1
Classification: PHYSICAL SCIENCES (Atmospheric Sciences).
2
Significant anthropogenic-induced changes of climate classes since 1950
3
4
Duo Chan1, Qigang Wu1*
5
1
6
School of Atmospheric Sciences, Nanjing University
7
8
9
10
11
*Corresponding author address: Prof. Qigang Wu, School of Atmospheric Sciences, Nanjing
12
University, Hankou Road #22, Nanjing, Jiangsu, China, 210093
13
E-mail: [email protected]
14
15
16
Keywords: Köppen-Geiger climate classification, anthropogenic climate change, CMIP5， global
17
warming
18
19
20
21
22
1
23
Abstract Anthropogenic forcings have contributed to the global and regional warming in last few
24
decades (1-5) and likely affected the terrestrial precipitation (5-9). Ecosystem stress from recent and
25
projected climate change is likely to be most severe in areas where the climate classification changes.
26
Here we calculate major Köppen climate classes from gridded data and show that 5.7% of the global
27
land area has shifted toward warmer and drier climate types from 1950-2010. Significant changes
28
include expansion of arid and high-latitude continental climate zones, shrinkage in polar and
29
midlatitude continental climates, poleward shifts in temperate, continental and polar climates, and
30
increasing average elevation of tropical and polar climates. Using Coupled Model Intercomparsion
31
Project 5 (CMIP5) multi-model averaged historical simulations forced by observed anthropogenic
32
and natural, or natural only, forcing components, we find that climate class changes cannot be
33
explained as natural variations but are driven by anthropogenic factors. Results here provide strong
34
evidence that anthropogenic climate change has already generated significant impacts on ecosystems
35
over the global land area.
36
Significance Statement In this study, we detect significant changes in major Köppen climate types
37
since 1950 and attribute these changes to anthropogenic forcing factors. The Köppen climate
38
classification was designed to explain observed biome distributions, has only been slightly revised in
39
about 100 years, and is understood to a considerable extent by the general public, so our results
40
provide strong evidence that anthropogenic climate change has already generated significant impacts
41
on ecosystems over the global land area. Understanding and publicly communicating such impacts
42
has been an increasingly important interdisciplinary research target of the scientific community.
43
\body The Köppen climate classification scheme was developed (10) to explain observed biome
2
44
distributions, which have many sharp boundaries due to plant sensitivity to threshold values of
45
average monthly temperature and precipitation and their annual cycle (11). Long-term changes in
46
climate classes are more general indicators for climatic changes than are temperature and
47
precipitation trends alone. Köppen or similar classifications have been used to estimate the potential
48
impacts of past and projected future climate on prevalent ecoregions on regional and global scales
49
(12-20). Köppen climate types are projected to shift strongly toward warmer and drier climate types
50
(temperate, tropical and arid), with climate types in 31.4% and 46.3% of the global land area
51
projected to change by 2100 under RCP4.5 and RCP8.5 scenarios, respectively (19).
52
However, it is still not clear whether significant changes of climate types are already detectable in
53
observations, and whether such changes can be attributed to external anthropogenic forcing. This
54
study uses updated thresholds (11) for five major climate classes (30 subtypes are not of concern
55
here) based on climate data at a location, with tests performed in the following order to resolve
56
conflicts: Arid (class B), Tropical (class A), Polar (class E), Temperate (class C), Continental (class
57
D). We compute Köppen climate classifications from gridded observation and model data compared
58
to a base period centered on 1950 to develop four indices describing the distribution of climate types:
59
1) percentage of world land area with a major climate type change from 1950; 2) total area occupied
60
by each major climate type; 3) averaged absolute latitude of each major climate type; and 4)
61
averaged elevation of each major climate type. Changes of these indices and their statistical
62
significance are first evaluated, and the relative roles of external anthropogenic and natural forcing in
63
these changes are assessed.
64
We use the University of Delaware global land 0.5°gridded monthly temperature and precipitation
3
65
dataset (21-22), updated through 2010, and calculate the Köppen climate type in each grid box.
