ArticlePDF Available

Global warming induced changes of climate zones based on CMIP5 projections


Abstract and Figures

Climate classifications can provide an effective tool for integrated assessment of climate model results. We present an analysis of future global climate projections performed in the framework of the Coupled Model Intercomparison Project Phase 5 (CMIP5) project by means of Köppen-Trewartha classification. Maps of future climate type distributions were created along with the analysis of the ensemble spread. The simulations under scenarios with representative concentration pathway (RCP) 4.5 and RCP8.5 showed a substantial decline in ice cap, tundra, and boreal climate in the warming world, accompanied by an expansion of temperate climates, dry climates, and savanna, nearly unanimous within the CMIP5 ensemble. Results for the subtropical climate types were generally not conclusive. Changes in climate zones were also analyzed in comparison with the individual model performance for the historical period 1961-1990. The magnitude of change was higher than model errors only for tundra, boreal, and temperate continental climate types. For other types, the response was mostly smaller than model error, or there was considerable disagreement among the ensemble members. Altogether, around 14% of the continental area is expected to change climate types by the end of the 21st century under the projected RCP4.5 forcing and 20% under the RCP8.5 scenario.
Content may be subject to copyright.
Clim Res
Vol. 71: 17–31, 2016
doi: 10.3354/cr01418 Published online November 14
The outputs of state-of-the-art global climate mod-
els are currently available within the Coupled Model
Intercomparison Project Phase 5 (CMIP5, Taylor et al.
2012b), which served as the basis for the IPCC’s Fifth
Assessment Report, published in September 2013
(available online at Besides the global
climate models (GCMs) themselves, which were
improved, e.g. toward higher resolution and in some
cases by including new processes and interactions
such as so-called Earth System Models (ESMs), the
methodology re garding the construction of projec-
tion scenarios also changed in comparison to previ-
ous GCM experiments (CMIP3 GCMs, Meehl et al.
2007). For the core CMIP5 GCM experiments, 4 rep-
resentative concentration pathways (RCPs) with
radiative forcing ranging from 2.6 to 8.5 W m−2 in the
year 2100 were chosen, designated as RCP2.6,
RCP4.5, RCP6.0, and RCP8.5 (Moss et al. 2010).
Each generation of climate models must in evitably
be subject to tests of how realistic the models are in
simulating the observed climate characteristics in the
recent past. The climate classifications can serve,
inter alia, as effective tools for analysis of model per-
formance. The Köppen classification (Köppen 1923,
1931, 1936, Geiger 1954) or Köppen-Trewartha classi-
fication (KTC, Trewartha 1968, Trewartha & Horn
1980) have most often been used for this purpose. The
climate types are based on long-term climato logical
means of near-surface air temperature and precipita-
tion that are easily obtained from the outputs of
GCMs. The KTC provides a slightly more detailed de-
scription of climate type distributions than the original
© The authors 2016. Open Access under Creative Commons by
Attribution Licence. Use, distribution and reproduction are un -
restricted. Authors and original publication must be credited.
Publisher: Inter-Research ·
*Corresponding author:
Global warming-induced changes in climate zones
based on CMIP5 projections
Michal Belda*, Eva Holtanová, Jaroslava Kalvová, Tomáš Halenka
Charles University in Prague, Dept. of Meteorology and Environment Protection, 18200, Prague, Czech Republic
ABSTRACT: Climate classifications can provide an effective tool for integrated assessment of
climate model results. We present an analysis of future global climate projections performed in the
framework of the Coupled Model Intercomparison Project Phase 5 (CMIP5) project by means of
Köppen-Trewartha classification. Maps of future climate type distributions were created along
with the analysis of the ensemble spread. The simulations under scenarios with representative
concentration pathway (RCP) 4.5 and RCP8.5 showed a substantial decline in ice cap, tundra, and
boreal climate in the warming world, accompanied by an expansion of temperate climates, dry
climates, and savanna, nearly unanimous within the CMIP5 ensemble. Results for the subtropical
climate types were generally not conclusive. Changes in climate zones were also analyzed in
comparison with the individual model performance for the historical period 1961−1990. The mag-
nitude of change was higher than model errors only for tundra, boreal, and temperate continental
climate types. For other types, the response was mostly smaller than model error, or there was con-
siderable disagreement among the ensemble members. Altogether, around 14% of the
continental area is expected to change climate types by the end of the 21st century under the
projected RCP4.5 forcing and 20% under the RCP8.5 scenario.
KEY WORDS: Köppen-Trewartha climate classification · Coupled Model Intercomparison Project
Phase 5 · CMIP5 · Global climate model · Climate type change · Representative concentration
Clim Res 71: 17–31, 2016
Köppen scheme (de Castro et al. 2007). Belda et al.
(2014) reviewed the KTC and its differences from the
original Köppen scheme, and analyzed ob served pat-
terns in climate types and their changes during the
20th century.
Climate types derived from GCM projections of
future climate are useful for a variety of sectors and
scientific fields. They provide an idea of what
changes can be expected in the areas of individual
climate types. Due to their strong relationship with
the distribution of natural vegetation zones (e.g. Tre-
wartha & Horn 1980, Bailey 2009), it is possible to
assess the development of different ecoregions, even
though further information on e.g. edaphic and topo -
grafic properties (Baker et al. 2010, Hargrove & Hoff-
man 2004) is needed for such assessments.
Projected changes in climate types have previously
been analyzed in various studies using different cli-
mate models and emission scenarios. Lohmann et al.
(1993) assessed the outputs of the atmospheric gen-
eral circulation model ECHAM3 using the Köppen
classification, and derived shifts in climate zones in
greenhouse gas warming simulations over 100 model
years. They projected a retreat of the permafrost
climate and an extension of both the tropical rainy
climate and dry climate.
Kalvová et al. (2003) applied the Köppen classifica-
tion to simulations of 4 GCMs, namely HadCM2,
ECHAM4, CSIRO-Mk2b, and CGCM1, for the pres-
ent and future periods. They confirmed the results
described by Lohmann et al. (1993) regarding tropi-
cal and dry climates and described a decline in the
area of boreal and cold climates.
More recently, Rubel & Kottek (2010) created a se-
ries of digital world maps of Köppen climate types for
the period 1901−2100 based on observed data (CRU
TS2.1, GPCC Version 4) and 20 simulations of
5 GCMs (each GCM with 4 Special Report on Emis-
sion Scenarios [SRES] emission scenarios). In the case
of the emission scenario with the highest rate of emis-
sion increase (A1FI), the results showed an in crease in
the areas covered by tropical, dry, and temperate cli-
mates, and a decrease in the coverage of cold and bo-
real climates. Projected changes for the milder emis-
sions scenario (B1) were significantly smaller.
Baker et al. (2010) compared KTC types over China
for historical (1961−1990) and projected future cli-
mates (2041−2070) simulated by HadCM3 for the
SRES A1FI scenario. They showed that the spatial
patterns of climate change resulted in a northern
migration of warmer climatic types as well as a slight
expansion in the high-latitude desert and arid shrub-
land regions in northwestern China.
Mahlstein et al. (2013) used simulations of 13
CMIP5 GCMs for determination of Köppen-Geiger
climate types and analyzed their changes during
1900−2098. They found that under the RCP8.5 forc-
ing, for which the mean warming reaches about
4.5°C by the end of the 21st century (Rogelj et al.
2012), approximately 20% of the global land area
would undergo a shift in the original climate zones.
Frost climates are projected to largely decline, some
arid climatic zones are expected to expand, and large
parts of the global land area with cool summers will
experience a change to climates with hot summers.
However, Mahlstein et al. (2013) also emphasized
large model uncertainties and reported that the pace
of the climate type shifts increases with increasing
global mean temperature.
Feng et al. (2012) analyzed observed and projected
climate changes and their impact on vegetation for
the area north of 50° N over the period of 1900−2099
using the KTC scheme. To estimate the future
changes, they used the simulations of 16 CMIP3
GCMs for 3 SRES emission scenarios (B1, A1B, and
A2). Their results showed a decrease in areas classi-
fied as tundra, ice cap, and subarctic continental cli-
mates, and an expansion of the temperate and boreal
oceanic climates. Moreover Feng et al. (2012) pro-
jected that arid, warm temperate, and snow and
polar climates will successively shift to the north in
the northern hemisphere.
Feng et al. (2014) focused on shifts in KTC climate
types in 1900−2100. In contrast to Feng et al. (2012),
the analysis was done for the whole global land area
and model simulations of 20 CMIP5 GCMs for
RCP4.5 and RCP8.5 pathways. Feng et al. (2014)
found that during the 21st century, the KTC types
would shift toward warmer and drier types, with the
largest changes in the northern hemisphere north
of 30° N. They also concluded that temperature
changes are the dominant factor causing the projec -
ted shifts in climate types during the 21st century.
Here we used the KTC to assess changes in climate
type areas simulated by a suite of 30 CMIP5 GCMs
for the period of 2006−2100 and 2 RCPs (RCP4.5
and RCP8.5). Our study follows previous papers, i.e.
Belda et al. (2014) mentioned above and Belda et al.
