca- bership value). Hence, a group of contiguous pixels with the same class
tions which are the most suitable given the objective of the evaluation. value would define a ridge shape.
This method is capable of integrating expert knowledge and spatial data
to model the suitability of farm terrain for growth of plants with medic- 2.2.4.2. Botanical knowledge linking plant growing conditions with specific
inal properties. In particular, the method does not require a priori spe- topographic attributes. All plants have specific growth requirements. Be-
cies growth conditions to be identified within the study area and is fore the land suitability model could be built for the chosen species

KNOWLEDGE
FROM
BOTANISTS
AND
ECOLOGISTS MULTI-
CRITERIA
ANALYSIS RECOMMENDED
FOR PLANT RONGOA
SITING – PLANTING SITES
see figure 4

CONTEXTUAL
AND DOMAIN-
SPECIFIC
DATA

Fig. 3. Schematic of the project.

 A. Moore et al. / Ecological Informatics 31 (2016) 49–58 53

Fig. 4. Process-flow diagram of the medicinal plant spatial analysis.

(Mapou, Kahikatea, Totara, Kawakawa, Harakeke, Manuka and Ti The Green Toolbox, a web based decision support tool for land man-
Kouka), their environmental requirements had to be ascertained. Plant agers, was used to elucidate the topographic and climatic requirements
growth can be affected by a variety of factors including, among others, of each species. The Green Toolbox (http://www.landcareresearch.co.
climate, aspect and the presence of other species. Since detailed soil nz/resources/tools/green-toolbox) was developed by Landcare Re-
data were unavailable for Te Kaio farm, topographic factors only were search, the NZ Crown Research Institute for land-based natural science
considered in the analysis as they are known to influence microclimates (see also http://natureservices.landcareresearch.co.nz/about/). The
that create areas suitable for growth within the broader farm terrain. tool box was developed to provide land managers support in selecting

Table 1
Topographic indicators for environmental factors affecting the growth of the selected plants.
Source: Green Toolbox, Landcare Research.

Species Drought Wind Frost Water logging Salt wind

Elevation Aspect Slope Position Aspect Position Distance Aspect Slope Position Distance Aspect Slope Sheltered Exposed
on slope on slope from on slope from grid dist. grid dist.
coast (m) coast (m) (m) (m)

Kawakawa Mid–Hi E/W Med Mid Not NW Mid N1000 N Steep Mid b100 N Steep b10 b25
Mapou Mid–Hi E/W Med Mid NW/E Ridge b1000 N Steep Mid b100 E/W Med 10–50 25–100
Harakeke Lo–Mid N Steep NW/E Ridge b1000 S Gentle/Steep Pit N100 S Gentle b10 b25
Kahikatea/white Hi S Gentle NW/E Ridge b1000 E/W Med Mid N100 S Gentle N50 N100
pine
Totara Lo–Mid N Steep NW/E Ridge b1000 S Gentle/Steep Pit N100 E/W Med N50 N100
Manuka/Kanuka Lo–Mid N Steep NW/E Ridge b1000 S Gentle/Steep Pit N100 S Gentle 10–50 25–100
Ti Kouka/cabbage Lo–Mid N Steep Mid NW/E Ridge b1000 S Gentle/Steep Pit N100 S Gentle 10–50 25–100
tree
54 A. Moore et al. / Ecological Informatics 31 (2016) 49–58