66
Model runs are selected from the CMIP5 (23), including preindustrial control (PI-CTL) runs to
67
estimate natural variability statistics, historical runs to distinguish natural and anthropogenic factors,
68
and projection runs to estimate future climate type changes (Supplementary Table S1). For each
69
major climate type, we apply a one-sided local significance test to identify whether the observed
70
trends of four indices is significantly different from zero at the 5% significance level. Hereafter, all
71
time series and derived statistics are based on 15-year centered averages (e.g., “1950” refers to the
72
1943-1957 average, and “2003” indicates 1996-2010).
73
Observed changes of climate type indices
74
Figure 1a compares the percentage of the world land area (black line) that has experienced a climate
75
type change from 1950-2003 to the 95% cumulative probability boundary of such changes (dark gray
76
shading) in a hypothetical unchanging but naturally varying climate based on 541 PI-CTL
77
simulations (Fig. 1b). The area with major climate type changes becomes significantly greater than
78
zero with 5% significance level beginning around 1980. This suggests that significant climate shifts
79
were detectable before the recent dramatic and accelerated warming. It is possible that control runs
80
underestimate natural variability, so a more conservative test is to double variations from the mean in
81
Fig. 1b, giving an alternative 95% cumulative probability in Fig. 1a shown by light gray shading, and
82
this level is consistently exceeded starting 1996. About 5.7% of the land surface has experienced
83
shifts in major climate types by 2003, and changes are scattered in all major types rather than being
84
constrained in only one or two. Such ubiquitous shifts are already affecting agriculture and
85
endangered species over the entire globe.
4
86
Figure 2 shows linear trends of area, latitude and elevation indices from 1950 to 2003. Expansion
87
(shrinkage) exceeding 5% significance is found in arid (polar) climate at a rate of 4.8×105 (-2.8×105)
88
km2 decade-1. Significant poleward shifts are detected in temperate, continental and polar climate
89
averaging 35.0, 16.0 and 12.5 km decade-1, respectively, and significant elevation shifts in tropical
90
and polar climate averaging 3.0 and 14.3 m decade-1, respectively. Trends of total areas and averaged
91
elevations of temperate and continental climate are negative and not significant, but both shrinkage
92
of continental climate over regions south of 55°N (-2.9×105 km2 decade-1) and expansion of
93
continental climate north of 55°N (2.2×105 km2 decade-1) are statistically significant. All climate
94
types show net poleward movement (Fig. 2b) due to poleward expansion of A and B climates and
95
poleward shrinkage of C, D, and E climates.
96
Figures 2d-e show grid boxes with disappearing and emerging climate types from 1950-2003. The
97
most conspicuous feature is a worldwide expansion of B climate (mainly semiarid) at the expense of
98
C and midlatitude D climate. Rising temperature and decreasing precipitation are about equally
99
important in causing the expansion of semiarid climate in Asia and western North America, while the
100
contribution of decreasing precipitation to the increasing semiarid climate is much larger than that of
101
temperature over North Africa, South Africa and South America. Overall, temperature and
102
precipitation play similar roles in the expansion of B climate (Supplementary Fig. S1). In the tropics,
103
B replaces A climate over northern India and the southern ecotome of Sahara due to increasing
104
drought, but emerges in southern India and higher elevations over South America, northern South
105
Africa and northern Australia mainly due to increasing temperature. The above changes raise the
106
average elevation of A climate. Over the low level regions north of 55°N, the current E climate
107
distribution is more likely to be affected first by increasing warming, but higher elevations remain
5
108
cold enough to maintain the existing climate zones. The replacement of E by D climate is found over
109
Alaska, northern Canada, Siberia and Far East regions in Asia, and over the Tibetan Plateau, leading
110
to a significant shrinkage and higher elevation of E climate, and a poleward shift of D and E climate.
111
The rising elevation of A and E climates was reported in a modeling study (20). The reduction of the
112
area of C climate is mainly caused by the shift to A or B climate over large regions of South Africa
113
and South America driven by both temperature and precipitation changes. The shift of D to C climate
114
in large areas over Europe and East Asia due to increasing temperature and the shift of B to C climate
115
over South America due to increasing precipitation contribute to a significant poleward shift of C
116
climate.
