hoerling pdsi drought revisedfinal jclim...! 6! 140! because! projected! increases! in!...
TRANSCRIPT
1
Is a Transition to Semi-Permanent Drought Conditions Imminent in the U.S. 1 Great Plains? 2
3 4 5
Martin P. Hoerling1, Jon K. Eischeid2, Xiao-‐Wei Quan2, Henry F. Diaz1, Robert S. 6 Webb1, Randall M. Dole1, and David Easterling3 7
8 9 10
1NOAA Earth System Research Laboratory, Boulder, Colorado 11 12
2 University of Colorado, Cooperative Institute for Research in Environmental Sciences, 13 Boulder, Colorado 14
3 NOAA National Climatic Data Center, Asheville, North Carolina 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
Submitted to J. Climate 30 31
Revised 32 33
October 11 2012 34 35 36 37 38
39 Corresponding author address: M. Hoerling, NOAA/Earth System Research 40 Laboratory, 325 Broadway, Boulder CO 80305. 41 E-‐mail: [email protected] 42 43 44 45 46
2
47 ABSTRACT 48
49 50
How Great Plains climate will respond under global warming continues to be 51
a key unresolved question. There has been, for instance, considerable 52
speculation that the Great Plains is embarking upon a period of increasing 53
drought frequency and intensity that will lead to a semi-‐permanent Dust 54
Bowl in coming decades. This view draws on a single line of inference of how 55
climate change may affect surface water balance based on sensitivity of the 56
Palmer Drought Severity Index (PDSI). A different view foresees a more 57
modest climate change impact on Great Plains surface moisture balances. 58
This draws on direct lines of analysis using land surface models to predict 59
runoff and soil moisture, the results of which do not reveal an ominous fate 60
for the Great Plains. Our study presents a parallel diagnosis of projected 61
changes in drought as inferred from PDSI and soil moisture indicators in 62
order to understand causes for such a disparity, and to shed light on the 63
uncertainties. PDSI is shown to be an excellent proxy indicator for Great 64
Plains soil moisture in the 20th Century; however, its suitability breaks down 65
in the 21st Century with the PDSI severely overstating surface water 66
imbalances and implied agricultural stresses. Several lines of evidence and 67
physical considerations indicate that simplifying assumptions regarding 68
temperature effects on water balances especially concerning 69
evapotranspiration in Palmer’s formulation compromise its suitability as 70
drought indicator in a warming climate. We conclude that projections of 71
3
acute and chronic PDSI decline in the 21st Century are likely an exaggerated 72
indicator for future Great Plains drought severity. 73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
4
1. Introduction 95
96
The future trajectory of climate change in the U.S. Great Plains (30°N-‐50°N, 105°W-‐97
95°W) and in particular the frequency and intensity of future drought is a matter of 98
continuing interest and debate given its past history of severe drought episodes 99
such as the 1930’s “Dust Bowl”. While some have raised the specter of a shift to 100
semi-‐permanent 1930’s type drought conditions on the Great Plains due to human-‐101
induced global warming, the Special Report of the Intergovernmental Panel on 102
Climate Change regarding extreme events (IPCC 2012) expresses only low 103
confidence in a projected change in drought over the U.S. Great Plains as a whole, 104
and medium confidence for some increased dryness across the southern portion of 105
the domain. 106
107
Why such substantial differences in expectations for drought in the Great Plains, and 108
why the overall low level of confidence in drought projections for this region as a 109
whole? A possible explanation is an inconsistency in trends for surface water 110
balances among analyses that use different drought indices. One approach has been 111
to examine the sensitivity of the Palmer Drought Severity Index (PDSI; Palmer 112
1965) to projected changes in monthly temperature and precipitation (e.g., Rind et 113
al. 1990; Jones et al. 1996; Burke et al. 2006; Burke and Brown 2008; Dai 2011, 114
2012; Burke 2011; Wehner et al. 2011). Results from these studies, though derived 115
from widely different generations of climate models, share the common feature that 116
5
PDSI quickly approaches severe drought values with increasing greenhouse gas 117
forcing. 118
119
Rind et al. (1990) suggested that increasing drought over the U.S. would become 120
apparent in the 1990s, while Dai (2011) argues that the U.S. might see persistent 121
droughts in the next 20-‐50 years. The spatial pattern of U.S. drought increase was 122
found to be highly reproducible among the 19 models contributing to CMIP3, largely 123
because the response of PDSI to future temperature increase is very robust (Dai 124
2011, 2012; Wehner et al. 2011). Wehner et al. demonstrate that climate models 125
having the largest projected temperature increase yield the largest increase in 126