plants for particular sites and draws upon the human expertise at a
number of land-based research institutes in Aotearoa New Zealand.
Specifically, data on aspect, slope, elevation, position on slope and prox-
imity (distance from coast) thresholds for each species were collected.
These data indicate the likelihood of the presence for environmental
conditions to which the plants might be exposed, namely wind, salt–
wind from seaborne locations, frost, drought or waterlogging of soil.
For example, north facing, steep slopes with mid-low elevations might
be prone to drought and Harakeke would not grow well there. The indi-
cators, derived from the “Green Toolbox” affecting the growth of the se-
lected plants, are given in Table 1.
The Green Toolbox was limited in that elevation and slope are only
loosely described. However, consultation with the toolbox developer
and other botanists was undertaken to translate linguistic thresholds
(e.g. “steep” slope) into numerical thresholds for each of the criteria. As-
pect did not need this interpretation as there was already enough infor-
mation to assign numerical limits i.e. as only north, south, east and west
are given as aspect descriptors, any one of these can be assigned 90°
ranges, e.g. north is modelled as between −45 (315) and 45°. The dis-
tance to coast criterion also had explicit numerical thresholds (e.g.
N100 m from the coast), as did the position on slope criterion. (The
Green Toolbox definitions “ridge”, “mid”, and “pit” were assumed to
match the fuzzy modelling output for “ridge”, “planar” and “pit” respec-
tively). A distance from structures criterion was also added, based on a
distance raster calculated for all pixels from buildings, roads and tracks.
This was derived from a farm management perspective, which suggests
that plants should not be placed where they are likely to interfere with
farm operations.
All criteria were modelled spatially using either the DEM, slope, as-
pect, the fuzzy position on slope rasters, coastline and buildings/roads/
tracks (the latter two used as the basis for the distance raster) as
input. As an example, for Harakeke and drought, raster reclassification
was used to isolate groups of pixels with northerly aspect, steep slope
and mid-low elevation. All outputs were normalised to a suitability
scale from 0 to 1, with the specified conditions for plant susceptibility
having the lowest suitability (Fig. 5).

2.2.4.3. Multi-criteria analysis. As noted earlier, MCA uses analytical hier-

archy processing (AHP) to build a hierarchical model that satisfies the
overall goal of the analysis by combining objectives that are made up
of criterion layers. To build the model, the criteria specified were used
as objectives in the analysis and these were satisfied by the combination
of the attribute (criterion) data layers through which the objective is
measured. The objectives were then combined to meet the overall
goal of finding areas for the minimum suitable growth requirements
of the specific plants. Fig. 6 shows this hierarchical structure. Formally,
for a goal G there is a set of p objectives Oq for q = 1,2, …, p. For the p
objectives there is a set of n criteria, Aj for j = 1,2, …, n. Therefore, for
the qth objective the set is Ak(q) for k = 1,2, …, l where l ≤ n. Two sets
of weights are associated with the objectives and criteria, respectively
(adapted from Boroushaki and Malczewski, 2008). Thus, in Fig. 6,
there are p = 6 objectives and where q = 1, for example, objective O1
(= water logging) has l = 2 criteria Ak(1), where A1(1) = aspect and
A2(1) = slope.
In this paper, the output rasters were combined using Idrisi GIS soft-
ware. The MCA toolbox in Idrisi uses a set of criterion map layers (the
rasters) for each objective and combines them using weightings derived Fig. 5. Some of the maps (normalised) used in the study for the conditions affecting
from pairwise comparison matrices. A similar combination process is growth of Harakeke: a) coastal elevation; b) northerly aspect; and c) steep slope.
then used to weight and combine the objective scores for the final
goal of the analysis. values were developed to provide the numerical definitions given to
the combination weightings to establish how important criteria were
2.2.4.4. Pairwise comparison matrices. Botanical information is rarely re- compared to each other. In this context, a value of 1 meant equal impor-
corded in a form suitable for use in modelling systems. For example a tance (i.e. criterion 1 is 1 times as important as criterion 2), 3 was mod-
“mid-position on a medium slope” is hard to enumerate. Hence, three erately more important, 5 was strongly more important, 7 was very
botanical expert botanists were consulted (separately) to formalise strongly more important, and 9 was extremely more important (effec-
the variables. Their information was consistent and a robust set of tively saying that criterion 1 is 9 times more important than criterion
 A. Moore et al. / Ecological Informatics 31 (2016) 49–58 55

Suitability for
GOAL
Growth

Distance from
OBJECTIVES Water Logging Salt Wind Frost Drought Wind
Structures

Distance to
ATTRIBUTES Aspect Distance to Coast Aspect Elevation Position on Slope
nearest structure

Slope Slope Aspect Distance to Coast

Position on Slope Slope Aspect

Distance to Coast Position on Slope

Fig. 6. Multi-criteria analysis hierarchical structure of criteria, objectives (factors) and goal.

