Atmospheric Research 228 (2019) 107–121

Contents lists available at ScienceDirect

Atmospheric Research
journal homepage: www.elsevier.com/locate/atmosres

Climate classifications from regional and global climate models: T

Performances for present climate estimates and expected changes in the
future at high spatial resolution
⁎
Francisco J. Tapiadora, , Raúl Morenoa, Andrés Navarrob, José Luis Sánchezb,
Eduardo García-Ortegab
a
Institute of Environmental Sciences (ICAM), Faculty of Environmental Sciences and Biochemistry, Earth and Space Sciences (ess) Group, University of Castilla-La Mancha
(UCLM), Avda. Carlos III s/n, Toledo 45071, Spain
b
Atmospheric Physics Group (GFA), Environmental Institute, University of León (ULE), León 24071, Spain

A R T I C LE I N FO A B S T R A C T

Keywords: Climate classifications based on temperature and precipitation measurements are increasingly being used for
Climate classifications environmental and climate change studies. Using three classification methods (Köppen, Extended Köppen, and
Earth System Models Holdridge) and one observational dataset for present climate (CRU, Climate Research Unit), we show that GCMs
Global Climate Models have bridged the gap that led to the emergence of RCMs thirty years ago, as GCMs can now provide global
Regional Climate Models
climate classifications whose accuracy and precision are comparable to those of regional outputs of the RCMs.
Projections of high-resolution GCMs for future climates under the assumptions of three Representative
Concentration Pathways (RCP26, RCP45 and RCP85) can therefore be used as a primary source for climate
change and global warming studies at high resolution. This paper provides comprehensive, model-derived cli-
mate classifications for the entire planet, using RCMs and two GCMs for present and future climate-change
scenarios, and discusses how well the models actually represent the climates of the world when compared with
reference, ground validation data. It turns out that both GCMs and RCMs appear still limited to provide practical
estimates of the world climates even for present climate conditions. The modeling of precipitation remains the
Achilles' heel of models and thus of multidimensional indices, which are very sensitive to this variable. The
conclusion is that model outputs at regional scale need to be taken with extreme caution without venturing into
informing policies presenting potentially large societal impacts. Nonetheless, the role of models as privileged
tools to advance our scientific knowledge of the Earth's system remains undisputed.

1. Introduction Climate classifications are techniques used to perform a dimensional

reduction of physical variables into index-classes that can be related to
Classifications of atmospheric variables have been widely used to the biota. They have been applied to Arctic research (Wang and
identify structural relationships in weather and climate model outputs. Overland, 2004), studies of ecosystem impacts (Roderfeld et al., 2008),
They have been used for research in extreme weather and climatic biome distribution (Leemans et al., 1996) and biodiversity analyses
events (Chu and Zhao, 2011; Tramblay and Hertig, 2018), radiation (Garcia et al., 2014), hydrological cycle studies (Manabe and Holloway,
(Rozwadowska, 2004; Vindel et al., 2015), convection (Dimitrova et al., 1975), comparison of vegetation distribution (Monserud, 1990), ana-
2009; Aran et al., 2011; Lack and Fox, 2012), and for deriving cli- lysis of precipitation metrics (Tang and Hossain, 2012), analysis of
matologies of processes such as fog (Cereceda et al., 2008) and torna- future vegetation changes (Jiang et al., 2013), provision of input to
does (Giaiotti et al., 2007). Other applications of such classifications global models (Prentice, 1990), and visualization of climate change
include investigation into urbanization effects in weather and climate (Jylhä et al., 2010). Recently, Tapiador et al. (2019a) used three cli-
(Lin et al., 2018), and evaluation of products of precipitation (Ramos, mate classifications to evaluate the consensus of the outputs of 47
2001; Serra et al., 2014; Miró et al., 2017; Wen et al., 2017; Kim et al., Global Climate Models (GCMs) of different physical parameterizations
2017; Sharifi et al., 2018) and temperature (Peña-Angulo et al., 2016). and varied grid size. Increased computing capabilities over the past

⁎
Corresponding author.
E-mail address: [email protected] (F.J. Tapiador).

https://doi.org/10.1016/j.atmosres.2019.05.022
Received 8 April 2019; Received in revised form 17 May 2019; Accepted 24 May 2019
Available online 27 May 2019
0169-8095/ © 2019 Published by Elsevier B.V.
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Table 1
List of CORDEX simulations used in this study.
CORDEX domain Available simulations

EURO: Europe (40 simulations) 1. 11_CNRM-CERFACS-CNRM-CM5_CLMcom-CCLM4-8-17, 2. 11_CNRM-CERFACS-CNRM-CM5_RMIB-UGent-ALARO-0, 3.