(2015), wherein we assessed the performance of 43
CMIP5 GCMs in simulating the KTC climate types in
the reference period 1961−1990. One of the main
conclusions of Belda et al.’s (2015) analysis was that
models generally had problems capturing the rain-
forest climate type Ar (see Table 2 for climate types),
mainly in Amazonia. The desert climate type BW was
underestimated by half of the models. Boreal climate
Belda et al.: Warming-induced changes in climate zones
type Ewas overestimated by many models, mostly
spreading over to the areas of observed tundra type
Ft. Further, Belda et al. (2015) indicated that CMIP5
GCMs did not show any clear tendency to improve
the representation of climate types with increasing
spatial resolution.
In addition to previous analyses of CMIP5 models
in terms of Köppen classification by Mahlstein et al.
(2013) and Feng et al. (2014), here we use the largest
possible set of GCMs, describe the temporal evolu-
tion of KTC types for individual GCMs, and present
simulated changes in the context of model perform-
ance for the present climate. We also add an analysis
of future climate uncertainty in terms of ensemble
spread throughout the scenario simulations.
Various supplementary graphical products, includ-
ing figures describing the model performance of
CMIP5 GCMs used by Belda et al. (2015) are avail-
able at
2.1. Data
A suite of CMIP5 GCM simulations is employed
here, selected based on the availability of data for
both RCP4.5 and RCP8.5 scenarios. Basic information
on all model simulations incorporated here is pre-
sented in Table 1. The data are available at http://; we used monthly mean
surface air temperature and precipitation to classify
the KTC types. The outputs from the experiment
denoted as ‘historical’ were used for the reference
period 1961−1990. For the future time period
2006−2100, we considered 2 alternative simulations,
RCP4.5 and RCP8.5. RCP4.5 assumes radiative forc-
ing of 4.5 W m−2 at stabilization after 2100, whereas
RCP8.5 represents a ‘rising pathway’ with radiative
forcing higher than 8.5 W m−2 after 2100. For more
details on RCPs, see Moss et al. (2010). Where more
ensemble members were available, we chose the
ensemble member r1i1p1 (considered a baseline sim-
ulation of the subensemble for the puposes of this
analysis) (Taylor et al. 2012a).
As one of the indicators of uncertainty in the cli-
mate signal, errors in the historical experiment dur-
ing the reference period were considered in terms of
KTC types based on monthly mean surface air tem-
perature and precipitation provided by the Climatic
Research Unit (CRU) TS 3.22 dataset (Harris et al.
2014, hereafter TS3) available in spatial resolution of
0.5° × 0.5° over global land areas excluding Ant -
arctica. As a part of the uncertainty analysis, a com-
parison of the classification based on 2 versions of
CRU (TS 3.22 and TS 3.1.10) and the University of
Delaware dataset version 4.01 (UDEL; Willmott &
Matsuura 2001) was performed with the conclusion
that the differences between these datasets are con-
siderably smaller than the spread of the model simu-
lations, and thus the impact of the choice of the
observational dataset on GCM performance evalua-
tion is negligible.
2.2. Methods
The KTC system (Trewartha & Horn 1980, Belda et
al. 2014) has 6 main climate groups. Five of them (A,
C, D, E, and F) are basic thermal zones. The sixth
group, B, is the dry climatic zone that cuts across the
other climate types, except for the polar climate F.
Similarly to original Köppen classification scheme,
the main climate types are determined according to
long-term annual and monthly means of surface air
temperature and precipitation amounts. The dryness
threshold distinguishing group Bis based on the def-
inition by Patton (1962). A brief summary of climate
types and subtypes is provided in Table 2.
The KTC climate types were calculated in the orig-
inal model grids for the reference period 1961−1990
and for running 30 yr periods during the 21st century,
beginning with 2006−2035 until 2071−2100 or 2070−
2099 (as data for some of the model runs are only
available until the year 2099). Land areas falling into
each climate type/subtype were expressed in terms
of relative areas, i.e. as a percentage of the whole
global land area (excluding Antarctica). Simulated
changes of KTC types for both RCP4.5 and RCP8.5
were assessed in several different ways. An overall
picture of the multi-model ensemble evolution in
time is provided as medians and 10th and 90th
percen tiles of changes of relative areas with respect
to the values simulated for the reference period
Further, we pay special attention to 3 selected time
periods denoted as near future (2006−2035), mid-
century (2020−2050), and far future (2071−2100 or
2070−2099 based on the simulation period). We de -
monstrate changes in selected climate type areas for
each of these periods simulated by individual GCMs
together with model errors in the reference period
indicating the reliability of the climate change signal.
All changes are expressed in percentage of area sim-
ulated by the respective GCMs in the reference pe -
riod. The model errors are defined as differences in
Clim Res 71: 17–31, 2016
No. CMIP5 model Resolution Modeling center/model versions
1 ACCESS1.3 1.88° × 1.24° Commonwealth Scientific and Industrial Research Organisation (CSIRO)
and Bureau of Meteorology, Australia
2 CanESM2 2.8° × 2.8° Canadian Centre for Climate Modelling and Analysis
3 CCSM4 1.25° × 0.94° National Center for Atmospheric Research
4 CESM1-BGC 1.25° × 0.94° Community Earth System Model Contributors
BGC: BioGeoChemistry
CAM5: Community Atmospheric Model v5
FV2: Finite volume 2degree
5 CESM1-CAM5 1.25° × 0.94°
6 CESM1-CAM5.1-FV2 2.50° × 1.88°
7 CNRM-CM5 1.4° × 1.4° Centre National de Recherches Météorologiques; Centre Européen de
Recherche et Formation Avancées en Calcul Scientifique
8 CSIRO-Mk3.6.0 1.9° × 1.9° CSIRO; Queensland Climate Change Centre of Excellence
9 FGOALS-g2 2.81° × 3.00° LASG, Institute of Atmospheric Physics, Chinese Academy of Sciences and
CESS,Tsinghua University
10 GFDL-CM3 2.5° × 2° Geophysical Fluid Dynamics Laboratory
11 GFDL-ESM2G 2.5° × 2°
12 GFDL-ESM2M 2.5° × 2°
13 GISS-E2-H 2.5° × 2° NASA Goddard Institute for Space Studies
H: Hycom Ocean Model
R: Russell Ocean Model
CC: interactive terrestrial carbon cycle, ocean biogeochemistry
14 GISS-E2-H-CC 2.5° × 2°
15 GISS-E2-R 2.5° × 2°
16 GISS-E2-R-CC 2.5° × 2°
17 HadGEM2-AO Met Office Hadley Centre
AO: aerosols, ocean & sea-ice
CC: AO+terrestrial carbon cycle, ocean biogeochemistry
ES: CC+chemistry
18 HadGEM2-CC 1.875° × 1.25°
19 HadGEM2-ES 1.875° × 1.25°
20 INM-CM4 2° × 1.5° Institute for Numerical Mathematics
21 IPSL-CM5A-MR 2.5° × 1.3° Institut Pierre-Simon Laplace
MR: Medium resolution
LR: Low resolution
22 IPSL-CM5B-LR 3.75° × 1.9°
23 MIROC5 1.4° × 1.4° Atmosphere and Ocean Research Institute (The University of Tokyo),
National Institute for Environmental Studies, and Japan Agency for Marine-
Earth Science and Technology
24 MIROC-ESM 2.8° × 2.8° Japan Agency for Marine-Earth Science and Technology,
Atmosphere and Ocean Research Institute (The University of Tokyo), and
National Institute for Environmental Studies
CHEM: added atmospheric chemistry
25 MIROC-ESM-CHEM 2.8° × 2.8°
26 MPI-ESM-LR 1.9° × 1.9° Max Planck Institute for Meteorology
MR: Medium resolution
LR: Low resolution
27 MPI-ESM-MR 1.9° × 1.9°
28 MRI-CGCM3 1.125° × 1.125° Meteorological Research Institute
29 NorESM1-M 2.5° × 1.9° Norwegian Climate Centre
M: intermediate resolution
ME: M+carbon cycle
30 NorESM1-ME 2.5° × 1.9°
Table 1. CMIP5 global climate models analyzed in this study with model versions explained (where applicable)
Belda et al.: Warming-induced changes in climate zones
simulated and observed CRU TS3.22 areas expressed
as a percentage of observed values.
For illustration of the geographical distribution of
changes simulated for the far-future period, maps of
projected distributions of KTC types are shown for
both RCPs. Climate zones in the future period were
calculated based on temperature and precipitation
scenarios constructed using the delta method (Deque
2007). En semble mean values in the periods 2070−
2099 and 1961−1990 were used to calculate deltas
that were then added (multiplied) to the present cli-
mate state represented by the temperature (precipi-
tation) from the CRU TS3.22 database. KTC was then
applied, which provided spatial distributions of cli-
mate zones in the scenarios.