117
Attribution of significant changes of climate type indices
118
To determine possible roles of external anthropogenic and natural radiative forcings in the above
119
climate shifts, the four indices are calculated from the multi-model averaged historical CMIP5
120
simulations forced by observed atmospheric composition changes (including anthropogenic forcings
121
such as greenhouse gases and sulfate aerosols and natural forcings such as volcanic eruptions and
122
solar output changes, termed HIST-ALL), by greenhouse gas forcings only (HIST-GHG), or by
123
natural forcings only (HIST-NAT). Supplementary Table S1 lists the selected model runs (23), which
124
use historical data ending 2005 (the last 15-year average is centered on 1998). Figure 1a shows
125
multi-model ensemble means of the percentage of the world land area with a climate type change
126
from 1950. By 1998, about 4.5%, 6.0% and 3.7% of the land surface has experienced shifts in major
127
climate types in HIST-ALL, HIST-GHG and HIST-NAT simulations, respectively. Both HIST-ALL
128
and HIST-GHG simulations fairly well reproduce the broad scale pattern of observed climate type
6
129
changes (Supplementary Fig. S1), including emergence of tropical climate over northern India and
130
Southern Hemisphere, expansion of B climate in the midlatitude Northern Hemisphere, and a shift of
131
E to D climate in high-latitudes.
132
For significant trends in Fig. 2, Fig. 3 shows corresponding trends for HIST-ALL, HIST-GHG, and
133
HIST-NAT multi-model averages. For each major climate type, the HIST-ALL and HIST-GHG
134
experiments qualitatively reproduce all significant observed trends. The observed trends are
135
consistent with that in the HIST-ALL run, except that the simulated B climate expansion is smaller
136
than observed, which is explained by the finding that the models underestimate the observed
137
precipitation trends (8, 24). By consistent, we mean that the observed trend lies within the 90%
138
confidence interval obtained by combining the uncertainty for the ensemble-mean forced model
139
trend with the uncertainty estimated from control runs. HIST-NAT trends are small and have the
140
opposite sign of all significant observed trends. In Fig. 3, increases in well mixed greenhouse gases
141
(based on HIST-GHG) are the main driver of significant changes in major climate types, but these
142
runs overstate most trends because they omit offsetting cooling factors such as sulfate aerosols.
143
The increase of arid and tropical climates squeezes the areas occupied by C and D climates in the
144
subtropics and midlatitudes. Although observations show weak contraction in C and weak expansion
145
of A climates, projections (Supplementary Fig. S2) indicate acceleration of both trends after about
146
2006 to significant rates by 2020, while significant expansion of B and shrinkage of midlatitude D
147
climates continue.
148
Conclusions
149
In summary, significant changes in major Köppen climate types since 1950 are detected and almost
7
150
completely attributed to anthropogenic factors. Our results demonstrate that anthropogenic climate
151
changes are already having a significant impact on ecosystems, which strengthens existing detection
152
and attribution evidence. When a climate threshold is passed, the existing vegetation is less suited for
153
the changed environment, although responses to a climate change have varying lag times (25-28).
154
The change in vegetation also can jeopardize the habitats of the wildlife that depends on those plant
155
communities.
156
Methods
157
This study uses updated thresholds (11) for five major climate classes (30 subtypes are not of
158
concern here) based on climate data at a location, with tests performed in the following order to
159
resolve conflicts: Arid (class B; MAP (mean annual precipitation, mm) <10×Pthreshold, where the
160
aridity threshold Pthreshold = 2 x MAT (mean annual temperature, °C) if 70% of MAP occurs in winter;
161
and Pthreshold = 2 x MAT + 28 if 70% of MAP occurs in summer; otherwise Pthreshold = 2 x MAT + 14),
162
Tropical (class A; Tcold (temperature of the coldest month) ≥ 18°C), Polar (class E; Thot (temperature
163
of the hottest month) < 10°C), Temperate (class C; Thot>10 and
164
Thot>10 and Tcold≤0).
165
For detection, attribution, and projection studies, monthly temperature and precipitation grids are
166
downloaded from CMIP5 model runs (23) using PI-CTL, HIST-ALL, HIST-GHG, HIST-NAT,
167
RCP4.5, and RCP8.5 scenarios (Supplementary Table S1). Years are arbitrary for PI-CTL runs.