11_ICHEC-EC-EARTH_KNMI-RACMO22E_KOPPEN, 4. 11_IPSL-IPSL-CM5A-LR_GERICS-REMO2015, 5. 11_IPSL-IPSL-CM5A-
MR_IPSL-INERIS-WRF331F, 6. 11_MOHC-HadGEM2-ES_CLMcom-CCLM4-8-17, 7. 11_MOHC-HadGEM2-ES_DMI-HIRHAM5, 8.
11_MPI-M-MPI-ESM-LR_CLMcom-CCLM4-8-17, 9. 11_MPI-M-MPI-ESM-LR_MPI-CSC-REMO2009, 10. 11_NOAA-GFDL-GFDL-
ESM2G_GERICS-REMO2015, 11. 44_CCCma-CanESM2_SMHI-RCA4, 12. 44_CNRM-CERFACS-CNRM-CM5_CLMcom-CCLM5-0-6
13. 44_CNRM-CERFACS-CNRM-CM5_CNRM-ALADIN53, 14. 44_CN- RM-CERFACS-CNRM-CM5_HMS-ALADIN52, 15. 44_CNRM-
CERFACS-CNRM-CM5_RMIB-UGent-ALARO-0, 16. 444_CNRM-CERFACS-CNRM-CM5_SMHI-RCA4, 17. 44_CSIRO-QCCCE-
CSIRO-Mk3-6-0_SMHI-RCA4, 18. 44_ICHEC-EC-EARTH_KNMI-RACMO22E, 19. 44_IPSL-IPSL-CM5A-MR_SMHI-RCA4, 20.
44_MIROC-MIROC5_CLMcom-CCLM5-0-6, 21. 44_MIROC-MIROC5_SMHI-RCA4, 22. 44_MOHC-HadGEM2-ES_CLMcom-CCLM5-
0-6 23. 44_MOHC-HadGEM2-ES_SMHI-RCA4, 24. 44_MPI-M-MPI-ESM-LR_CLMcom-CCLM4-8-17, 25. 44_MPI-M-MPI-ESM-
LR_CLMcom-CCLM5-0-6, 26. 44_MPI-M-MPI-ESM-LR_MPI-CSC-REMO2009, 27. 44_MPI-M-MPI-ESM-LR_SMHI-RCA4, 28.
44_NCC-NorESM1-M_SMHI-RCA4, 29. 44_NOAA-GFDL-GFDL-ESM2M_SMHI-RCA4, 30. 44i_CCCma-CanESM2_SMHI-RCA4, 31.
44i_CNRM-CERFACS-CNRM-CM5_HMS-ALADIN52, 32. 44i_CNRM-CERFACS-CNRM-CM5_SMHI-RCA4, 33. 44i_CSIRO-QCCCE-
CSIRO-Mk3-6-0_SMHI-RCA4, 34. 44i_ICHEC-EC-EARTH_KNMI-RACMO22E, 35. 44i_IPSL-IPSL-CM5A-MR_SMHI-RCA4, 36.
44i_MIROC-MIROC5_SMHI-RCA4, 37. 44i_MOHC-HadGEM2-ES_SMHI-RCA4, 38. 44i_MPI-M-MPI-ESM-LR_SMHI-RCA4, 39.
44i_NCC-NorESM1-M_SMHI-RCA4, 40. 44i_NOAA-GFDL-GFDL-ESM2M_SMHI-RCA4.
MENA: Middle East North Africa (6 1. 22_NOAA-GFDL-GFDL-ESM2M_SMHI-RCA4, 2. 22i_NOAA-GFDL-GFDL-ESM2M_SMHI-RCA4, 3. 44_CNRM-CERFACS-CNRM-
simulations) CM5_SMHI-RCA4, 4. 44_NOAA-GFDL-GFDL-ESM2M_SMHI-RCA4, 5. 44i_CNRM-CERFACS-CNRM-CM5_SMHI-RCA4, 6.
44i_NOAA-GFDL-GFDL-ESM2M_SMHI-RCA4.
AFR: Africa (31 simulations) 1. 44_CCCma-CanESM2_SMHI-RCA4, 2. 44_CCCma-CanESM2_UQAM-CRCM5, 3. 44_CNRM-CERFACS-CNRM-CM5_CLMcom-
CCLM4-8-17, 4. 44_CNRM-CERFACS-CNRM-CM5_SMHI-RCA4, 5. 44_CSIRO-QCCCE-CSIRO-Mk3-6-0_SMHI-RCA4, 6. 44_ICHEC-
EC-EARTH_KNMI-RACMO22T, 7. 44_IPSL-IPSL-CM5A-LR_GERICS-REMO2009, 8. 44_IPSL-IPSL-CM5A-MR_SMHI-RCA4, 9.
44_MIROC-MIROC5_GERICS-REMO2009, 10. 44_MIROC-MIROC5_SMHI-RCA4, 11. 44_MOHC-HadGEM2-ES_CLMcom-CCLM4-8-
17, 12. 44_MOHC-HadGEM2-ES_GERICS-REMO2009, 13. 44_MOHC-HadGEM2-ES_SMHI-RCA4, 14. 44_MPI-M-MPI-ESM-
LR_CLMcom-CCLM4-8-17, 15. 44_MPI-M-MPI-ESM-LR_MPI-CSC-REMO2009, 16. 44_MPI-M-MPI-ESM-LR_SMHI-RCA4, 17.
44_MPI-M-MPI-ESM-LR_UQAM-CRCM5, 18. 44_NCC-NorESM1-M_DMI-HIRHAM5, 19. 44_NCC-NorESM1-M_SMHI-RCA4, 20.
44_NOAA-GFDL-GFDL-ESM2G_GERICS-REMO2009, 21. 44_NOAA-GFDL-GFDL-ESM2M_SMHI-RCA4, 22. 44i_CCCma-
CanESM2_SMHI-RCA4, 23. 44i_CNRM-CERFACS-CNRM-CM5_SMHI-RCA4 24. 44i_CSIRO-QCCCE-CSIRO-Mk3-6-0_SMHI-RCA4
25. 44i_ICHEC-EC-EARTH_KNMI-RACMO22T 26. 44i_IPSL-IPSL-CM5A-MR_SMHI-RCA4 27. 44i_MIROC-MIROC5_SMHI-RCA4
28. 44i_MOHC-HadGEM2-ES_SMHI-RCA4, 29. 44i_MPI-M-MPI-ESM-LR_SMHI-RCA4, 30. 44i_NCC-NorESM1-M_SMHI-RCA4, 31.
44i_NOAA-GFDL-GFDL-ESM2M_SMHI-RCA4.
NAM: North America (4 simulations) 1. 44_CCCma-CanESM2_SMHI-RCA4, 2. 44_CCCma-CanESM2_UQAM-CRCM5, 3. 44_MPI-M-MPI-ESM-LR_UQAM-CRCM5, 4.
44i_CCCma-CanESM2_SMHI-RCA4.
CAM: Central America (18 simulations) 1. 44_CCCma-CanESM2_SMHI-RCA4, 2. 44_CNRM-CERFACS-CNRM-CM5_SMHI-RCA4, 3. 44_CSIRO-QCCCE-CSIRO-Mk3-6-
0_SMHI-RCA4, 4. 44_IPSL-IPSL-CM5A-MR_SMHI-RCA4, 5. 44_MIROC-MIROC5_SMHI-RCA4, 6. 44_MOHC-HadGEM2-ES_SMHI-
RCA4, 7. 44_MPI-M-MPI-ESM-LR_SMHI-RCA4, 8. 44_NCC-NorESM1-M_SMHI-RCA4, 9. 44_NOAA-GFDL-GFDL-ESM2M_SMHI-
RCA4, 10. 44i_CCCma-CanESM2_SMHI-RCA4, 11. 44i_CNRM-CERFACS-CNRM-CM5_SMHI-RCA4, 12. 44i_CSIRO-QCCCE-CSIRO-
Mk3-6-0_SMHI-RCA4, 13. 44i_IPSL-IPSL-CM5A-MR_SMHI-RCA4, 14. 44i_MIROC-MIROC5_SMHI-RCA4, 15. 44i_MOHC-
HadGEM2-ES_SMHI-RCA4, 16. 44i_MPI-M-MPI-ESM-LR_SMHI-RCA4, 17. 44i_NCC-NorESM1-M_SMHI-RCA4, 18. 44i_NOAA-
GFDL-GFDL-ESM2M_SMHI-RCA4.
SAM: South America (4 simulations) 1. 20_CCCma-CanESM2_INPE-Eta, 2. 20_MIROC-MIROC5_INPE-Eta, 3. 20_MOHC-HadGEM2-ES_INPE-Eta, 4. 44_MPI-M-MPI-
ESM-LR_MPI-CSC-REMO2009.
WAS: South Asia (2 simulations) 1. 44_MPI-M-MPI-ESM-LR_MPI-CSC-REMO2009, 2. 44i_MPI-M-MPI-ESM-LR_MPI-CSC-REMO2009.
SEA: South-East Asia (1 simulation) 1. 22_MOHC-HadGEM2-ES_SMHI-RCA4
AUS: Australasia (12 simulations) 1. 44_CSIRO-BOM-ACCESS1-0_UNSW-WRF360J, 2. 44_CSIRO-BOM-ACCESS1-0_UNSW-WRF360K, 3. 44_CSIRO-BOM-ACCESS1-
0_UNSW-WRF360L, 4. 44_CSIRO-BOM-ACCESS1-3_UNSW-WRF360J, 5. 44_CSIRO-BOM-ACCESS1-3_UNSW-WRF360K, 6.
44_CSIRO-BOM-ACCESS1-3_UNSW-WRF360L, 7. 44i_CSIRO-BOM-AC- CESS1-0_UNSW-WRF360J, 8. 44i_CSIRO-BOM-ACCESS1-
0_UNSW-WRF360K, 9. 44i_CSIRO-BOM-ACCESS1-0_UNSW-WRF360L, 10. 44i_CSIRO-BOM-ACCESS1-3_UNSW-WRF360J, 11.
44i_CSIRO-BOM-ACCESS1-3_UNSW-WRF360K, 12. 44i_CSIRO-BOM-ACCESS1-3_UNSW-WRF360L.
ARC: Artic (10 simulations) 1. 44_CCCma-CanESM2_SMHI-RCA4, 2. 44_MPI-M-MPI-ESM-LR_MGO-RRCM, 3. 44_MPI-M-MPI-ESM-LR_SMHI-RCA4-SN, 4.
44_MPI-M-MPI-ESM-LR_SMHI-RCA4, 5. 44_NCC-NorESM1-M_SMHI-RCA4 6. 44i_CCCma-CanESM2_SMHI-RCA4, 7. 44i_MPI-M-
MPI-ESM-LR_MGO-RRCM, 8. 44i_MPI-M-MPI-ESM-LR_SMHI-RCA4-SN, 9. 44i_MPI-M-MPI-ESM-LR_SMHI-RCA4, 10. 44i_NCC-
NorESM1-M_SMHI-RCA4.
ANT: Antarctica (2 simulations) 1. 44_ICHEC-EC-EARTH_KNMI-RACMO21P, 2. 44i_ICHEC-EC-EARTH_KNMI-RACMO21P