The geographical distribution of observed KTC
types and its simulated changes are illustrated in
Figs. 1−3. The climate change signal patterns are
similar for both scenarios, with stronger manifesta-
tion under stronger forcing of RCP8.5. In the north-
ern hemisphere, the most remarkable feature is the
northward shift of the border between Dc and E
types, with an increase in the area of Dc and shrink-
ing of the Etype. The shift of the southern border of
Dc is not as evident; only in Europe, an eastward shift
of the Dc−Do border is projected, inducing expansion
of the Do area over western and central Europe. Fur-
ther, a global feature is shrinking of the Ft area, not
only in the high latitudes, but also in high-elevation
regions of the Himalayas and the Andes. In South
America, the Ft type is projected to dissapear by the
end of the century under both RCPs. Another distinct
pattern of change in South America is the expansion
of the dry types BW and BS. In Africa and Australia,
the GCMs project an increase in the BW area and
shrinking of the Ctypes. In southeastern Asia, our
results suggest an expansion of the Aw type, which
might be connected to increased strength of the
Indian summer monsoon as documented e.g. by
Menon et al. (2013).
The values of multi-model medians of simulated
changes, 10th and 90th percentiles, and the range
be tween them for the period 2071−2100 under both
RCP4.5 and RCP8.5 forcings are summarized in
Table 3.
The KTC climate types can be divided into 3
groups (decreasing area, increasing area and no con-
clusive change) based on the temporal behavior of
simulated continental areas belonging to respective
KTC types during the 21st century under the RCP4.5
and RCP8.5 forcing. The first group comprises boreal
climate E, tundra Ft, and ice cap climate Fi that,
according to the GCMs analyzed in our study, are
expected to retreat. These 3 types occur at high lati-
tudes or altitudes.
All GCMs simulate a decrease in the continental
coverage (Antarctica not included in the analysis) of
ice cap climate Fi (Fig. 4), which is clearly seen for
the multi-model median (M-MED). Under the
RCP4.5 forcing, the relative area of Fi decreases to
73% (Table 3) of the value simulated for the refer-
ence period 1961−1990. In the case of RCP8.5, the
decrease is even stronger, as the Fi area decreases to
52% of its reference value. The decrease to less than
90% is already expected in the period 2006−2035 for
both scenarios. The multi-model spread of simulated
Type Criteria
Subtype Precipitation/temperature regime
A Tcold > 18°C; Pmean > R
Ar 10 to 12 mo wet; 0 to 2 mo dry
Aw Winter (low-sun period) dry; >2 mo dry
As Summer (high-sun period) dry;
rare in type Aclimates
B Pmean < R
BS R/2 < Pmean < R
BW Pmean < R/2
C Tcold < 18°C; 8 to 12 mo with Tmo > 10°C
Cs Summer dry; at least 3 times as much precipi-
tation in winter half-year as in summer half-
year; Pdry < 3 cm; total annual precipitation
< 89 cm
Cw Winter dry; at least 10 times as much precipita-
tion in summer half-year as in winter half-year
Cf No dry season; difference between driest and
wettest month less than required for Cs and
Cw; Pdry > 3 cm
D 4 to 7 mo with Tmo > 10°C
Do Tcold > 0°C
Dc Tcold < 0°C
E 1 to 3 mo with Tmo > 10°C
F All months with Tmo < 10°C
Ft Twarm > 0°C
Fi Twarm < 0°C
Table 2. Definition of Köppen-Trewartha classification (KTC)
climate types according to Trewartha & Horn (1980), with
dryness threshold defined by Patton (1962). Tmo: long-term
monthly mean air temperature; Tcold (Twarm): monthly
mean air temperature of the coldest (warmest) month;
Pmean: mean annual precipitation (cm); Pdry: mean precipi-
tation of the driest summer month; R: Patton’s precipitation
threshold, defined as R= 2.3T− 0.64Pw + 41, where Tis
mean annual temperature (°C) and Pw is the percentage of
annual precipitation occurring in winter
Clim Res 71: 17–31, 2016
changes is quite large and is the same for both RCPs.
In the case of RCP4.5, the decrease is often fastest
during the first half of this century; in the second half
it is rather slow, whereas the stronger forcing of
RCP8.5 leads to more pronounced decline during the
whole century. The decrease in relative area occu-
pied by type Fi is solely due to transformation to tun-
dra, Ft. Regarding the comparison of simulated
changes to model errors in the reference period, all
GCMs (except for CSIRO-Mk3-6-0 and MIROC-
Fig. 1. Köppen-Trewartha climate types derived from observations (CRU TS3.22) for the period 1961−1990
Fig. 2. Köppen-Trewartha climate types for the period 2070−2099, derived from the CRU TS3.22 observational dataset and the
CMIP5 ensemble RCP4.5 scenario using the delta method
Belda et al.: Warming-induced changes in climate zones
ESM) overestimate the observed area of Fi. Only in
37 out of 180 cases (3 periods, 2 RCPs, 30 GCMs) are
the projected changes larger (in absolute value) than
model errors (Fig. 5).
Similarly, all GCMs simulate a decrease in tundra
climate type Ft, and under the RCP4.5 forcing, a
faster rate of change occurs in the first half of the cen-
tury (Fig. 4). According to M-MED, the relative area
of Ft decreases by the end of the 21st century to 63%
(42%) for RCP4.5 (RCP8.5), and the multi-model
range is larger for RCP8.5 (Table 3). Models MIROC-
ESM, MIROC-ESM-CHEM, and INMCM4 simulate
the slowest decline in Ft (Fig. 5), even though the first
2 of these GCMs are the most sensitive to Fi changes.
The largest change in Ft for the far future period is
simulated by GFDL-CM3, which shows a decrease to
30% of the reference under RCP4.5 and 19% under
RCP8.5. Projected changes are larger than model
errors in 60% of all cases for RCP4.5, and
in 90% for RCP8.5 at the end of the cen-
tury. The Ft climate type is expected to
transform into boreal climate E, al though
under RCP8.5, transitions of smaller
areas to Dc and Do climate types are also
According to the outputs of all analyzed
CMIP5 GCMs (except for GFDL-ESM2G
and NorESM1-ME), the continental area
occupied by boreal climate Eis also ex-
pected to decrease (Fig. 5, Table 3). Time
evolution of Etype area in the running
30 yr periods according to individual
GCMs shows a gradual mono tonic de-
crease or only small fluctuations (Fig. 4).
The exception is model CESM1-CAM5-
1-FV2, which for both RCPs shows a neg-
ative peak around the year 2061 preceded
by a steep decrease after 2045 and fol-
type M-MED p10 p90 Range M-MED p10 p90 Range
Fi 73 57 85 28 52 40 68 28
Ft 63 45 77 32 42 28 66 38
E83 67 97 30 64 36 86 50
Dc 115 106 127 21 130 116 145 29
Do 112 96 122 26 115 100 134 34
BW 108 103 117 14 113 108 123 15
BS 108 100 120 20 113 101 131 30
Aw 117 100 125 25 120 101 137 36
Ar 103 96 109 13 103 89 116 27
Cf 95 85 100 15 98 80 106 26
Cw 4177972 2075952
Cs 122 77 156 79 121 52 214 162
Table 3. Multi-model statistics of the percentage changes of Köppen-Tre-
wartha classification (KTC; see Table 2 for definitions) climate type areas in
the future with respect to the reference period (1961–1990) for the RCP4.5
and RCP8.5 scenarios. M-MED: multi-model median, p10: 10th percentile,
p90: 90th percentile, range: range between p10 and p90
Fig. 3. Köppen-Trewartha climate types for the period 2070−2099, derived from the CRU TS3.22 observational dataset and the
CMIP5 ensemble RCP8.5 scenario using the delta method
Clim Res 71: 17–31, 2016
Fig. 4. Temporal evolution of continental area belonging to selected climate types (Fi, Ft, E, Dc, BW, and Ar; see Table 2 for
definitions) for moving 30 yr periods throughout the 21st century relative to the reference period 1961−1990 (100% means no
change); x-axis: 30 yr periods (period 1 is 2006−2035, period 66 is 2071−2100); squares: multi-model medians calculated from
the ensemble of 30 selected CMIP5 GCMs (green for RCP4.5, black for RCP4.5); green area (diagonal hatching): values
between the 10th and 90th percentiles of the multi-model ensemble for RCP4.5 (RCP8.5)
Belda et al.: Warming-induced changes in climate zones
lowed by a steep rise until 2073 and a moderate de-
crease afterwards (not shown). This pattern both af-
fects the spread of the results and is reflected in some
other types (Do, Dc, Ar). Even though all GCMs simu-
lated the observed area of type Ewith the smallest er-
rors, their reactions to radiative forcing are quite di-
verse. The spread of the multi-model ensemble is
larger for the stronger forcing of RCP8.5 than for
RCP4.5 (Table 3). The de crease in continental area for
type Eby the end of the century seems to be the most
convincing (in comparison to model errors, Fig. 5) of
all climate types that are expected to decrease. Boreal
climate transforms mainly to temperate continental
climate Dc. For RCP8.5 the losses, generally from the
southern extent of type E in the northern hemisphere,
are >4 times larger than the gains of the area from
tundra climate Ft.
The second group of KTC types consists of Dc, Do,
BW, BS, and Aw that are all expected to increase
their relative continental areas, according to most of
the GCMs and both RCPs. All GCMs considered in
our study (except for CESM1-CAM5-1-FV2) give a
gradual expansion of continental temperate climate
Dc during 2006−2100. Based on M-MED, the relative
area occupied by Dc for RCP4.5 (RCP8.5) increases
to approx. 115% (130%) of the area in the reference
period by the end of the century (Fig. 5, Table 3). The
stronger forcing of RCP8.5 leads to a higher increase
in Dc area but also to a somewhat larger multi-model
spread (Fig. 4, Table 3). Regarding the model errors,
the GCMs tend to overestimate the observed area of
Dc, but the errors are generally smaller in compari-
son to other KTC types. In the far-future period under
RCP8.5, most of the simulated changes are larger
than corresponding model errors (Fig. 5). The expan-
sion of Dc is given mainly by the transition from E; for
RCP8.5, a small portion also comes from Ft.