2). The reciprocal also holds (e.g. if criterion 1 is 9 times more important everything with a suitability of 0.8 or higher was considered to be a suit-
than criterion 2, this was recorded, along with the information that cri- able location to grow the different plants. This threshold was decided on
terion 2 is one ninth as important as criterion 1). by using trial and error to find the best value that suited all of the differ-
Further dialogue with the botanists was needed to fill in the pairwise ent species. The thresholded maps are shown individually in Fig. 7 and
comparison matrices to rank the criteria by importance relative to the combined in Fig. 8.
growth requirements for each species. This information facilitated the The model outputs suggest that Totara can grow in most areas of the
translation of generic text descriptions (which is how botanical knowl- farm, except on the ridge tops, which are exposed to the prevailing
edge is expressed by experts) into specific numbers that could be used winds. Evidence of aged Totara fence posts was noted earlier, and anec-
for the analysis to find the suitable areas for the growth of each plant. dotal stories abound of the existence of Totara on the farm before it was
These values were then placed in each cell in the pairwise comparison cleared for agriculture. The model indicates that Ti Kouka will grow best
tables for subsequent analysis. Table 2 shows an example of the table on level or gently sloping areas particularly in valley floors. Mapou
for Harakeke. As examples of its content, waterlogging is moderately should be easy to establish on most areas of the farm. Like Totara, it
less important than salt–wind, and drought is extremely more impor- does not grow well when exposed to wind and would therefore be dif-
tant than waterlogging. ficult to establish on the ridge tops. Manuka has a strong preference for
This value scale was also used when combining at the criterion level north facing slopes, as it tends to favour drier conditions.
as shown in Table 3. This table reveals that frost, position on slope and Kawakawa, a plant highly valued by Māori, will only establish in lim-
elevation were the major considerations for the successful growth of ited areas of the farm. Kawakawa requires shelter and moisture so
Harakeke. This table is the same format for all of the plants but the cri- should grow well in the damper valleys and gullies. Planting possibili-
terion layers. It is applied to vary depending on how prone each species ties for Kahikatea showed a strong correlation to south-east facing
is to the particular conditions. slopes which are less prone to drought and to frost, both of which can
The weightings derived from the pairwise comparison matrices limit the establishment of this tree species. The model indicates that
were then combined using linear weighted regression to give a single Harakeke will be easiest to establish on the northern slopes of Te Kaio.
weighting for each criterion layer to produce the final suitability The suggested planting scheme was produced by giving the plants
indexes. with the lowest number of acceptable pixels priority over all of those
with more. The scheme still leaves about a third of the farm unplanted.
3. Results However, these areas will provide open grazing land for livestock. The
analysis defines unique growth sites for most species except Mapou
Suitability maps for the seven medicinal plant species modelled and Totara, which have nearly identical growth requirements. Hence,
were derived. A threshold was applied to the suitability indices so that they were combined for the suggested areas (Table 4).

Table 2
Pairwise comparison table of the objectives for Harakeke.

Water Salt Frost Drought Wind Distance from Table 3

logging wind structures Harakeke pairwise comparison table for frost.
Harakeke (NZ FLAX)
Elevation Aspect Slope Position on slope
Waterlogging 1 0.33 0.33 0.14 0.2 4
Salt–wind 3 1 1 0.14 3 5 Frost
Frost 3 1 1 0.14 2 6 Elevation 1 3 3 0.2
Drought 7 7 7 1 5 9 Aspect 0.33 1 0.2 0.2
Wind 5 0.33 0.5 0.2 1 7 Slope 0.33 5 1 0.2
Distance from structures 0.25 0.2 0.17 0.11 0.14 1 Position on slope 5 5 5 1
56 A. Moore et al. / Ecological Informatics 31 (2016) 49–58

Fig. 7. Thresholded suitability for plant growth maps. The marked areas are suitable for the following plants and can be related to the 3D terrain: a) Mapou/Totara, b) Ti Kouka, c) Manuka,
d) Kawakawa, e) Kahikatea, and f) Harakeke.