decades have resulted in Earth System Models (ESMs) and GCMs cap- temperature and precipitation thresholds, a tripartite climate classifi-
able of deriving outputs below 25 km grid size. It has even been argued cation is defined, with a 3-letter key identifying the major climate, the
that high resolution ESMs have replaced Regional Climate Models precipitation cycles, and the summer temperature. While his intention
(RCMs) for informing policy decisions about climate change adaptation was to help define climates using vegetation distribution as a surrogate
and mitigation strategies (Tapiador et al., 2019b). Building on this for the scarce meteorological data of the time, data availability is no
previous work, this follow-up paper extends such research to the out- longer an issue in 2019 (Tapiador et al., 2019a). Current computing
puts of RCMs and provides a critical evaluation of the relative merits of power makes it possible for ESMs to include biogeochemical feedbacks
each category. (Bradford et al., 2016), cryosphere dynamics (Mioduszewski et al.,
Climate classifications stem from the work of the Russian scholar 2019), and human population interactions (Navarro et al., 2018).
Vladimir Köppen in the late 19th century and were not intended to Drawing upon Köppen's theory, incremental advances, such as those
explain environmental factors in terms of climate variables. Rather, in Netzel and Stepinski (2016), have allowed a more detailed char-
Köppen's specific aim was to be able to define climates based on the acterization of climates. The use of objective, clustering methods in
observed distribution of vegetation at the time (Köppen, 1900). His global (Hoffman et al., 2005; Netzel and Stepinski, 2016) and regional
method is well-known and simple (Köppen's method is standard content (Tapiador et al., 2011) climate model outputs yields ‘model climatol-
in basic courses such as Geography 101): using a number of ogies’ that compare well with observations and with satellite estimates

108
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Table 2
Climatic classes for Köppen, Extended-Köppen and Holdridge.
Köppen (11 classes) Extended-Köppen (28 classes) Holdridge (33 classes)

Af: tropical rainforest climate, Aw: tropical savanna Af: tropical rainforest, Aw: tropical (wet savanna), Am: Polar Desert, Polar Desert, Polar Desert, Dry Tundra,
climate, BS: steppe climate, BW: desert climate, Cs: tropical monsoon, As: tropical (dry savanna), Ai: tropical Moist Tundra, Wet Tundra, Rain Tundra, Boreal
warm climate with dry summer, Cw: warm climate (range of temperature between the warmest and coldest Desert, Dry Scrub, Wet Forest, Rain Forest, Montane
with dry winter, Cf: humid temperate climate, Dw: months < 5 °C), Ag: Ganges types of annual march of Desert, Desert Scrub, Steppe, Moist Forest, Wet Forest,
cold climate with dry winter, Df: cold climate with temperature (hottest month comes before the solstice and Rain Forest, Subtropical Desert, Desert Scrub, Thorn
moist winter, ET: tundra climate, EF: permafrost the summer rainy season), BWh-BWk-BWn: desert (arid), Steppe, Dry Forest, Moist Forest, Wet Forest, Rain
climate. BWw: arid (winter drought), BWs: arid (summer drought). Forest, Tropical Desert, Desert Scrub, Thorn
BSk-BSh-BSn: steppe (semi-arid), BSs: semi-arid (summer Woodland, Very Dry Forest, Dry Forest, Moist Forest,
drought), BSw: semi-arid (winter drought), Csa-Csb: Mild Wet Forest, Rain Forest.
temperate (Mediterranean), Cfa-Cwa: mild temperate
(Humid subtropical), Cfb- Cwb: mild temperate (Oceanic),
Df: cold climate (moist winter), Dw: cold climate (dry
winter), Dfd, Dwd: cold climate (Subartic), ET: polar
(Tundra), EF: polar (permafrost).

of vegetation vigor, such as the NDVI index (Normalized Difference (2017), Rotstayn (2000), and Suzuki et al. (2015) and do not affect the
Vegetation Index). The problem with these objective classification results presented in this paper.
methods is that it is difficult to relate the automatic classes derived by
the algorithm to the biota, and this shortcoming reduces their applic-
ability to atmospheric and other environmental studies. 3. Methods