The expected increase in dry climate types BW and
BS is not as convincing and well-marked as the in-
crease in type Dc. According to M-MED, the relative
continental area of desert climate BW grows by the
end of the century to ~108% (113%) for RCP4.5
(RCP8.5) (Fig. 4, Table 3). Some of the GCMs, e.g.
FGOALS-g2 and MRI-CGCM3, give a similar relative
continental area for BW at the end of the 21st century
as in the reference period (Fig. 5). The patterns in
temporal behavior differ considerably among GCMs.
Some models simulate a steady rise in BW area, others
project a slight decrease during first decades followed
by an increase or an increase followed by a short de-
cline and a final rise. However, the multi-model
spread of changes simulated for the end of the 21st
century is one of the lowest of all KTC types. Both the
simulated increase and the multi-model spread are
larger for the RCP8.5 scenario. The simulated changes
are larger than model errors for 50% (30%) of the
GCMs for RCP8.5 (RCP4.5) in 2071−2100. Regarding
the transitions between climate types, the BW gains
the area mainly from BS. However, a small part of the
BW area transforms into BS.
Our findings for steppe climate type BS are similar
to BW. Most of the GCMs simulate a larger or similar
relative continental area for BS at the end of the cen-
tury with respect to the reference period (see Fig. S15
in the Supplement at com/ articles/ suppl/
c071 p017 _ supp. pdf). The multi-model median of
changes represents an increase to 108 % (113 %) for
RCP4.5 (RCP8.5) (Fig. S3, Table 3). For about half of
the GCMs, under RCP8.5 in the far future, the
expected change is greater than the model error. The
expected climate changes lead to transition of Cf and
Aw into BS and from BS into BW.
For savanna climate type Aw, most GCMs project a
moderate expansion with M-MED of 117% (120%)
for RCP4.5 (RCP8.5) (Fig. S2, Table 3). An exception
is the model CanESM2, which projects a slight de -
crease in Aw area. Model errors are smaller than
simulated changes for 2071−2100 according to 30%
(50%) of simulations under RCP4.5 (RCP8.5). Similar
to the case of boreal climate E, even though model
performance in simulating Aw in the reference
period is relatively good, the reactions to radiative
forcing differ considerably among models. Part of the
continental area occupied by Aw undergoes a transi-
tion to BS and a part of Cf area transforms into Aw.
Expected temporal evolution of relative continental
area occupied by oceanic temperate climate Do dif -
fers between individual GCMs. Some of them project
an increase in the area, others project an initial de -
crease and then a slow rise to approximately the same
Do extent as simulated for the reference period. The
time development of the 10th percentile (Fig. S8)
shows that some GCMs even project a decrease in Do
area in the far future, especially for RCP4.5. M-MED
shows an overall change to 112% of the reference
area for RCP4.5 and 115% for RCP8.5. Simulated
changes in Do are smaller than model errors (Fig. S20),
except for IPSL-CM5A-MR and Had GEM2- AO. Re-
garding the transitions between climate types, Do is
expected to transform mainly into Cf.
Until now we have dealt with KTC types that are ex-
pected to decline or increase their area according to
most CMIP5 GCMs, even though the sensitivity of the
models was different and multi-model spread was
quite large in some cases. Results for the remaining
KTC types are less conclusive. Regarding the tropical
Clim Res 71: 17–31, 2016
Fig. 5. Changes in relative continental areas of selected Köppen-Trewartha classification (KTC) climate types (Fi, Ft, E, Dc,
BW, Ar; see Table 2 for definitions) projected for the periods 2006−2035 (P1), 2021−2050 (P2), and 2071−2100 (P3) relative to
the reference period 1961−1990 (100% means no change) based on the ensemble of 30 selected CMIP5 GCMs for RCP4.5
(green) and RCP8.5 (red); error: model error in the reference period expressed as the difference between simulated and
observed (based on CRU TS3.22) relative area in the percentage of the observed value
Belda et al.: Warming-induced changes in climate zones 27
Fig. 5. (continued)
Clim Res 71: 17–31, 2016
rainforest climate Ar, the ensemble does not show any
significant signal, with ambiguous signs of change for
individual GCM simulations (Fig. 5, Table 3). For both
RCPs, the spread is rather small, similar to BW and Cf.
The model errors are larger than simulated changes
for all GCMs. The changes of Ar are given by transi-
tions from Aw and Cf and into Aw.
Most of the GCMs simulate a decline in the area
occupied by the subtropical humid climate Cf in
2006−2035, to ca. 95% of reference value according
to M-MED. Thereafter, M-MED does not vary con-
siderably, even though the multi-model spread
grows throughout the century (Fig. S5). Simulated
changes are mostly smaller than model errors, ex -
cept for GISS-E2-R, GISS-E2-R-CC, and CanESM2
(Fig. S17). The subtropical humid climate Cf trans-
forms mainly to Aw and BS. The area of type Cf
increases due to gains from Do, Dc, and BS.
We do not discuss the results for Cw and Cs, as
they both occupy a small fraction of global land area,
and the spread of the model results is quite large.
Therefore, it is difficult to draw any conclusions
about their projected changes.
Overall, the GISS models and MRI-CGCM3, AC-
CESS1-3, GFDL-ESM2M, and NorESM1 have the
least pronounced response to radiative forcing. For
RCP8.5, these models simulate changes of ca. 16%
of continental area (not including Antarctica). On
the other hand, MIROC-ESM, MIROC-ESM-CHEM,
GFDL-CM3, and CanESM2 show the largest KTC
type changes. According to these GCMs, more than
30% of the considered land area will undergo a
change of KTC type by the end of the 21st century.
However, for individual KTC types, the models simu-
lating the largest or smallest changes differ. For ex-
ample, MIROC-ESM and MIROC-ESM-CHEM show
the largest reduction in Fi but the slowest decline of
Ft. It is noteworthy that GCMs developed in the same
modeling center do not necessarily yield similar
results. For example, GFDL-CM3 shows the most
sensitive response of Dc area to radiative forcing,
where as GFDL-ESM2M gives a change of only 1 %
(7%) for RCP4.5 (RCP8.5) at the end of the century.
We assessed changes in the global distributions of
Köppen-Trewartha climate types throughout the 21st
century as simulated by a suite of 30 CMIP5 global
climate models for 2 representative concentration
pathways, RCP4.5 and RCP8.5. Ice cap climate Fi,
tundra Ft, and boreal climate Eare expected to
decline (Fig. 4). On the other hand, the relative conti-
nental area occupied by temperate climates Dc and
Do, dry climates BW and BS, and savanna climate
Aw will increase (with a few exceptions). The results
for 2 remaining climate types, Ar and Cf, are less
convincing; the changes are rather small, and the
models do not even agree on the sign of the changes.
Nevertheless, most of the GCMs simulate a slight
decrease or increase at the beginning of the 21st cen-
tury and very small changes thereafter. The types Cs
and Cw cover only a small portion of the total conti-
nental area, and simulated changes have a large
spread; therefore we will not discuss these types in
Our conclusions about a decrease in Fi and Ft area
and an increase in Dc and Do extent are consistent
with the expected rise in near-surface air tempera-
ture and are in agreement with results described by
other recent studies based on CMIP5 GCMs, e.g. by
Feng et al. (2014), and also by studies for the previous
generation of GCMs, e.g. Rubel & Kottek (2010).
Regarding the temporal evolution of relative conti-
nental areas covered by specific KTC types during
the 21st century based on M-MED of simulated
changes, a distinct difference in comparison to the
reference period is already apparent for the first 30 yr
time window of 2006−2035, and in most cases (except
for Ar and Cf ), the magnitude of simulated changes
increases throughout the century. This pattern is
more pronounced for RCP8.5. The course of simu-
lated changes is not always smooth; for example,
under RCP4.5 forcing, the decrease in area covered
by Ft is faster during the first half of the century,
while for RCP8.5 the decline is more stable. Similarly,
under RCP4.5, the rate of increase/decrease of BW,
Dc, and Ar is slower in the last third of the century.
This might be partly due to differences in the RCPs;
RCP4.5 represents a stabilization scenario with
radiative forcing reaching its maximum in the second
half of the 21st century; in contrast, under RCP8.5,
radiative forcing increases throughout the whole 21st
century (IPCC 2013). However, the influence of RCPs
cannot be simply generalized. For example, the
expansion of Do shows almost the same rate under
both RCPs.
Considering the projections given by individual
GCMs, for some KTC types the GCMs agree on the
sign and general pattern of changes; however, the
sensitivity of models to the radiative forcing differs
for different KTC types (Fig. 5). For example, the
models MIROC-ESM and MIROC-ESM-CHEM give
the smallest change in Ft and the largest change in Fi
(Fig. S24). The course of simulated changes is quite
Belda et al.: Warming-induced changes in climate zones
smooth for some of the GCMs, while for others it
exhibits wave-like behavior, breaks, and jumps even
when using 30 yr running means.