4. Discussion As noted earlier, data availability constrained some aspects of the

analysis, and time and effort need to be invested to help resolve this
All vegetation species modelled had their own niche areas within constraint. In particular, soil data are a notable omission from the
the farm extent, with very little competition for the same land. From a dataset used, as the soil maps for the area of Te Kaio farm were not de-
decision support perspective, this means that information is available tailed enough to be included in the analysis. If pre-existing data cannot
for a possible planting regime of species with medicinal properties. be found, then a detailed soil survey should be undertaken for future
Clearly, the actual planting of the rongoā species on the farm will analysis.
provide the ultimate validation of the results if the plants grow well in As with most research where expert knowledge is a key input to the
the areas indicated by the model. In future iterations, additional species analysis, there is a bottleneck in knowledge elicitation (e.g. see Jeng
with medicinal properties should be considered. et al., 1996). The information that was available was useful, however
The seven species modelled in this research were from a shortlist of it had a lack of specificity and needed refinement before the computer
eleven identified as being the most valued species by the owners of the software could process it. With this information in an incompatible for-
farm (non-availability of data meant that four were not processed). mat for analysis it meant that further consultation was needed with bot-
However, if all the medicinal species known were analysed, then the anists and ecologists to convert the linguistic descriptors into numerical
suitability maps could be enhanced as a basis for decision making on values that could be explicitly modelled. For some of the experts
the part of the farm managers. approached, this resulted in an unwillingness to commit knowledge
 A. Moore et al. / Ecological Informatics 31 (2016) 49–58 57

Fig. 8. Suggested planting scheme.

that is inherently expressed in language to a single value. As one expert effective way to model intangible local and botanical domain knowl-
put it, “… there is no recipe book for growing plants”. edge as well as associated data.
Finally, the uncertainties that are associated with the data, informa- The analysis has yielded very useful results, with positive feedback
tion and knowledge used in this study should be acknowledged. Like received from the Te Kaio farm community when the results were pre-
any data, spatial or otherwise, the data used are subject to spatial and sented to them in visual form. In fact, the farm trustees have resolved to
attribute uncertainty. For example, the raster DEM is subject to a resolu- use mapping and GIS in future management plans.
tion that has the potential to affect the results of the analysis operations As well as the need to verify these results empirically, there are plans
performed on it. Also, when basing an analysis procedure on the output to apply the approach to other similarly-managed farms elsewhere in
of another analysis procedure, the propagation of error from one to the New Zealand. However, in order to build a more comprehensive deci-
other needs to be explicitly modelled (Heuvelink, 1998). An example of sion support resource for indigenous farm managers, other species
this relates the modelled locations of medicinal plants relative to any fu- need to be modelled and informed by a wider data set, principally soil
ture stages of analysis. It is certainly important to make these sources of data, that were not available for this study. All other sources of uncer-
uncertainty explicit, not least in a decision making environment. tainty also need to be minimised, wherever possible.
The results of this research will be added to an ongoing initiative to
create a spatial data resource for agroecology (Pagan, 2013). The spatial
5 . Conclusions database is time-stamped so, although the agroecology spatial analysis
output is of one time only, it can house baseline data and spatiotemporal
This paper has successfully applied GIS techniques to a unique appli- agroecology monitoring data in current and future initiatives (e.g. biodi-
cation, namely indigenous agroecology farm management that uses a versity: invertebrates, vegetation, birds; soil conditions: structure,
combination of traditional Māori knowledge and biologically-led agri- chemistry; stream health: habitat, chemistry; cultural recordings:
cultural practices. The study reinforces the acknowledged value of GIS rongoā, and community; and farm management: seasonal use, and
to address local cultural and traditional values and concepts, and the stocking rate).
technique used (multi-criteria analysis) has been shown to be an
Acknowledgements
Table 4
Allocated plants by area. Thanks to Nga Pae o te Maramatanga for financial and studentship
support (Grant Number 11RF02-BHUOT), also to members of the Indig-
Plant Priority Number of pixels % coverage
enous Agroecology Project Group, including Hayden Hamilton, who
Non-suitable 1,487,416 34.33
performed part of the agroecology data collection in January–February
Kawakawa 1 54,181 1.25
Ti Kouka 2 190,677 4.40 2013. We are also grateful to the Te Kaio community, Michael Fletcher
Harakeke 3 201,228 4.64 and Environment Canterbury, Iain Gover, Alan Mark, Roger May, Colin
Manuka 4 326,958 7.55 Meurk and Robin Wybrow. Phil Rhodes, Tim Hastings and Riki Cam-
Kahikatea 5 1,034,428 23.88 bridge from the University of Otago are thanked for their baseline data
Mapou or Totara 6 1,037,674 23.95
survey in 2011.
58 A. Moore et al. / Ecological Informatics 31 (2016) 49–58