Balling (1984) divided the numerical procedures to determine cli-

2. Data matic classification into two major groups: similarity matrices and
grouping. Similarity matrices include the methods of correlation coef-
Two GCMs from those in the Coupled Model Intercomparison ficients, distance measure and principal component transformation.
Project (CMIP5, Taylor et al., 2012) were used to generate climate Grouping comprises the methods of taxonomic structure evaluation,
classifications: MIROC4h (Sakamoto et al., 2012) and CMCC-CM cluster selection, testing canonical correlation and cophenetic correla-
(Scoccimarro et al., 2011) at spatial resolutions of 0.56° and 0.75° re- tion –a technique to gauge how faithfully a dendrogram preserves the
spectively. The CMIP5 dataset has been used for research on historical pairwise distances.
changes in precipitation (Sillmann et al., 2013a,b; Ren et al., 2013) and Three grouping indices are used here: the classic Köppen system (11
in terms of seasonal changes (Li et al., 2013). CMIP5 simulations have classes), the Köppen-Trewartha scheme (28 classes; called here
also been critical in improving our understanding of atmospheric dy- ‘Extended Köppen’) and the Holdridge classification (33 classes). The
namics (Allen and Landuyt, 2014; Lott et al., 2014; Kitoh et al., 2013), classes are described in Table 2. The algorithm followed for the Köppen
cloud-aerosol interactions (Ekman, 2014), present and future biases and approach follows the pioneering work by Lohmann et al. (1993), who
uncertainties (Cattiaux et al., 2013) greenhouse gas forcing (Bellouin first used climate classifications as a diagnostic tool for climate models.
et al., 2011; Cook and Seager, 2013), and attribution to human activ- For Extended Köppen, we used the same method as Baker et al. (2009),
ities of the disruptions in precipitation cycles (Tapiador et al., 2016). which classifies the ecoregions in six main climate groups, and in which
In addition to the CMIP5 database, the CORDEX (Coordinate every subgroup is clearly defined. For these two climate classifications,
Regional Climate Downscaling Experiment) dataset was used. CORDEX the methods are the same as those used in Tapiador et al. (2019a,b,c).
is an international climate downscaling initiative that provides quality- As a novelty compared to the previous work, the Holdridge classifica-
controlled datasets for the recent historical past and projections into the tion method was used. It was implemented following the algorithmic
21st century (Giorgi et al., 2006; Jones et al., 2011). It uses the dyna- description in the original paper (Holdridge, 1947, 1967).
mical downscaling approach, which consists of RCMs driven by The Holdridge Life Zone (HLZ) classification is an empirically based
boundary conditions derived from GCMs. This technique allows higher system that characterizes vegetation types in equilibrium with their
spatial resolutions (0.44° and 0.11°) than the standard resolution of climate. For this purpose, HLZ divides the system in ecological units
GCMs used in CMIP5 (Zou et al., 2016). named as Life Zones. These units are delimited by biotemperature,
CORDEX divides the globe into 14 different domains covering the precipitation, potential evapotranspiration ratio, and elevation
majority of populated land regions. To facilitate inter-model compar- (Holdridge, 1967).
isons, a common domain for each region was defined. By September Holdridge's first classification was not exempt from criticism.
2018, the project had produced enough data to allow a sensible use of Several authors, for instance, pointed out that the first version of HLZ
their outputs for validation. The number of simulations used for each did not consider the seasonality in climatic parameters (Peng, 2000;
region is shown in Table 1. Smith et al., 1992; Yates et al., 2000). A subsequent revision modified
The observation data for the analyses are taken from the Climate this classification scheme in an attempt to account for seasonal varia-
Research Unit database (CRU, Harris et al., 2014). A major reason to bility (Holdridge et al., 1971) but that is not used here. The original
use CRU instead of other precipitation databases such as the Global HLZ classification has been widely used in ecological studies
Precipitation Climatology Project (GPCP, Adler et al., 2018) or the (Chakraborty et al., 2013; Sawyer and Lindsey, 1964; Yue et al., 2001),
Global Precipitation Climatology Centre (GPCC, Schneider et al., 2018) in regional analyses (Lugo, 1999; Szelepcsényi et al., 2014; Tatli and
estimates of precipitation is that most models have been nudged against Dalfes, 2016), and in climate change studies (Chen et al., 2003; Khatun
the CRU dataset, so in principle the comparisons here are the best-case et al., 2013; Szelepcsényi et al., 2018).
scenario in terms of agreement with observational datasets. In any case, In order to quantitatively validate the estimates of the projected
CRU and GPCC estimates of precipitation are quite similar (R2 = 0.914, climates, the Kappa statistic method (Cohen, 1960) was used. The
linear regression of CRU = 0.922·GPCC + 0.171), owing to the use of reader can find a full description of the rationale and calculation details
similar sets of rain gauges. The practice and consequences of model of the method in the previously cited work. The method was devised to
nudging and tuning have recently been explored by Hourdin et al. objectively compare nominal scales and has already been applied to

109
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

(caption on next page)

110
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Fig. 1. A comparison of observational-based climate classifications (first columns), RCMs-based classifications (second columns) and GCM-based classifications (third
and fourth columns), for present climate and for three classification methods: Köppen (first rows), Extended-Köppen (second rows) and Holdridge (third rows). The
domains are Europe (a), Middle-East and North Africa (b) and Africa (c).

gauge the performances of the Holdridge classification (Monserud and however, the latter is better than the former in the Holdridge classifi-
Leemans, 1992). By construction, the Kappa statistic (k) ranges from 0.0 cation, and the only model capable of correctly classifying the French
to 1.0. The evaluation thresholds are: [0.0, 0.1): No agreement; departments of Landes and Pyrénées-Atlantiques.
[0.1,0.2): Very poor agreement; [0.2,0.4): Poor agreement [0.4,0.5): For the Middle-East and North Africa (Fig. 1b), the < RCM >
Fair agreement; [0.6,0.7): Good agreement; [0.7,0.8): Very Good severely overestimates the extension of the Sahara and Arabian deserts.
agreement; [0.8,0.9): Excellent agreement; [0.9,1.0]: Perfect agree- The CMCC-CM simulation suffers from the same problem, but the
ment. MIROC4h is realistic in that key aspect. The interest of the Holdridge
classification is well exemplified in this region by the detail it provides
in the Sub-Saharan region. In Equatorial Africa, the RCMs show a sur-
4. Results and discussion
prising lack of capability. The simulations for the whole continent
(Fig. 1c), which include a different number of models and a different
Figs. 1-3 provide a visual account of the commonalities and differ-
domain, improve very slightly. The classifications for the global models
ences between the climates derived from the observational data (CRU),
in this area are of course the same as in the Middle-East and North
a “poor-man's ensemble” average of the available RCMs (< RCM >) for
Africa region (Fig. 1b).
each region, and the two GCMs: MIROC4h and CMCC-CM. While it is
The results for the Americas (Fig. 2) also show differences across
easy to combine and elaborate these results into aggregated graphs, it
observations, but, in this case, the models are more alike. In North
is, however, instructive to evaluate the outputs of the models directly
America (Fig. 2a), the < RCM > is close to the CMCC-CM, and both
over the maps as aggregation can mask important discrepancies at re-
compare reasonably well with the three reference, ground-truth clas-
gional scale. The supplementary information attached to this paper
sifications. The RCMs in Central America (Fig. 2b) show the effects of
contains maps for all the individual models, classification methods, and
Amazonia being too close to the boundary of the domain (evident in the
time periods (present and future climates), so the effect of the ensemble
Holdridge map) and thus performances are discrete, but there are dif-
averaging can be discounted for. It is outside the scope of the present
ferences even for the intended region of interest. The climate of the
paper either to pinpoint specific shortcomings of any model, compare
Amazonia is better captured when the domain extends further south
the relative merits of each RCM, or to provide details on specific per-
(Fig. 2c).
formances. The focus here is in the comparison between high resolution
Fig. 3 shows the consensus between GCMs and RCMs in hot humid
ESMs and the RCM consensus. Nonetheless, modelers can benefit from a
regions for the Köppen and Extended-Köppen classifications. However,
close inspection of how its model compares with the others in terms of
they disagree in the more detailed method (Holdridge, Fig. 3b). The
the aggregated metric that climate classification is.
models' climate classifications also fail in complex areas of South Asia
The results show that there are significant mismatches between the
and Australasia (Fig. 3a-c). Thus, for instance, CMCC-CM and <
models and the reference data in, for instance, the case of Europe
RCM > misrepresent the temperate climate of Australia's east coast
(Fig. 1a). This is more apparent in the fringes of the climate regions,
while MIROC4h overestimates the semi-arid region in central Australia.
where a small difference in the annual mean of the variables may de-
In other regions, such as peninsular India, models have difficulties in
cide the class to which they belong. Such a clear-cut characteristic
identifying the extension of climate classes.
implicit in hierarchical climate classification methods, i.e. their limited
The quantitative results of these comparisons are shown in Figs. 4-7.
tolerance to minor errors, is in principle a good feature in terms of
In these correlogram plots, colorized ellipses serve as visual indicators
validation. The implacability of the classification method towards
of correlation between models. Thus, the ellipse collapses diagonally as
minor differences in the simulated variables increases with the number
the agreement between two models approaches k = 1. The ellipse be-
of classes, so the more detailed the Extended Köppen and Holdridge
comes a circle when the two models completely disagree (k = 0).
systems are, the less indulgent they are with model errors, whereas
Comparison with CRU reference data (validation) for Europe using the
Köppen allows more leeway.
Köppen classification generates the best results, with both the <
It is worth remembering that Europe is a region with a large number
RCM > and the MIROC4h presenting excellent performance. In con-
of RCMs (40), which were able to be adjusted and fine-tuned in the
trast, all the models perform only adequately in Australasia and South
development stage towards already high-quality reference data (CRU).
East Asia. Over this region, the kappa index for Holdridge climates is
While other regions -notably the polar regions- do not have lengthy
poor.
precipitation records, models in Europe have benefited from such re-
What are the reasons for such modest capacity? Detailed inspection
cords and from decades of experience and local knowledge applied to
of the cross-correlations between observations and simulations for
improving the codes. Indeed, no RCMs would have survived in the
precipitation and temperature shows a shared pattern for all the regions
community without managing to realistically capture the European
(Figs. 8 and 9). While, in general, temperature is well modeled, with R2
climate in the historical era. For other areas, where some RCMs have
correlations above 0.80 for all but one case (the MIROC4h model over
been applied for the first time, there is still room for future adjustments
South-East Asia), the precipitation is still poorly represented in the
and fine-tuning, but Europe (and North America) represents, in a sense
models. This is unsurprising given the difficulties in not only model but
the best-case setup for models to show off their actual capabilities.
also measure precipitation (Kidd et al., 2017; Tan et al., 2015; Tapiador
Southern Spain is particularly poorly modeled (Fig. 1a). The dry
et al., 2017; Xue et al., 2013). The large dispersion in the scatterplots in
semiarid, steppe, class (BS) in the Köppen classification do not feature
Figs. 8 and 9 shows how easily precipitation is severely over- or under-
in the < RCMs > , and is minimal in the MIROC4h. The CMCC-CM
estimated even at spatially-averaged and temporally-aggregated values.
model correctly simulates the actual class of this area but only for the
These results suggest that both GCMs and the RCMs are still too
more arid sector in the Spanish southeast. In terms of the Extended
limited to extract definite estimates of the world climates, even for
Köppen, the south of the Iberian Peninsula, which belongs to the dry
present climate conditions. Moreover, a closer view of the agreement
climates with deficient precipitation during most of the year (B in
between modeled and observed precipitation in South and Central
Table 2), is erroneously assigned to the C classes (moist Mid-latitude
America, Asia, Australasia and the Arctic raises serious doubts about
climates with mild winters) by the RCMs. The CMCC-CM is better, while
whether or not such estimates can be used in practice. While the maps
the MIROC4h is less precise but better than the RCMs. Interestingly,