The magnitude of changes for 3 selected time pe -
riods (near future, mid-century, far future) were com-
pared to model errors in the reference period 1961−
1990 (Fig. 5). The errors were evaluated from com-
parison of relative land areas of KTC types derived
from GCM simulations and from CRU TS3.22 obser-
vations; in this way, they could be interpreted as bi-
ases as well. Regarding the end of the 21st century
(far future), only for 3 out of 12 KTC types, viz. Ft, E,
and Dc, the changes are higher than the model errors
(according to most of GCMs under RCP8.5; under
RCP4.5, half of the GCMs show that changes are
higher than errors for Dc, and ca. 75% of GCMs indi-
cate that this is the case for Eand Ft). Thus, consider-
ing the model errors, the simulated decrease in rela-
tive continental area is clearly pronounced in the case
of boreal climate Eand tundra climate Ft, and the in-
crease is pronounced in the case of continental tem-
perate climate Dc. Regarding the expected de crease
in Fi area, the simulated changes are larger than
model errors according to only one-third of the GCMs
for both RCP 8.5 and RCP 4.5. Further, in case of sa-
vanna climate Aw, dry climates BW, and steppe cli-
mate BS, the simulated changes in the far future are
larger than model errors according to about half of
the GCMs for both RCPs, except for dry climate BW
under RCP4.5 (one-third of the models) and steppe
climate BS under the same scenario (only 13%).
For Cf and Do, the simulated changes are larger
than model errors according to only 6 and 4 out of 30
GCMs, respectively. The type Ar is the only KTC
type for which the simulated changes are smaller
than model errors for both RCPs and all 3 time peri-
ods. We found no straightforward relationship be -
tween the model performance and the strength of the
climate signal in projected changes.
Besides a comparison of simulated changes to
model errors, we assessed the uncertainty stemming
from necessary choices in GCM structure. We used
the range between the 10th and 90th percentile of
the multi-model ensemble to assess this uncertainty.
The smallest multi-model spread of simulated
changes is seen for the BW type (Table 3), the largest
for Eand Ft. The simulated changes of Ar, Cf, and Do
types are ambiguous in the sense that the multi-
model ranges include a ‘zero change’.
Considering the changes in relative areas for the
KTC types all together, the lowest sensitivity to radia-
tive forcing under RCP4.5 is seen for MRI-CGCM3,
with 8% of total continental area undergoing a KTC
type change until the end of the 21st century. Then
follows a group of models with simulated changes of
<12% (3 GISS GCMs, Nor-ESM1-M, Nor-ESM1-ME,
GFDL-ESM2M, ACCESS1-3). For RCP8.5, the lowest
sensitivity was found for all GISS models and
MRI-CGCM3, with changes of 16−17%. The largest
sensitivity was found for MIROC-ESM and MIROC-
ESM-CHEM, with nearly a quarter of the global con-
tinental area (without Antarctica) showing changed
KTC types under RCP4.5 and about 35% under
RCP8.5. According to M-MED for RCP4.5 (RCP8.5),
14% (22%) of the continental area is expected to
change its climate type by the end of the century. Our
results are in agreement with Mahlstein et al. (2013),
who projected that approximately 20% of global land
will experience a change in climate type until 2100
under RCP8.5 forcing, although their study was
based on a smaller number of models than ours.
According to Feng et al. (2014), a larger portion of the
continental area is expected to undergo a change in
climate type (31% for RCP 4.5 and 46% for RCP8.5).
Regarding the shifts in the 6 major climate types,
the changes projected for the far-future period under
RCP8.5 based on our results and the study of Feng et
al. (2014) are summarized in Table 4. The values are
shown as a percentage of the respective KTC type
area in the reference period 1961−1990. The simu-
lated changes are more distinct for types D, E, and F
than for other KTC types. For these 3 types, 25−50%
of the reference area is expected to shift to another
KTC type. This likely points to a more important in-
fluence of air temperature changes than precipitation
changes on the KTC type shifts, which was shown by
Feng et al. (2014) and Mahlstein et al. (2013).
1 M-MED 17.4 13.2 −11.2 24.7 −37.4 −50.0
2 M-mean 14.9 15.6 −13.5 26.3 −39.4 −61.0
3 SD 10.1 6.3 12.7 9.3 22.3 27.4
4 M-mean SD 4.9 9.3 −26.2 17.0 −61.7 −88.5
5 M-mean + SD 25.0 21.9 −0.8 35.5 −17.1 −33.6
6 Fmean 11.6 15.9 −13.4 40.0 −50.4 −59.2
7 FSD 4.0 5.3 7.9 13.7 16.9 10.8
8 Fmean FSD 7.6 10.6 −21.3 26.3 −67.3 −70.0
9 Fmean + FSD 15.6 21.2 −5.5 53.7 −33.5 −48.4
Table 4. Comparison of multi-model statistics aggregated for
the main Köppen-Trewartha classification (KTC) climate
types (AF, see Table 2 for definitions) in this study (rows
1−5) and in Feng et al. (2014, rows 6−9). Values are given as
a percentage of the respective KTC type area in the refer-
ence period 1961−1990. M-MED: multi-model median, M-
mean: multi-model mean, SD: standard deviation, F: Feng
et al. (2014) values
Clim Res 71: 17–31, 2016
For the purpose of comparing our results to the
study of Feng et al. (2014), the multi-model mean (M-
mean) and standard deviation (SD) were calculated
for the 6 main climate types (Table 4). M-mean dif-
fers from M-MED most prominently for the Fclimate
type, for which the SD has also the highest value. The
values of M-mean according to our results and Feng
et al. (2014) are fairly similar except for Dand E,
where Feng et al. (2014) found larger changes than
presented in our analysis. The SD values according
to Feng et al. (2014) are all smaller than our SD val-
ues, except for type D. In our study, we prefer the
median and range between the 10th and 90th per-
centile to characterize the distribution of the multi-
model ensemble, as the distribution of simulated
KTC type changes is generally not symmetrical.
There are several possible reasons for the differ-
ences between our results and the results of Feng et
al. (2014). The analyses are based on different groups
of GCMs which may significantly influence the re -
sults, as individual models show different changes in
KTC types in reaction to a given forcing. Further, we
used model outputs in the original model grids, but
Feng et al. (2014) applied a downscaling procedure
to the GCM outputs. The fact that we investigated all
continental areas excluding Antarctica, whereas
Feng et al. (2014) only considered global continents
north of 60° S, and the different observational data -
sets used could also play a role, although, as discus -
sed previously, the differences are very small com-
pared to the ensemble spread.
A change in relative continental areas of climate
types is not the only expected impact of climate
change. Potential geographical shifts are also very
important. Our results indicate a poleward shift of Ft,
E, Do, Dc, and Cf types (not shown). On the other
hand, Ar and Aw types, which are found near the
equator, did not experience any latitudinal shift.
Regarding the dry climates, the GCMs do not agree
entirely, especially in the case of BW. A detailed
analysis of these shifts, however, is beyond the scope
of this study and will be the subject of future investi-
Acknowledgement. We acknowledge the World Climate
Research Programme's Working Group on Coupled Model-
ling, which is responsible for CMIP, and we thank the cli-
mate modeling groups for producing and making available
their model outputs. For CMIP the US Department of
Energy's Program for Climate Model Diagnosis and Inter-
comparison provides coordinating support and led develop-
ment of software infrastructure in partnership with the
Global Organization for Earth System Science Portals. The
CRU TS 3.22 dataset was provided by the Climatic Research
Unit, University of East Anglia. This study was supported by
project UNCE 204020/2012, funded by Charles University in
Prague, and by research plan no. MSM0021620860 funded
by the Ministry of Education, Youth and Sports of the Czech
Republic. In addition, the work is part of the activity under
the Program of Charles University PRVOUK No. 02 ‘Envi-
ronmental Research’.
Bailey RG (2009) Ecosystem geography: from ecoregions to
sites, 2nd edn. Springer, New York, NY
Baker B, Diaz H, Hargrove W, Hoffman F (2010) Use of the
Köppen-Trewartha climate classification to evaluate cli-
matic refugia in statistically derived ecoregions for the
People’s Republic of China. Clim Change 98: 113−131
Belda M, Holtanová E, Halenka T, Kalvová J (2014) Climate
classification revisited: from Köppen to Trewartha. Clim
Res 59: 1−13
Belda M, Holtanová E, Halenka T, Kalvová J, Hálvka Z
(2015) Evaluation of CMIP5 present climate simulations
using the Köppen-Trewartha climate classification. Clim
Res 64: 201−212
de Castro M, Gallardo C, Jylha K, Tuomenvirta H (2007) The
use of a climate-type classification for assessing climate
change effects in Europe from an ensemble of nine
regional climate models. Clim Change 81: 329−341
Deque M (2007) Frequency of precipitation and tempera-
ture extremes over France in an anthropogenic sce-
nario: model results and statistical correction according
to observed values. Global Planet Change 57: 16–26
Feng S, Ho CH, Hu Q, Oglesby RJ, Jeong SJ, Kim BM (2012)
Evaluating observed and projected future climate
changes for the Artic using the Köppen−Trewartha cli-
mate classification. Clim Dyn 38: 1359−1373
Feng S, Hu Q, Huang W, Ho CH, Li R, Tang Z (2014) Pro-
jected climate regime shift under future global warming
from multi-model, multi-scenario CMIP5 simulation.