drought severity due to the strong dependency of evapotranspiration on surface air 127
temperature in PDSI. Although this temperature sensitivity depends qualitatively on 128
the formula used to compute potential evapotranspiration (e.g. Chen et al. 2005; 129
Burke et al. 2006), Wehner et al. (2011) surmise that the rapid and severe decline of 130
PDSI under the influence of future warming should be indicative of decreases in soil 131
moisture, regardless of how mean precipitation may change. However, their 132
supposition has not been confirmed by analyses of projected soil moisture changes 133
simulated directly by climate model land surface schemes. 134
135
Alternatively, annual mean soil moisture in CMIP3 models is projected to decline in 136
some areas by the end of the 21st century, though not significantly so over the 137
northern Great Plains (Sheffield and Wood, 2008; Orlowsky and Seneviratne 2011). 138
In some regions, such as the northern Great Plains, soil moisture does not change 139
6
because projected increases in precipitation are offset by a warming-‐induced 140
increase in evapotranspiration. Consistent with these changes, surface runoff 141
projections do not exhibit appreciable change (2041–60 compared to 1901–70) for 142
the Missouri and upper Mississippi River drainage basins in CMIP3 models (Milly et 143
al. 2005). For the Great Plains as a whole, Sheffield and Wood’s (2008) diagnosis of 144
simulated soil moisture reveals no statistically significant change in the frequency of 145
short-‐term drought (4-‐6 month duration) for any emissions scenario (B1, A1B, A2) 146
through 2100, and the earliest year for detecting a statistically significant change in 147
the frequency of long-‐term (>12 months) drought is about 2050. More recently, 148
Winter and Eltahir (2012) analyzed regional climate model projections for the 149
Midwest U.S. (an area slightly east of the Great Plains region considered herein) and 150
found no reductions in future summer soil moisture, with increased 151
evapotranspiration balanced by increased precipitation. Thus, questions remain on 152
how Great Plains drought will respond to global warming. To better understand 153
uncertainty sources, we present a side-‐by-‐side analysis of projected changes in 154
drought using both a PDSI and soil moisture metric. 155
156
2. Data and Methods 157
a. Climate model simulations 158
We use historical simulations and climate projections based on the fourth version of 159
the Community Climate System Model (CCSM4; Gent et al. 2011). CCSM4 is one of 160
numerous models that comprise the Coupled Model Intercomparison Project-‐Phase 161
5 experiments (CMIP5). We analyze monthly output from an ensemble of three 162
7
CCSM4 simulations forced with variations in greenhouse gases, aerosols, time-‐163
varying solar irradiance, and the radiative effects of volcanic activity for 1850–2005 164
(Taylor et al. 2012). Projections of climate conditions for 2006–2100 are based upon 165
the Representative Concentration Pathway 8.5 for individual greenhouse gases and 166
aerosols (Moss et al. 2010). 167
168
The Community Land Model Version 4 (CLM4) used in CCSM4 represents a 169
significant advance over prior versions (Lawrence et al. 2011; Lawrence et al. 2012). 170
Advances include improved canopy treatment, a plant functional type dependency 171
on soil moisture stress function, improved surface hydrology and runoff, and 172
improved representation of evapotranspiration that provides more realistic 173
treatment of plant transpiration, evaporation from soils, and canopy evaporation. 174
The land model includes a prognostic carbon-‐nitrogen cycle with time varying 175
vegetation phenology. Soil water is predicted from a multi-‐layer model in which 176
time variability in soil moisture is a function of infiltration, runoff, gradient 177
diffusion, gravity, canopy transpiration through root extraction, and interactions 178
with ground water (Oleson et al. 2010). Over the U.S. Great Plains region, the land 179
model resolves 10 soil layers to ~3m, with five additional bedrock layers to a depth 180
of ~35m. 181
182
b. Drought indices 183
The monthly Palmer Drought Severity Index (PDSI, Palmer 1965) is calculated from 184
the CCSM4 archive of monthly temperature and precipitation. The potential 185
8
evapotranspiration (PET) is calculated using the Thornthwaite (1948) formulation, 186
which depends upon only temperature and latitude. Similar approaches were used 187
in the recent studies of Dai (2011, 2012) and Wehner et al. (2011). Burke et al. 188
(2006) and Dai (2011, 2012) also calculated PET using the Penman-‐Monteith 189
equation (Shuttleworth 1993), which while providing a more realistic estimate of 190
the meteorological control on PET, continues to be strongly temperature driven 191
(Fennessey and Kirshen, 1994). A recognized limitation of PDSI as an index for 192