111
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Fig. 2. As Fig. 1, for the Americas.

112
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Fig. 3. As Fig. 1, for Asia, South East Asia and Australasia. The feature at 180E is a plotting artifact.

113
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Fig. 4. Class agreement for three climate classifications: Köppen (left), Extended-Köppen (center) and Holdridge (right). The upper-left triangles of the matrices
display the correlation ellipsoids between classes and the bottom-right triangle shows the Kappa coefficient matrix (k). k = 1 means a perfect agreement and k = 0 no
agreement at all. Notice that there are a different number of RCMs for each region. < RCM > is the ensemble average of the available models. Plots including every
RCM are available in the supplementary information.

may in some way blur the actual differences, the scatterplots reveal delineate.
large differences for this climate variable, which is critical for life and The confidence in models being capable of capturing the pre-
human activities. cipitation of the future climates is also a matter of further discussion
Notwithstanding the limited skill exhibited by the models, with considering the modest performances exhibited in the simulation of
k < 0.5 in many cases, it is worth calculating the expected climates of present climates, as presented above.
the future under the three more common Representative Concentration
Pathways (RCP 2.6, RCP 4.5 and RCP 8.5, van Vuuren et al., 2011) for
several regions. What Figs. 10 and 11 show is the remarkable con- 5. Conclusions
sistency between the GCM used here and the RCMs, and also the small
changes expected in some regions using this metric. Thus, for Europe, Climate classifications have proved useful in reducing the di-
the most important changes are those under RCP 8.5 (Fig. 10a) as one mensionality of global and regional climate models' outputs into vari-
can expect. A closer look at specific regions such as the Anatolian ables that can be directly related with the biota. Index-based methods
plateau and the Iberian Peninsula reveals an increase in the extent of such as Köppen, Köppen-Trewartha and Holdridge have the advantage
dry climates (BS in Table 2). A noticeable reduction of cold climates of being more easily interpreted than objective classifiers such as the K-
over Central and Eastern Europe is also expected in RCP 8.5 (Fig. 10a, Means algorithm, and are more suitable for a direct application in en-
Extended-Köppen). vironmental studies. Thus, a major advantage of the Köppen-Trewartha
Even though major changes are projected in Amazonia (Fig. 11b, method (Extended Köppen in this paper) is the immediacy in the in-
Holdridge), the Indian subcontinent (Fig. 11c, Ext-Köppen and Hol- terpretation of the changes, as the classes can be more easily related to
dridge) and the southern region of the Arabian peninsula (Fig. 10b, Ext- climatological categories used in most biogeographical studies
Köppen), these should be taken with caution because of the difficulties (Tapiador et al., 2019a).
in the precise estimation of precipitation at ground in those areas. In the An additional and useful application of such classifications is the
first case, the Amazonia, the number of rain gauges is far too small for comparison and validation of climate models. Here, using several RCMs
such a large area. In the Indian subcontinent, both the monsoons and and two high-resolution GCMs, it has been shown that finer (more
the Himalayas affect the precise estimation of precipitation at ground, detailed) classification methods can help to reveal model shortcomings
while for the Arabian Peninsula, the desert boundary is difficult to and thus to advance the field. Indeed, through bias correction or nud-
ging towards reference data, models and model outputs can be forced to

114
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Fig. 5. As Fig. 4, for the Americas.

perform better in the validations. However, a major driver of validation et al. (2019c), precipitation is the ultimate test of model performance.
is precisely to identify issues so they can be corrected. Since errors are The use of climate classifications such as those shown here as a vali-
usually region-dependent, identification of the mismatches can assist dation metric is conceptually interesting because it has the potential to
model developers to focus on the processes still not well represented in provide valuable information for environmental applications about the
models, such as shallow convection or mixed-phase precipitation. likely climates of the future. However, the results of the comparisons
Indeed, the deficiencies in the physics of the models greatly affects show that the modeling of precipitation remains the Achilles' heel of
precipitation simulations and therefore climate classifications such elaborated, multidimensional indices. What is clear is that the
(Tapiador et al., 2019a). weight given to a climate class in a geographical location is dependent
From the results of the calculations in this paper, it can be con- on how well precipitation is modeled in that particular place, and the
cluded that while there is an overall agreement between models and extent of the observational issues affecting the reference database.
observations, large uncertainties persist for large parts of the world for Only better and more direct observations, such as those provided by
both GCMs and RCMs. Assuming that the models can simulate the fu- the Global Precipitation Measuring (GPM) mission, could provide a
ture as well as they can model the present, future climate classifications more detailed picture of the actual precipitation on Earth and reduce
with RCMs were derived for three RCPs (RCP 2.6, RCP 4.5 and RCP current uncertainties in climate classifications (Tapiador et al., 2019a).
8.5). The extent to which they can be considered a good proxy for the The community objective for the future is to be able to observe total
climate types of the future and be used to provide useful input for en- precipitation to an average accuracy of 15% over oceans and/or 25%
vironmental research remains uncertain given their performance for over land and ice surfaces averaged over a 100 × 100 km2 region and
present climate conditions. However, the host of RCM and GCM models two to three day time period (National Academies of Sciences, 2018).
show a reasonable consensus regardless of the classification method Achieving such an important goal would undoubtedly improve the
used, and incrementing GHG concentrations yields increasingly more classification of present and future climates and would make them more
profound changes, with large areas of the world shifting towards dif- suitable for environmental applications.
ferent climates from those in the present. As a final conclusion of the results of the research shown here, it
Reference data is seldom perfect, especially regarding precipitation, seems clear that the applicability of model outputs has to be taken with
and this represents a challenge for validation. This is particularly im- extreme caution and within the realm of pure research, without ven-
portant in regions with a low density of rain gauges and/or a limited turing into informing policies with a potentially large societal impact.
record of historical measurements. Temperature records, while more Ultimately, physical models are merely simplified tools to increase our
consistent, have the disadvantage of being easier to model owing to understanding of the world. Taking their outputs as face value and
their smaller spatial and temporal variability. As argued by Tapiador mistakenly considering the results of careful simulations as a 1-to-1

115
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Fig. 6. As Fig. 4, for the Asia, South East Asia and Australasia.