Global Planet Change 112: 41−52
Geiger R (1954) Klassifikationen der Klimate nach W. Köp-
pen. In: Bartels J, ten Bruggencate P (eds) Landolt-
Börnstein: Zahlenwerte und Funktionen aus Physik,
Chemie, Astronomie, Geophysik und Technik (alte
Serie), Vol 3. Springer, Berlin, p 603−607
Hargrove WW, Hoffman FM (2004) Potential of multivariate
quantitative methods for delineation and visualization of
ecoregions. Environ Manag 34: S39−S60
Harris I, Jones PD, Osborn TJ, Lister DH (2014) Updated
high-resolution grids of monthly climatic observations —
the CRU TS3.10 dataset. Int J Climatol 34: 623−642
IPCC (Intergovernmental Panel on Climate Change) (2013)
Climate Change 2013: the physical science basis. Cam-
bridge University Press, Cambridge
Kalvová J, Halenka T, Bezpalcová K, Nemešová I (2003)
Köppen climate types in observed and simulated cli-
mates. Stud Geophys Geod 47: 185−202
Köppen W (1923) Die Klimate der Erde. Grundriss der
Klima kunde. Walter de Gruyter & Co., Berlin
Köppen W (1931) Grundriss der Klimakunde. Walter de
Gruyter & Co., Berlin
Köppen W (1936) Das geographische System der Klimate.
In: Köppen W, Geiger R (eds) Handbuch der Klimato -
logie. Gebrüder Borntraeger, Berlin, p C1−C44
Lohmann U, Sausen R, Bengtsson L, Cubasch U, Perlwitz J,
Belda et al.: Warming-induced changes in climate zones
Roeckner E (1993) The Köppen climate classification as a
diagnostic tool for general circulation models. Clim Res
3: 177−193
Mahlstein I, Daniel JS, Solomon S (2013) Pace of shifts in
climate regions increases with global temperature.
Nat Clim Change 3: 739−743
Meehl GA, Covey C, Delworth T, Latif M and others (2007)
The WCRP CMIP3 multi-model dataset: a new era in
climate change research. Bull Am Meteorol Soc 88:
Menon A, Levermann A, Schewe J, Lehmann J, Frieler K
(2013) Consistent increase in Indian monsoon rainfall
and its variability across CMIP-5 models. Earth Syst Dyn
4: 287−300
Moss RH, Edmonds JA, Hibbard KA, Manning MR and
others (2010) The next generation of scenarios for cli-
mate change research and assessment. Nature 463:
Patton CP (1962) A note on the classification of dry climate
in the Köppen system. Calif Geogr 3: 105−112
Rogelj J, Meinshausen M, Knutti R (2012) Global warming
under old and new scenarios using IPCC climate sensi-
tivity range estimates. Nat Clim Change 2: 248−253
Rubel F, Kottek M (2010) Observed and projected climate
shifts 1901−2100 depicted by world maps of the Köppen-
Geiger climate classification. Meteorol Z 19: 135−141
Taylor K, Balaji V, Hankin S, Juckes M, Lawrence B, Pascoe
S (2012a) CMIP5 Data Reference Syntax (DRS) and con-
trolled vocabularies. http:// pcmdi-cmip. llnl. gov/ xmip5/
docs/ cmip5_data_refernce_syntax. pdf (accessed 7 Nov
Taylor K, Stouffer RJ, Meehl GA (2012b) An overview of
CMIP5 and the experiment design. Bull Am Meteorol
Soc 93: 485−498
Trewartha GT (1968) An introduction to climate. McGraw-
Hill, New York, NY
Trewartha GT, Horn LH (1980) Introduction to climate, 5th
edn. McGraw Hill, New York, NY
Willmott CJ, Matsuura K (2001) Terrestrial air temperature
and precipitation: monthly and annual time series (1950
1999). html_ pages
/ README.ghcn_ts2.html
Editorial responsibility: Tim Sparks,
Cambridge, UK
Submitted: February 22, 2016; Accepted: July 24, 2016
Proofs received from author(s): November 3, 2016
... Otto-Bliesner et al., 2017) and climate change projections (e.g. Belda et al., 2017;Coppola et al., 2021a;Thomas et al., 2022). However, model simulations of climate are subject to many uncertainties. ...
... Continentality of climate generally characterizes the influence of the distance of ocean on the climate of a place (Driscoll and Yee Fong, 1992). Previous studies based on Köppen-Trewartha climatic classification showed projected transition from continental temperate climate to oceanic temperate climate over Central Europe (Feng et al., 2012;Belda et al., 2017). This is seemingly in contrast to our results. ...
Full-text available
The multi-model ensembles like CMIP5 or CMIP6 provide a tool to analyze structural uncertainty of climate simulations. Currently developed regional and local climate change scenarios for the Czech Republic assess the uncertainty based on state-of-the-art Global Climate Model (GCM) and Regional Climate Model (RCM) ensembles. Present study focuses on multi-model spread of projected changes in long-term monthly means and inter-annual variability of monthly mean minimum, mean and maximum daily air temperature and monthly mean precipitation. We concentrate in more detail on the simulation of CNRM-ESM2-1, the driving GCM for the convection permitting ALADIN-Climate/CZ simulation contributing to the local scenarios in very high resolution. For this GCM, we also analyze a mini-ensemble with perturbed initial conditions to evaluate the range of internal climate variability. The results for the Czech Republic reveal minor differences in model performance in the reference period whereas quite substantial inter-generation shift in projected future change towards higher air temperature and lower summer precipitation in CMIP6 comparing to CMIP5. One of the prominent features across GCM generations is the pattern of summer precipitation decrease over central Europe. Further, projected air temperature increase is higher in summer and autumn than in winter and spring, implying increase of thermal continentality of climate. On the other hand, slight increase of winter precipitation and tendency towards decrease of summer precipitation lead to projected decrease of ombric continentality. The end of 21st century projections also imply higher probability of dry summer periods, higher precipitation amounts in the cold half of the year and extremely high temperature in summer. Regarding the CNRM-ESM2-1, it is often quite far from the multi-model median. Therefore, we strictly recommend to accompany any analysis based on the simulation of nested Aladin-CLIMATE/CZ with proper uncertainty estimate. The range of uncertainty connected to internal climate variability based on one GCM is often quite large in comparison to the range of whole CMIP6 ensemble. It implies that when constructing climate change scenarios for the Central Europe region, attention should be paid not only to structural uncertainty represented by inter-model differences and scenario uncertainty, but also to the influence of internal climate variability.
... Applications of CMIP models to environmental studies began in the early stages of the project. Some common research topics have been: landscape analysis Guan et al., 2021;Zhou et al., 2021), climate zonation (Rubel and Kottek, 2010;Kriticos et al., 2012;Tapiador et al., 2019a), changes in environmental conditions (Belda et al., 2016;Akinsanola et al., 2020a;Dong et al., 2021), and adaptation to climate change (Läderach et al., 2017;Chen et al., 2020;Mondal et al., 2021). Table 1 lists CMIP6 models used in our work. ...
... Overall, the shifts were at the limits of climate zones and were clearer over the Arctic Ocean. The shrinkage of polar climates was not restricted to high latitudes but were also in high-elevation areas, as suggested by previous studies (Belda et al., 2016;Cui et al., 2021). As expected, changes towards wetter climate types were testimonial. ...
Climate classifications are useful to synthesize the physical state of the climate with a proxy that can be directly related to biota. However, they present a potential drawback, namely a strong sensitivity because of the use of hard thresholds (step functions). Thus, minor discrepancies in the base data produce large differences in the type of climate. However, such an a priori limitation is also a strength because such sensitivity can be used to better gauge model performance. Although previous attempts of classifying climates of the world using global climate model outputs were encouraging, the applicability of their classifications to impact studies has been limited by past model issues. Notwithstanding the persistence of certain imperfections and limitations in current models, the high‐quality physical simulations from phase six of the Coupled Intercomparison Project (CMIP6) has opened new possibilities in the field, thanks to their improved representation of atmospheric and oceanic physics. The purpose of this paper is twofold: to show that climate classifications from CMIP6 are sufficiently robust for use in impact studies, and to use those classifications for identifying error sources and potential issues that deserve further attention in models. Thus, 52 CMIP6 climate models were evaluated by using three climate classifications schemes, classical Köppen, extended‐Köppen, and modified Thornthwaite. We first assessed model ability to reproduce present climate types by comparing their outputs with observational data. Models performed best for the Köppen and extended‐Köppen classification methods (Cohen's kappa κ = 0.7), and had moderate scores for the Thornthwaite climate classification (κ = 0.4). By tracing back the observed biases, we were able to pinpoint the misrepresentation of dry climates as a major source of error. Another finding was that most models still had some difficulties in representing the seasonal variability of precipitation over several monsoonal regions, thereby yielding the wrong climate type there. Models were also evaluated for future climate. Substantial changes in climate types are projected in the SSP5‐8.5 scenario. These changes include a shrinkage of polar/frigid climates (22%) and an increase of dry climates (7%). Simulations arising from global climate models can be directly used to understand the global climate. They are however in the form of multidimensional matrices, which makes the outputs difficult to compare and validate. Conversely, climate classifications simplify the complex interactions of the climate system and serve as a single, aggregated parameter for environmental applications. The purpose of this work is to show that climate classifications from GCMs are robust enough to be used in impact studies, and use those classifications to identify potential issues deserving further attention in models.