References Malczewski, J., 1999. GIS and Multicriteria Decision Analysis. John Wiley & Sons, New
York.
Boroushaki, S., Malczewski, J., 2008. Implementing an extension of the analytical hierar- Mark, G.T., Lyons, A.C., 2010. Māori healers' views on wellbeing: the importance of mind,
chy process using ordered weighted averaging operators with fuzzy quantifiers in body, spirit, family and land. Soc. Sci. Med. 70, 1756–1764.
ArcGIS. Comput. Geosci. 34, 399–410. Norton, D.A., Fuller, P.F., 1994. Restoration of lowland totara forest in Canterbury. N. Z.
Cassel, S.F., 2007. Soil salinity mapping using ArcGIS. In: Pierce, F.J., Clay, D. (Eds.), GIS J. For. 21–22 (May).
Applications in Agriculture. CRC Press, Boca Raton, FL, pp. 141–162. Pacey, H.A., 2005. The benefits and barriers to GIS for Māori. Submitted With Partial Ful-
Clayton, D.H., Wolfe, N.D., 1993. The adaptive significance of self-medication. Trends Ecol. filment of the Requirements for the Degree of Master of Indigenous Planning and De-
Evol. 8, 60–63. velopment. Lincoln University, Christchurch.
Durie, M., 2001. Mauri Ora: The Dynamics of Māori Health. Oxford University Press, Auck- Pagan, M., 2013. Taiwhenua and the collaboration of GIS technology and traditional Māori
land, New Zealand. knowledge: establishing a space–time dataset as the basis for assessing future spatial
Eastman, J.R., 1999. Multi-criteria evaluation and GIS. In: Longley, P., Goodchild, M., changes in indigenous agroecology practice. Report to Nga Pae o te Maramatanga.
Maguire, D., Rhind, D. (Eds.), Geographic Information Systems: Principles, Pere, R., 1995. Te Wheke: A Celebration of Infinite Wisdom. 2nd edition. AoAko, Gisborne,
Techniques, Management and Applications. John Wiley & Sons, pp. 493–502. New Zealand.
Elaalem, M., Comber, A., Fisher, P., 2011. A comparison of fuzzy AHP and ideal point Perry, N.B., Brennan, N.J., Van Klink, J.W., Harris, W., Douglas, M.H., McGimpsey, J.A.,
methods for evaluating land suitability. Trans. GIS 15 (3), 329–346. Smallfield, B.M., Anderson, R.E., 1997. Essential oils from New Zealand manuka and
Fisher, P., Wood, J., Cheng, T., 2004. Where is Helvellyn? Fuzziness of multi-scale kanuka: chemotaxonomy of Leptospermum. Phytochemistry 44, 1485–1494.
landscape morphometry. Trans. Inst. Br. Geogr. 29 (1), 106–128. Pierce, F.J., Clay, D., 2007. GIS Applications in Agriculture. CRC Press.
Githiori, J.B., Hoglund, J., Waller, P.J., Baker, R.L., 2004. Evaluation of anthelmintic proper- Pieroni, A., Howard, P., Volpato, G., Santoro, R.F., 2004. Natural remedies and nutraceuticals
ties of some plants used as livestock dewormers against Haemonchus contortus infec- used in ethnoveterinary practices in inland Southern Italy. Vet. Res. Commun. 28, 55–80.
tions in sheep. Parasitology 129, 245–253. Porritt, A., 1967. The history of medicine in New Zealand. Med. Hist. 11, 334–344.
Harmsworth, G., Park, M., Walker, D., 2005. Report on the Development and Use of GIS for Porter, N.G., Wilkins, A.L., 1998. Chemical, physical and antimicrobial properties of essen-
Iwi and Hapu: Motueka Case Study, Aotearoa—New Zealand. Landcare Research NZ Ltd. tial oils of Leptospermum scoparium and Kunzea ericoides. Phytochemistry 50,