Fig. 7. As Fig. 4, for the polar regions.

116
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Fig. 8. Scatterplots of the monthly mean precipitation and annual mean temperature at every grid point against the ground reference data (CRU). Each panel depicts
the ensemble of RCMs and the two GCMs (MIROC4h and CMCC-CM). Pixels from coastal areas were removed to avoid the negative effects of interpolation (only for
temperature).

117
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Fig. 9. As Fig. 8 for specific sectors in the RCM domains: Australia (left), Central America (center) and the Arctic (right).

Fig. 10. A comparison of GCM-based climate classifications (CMCC-CM) and RCM-based classifications, for present climate and future scenarios (RCP 2.6, RCP 4.5
and RCP 8.5) for three classification methods: Köppen (left), Extended-Köppen (center) and Holdridge (right).

118
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Fig. 11. As Fig. 10 for North America, Central America and Asia.

relationship with observed nature is not only a pervading error in the References
history of science but also a practice that has consistently proved to
hinder further progress. Adler, R.F., Sapiano, M.R.P., Huffman, G.J., Wang, J.J., Gu, G., Bolvin, D., Chiu, L.,
Schneider, U., Becker, A., Nelkin, E., Xie, P., Ferraro, R., Shin, D.B., 2018. The Global
Precipitation Climatology Project (GPCP) monthly analysis (New Version 2.3) and a
Acknowledgements review of 2017 global precipitation. Atmos. 9, 138.
Allen, R.J., Landuyt, W., 2014. The vertical distribution of black carbon in CMIP5 models:
Funding from projects CGL2013-48367-P, CGL2016-80609-R Comparison to observations and the importance of convective transport. J. Geophys.
Res. Atmos. 119, 4808–4835.
(Spanish Ministry of the Economy and Competitiveness, MINECO) is Aran, M., Pena, J.C., Torà, M., 2011. Atmospheric circulation patterns associated with
gratefully acknowledged. AN, EGO, JLS acknowledge grant LE240P18. hail events in Lleida (Catalonia). Atmos. Res. 100 (4), 428–438.
RM acknowledges a FPI grant from the MINECO. FJT acknowledges Baker, B., Diaz, H., Hargrove, W., Hoffman, F., 2009. Use of the Köppen-Trewartha cli-
mate classification to evaluate climatic refugia in statistically derived ecoregions for
grant 1365002970/KMA2018-00721 (Development of Numerical the People's Republic of China. Clim. Chang. 98, 113–131.
Weather Prediction and Data Application Technique) from the Korean Balling, R.C., 1984. Classification in climatology. In: Gaile, G.L., Willmott, C.J. (Eds.),
Meteorological Agency. Spatial Statistics and Models. Springer, Dordrecht, pp. 81–108.
Bellouin, N., Rae, J., Jones, A., Johnson, C., Haywood, J., Boucher, O., 2011. Aerosol
forcing in the climate model intercomparison project (CMIP5) simulations by
Appendix A. Supplementary data HadGEM2-ES and the role of ammonium nitrate. J. Geophys. Res. Atmos. 116,
D20206.
Bradford, M.A., Wieder, W.R., Bonan, G.B., Fierer, N., Raymond, P.A., Crowther, T.W.,
Supplementary data to this article can be found online at https:// 2016. Managing uncertainty in soil carbon feedbacks to climate change. Nat. Clim.
doi.org/10.1016/j.atmosres.2019.05.022.

119
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Chang. 6, 751–758. to ecosystem mapping. J. Biogeogr. 26, 1025–1038.