... Köppen's climate classification method (Köppen 1936) was implemented in this study to classify the climate zone and identify the climate shift to arid conditions in Asia. The method is popularly used to detect regional climates depending on their temperature and precipitation patterns due to the availability and simplicity of this method (Belda et al. 2014(Belda et al. , 2016 ...
Full-text available
A warming climate cause changes in regional climate and climate extremes. However, understanding how future warming can affect the arid climate and drought in Asia is still unclear. Global warming (GW) impacts on shifts in arid regions and change patterns of drought in the Asia region are investigated using the shared socioeconomic pathway 5–8.5 of the Coupled Model Intercomparison Project Phase 6. The variations in the arid region are explored under potential GW targets above preindustrial levels (+ 0.66 °C for the reference period and + 1.5 °C, + 2.5 °C, + 3.0 °C for future periods). We examine changes in precipitation and temperature indices related to drought, the drought patterns corresponding to consistent arid zones and arid-wet shifting zones based on GW scenarios. The results indicate that warming climates influence temperature and precipitation, and consequently alter climate distribution toward an expansion of the arid region. Due to shifts in climate zones, arid regions may expand by 2.6% relative to the reference climate in response to 3.0 °C of warming. Changes in precipitation indices contribute to the spatial patterns and magnitude of drought changes. Future drought features induced by meteorological conditions exhibit different behaviors according to arid-wet shifts and drought events. Especially, drought features throughout Asia are likely to be aggravated in terms of both severity and frequency in response to a greater degree of GW, especially in arid regions and regarding extreme droughts. This study provides valuable information on the relationship between regional climate and drought features in a world affected by GW.
... This algorithm sorts the data according to the climate values of the biome (temperature and precipitation) and sets the borders between them (Fick & Hijmans, 2017;Gotway et al., 1996;Orus et al., 2005;Timpano et al., 2011;Valjarević et al., 2018;Zabel et al., 2014). The main four maps present the climate properties for the four future periods: 2021-2040, 2041-2060, 2061-2080and 2081-2100(Belda et al., 2016. The overlap of the two grids was made possible by the proximity algorithm. ...
The Updated Trewartha climate classification (TWCC) at global level shows the changes that are expected as a consequence of global temperature increase and imbalance of precipitation. This type of classification is more precise than the Köppen climate classification. Predictions included the increase in global temperature (T in °C) and change in the amount of precipitation (PA in mm). Two climate models MIROC6 and IPSL-CM6A-LR were used, along with 4261 meteorological stations from which the data on temperature and precipitation were taken. These climate models were used because they represent the most extreme models in the CMIP6 database. Four scenarios of climate change and their territories were analysed in accordance with the TWCC classification. Four scenarios of representative concentration pathway (RCP) by 2.6, 4.5, 6.0 and 8.5 W/m2 follow the increase of temperature between 0.3°C and 4.3°C in relation to precipitation and are being analysed for the periods 2021–2040, 2041–2060, 2061–2080 and 2081–2100. The biggest extremes are shown in the last grid for the period 2081–2100, reflecting the increase of T up to 4.3°C. With the help of GIS (geographical information systems) and spatial analyses, it is possible to estimate the changes in climate zones as well as their movement. Australia and South East Asia will suffer the biggest changes of biomes, followed by South America and North America. Climate belts to undergo the biggest change due to such temperature according to TWCC are Ar, Am, Aw and BS, BW, E, Ft and Fi. The Antarctic will lose 11.5% of the territory under Fi and Ft climates within the period between 2081 and 2100. The conclusion is that the climates BW, Bwh and Bwk, which represent the deserts, will increase by 119.8% with the increase of T by 4.3°C.
... 气候模式(包括全球气候模式和区域气候模式)是 气候预估的基础工具. 基于参加耦合模式比较计划第 五阶段(CMIP5)全球模式集合结果, 使用柯本气候分类 法, Mahlstein等人 [17] 指出全球气候带随全球气温的升 高而线性变化; Feng等人 [18] 预估全球的温带、热带和 干旱带的面积将扩张, 而极地带、亚寒带和亚热带的 面积将减少; Belda等人 [13] 预估到21世纪末, 在RCP4.5 ...
... The Köppen climate classification has been used to set up dynamic global vegetation models (Poulter et al., 2011(Poulter et al., , 2015, to characterize species composition (Brugger and Rubel, 2013), to model the species range distribution (Tererai and Wood, 2014;Brugger and Rubel, 2013;Webber et al., 2011), and to analyze the species growth behavior (Tarkan and Vilizzi, 2015). The Köppen classification has also been applied to detect the shifts in geographical distributions of climate zones (Belda et al., 2016;Chan and Wu, 2015;Feng et al., 2014;Mahlstein et al., 2013). It also has the potential to aggregate climate information on warmth and precipitation seasonality into ecologically important climate classes, thereby simplifying spatial variability. ...
Full-text available
The Köppen–Geiger classification scheme provides an effective and ecologically meaningful way to characterize climatic conditions and has been widely applied in climate change studies. Significant changes in the Köppen climates have been observed and projected in the last 2 centuries. Current accuracy, temporal coverage and spatial and temporal resolution of historical and future climate classification maps cannot sufficiently fulfill the current needs of climate change research. Comprehensive assessment of climate change impacts requires a more accurate depiction of fine-grained climatic conditions and continuous long-term time coverage. Here, we present a series of improved 1 km Köppen–Geiger climate classification maps for six historical periods in 1979–2013 and four future periods in 2020–2099 under RCP2.6, 4.5, 6.0, and 8.5. The historical maps are derived from multiple downscaled observational datasets, and the future maps are derived from an ensemble of bias-corrected downscaled CMIP5 projections. In addition to climate classification maps, we calculate 12 bioclimatic variables at 1 km resolution, providing detailed descriptions of annual averages, seasonality, and stressful conditions of climates. The new maps offer higher classification accuracy than existing climate map products and demonstrate the ability to capture recent and future projected changes in spatial distributions of climate zones. On regional and continental scales, the new maps show accurate depictions of topographic features and correspond closely with vegetation distributions. We also provide a heuristic application example to detect long-term global-scale area changes of climate zones. This high-resolution dataset of the Köppen–Geiger climate classification and bioclimatic variables can be used in conjunction with species distribution models to promote biodiversity conservation and to analyze and identify recent and future interannual or interdecadal changes in climate zones on a global or regional scale. The dataset referred to as KGClim is publicly available via (Cui et al., 2021d) and can also be downloaded at (Cui et al., 2021c) for historical climate and (Cui et al., 2021b) for future climate.
A possible shift in climate zones in Southeast Asia (SEA) for different shared socioeconomic pathways (SSPs) is evaluated in this study. The ability of 19 Coupled Model Intercomparison Project (CMIP6) global climate models (GCMs) in reconstructing the Köppen-Geiger climate zones in SEA, estimated using reanalysis data (ERA5) for the period 1979-2014, was analysed using five categorical evaluation metrics. The best-performing models were selected to prepare an ensemble to project possible shifts in climate zones for different SSP scenarios in the future. Besides, future projections in climate variables were evaluated to assess the driving factor of climate shifts in the future. The results showed that three CMIP6 GCMs, EC-Earth3-Veg-LR, CMCC-ESM2 and CanESM5, had a higher skill in classifying the observed climate of SEA. Selected GCMs showed climate shifting in 3.4 to 12.6% of the total area of SEA for different SSPs. The highest geographical shift in climate was projected in the north, from dry winter and hot summer (Cwa) to tropical with dry winter (Aw), followed by Aw to tropical monsoon (Am) in the north and south, and tropical without dry season (Af) to Am in the middle and southwest of SEA. An increase in minimum temperature was the key to climate shifting from Cwa to Aw in the north, while increased rainfall was a reason for Aw to Am transition in the north and south. Overall, climatic shifting was higher for high emission scenarios. The maps of future climate zones generated in this study can help to identify the hotspots of ecologically vulnerable areas in SEA due to climate change.
The effect of air-sea coupling in the simulation of the European climate is assessed through a climate type classification that uses surface temperature and precipitation from a regional atmosphere-ocean coupled model and from its atmospheric component. The atmospheric setup in both models is the same, differing only in the representation of the oceanic fields. The simulations cover the present and future-time climate under the RCP8.5 CMIP5 scenario. Climate type distributions obtained from both coupled and uncoupled simulations are similar to those obtained from ERA5 for the 1976–2005 period. Both models simulate colder climate types for present-time in southern and northeastern regions compared to ERA5, possibly due to a weaker influence of the Atlantic circulation, and drier climate types in some western Mediterranean areas. Also, for present-time coupling leads to more humid winters (relatively drier summers) in some zones of north Spain and south France, and drier climates in some western Mediterranean spots. Based on simulations with these models under the RCP8.5 scenario, we find that by the end of the 21st century (2070–2099) the climate type distribution changes in more than 50% of the domain. While both models project the reduction of regions with cold climate types and the expansion of those with hot summers and hot arid climate types, these changes affect a larger area in the coupled simulation. These differences may be related to a drier signal in the coupled simulation, especially during summer, due to the influence of colder surface water in the North Atlantic Ocean and the Mediterranean Sea. Using a climate classification to evaluate the annual cycles of the simulated temperature and precipitation data provides a novel insight into the impact of air-ocean coupling on the representation of the climate, and consequently into the simulated impact on ecosystems and human activities in Europe.