Heuvelink, G.B.M., 1998. Error Propagation in Environmental Modelling with GIS. CRC 407–415.
Press, Boca Raton, FL. Reshmidevi, T.V., Eldho, T.I., Jana, R., 2009. A GIS-integrated fuzzy rule-based inference
Huffman, M.A., 2003. Animal self-medication and ethno-medicine: exploration and ex- system for land suitability evaluation in agricultural watersheds. Agric. Syst. 101
ploitation of the medicinal properties of plants. Proc. Nutr. Soc. 62, 371–381. (1–2), 101–109.
Huffman, M., Seifu, M., 1989. Observations on the illness and consumption of a possibly Riley, M., 1994. Māori Healing and Herbal New Zealand Ethnobotanical Source Book. 3rd
medicinal plant Vernonia amygdalina, by a wild chimpanzee in the Mahale Mountains edition. Viking Sevenseas N.Z. Ltd., Paraparaumu, New Zealand (528 pp.).
National Park, Tanzania. Primates 30, 51–63. Saaty, T.L., 1980. The Analytical Hierarchy Process: Planning, Priority Setting and Resource
Jeng, B., Liang, T.-P., Hong, M., 1996. Interactive induction of expert knowledge. Expert Allocation. McGraw-Hill, New York.
Syst. Appl. 10 (3–4), 393–401. Sharma, P.K., Singh, V., 1989. Ethnobotanical studies in northwest and Trans-
Johnson, M., Ataria, J., Champeau, O., Hudson, M., Lord, J., Wehi, P., Whaanga-Schollum, D., Himalaya. V. Ethno-veterinary medicinal plants used in Jammu and Kashmir,
Moore, A., 2013. Indigenous agroecology. Poster. VII Southern Connection Congress, India. J. Ethnopharmacol. 27, 63–70.
January 21–25, 2013, Dunedin, NZ. Simpson, P., 2000. Dancing Leaves. The Story of New Zealand's Cabbage Tree, Ti Kouka.
Landcare Research, New Zealand, Ltd., 2013. Methods for Recording Māori Values on GIS Canterbury University Press, Christchurch.
Retrieved June 21st 2013, from http://www.landcareresearch.co.nz/science/living/ Villalba, J.J., Provenza, F.D., Shaw, R., 2006. Sheep self-medicate when challenged with
indigenous-knowledge/land-use/recording-values. illness-inducing foods. Anim. Behav. 71, 1131–1139.
LandSerf, 2009. LandSerf. Jo Wood, City University, London (Retrieved June 25th 2015, Weiner, D., Harris, T.M., Craig, W.J., 2002. Community participation and geographic
from, http://www.landserf.org/). information systems. In: Craig, W.J., Harris, T.M., Weiner, D. (Eds.), Community Partic-
Lans, C., Brown, G., 1998. Ethnoveterinary medicines used for ruminants in Trinidad and ipation and Geographic Information Systems. Taylor & Francis, London, pp. 3–16.
Tobago. Prev. Vet. Med. 35, 149–163. Wilson, J.P., 1999. Local, national, and global applications of GIS in agriculture. In: Longley,
Longley, P.A., Goodchild, M.F., Maguire, D.J., Rhind, D.W., 2011. Geographic Information P.A., Goodchild, M.F., Maguire, D.J., Rhind, D.W. (Eds.), Geographical Information
Systems and Science. 3rd ed. John Wiley & Sons, Ltd., Chichester. Systems: Management Issues and Applications. John Wiley & Sons, pp. 981–998.
Maddocks-Jennings, W., Wilkinson, J.M., Shillington, D., Cavanagh, H., 2005. A fresh look
at manuka and kanuka essential oils from New Zealand. Int. J. Aromather. 15,
141–146.