Cattiaux, J., Douville, H., Peings, Y., 2013. European temperatures in CMIP5: origins of Manabe, S., Holloway, J.L., 1975. The seasonal variation of the hydrologic cycle as si-
present-day biases and future uncertainties. Clim. Dyn. 41, 2889–2907. mulated by a global model of the atmosphere. J. Geophys. Res. 80, 1617–1649.
Cereceda, P., Larrain, H., Osses, P., Farías, M., Egaña, I., 2008. The climate of the coast Mioduszewski, J.R., Vavrus, S., Wang, M., Holland, M., Landrum, L., 2019. Past and fu-
and fog zone in the Tarapacá Region, Atacama Desert, Chile. Atmos. Res. 87 (3–4), ture interannual variability in Arctic Sea ice in coupled climate models. Cryosphere
301–311. 13, 113–124.
Chakraborty, A., Joshi, P.K., Ghosh, A., Areendran, G., 2013. Assessing biome boundary Miró, J.J., Caselles, V., Estrela, M.J., 2017. Multiple imputation of rainfall missing data in
shifts under climate change scenarios in India. Ecol. Indic. 34, 536–547. the Iberian Mediterranean context. Atmos. Res. 197, 313–330.
Chen, X., Zhang, X.S., Lia, B.L., 2003. The possible response of life zones in China under Monserud, R.A., 1990. Methods for Comparing Global Vegetation Maps. IIASA Working
global climate change. Glob. Planet. Chang. 38, 327–337. Paper, Laxenburg.
Chu, P.S., Zhao, X., 2011. Bayesian analysis for extreme climatic events: a review. Atmos. Monserud, R.A., Leemans, R., 1992. Comparing global vegetation maps with the Kappa
Res. 102 (3), 243–262. statistic. Ecol. Model. 62, 275–293.
Cohen, J., 1960. A coefficient of agreement for nominal scales. Educ. Psychol. Meas. 20, National Academies of Sciences, Engineering, and Medicine, 2018. Thriving on Our
37–46. Changing Planet: A Decadal Strategy for Earth Observation from Space. The National
Cook, B.I., Seager, R., 2013. The response of the North American Monsoon to increased Academies Press, Washington, DC.
greenhouse gas forcing. J. Geophys. Res. Atmos. 118, 1690–1699. Navarro, A., Moreno, R., Tapiador, F.J., 2018. Improving the representation of anthro-
Dimitrova, T., Mitzeva, R., Savtchenko, A., 2009. Environmental conditions responsible pogenic CO2 emissions in climate models: impact of a new parameterization for the
for the type of precipitation in summer convective storms over Bulgaria. Atmos. Res. Community Earth System Model (CESM). Earth Syst. Dyn. 9, 1045–1062.
93, 30–38. Netzel, P., Stepinski, T., 2016. On using a clustering approach for global climate classi-
Ekman, A.M.L., 2014. Do sophisticated parameterizations of aerosol-cloud interactions in fication. J. Clim. 29, 3387–3401.
CMIP5 models improve the representation of recent observed temperature trends? J. Peña-Angulo, D., Trigo, R.M., Cortesi, N., González-Hidalgo, J.C., 2016. The influence of
Geophys. Res. 119, 817–832. weather types on the monthly average maximum and minimum temperatures in the
Garcia, R.A., Cabeza, M., Rahbek, C., Araújo, M.B., 2014. Multiple dimensions of climate Iberian Peninsula. Atmos. Res. 178-179, 217–230.
change and their implications for biodiversity. Science 344, 486–496. Peng, C., 2000. From static biogeographical model to dynamic global vegetation model: a
Giaiotti, D.B., Giovannoni, M., Pucillo, A., Stel, F., 2007. The climatology of tornadoes global perspective on modelling vegetation dynamics. Ecol. Model. 135, 33–54.
and waterspouts in Italy. Atmos. Res. 83 (2–4), 534–541. Prentice, K.C., 1990. Bioclimatic distribution of vegetation for general circulation model
Giorgi, F., Jones, C., Asrar, G.R., 2006. Addressing climate information needs at the re- studies. J. Geophys. Res. 95, 11811–11830.
gional level: the CORDEX framework. Bull. World Meteorol. Organ. 58 (3), 175–183. Ramos, M.C., 2001. Divisive and hierarchical clustering techniques to analyse variability
Harris, I., Jones, P.D.D., Osborn, T.J.J., Lister, D.H.H., 2014. Updated high-resolution of rainfall distribution patterns in a Mediterranean region. Atmos. Res. 57 (2),
grids of monthly climatic observations—the CRU TS3.10 Dataset. Int. J. Climatol. 34, 123–138.
623–642. Ren, L., Arkin, P., Smith, T.M., Shen, S.S.P., 2013. Global precipitation trends in
Hoffman, F.M., Hargrove, W.W., Erickson, D.J., Oglesby, R.J., 2005. Using clustered 1900–2005 from a reconstruction and coupled model simulations. J. Geophys. Res.
climate regimes to analyze and compare predictions from fully coupled general cir- Atmos. 118, 1679–1689.
culation models. Earth Interact. 9, 1–27. Roderfeld, H., Blyth, E., Dankers, R., Huse, G., Slagstad, D., Ellingsen, I., Wolf, A., Lange,
Holdridge, L., 1947. Determination of world formulation from simple climatic data. M.A., 2008. Potential impact of climate change on ecosystems of the Barents Sea
Science 105, 367–368. Region. Clim. Chang. 87, 283–303.
Holdridge, L., 1967. Life Zone Ecology. Tropical Science Center, San José. Rotstayn, L.D., 2000. On the “tuning” of autoconversion parameterizations in climate
Holdridge, L., Grenke, W.C., Hatheway, W.H., Liang, T., Tosi, J.A., 1971. Forest models. J. Geophys. Res. Atmos. 105, 15495–15507.
Environments in Tropical Life Zones: A Pilot Study. Pergamon Press, New York. Rozwadowska, A., 2004. Optical thickness of stratiform clouds over the Baltic inferred
Hourdin, F., Mauritsen, T., Gettelman, A., Golaz, J.-C., Balaji, V., Duan, Q., Folini, D., Ji, from on-board irradiance measurements. Atmos. Res. 72, 129–147.
D., Klocke, D., Qian, Y., Rauser, F., Rio, C., Tomassini, L., Watanabe, M., Williamson, Sakamoto, T.T., Komuro, Y., Nishimura, T., Ishii, M., Tatebe, H., Shiogama, H.,
D., 2017. The art and science of climate model tuning. Bull. Am. Meteorol. Soc. 98, Hasegawa, A., Toyoda, T., Mori, M., Suzuki, T., Imada, Y., Nozawa, T., Takata, K.,
589–602. Mochizuki, T., Ogochi, K., Emori, S., Hasumi, H., Kimoto, M., 2012. MIROC4h—a
Jiang, X., Rauscher, S., Ringler, T.D., Lawrence, D.M., Park Williams, A., Allen, C.D., new high-resolution atmosphere-ocean coupled general circulation model. J.
Steiner, A.L., Cai Michael, D., Mcdowell, N.G., 2013. Projected future changes in Meteorol. Soc. Jpn. 90 (3), 325–359.
vegetation in western North America in the twenty-first century. J. Clim. 26, Sawyer, J.O., Lindsey, A.A., 1964. The Holdridge bioclimatic formations of the eastern
3671–3687. and central United States. Proc. Indian Acad. Sci. 73, 105–112.
Jones, C., Giorgi, F., Asrar, G., 2011. The coordinated regional downscaling experiment: Schneider, U., Becker, A., Finger, P., Meyer-Christoffer, A., Ziese, M., 2018. GPCC mon-
CORDEX An international downscaling link to CMIP. CLIVAR Exchanges 16 (2), itoring product: near real-time monthly land-surface precipitation from rain-gauges
34–40. based on SYNOP and CLIMAT data. Theor. Appl. Climatol. 115, 15–40.
Jylhä, K., Tuomenvirta, H., Ruosteenoja, K., Niemi-Hugaerts, H., Keisu, K., Karhu, J.A., Scoccimarro, E., Gualdi, S., Bellucci, A., Sanna, A., Fogli, P., Manzini, E., Vichi, M., Oddo,
2010. Observed and projected future shifts of climatic zones in Europe and their use P., Navarra, A., 2011. Effects of tropical cyclones on ocean heat transport in a high
to visualize climate change information weather. Clim. Soc. 2, 148–167. resolution coupled general circulation model. J. Clim. 24, 4368–4384.
Khatun, K., Imbach, P., Zamora, J.C., 2013. An assessment of climate change impacts on Serra, C., Martínez, M.D., Lana, X., Burgueño, A., 2014. European dry spell regimes
the tropical forests of Central America using the Holdridge Life Zone (HLZ) land (1951-2000): Clustering process and time trends. Atmos. Res. 144, 151–174.
classification system. IForest 6, 183–189. Sharifi, E., Steinacker, R., Saghafian, B., 2018. Multi time-scale evaluation of high-re-
Kidd, C., Becker, A., Huffman, G.J., Muller, C.L., Joe, P., Skofronick-Jackson, G., solution satellite-based precipitation products over northeast of Austria. Atmos. Res.
Kirschbaum, D.B., 2017. So, how much of the Earth's surface is covered by rain 206, 46–63.
gauges? Bull. Am. Meteorol. Soc. 98, 69–78. Sillmann, J., Kharin, V.V., Zhang, X., Zwiers, F.W., Bronaugh, D., 2013a. Climate ex-
Kim, K., Park, J., Baik, J., Choi, M., 2017. Evaluation of topographical and seasonal tremes indices in the CMIP5 multimodel ensemble: part 1. Model evaluation in the
feature using GPM IMERG and TRMM 3B42 over Far-East Asia. Atmos. Res. 187, present climate. J. Geophys. Res. Atmos. 118, 1716–1733.
95–105. Sillmann, J., Kharin, V.V., Zwiers, F.W., Zhang, X., Bronaugh, D., 2013b. Climate ex-
Kitoh, A., Endo, H., Krishna Kumar, K., Cavalcanti, I.F.A., Goswami, P., Zhou, T., 2013. tremes indices in the CMIP5 multimodel ensemble: part 2. Future climate projections.
Monsoons in a changing world: a regional perspective in a global context. J. Geophys. J. Geophys. Res. Atmos. 118, 2473–2493.
Res. Atmos. 118, 3053–3065. Smith, T.M., Shugart, H.H., Bonan, G.B., Smith, J.B., 1992. Modelling the potential re-
Köppen, W., 1900. Versuch einer Klassifikation der Klimate, vorzugsweise nach ihren sponse of vegetation to global climate change. Adv. Ecol. Res. 22, 93–116.
Beziehungen zur Pflanzenwelt. Geogr. Z. 6, 657–679. Suzuki, K., Stephens, G., Bodas-Salcedo, A., Wang, M., Golaz, J.-C., Yokohata, T., Koshiro,
Lack, S.A., Fox, N.I., 2012. Development of an automated approach for identifying con- T., 2015. Evaluation of the warm rain formation process in global models with sa-
vective storm type using reflectivity-derived and near-storm environment data. tellite observations. J. Atmos. Sci. 72, 3996–4014.
Atmos. Res. 116, 67–81. Szelepcsényi, Z., Breuer, H., Sümegi, P., 2014. The climate of Carpathian Region in the
Leemans, R., Cramer, W., Van Minnen, J.G., 1996. Prediction of global biome distribution 20th century based on the original and modified Holdridge life zone system. Open
using bioclimatic equilibrium models. In: Breymeyer, A. (Ed.), Global Change: Effects Geosci. 6 (3), 293–307.
on Coniferous Forest and Grasslands. Wiley, New York, pp. 413–445. Szelepcsényi, Z., Breuer, H., Kis, A., Pongrácz, R., Sümegi, P., 2018. Assessment of pro-
Li, L., Li, W., Deng, Y., 2013. Summer rainfall variability over the southeastern United jected climate change in the Carpathian Region using the Holdridge life zone system.
States and its intensification in the 21st century as assessed by cmip5 models. J. Theor. Appl. Climatol. 131, 593–610.
Geophys. Res. Atmos. 118, 340–354. Tan, M.L., Ibrahim, A.L., Duan, Z., Creacknell, A.P., Chaplot, V., 2015. Evaluation of six
Lin, L., Ge, E., Liu, X., Liao, W., Luo, M., 2018. Urbanization effects on heat waves in high-resolution satellite and ground-based precipitation products over Malaysia.
Fujian Province, Southeast China. Atmos. Res. 210, 123–132. Remote Sens. 7, 1504–1528.
Lohmann, U., Sausen, R., Bengtsson, L., Cubasch, U., Perlwitz, J., Roeckner, E., 1993. The Tang, L., Hossain, F., 2012. Investigating the similarity of satellite rainfall error metrics as
Koppen climate classification as a diagnostic tool for general circulation models. a function of Köppen climate classification. Atmos. Res. 104-105, 182–192.
Clim. Res. 3, 177–193. Tapiador, F.J., Angelis, C.F., Viltard, N., Cuartero, F., de Castro, M., 2011. On the suit-
Lott, F., Denvilbutchart, S.N., Cagnazzo, C., Giorgetta, M.A., Hardiman, S.C., Manzini, E., ability of regional climate models for reconstructing climatologies. Atmos. Res. 101,
Krismer, T., Duvel, J.P., Maury, P., Scinocca, J.F., Watanabe, S., Yukimoto, S., 2014. 739–751.
Kelvin and rossby-gravity wave packets in the lower stratosphere of some high-top Tapiador, F.J., Behrangi, A., Haddad, Z.S., Katsanos, D., de Castro, M., 2016. Disruptions
CMIP5 models. J. Geophys. Res. 119, 2156–2173. in precipitation cycles: attribution to anthropogenic forcing. J. Geophys. Res. Atmos.
Lugo, A.E., 1999. The Holdridge life zones of the conterminous United States in relation 121, 2161–2177.