168
HIST-ALL runs are driven by annual forcing values reconstructed from observed data (such as
169
greenhouse gas concentrations and actual volcanic eruptions), and this study downloads 1940-2005
170
monthly grids for runs based on all forcing factors (HIST-ALL), greenhouse gases only (HIST-GHG),
8
0<Tcold<18), Continental (class D;
171
and natural factors only (HIST-NAT). RCP runs have specified annual forcings projected for
172
2006-2100, and each run is initiated from a HIST-ALL run.
173
Grid data pretreatment involves three steps: (1) Observation and model outputs are regridded onto a
174
1°×1°grid to guarantee the same resolution. (2) For the historical period (1940-2005), anomalies
175
relative to monthly climatology for 1940-1960 are computed and are then added to the observational
176
1940-1960 monthly means. In addition, for the prediction period (2006-2100), anomalies relative to a
177
15-year climatology (historical for 1996-2005 and RCP for 2006-2010) are added to the
178
observational 1996-2010 monthly means. These downscaling steps ensure the consistency between
179
observation and simulations. This is necessary since the Köppen-Geiger scheme is quite sensitive to
180
thresholds and models have problems in simulating the present-day distribution. (3) Third, a 15-yr
181
running smoothing is applied to all variables in both observation and model (including PI-CTL) data
182
to reduce the probabilistic influence of shorter-term climate variability and elongate the period with
183
available data. The 15-year averaging period is chosen as an optimal smoothing interval for Köppen
184
or related classifications (29), but results here are not highly sensitive to the length of the averaging
185
period. There is a 7 year loss on each side of the time interval, so the 1950 starting year represents
186
the 1943-1957 average. Analysis periods end with 2003 (1996-2010) for observations, 1998
187
(1991-2005) for historical runs and 2093 (2086-2100) for RCP runs.
188
The major climate zones (A - tropical, B - arid, C - temperate, D - cold, and E - polar) are then
189
defined using the Köppen-Geiger scheme with updated criteria (11). If tests are applied in the order
190
B, A, E, C, D, simplified criteria as stated in the main text will always assign the correct climate class.
191
Three indices are then computed to depict changes in climate regions. For each major climate type,
9
192
the area index is the total area with that climate type, the latitude index is the global averaged
193
absolute latitude (so an increase is a poleward average movement of the climate type), and the
194
elevation index is the average altitude. Since one climate type is assigned to each 11°land grid box
195
(111111 km at the equator), all grid boxes are weighted by the cosine of their latitude to ensure that
196
equal areas are afforded equal weight. In Fig. 2b, the average latitude trend (°decade-1) is multiplied
197
by 111 to be expressed as km decade-1.
198
To test whether a trend is significantly larger than internal variability, model outputs from
199
pre-industrial control (PI-CTL) runs (pretreated as above, with 15-year averaging) are used to
200
estimate the standard deviation of each trend in an unchanging but naturally fluctuating climate.
201
Steps of estimation are as follows. 1) Major climate types are computed for each PI-CTL run. 2) A
202
54-year sample is taken every 27 years from the first year in each model run, giving a total of 451
203
sample time series. 3) For each sample, we calculate the percentage of the land surface experiencing
204
climate shifts compared to the first year, as well as linear trends of the other three indices (changes in
205
area, average absolute latitude, and average elevation of each climate type). Figure 1b shows the
206
distribution of percent of land area with changed major climate types with 54-year time separation
207
(from the first to the last year of each sample) for the 451 samples, and the darker gray shading in
208
Fig. 1a shows the 95th percentile of percentages of land area with changed major climate types with
209
time separations from 1 to 54 years. A trend is nonzero with 95% confidence if the percentage of
210
changed climate types exceeds the 95th percentile. For the other indices, a 2-sided significance test is
211
used assuming the variable is normally distributed, and the change is outside the range of ( ±1.96)
212
for 95% confidence or 5% significance. 4) For Supplementary Fig. S2, the estimated internal
213
variability (standard deviation) for each variable for trends starting in 1950 and ending in all years
10
214
through 2093 is obtained by following the same procedure with PI-CTL samples of length 2 to 144
215
years, and the 95% confidence range is shown by gray shading in each inset panel. 5) For Fig. 3, the
216
90% confidence interval of the trend (units decade-1) for each HIST group of runs, climate type, and
217
index is the 5th to 95th percentile range of trends based on all model run samples, and the trend is
218
significantly different from the observed trend if this 90% range does not include the observed trend.