Eco-geographical regionalization involves dividing land into regions by considering both intra-regional consistency and interregional disparity and is based on the pattern of differentiation of eco-geographical elements. Owing to the complexity of the land surface, and the limitation of data and appropriate methods, regions in China have hitherto been mapped manually, meaning that the process of mapping was non-repeatable. To make the regionalization technique repeatable, this study aimed to extract and quantify the expert knowledge of regionalization using an automated method. The rough set method was adopted to extract rules of regionalization based on the existing eco-geographical regionalization map of China, as well as its corresponding meteorological and geological datasets. Then, the rules for regionalization were obtained hierarchically for each natural domain, each temperature zone, and each humidity region. Owing to differences in zonal differentiation, the rule extraction sequence for the eastern monsoon zone and Tibetan Alpine zone was temperature zone first followed by humidity region, with the reverse order being applied for the northwest arid/semi-arid zone. Results show that the extracted indicators were similar to those of the existing (expert-produced) regionalization scheme but more comprehensive. The primary indicator for defining temperature zones was the ≥10°C growing season, and the secondary indicators were the January and July mean temperatures. The primary and secondary indicators for identifying humid regions were aridity index and precipitation, respectively. Eco-geographical regions were mapped over China using these rules and the gridded indicators. Both the temperature zones and humidity regions mapped by the rules show ≥85% consistency with the existing regionalization, which is higher than values for mapping by the commonly used simplified method that uses the classification of one indicator. This study demonstrates that the proposed rough set method can establish eco-geographical regionalization that is quantitative and repeatable and able to dynamically updated.
Full-text available
Climate warming has a great impact on grain production in northeast China, but there are few studies on the temporal and spatial variation characteristics of annual Tmean (mean temperature), the impact of meteorological factors on Tmean, and the impact of Tmean increase on grain production in northeast China. This study found that annual Tmean decreased from southeast to northwest in Northeast China, and there were regional differences in spatial distribution. The annual Tmean isoline in Northeast China moves obviously from southeast to northwest. The annual warming trend of Tmean was significant from 1971 to 2000, and moderated from 2001 to 2020. In recent 50 years, Tmean had obvious periodic changes. In the mid-late 1980s, annual Tmean had a sudden warming change, and since then it has been rising continuously. Sunshine hours, average wind speed, evaporation and average air pressure had a very significant correlation with Tmean. In conclusion, the climate change in northeast China in the past 50 years has an obvious warming and drying trend, and there are regional differences in the warming and drying. The warming and drying climate has brought challenges to agricultural production and food security in Northeast China. However, the negative effects of grain production reduction caused by warming and drying climate can be avoided to a certain extent if we deal with it properly.
Full-text available
The possibility of an impact of global warming on the Indian monsoon is of critical importance for the large population of this region. Future projections within the Coupled Model Intercomparison Project Phase 3 (CMIP-3) showed a wide range of trends with varying magnitude and sign across models. Here the Indian summer monsoon rainfall is evaluated in 20 CMIP-5 models for the period 1850 to 2100. In the new generation of climate models, a consistent increase in seasonal mean rainfall during the summer monsoon periods arises. All models simulate stronger seasonal mean rainfall in the future compared to the historic period under the strongest warming scenario RCP-8.5. Increase in seasonal mean rainfall is the largest for the RCP-8.5 scenario compared to other RCPs. Most of the models show a northward shift in monsoon circulation by the end of the 21st century compared to the historic period under the RCP-8.5 scenario. The interannual variability of the Indian monsoon rainfall also shows a consistent positive trend under unabated global warming. Since both the long-term increase in monsoon rainfall as well as the increase in interannual variability in the future is robust across a wide range of models, some confidence can be attributed to these projected trends.
Full-text available
Human-induced climate change causes significant changes in local climates, which in turn lead to changes in regional climate zones. Large shifts in the world distribution of Köppen-Geiger climate classifications by the end of this century have been projected. However, only a few studies have analysed the pace of these shifts in climate zones, and none has analysed whether the pace itself changes with increasing global mean temperature. In this study, pace refers to the rate at which climate zones change as a function of amount of global warming. Here we show that present climate projections suggest that the pace of shifting climate zones increases approximately linearly with increasing global temperature. Using the RCP8.5 emissions pathway, the pace nearly doubles by the end of this century and about 20% of all land area undergoes a change in its original climate. This implies that species will have increasingly less time to adapt to Köppen zone changes in the future, which is expected to increase the risk of extinction.
Full-text available
This paper describes the construction of an updated gridded climate dataset (referred to as CRU TS3.10) from monthly observations at meteorological stations across the world's land areas. Station anomalies (from 1961 to 1990 means) were interpolated into 0.5° latitude/longitude grid cells covering the global land surface (excluding Antarctica), and combined with an existing climatology to obtain absolute monthly values. The dataset includes six mostly independent climate variables (mean temperature, diurnal temperature range, precipitation, wet-day frequency, vapour pressure and cloud cover). Maximum and minimum temperatures have been arithmetically derived from these. Secondary variables (frost day frequency and potential evapotranspiration) have been estimated from the six primary variables using well-known formulae. Time series for hemispheric averages and 20 large sub-continental scale regions were calculated (for mean, maximum and minimum temperature and precipitation totals) and compared to a number of similar gridded products. The new dataset compares very favourably, with the major deviations mostly in regions and/or time periods with sparser observational data. CRU TS3.10 includes diagnostics associated with each interpolated value that indicates the number of stations used in the interpolation, allowing determination of the reliability of values in an objective way. This gridded product will be publicly available, including the input station series ( and © 2013 Royal Meteorological Society
Full-text available
The fifth phase of the Coupled Model Intercomparison Project (CMIP5) will produce a state-of-the- art multimodel dataset designed to advance our knowledge of climate variability and climate change. Researchers worldwide are analyzing the model output and will produce results likely to underlie the forthcoming Fifth Assessment Report by the Intergovernmental Panel on Climate Change. Unprecedented in scale and attracting interest from all major climate modeling groups, CMIP5 includes “long term” simulations of twentieth-century climate and projections for the twenty-first century and beyond. Conventional atmosphere–ocean global climate models and Earth system models of intermediate complexity are for the first time being joined by more recently developed Earth system models under an experiment design that allows both types of models to be compared to observations on an equal footing. Besides the longterm experiments, CMIP5 calls for an entirely new suite of “near term” simulations focusing on recent decades and the future to year 2035. These “decadal predictions” are initialized based on observations and will be used to explore the predictability of climate and to assess the forecast system's predictive skill. The CMIP5 experiment design also allows for participation of stand-alone atmospheric models and includes a variety of idealized experiments that will improve understanding of the range of model responses found in the more complex and realistic simulations. An exceptionally comprehensive set of model output is being collected and made freely available to researchers through an integrated but distributed data archive. For researchers unfamiliar with climate models, the limitations of the models and experiment design are described.
We used the Köppen-Trewartha classification on CMIP5 family of global climate model (GCM) simulations and global Climatic Research Unit (CRU) data for comparison. This evaluation provides preliminary insight on GCM performance and errors. For the overall model intercomparison and evaluation, we used 2 simple statistical characteristics: normalized error, which assesses the total relative difference of the area classified by the individual model with respect to the area resulting from CRU data, and overlap, calculating relative area of matching grid boxes in model results and CRU data. With the additional analysis of the classification on world maps, we show that there are some common features in the model results. Many models have problems capturing the rainforest climate type Ar, mainly in Amazonia. The desert climate type BW is underestimated by as many as half of the models, with Australia being a typical example of a region where the BW is not well represented. The boreal climate type E is overestimated by many models, mostly spreading over to the areas of observed tundra type Ft. All applied metrics indicate that with the current generation of GCMs, there is no clear tendency for models to improve the representation of climate types with higher spatial resolution.
The analysis of climate patterns can be performed separately for each climatic variable or the data can be aggregated, for example, by using a climate classification. These classifications usually correspond to vegetation distribution in the sense that each climate type is dominated by one vegetation zone or eco-region. Thus, climate classifications also represent a convenient tool for the validation of climate models and for the analysis of simulated future climate changes. Basic concepts are presented on the global Climate Research Unit (CRU) data analysis. We focus on definitions of climate types according to the Köppen-Trewartha climate classification (KTC) with special attention given to the distinction of wet and dry climates. The distribution of KTC types is compared with the original Köppen classification (KCC) for the period of 1961-1990. In addition, we provide an analysis of the time development of the distribution of KTC types throughout the 20th century. There are observable changes identified in some subtypes, especially semi-arid, savanna and tundra.