120
F.J. Tapiador, et al. Atmospheric Research 228 (2019) 107–121

Tapiador, F.J., Navarro, A., Levizzani, V., García-Ortega, E., Huffman, G.J., Kidd, C., Vindel, J.M., Polo, J., Zarzalejo, L.F., Ramírez, L., 2015. Stochastic model to describe
Kucera, P.A., Kummerow, C.D., Masunaga, H., Petersen, W.A., Roca, R., Sánchez, J.- atmospheric attenuation from yearly global solar irradiation. Atmos. Res. 153,
L., Tao, W.-K., Turk, F.J., 2017. Global precipitation measurements for validating 209–216.
climate models. Atmos. Res. 197, 1–20. Wang, M., Overland, J.E., 2004. Detecting arctic climate change using Köppen climate
Tapiador, F.J., Moreno, R., Navarro, A., 2019a. Consensus in climate classifications for classification. Clim. Chang. 67, 43–62.
present climate and global warming scenarios. Atmos. Res. 216, 26–36. Wen, G., Xiao, H., Yang, H., Bi, Y., Xu, W., 2017. Characteristics of summer and winter
Tapiador, F.J., Navarro, A., Moreno, R., Sánchez, J.L., García-Ortega, E., 2019b. Regional precipitation over northern China. Atmos. Res. 197, 390–406.
climate models: 30 years of dynamic downscaling. Atmos. Res Under review. Xue, X.W., Hong, Y., Limaye, A.S., Gourley, J.J., Huffman, G.J., Khan, S.I., Dorji, C., Chen,
Tapiador, F.J., Roca, R., Genio, A.D., Dewitte, B., Petersen, W., Zhang, F., 2019c. Is S., 2013. Statistical and hydrological evaluation of TRMM-based Multisatellite
precipitation a good metric for model performance? Bull. Am. Meteorol. Soc. 100, Precipitation Analysis over the Wangchu Basin of Bhutan: are the latest satellite
223–233. precipitation products 3B42V7 ready for use in ungauged basins? J. Hydrol. 499,
Tatli, H., Dalfes, H.N., 2016. Defining Holdridge's life zones over Turkey. Int. J. Climatol. 91–99.
36, 3864–3872. Yates, D.N., Kittel, T.G.F., Cannon, R.F., 2000. Comparing the correlative Holdridge
Taylor, K.E., Stouffer, R.J., Meehl, G.A., 2012. An overview of CMIP5 and the experiment model to mechanistic biogeographical models for assessing vegetation distribution
design. Bull. Am. Meteorol. Soc. 93, 485–498. response to climatic change. Clim. Chang. 44, 59–87.
Tramblay, Y., Hertig, E., 2018. Modelling extreme dry spells in the Mediterranean region Yue, T., Liu, J., Jørgensen, S.E., Gao, Z., Zhang, S., Deng, X., 2001. Changes of Holdridge
in connection with atmospheric circulation. Atmos. Res. 202, 40–48. life zone diversity in all of China over half a century. Ecol. Model. 144, 153–162.
van Vuuren, D.P., Edmonds, J., Kainuma, M., Riahi, K., Thomson, A., Hibbard, K., Hurtt, Zou, L.W., Zhou, T.J., Peng, D.D., 2016. Dynamical downscaling of historical climate over
G.C., Kram, T., Krey, V., Lamarque, J.-F., Masui, T., Meinshausen, M., Nakicenovic, CORDEX East Asia domain: a comparison of regional ocean-atmosphere coupled
N., Smith, S.J., Rose, S.K., 2011. The representative concentration pathways: an model to stand-alone RCM simulations. J. Geophys. Res. 121, 1442–1458.
overview. Clim. Chang. 109, 5–31.

121