219
Acknowledgements: We acknowledge all participants of CMIP5 for producing and making
220
available model outputs. This work is funded by the National Key Scientific Research Plan of China
221
(Grant 2012CB956002) and the National Natural Science Foundation of China (Grant 41075052).
222
References
223
224
225
1.
Hegerl, G. C., et al. Understanding and Attributing Climate Change. Climate Change 2007: The
Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change, Cambridge University Press (2007).
226
227
2.
Stott, P. A., et al. Detection and attribution of climate change: A regional perspective. WIREs
Clim. Change, 1, 192–211 (2010).
228
229
230
3.
Gillett, N. P., Arora V. K., Flato, J. M., Scinocca, J. F., and K. von Salzen, K. Improved
constraints on 21st-century warming derived using 160 years of temperature observations.
Geophys. Res. Lett., 39, L01704 (2012).
231
232
4.
Knutson, T. R., Zeng, R., and Wittenberg, A. T. Multi-model assessment of regional surface
temperature trends. J. Clim., 26, 8709-8743 (2013).
233
234
235
236
5.
Bindoff, N. L., et al. Detection and Attribution of Climate Change: from Global to Regional.
Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth
Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University
Press (2013).
237
238
6.
Zhang, X. B., et al. Detection of human influence on twentieth-century precipitation trends.
Nature, 448, 461–465 (2007).
239
240
7.
Min, S.-K., Zhang, X. B., and Zwiers, F. Human-induced arctic moistening. Science, 320,
518–520 (2008).
241
8.
Noake, K., Polson, D., Hegerl, G., and Zhang, X. B. Changes in seasonal land precipitation
11
during the latter twentieth‐century. Geophys. Res. Lett. 39, L03706 (2012).
242
243
244
9.
Polson, D., Hegerl, G. C., Zhang, X. B., and Osborn, T. J. Causes of robust seasonal land
precipitation changes. J. Clim. 26, 6679–6697 (2013).
245
246
10. Köppen, W., 1936. Das geographisca systemder Klimate. In: Köppen,W., Geiger, G. (Eds.),
Handbuch der Klimatologie. 1. C. Gebr, Borntraeger, pp. 1–44.
247
248
11. Peel, M. C., Finlayson, B. L., and McMahon, T. A. Updated world map of the Köppen-Geiger
climate classification. Hydro. Earth Sys. Sci. 11, 1633-1644 (2007).
249
250
12. Fraedrich, K, Gerstengarbe, F. W., Werner P. C. Climate shifts during the last century. Climatic
Change 50, 405–417 (2001).
251
252
13. Wang, M., Overland, J.E. Detecting Arctic climate change using Köppen climate classification.
Climatic Change 67, 43–62 (2004).
253
254
255
14. Gnanadesikan, A., and Stouffer, R. J. Diagnosing atmosphere-ocean general circulation model
errors relevant to the terrestrial biosphere using the Köppen climate classification. Geophys. Res.
Lett. 33, L22701 (2006).
256
257
258
15. De Castro, M., Gallardo, C., Jylha, K., and Tuomenvirta, H. The use of a climate-type
classification for assessing climate change effects in Europe from an ensemble of nine regional
climate models. Climatic Change 81, 329-341 (2007).
259
260
16. Diaz, H. F., Eischeid, J. K. Disappearing “alpine tundra” Köppen climatic type in the western
United States. Geophys Res Lett 34, L18707 (2007).
261
262
17. Roderfeld, H., et al. Potential impact of climate change on ecosystems of the Barents Sea Region.
Climatic Change 87, 283-303 (2008).
263
264
18. Rubel, F. & Kottek, M. Observed and projected climate shifts 1901_2100 depicted by world
maps of the Köppen_Geiger climate classification. Meteorol. Z. 19, 135-141 (2010).
265
266
19. Feng, S. et al.. Projected climate regime shift under future global warming from multi-model,
multi-scenario CMIP5 simulations. Global Planet. Change 112, 41-52 (2014).
267
268
20. Mahlstein, I., Daniel, J. S., and Solomon, S. Pace of shifts in climate regions increases with
global temperature. Nature Clim. Change 3, 739-743 (2013).
269
270
21. Legates, D. R., and Willmott, C. J. Mean seasonal and spatial variability in global surface air
temperature. Theor. Appl. Climatol. 41, 11-21 (1990).
271
272
22. Legates, D. R., and Willmott, C. J. Mean seasonal and spatial variability in gauge‐corrected,
global precipitation. Int. J. Climatol. 10, 111-127 (1990).
273
274
23. Taylor, K. E., Stouffer, R. J., and Meehl, G. A. An overview of CMIP5 and the experiment design.
Bull. Am. Meteorol. Soc. 93, 485-498 (2012).
12
275
276
24. Liu, C., Allan, R. P., and Huffman, G. J. Co-variation of temperature and precipitation in CMIP5
models and satellite observations. Geophys. Res. Lett. 39, L13803 (2012).
277
278
25. Arft, A. M., et al. Responses of tundra plants to experimental warming: meta-analysis of the
international tundra experiment. Ecol. Monogr. 69, 491-511 (1999).
279
280
26. Elmendorf, S. C. et al. Plot-scale evidence of tundra vegetation change and links to recent
summer warming. Nature Clim. Change 2, 453-457 (2012).
281
282
27. Gottfried, M., et al. Continent-wide response of mountain vegetation to climate change. Nature
Clim. Change 2, 111-115 (2012).
283
284
28. Epstein, H. E., Yu, Q., Kaplan, J. O., and Lischke, H. Simulating future changes in arctic and
subarctic vegetation. Computing in Science and Engineering, 9, 12-23 (2007).
285
286
29. Fraedrich, K., Gerstengarbe, F. W., and Werner, P. C. Climate shifts during the last
century. Climatic Change 50, 405-417 (2001).
287
Figure Legends
288
Figure 1. (a) Percentage of world land area that has experienced a climate type change from 1950
289
from observations (black line) and HIST-ALL (yellow), HIST-GHG (red), HIST-NAT (blue) runs.
290
Dark shading shows the 95th percentile of changed climate types relative to 1950 based on PI-CTL
291
simulations as illustrated in (b), and light shading shows the estimated 95th percentile if the standard
292
deviation is doubled. (b) Distribution of 54-year changed area percentage (2003 compared with 1950)
293
based on 451 PI-CTL samples (gray bars); vertical dashed line is mean.
294
Figure 2. (a) Linear trends in areas of 5 major climate types for 1950-2003; asterisks denote
295
significant trends at the 5% significance level. A positive trend of high-latitude (north of 55°N) D
296
climate and a negative midlatitude (south of 55°N) D climate are over-plotted in blue with the net
297
negative trend of D climate in dark blue. (b) and (c) are the same as (a), but for trends in average
298
absolute latitude (positive indicates poleward) and elevation, respectively. (d) Map showing grid
299
boxes with major climate type in 1950 that “disappeared” (changed to another type) by 2003. Colors
300
are the same as in upper panels. (e) is the same as (d), but for “emerging” climate types (by 2003) in
301
the same grid boxes.
13
302
Figure 3. Significant observed trends (black bars) marked with * in Fig. 2 and the corresponding
303
simulated trends of indices for 1950-1998; yellow, red and blue bars denote HIST-ALL, HIST-GHG
304
and HIST-NAT runs, respectively. The black lines are 5-95% confidence intervals estimated from the
305
distribution of trends in the control. Simulated trends significantly different from the observation are
306
marked with diamonds. The units are 2×105 m2 decade-1 for area, 10 km decade-1 for latitude and 5
307
m decade-1 for elevation indices.
308
Figures
309
310
Figure 1
14
311
312
Figure 2
313
314
Figure 3
315
15
316
Supplementary Information
317
Supplementary Figure S1. Analysis of factors causing changes in climate types in historical period
318
from 1950 to 2003 using observed data (a to e) or to 1998 using HIST-ALL (f) or HIST-GHG (g)
319
model runs, with each stated year representing a 15-year centered average. (a to c) Same as in Fig. 2
320
with the left bar for each climate type the same as in Fig. 2, but the middle (red edge) bars are
321
temperature-induced changes (climate types derived with 2003 temperature and 1950 precipitation
322
values), and the right (blue edge) bars are precipitation-induced changes (2003 precipitation and
323
1950 temperature values). An asterisk denotes that a trend is significant at 95% confidence level. (d)
324
As in Fig. 2e, showing grid boxes with changed 2003 climate types when precipitation is held at
325
1950 values. (e) As in Fig. 2e, showing changed 2003 climate types when temperature is held at 1950
326
values. (f and g) Changed grid box climate types in 1998 from 1950 using multi-model averages of
327
HIST (f) and HIST-GHG (g) runs.
328
Supplementary Figure S2. Temporal evolution of each index and each major climate type, for the
329
historical period (1950-2003) and projections (2003-2093). In the historical period, the solid black
330
line is the observation; the dashed black line, solid red line and solid blue line denote multi-model
331
averaged indices derived from HIST-ALL, HIST-GHG and HIST-NAT experiments, respectively. In
332
the projection period, the solid orange and magenta lines are ensemble mean indices from RCP8.5
333
and RCP4.5 scenarios. Light orange and magenta shadings are corresponding estimated 95%
334
cross-model uncertainties. Units are km2, degrees (°) and m for area, absolute latitude and elevation
335
indices. The inset figure in each panel is the temporal evolution of the linear trend (units decade-1)
336
since 1950 over the entire observation and projection period. Orange and magenta lines denote
337
RCP8.5 and RCP4.5 predictions, respectively. Gray shadings are estimated range of long-term linear
338
trends in a natural fluctuating climate as a function of time span at the 95% confidence level.
16
339
340
Figure S1
17
341
342
Figure S2
343
18
344
345
346
347
348
Supplementary Table S1. CMIP5 model runs used in this study. Numbers are years of length of the
preindustrial control (PI-CTL) run, and number of ensemble members for historical runs with all
forcings (HIST-ALL), historical with greenhouse gas forcing (HIST-GHG), historical with natural
forcing (HIST-NAT), and RCP4.5 and RCP8.5 (radiative forcing of 4.5 or 8.5 W m -2 projected in
2100 relative to preindustrial) experiments.
349
MODELS
PI-CTL
(years)
ACCESS1.0
500
ACCESS1.3
500
BCC-CSM1.1
500
BCC-CSM1.1(m)
400
CanESM2
700
CCSM4
500
CESM1-CAM5
300
CMCC-CESM
200
CMCC-CM
200
CNRM-CM5
500
CSIRO-Mk3.6.0
500
FGOALS-g2
-FIO-ESM
-GFDL-CM3
500
GFDL-ESM2G
500
GFDL-ESM2M
500
GISS-E2-H
500
GISS-E2-R
500
HadGEM2-AO
-HadGEM2-CC
-HadGEM2-ES
300
INM-CM4
-IPSL-CM5A-LR
1000
IPSL-CM5A-MR
300
IPSL-CM5B-LR
300
MIROC-ESM
500
MIROC-ESM-CHEM
200
MIROC5
600
MPI-ESM-LR
1000
MPI-ESM-MR
1000
MRI-CGCM3
500
NorESM1-M
500
SUM
13500
HIST
-ALL
2
3
3
3
5
5
3
1
1
5
5
5
3
5
1
1
5
5
1
1
5
1
5
3
1
3
1
5
3
3
3
3
99
350
19
HIST
-GHG
-2
1
-5
3
---5
5
1
-3
-1
5
5
--4
-3
3
-3
1
---1
1
52
HIST
-NAT
--1
-5
4
---5
5
3
-3
-1
5
5
--4
-3
3
-3
1
---1
1
53
RCP4.5 RCP8.5
1
1
1
1
5
5
3
-1
1
5
1
1
1
1
1
5
5
1
1
4
1
4
1
1
1
1
3
3
3
1
1
65
1
1
1
1
5
5
3
1
1
3
5
1
1
1
1
1
2
2
1
1
4
1
4
1
1
1
1
3
3
1
1
1
60