drought over semi-‐arid regions including portions of the Great Plains is its non-‐193
Gaussian distribution with a tendency to yield more severe negative PDSI values. 194
We used the statistical normalization method of Wells et al. (2004) to re-‐calibrate 195
the PDSI which has the effect of altering the statistical distribution of extreme wet 196
and dry occurrences in some areas so as to be more consistent with the expected 197
frequency of rare events, to allow for more accurate comparisons of the index across 198
regions, and to facilitate intercomparison with other standardized drought indices. 199
Here, the parameters for calculating the PDSI values and the subsequent 200
calibrations are determined from the monthly model output of 1901–2000 for each 201
of the three simulations separately, and the calibrated PDSI values are obtained for 202
the entire 1850-‐2100 period for each run using the respective set of parameters. 203
204
The second drought index is derived from the CCSM4 monthly column integrated 205
soil moisture for 3 model layers having a total of ~10 cm depth, normalized, also 206
using the standard deviation of its annual mean during 1901-‐2000. Sensitivity 207
9
analyses indicated that our results do not differ materially when using a shallower 1 208
cm or deeper 1 m soil layer. 209
210
3. Results 211
Individual PDSI traces averaged over the Great Plains (30°N–50°N, 105°W–95°W) 212
for the period 1850–2100 for each of three model integrations (gray curves) and the 213
3-‐run ensemble average (yellow curve) are shown in Figure 1. Two significant 214
features of these time series are: i) a progressive reduction in PDSI commencing in 215
the late 20th Century and accelerating in the 21st Century occurring in all simulations 216
due to increasing GHG forcing, and ii) a randomly occurring decadal-‐long drought 217
event during the late 19th Century in one simulation due to internal model 218
variability. 219
The randomly occurring severe drought in the late 19th Century simulation has 220
many of the meteorological characteristics of severe drought over the Great Plains 221
during the 1930s. Figure 2 (left panels) shows the spatial patterns of decadally 222
averaged climate and land surface conditions, a period during which the simulated 223
PDSI (1865-‐1875) was consistently indicating drought. Negative index values of the 224
PDSI span the Great Plains region from North Dakota to Texas (top panel, left), 225
similar to the observed 1930s pattern (e.g. Hoerling et al. 2009). The pattern of soil 226
moisture deficits (second panel, left) largely mimics the indications for drought 227
provided by PDSI. Further, the spatial pattern and intensity of drought given by both 228
10
indicators appear largely determined by the simulated precipitation deficits (third 229
panel, left). Widespread reductions in rainfall with average annual deficits of 1 230
standardized departure span much of the Great Plains, similar to the magnitude of 231
precipitation deficits observed during the decade of the Dust Bowl (e.g. Schubert et 232
al. 2004; Hoerling et al. 2009). Figure 2 (lower left) also indicates that the drought 233
region experienced elevated surface temperatures, though generally less than 234
+0.5°C. While such warmth is consistent with an overall inverse relationship 235
between rainfall and temperature in CCSM4 (not shown) analogous to that 236
occurring in observations (e.g. Madden and Williams 1978), neither the pattern of 237
PDSI nor soil moisture indicators of this simulated drought event were determined 238
strongly by the temperature conditions. Overall, this randomly occurring moisture 239
deficiency over the Great Plains is well described by PDSI using only monthly 240
temperature and precipitation. 241
242
In contrast, under projected future climate change, modeled soil moisture and PDSI 243
respond very differently. Spatial maps of conditions by mid-‐21st century (Fig. 2, 244
right panels) reveal a North American-‐wide pattern of much reduced PDSI (top 245
panel) implying that drought conditions exceed both the scale and intensity of that 246
occurring during the Dust Bowl era. The PDSI pattern deviates radically from the 247
spatial pattern of projected soil moisture departures (second panel). Whereas the 248
soil moisture departures are more coherent with the simulated pattern of rainfall 249
change (third panel), the PDSI values are more coherent with the simulated pattern 250
11
of temperature change (bottom panel). 251
252
In addition, the simulated time series of Great Plains meteorological and land 253
surface conditions (Fig. 3) reveal that soil moisture exhibits a decline during the 21st 254
century, but the magnitude of land surface drying is less than a standard deviation 255
of typical interannual variability. By contrast, PDSI declines on order of 5 to 10 256
times its standard deviation by the latter decades of the 21st century. The 257
comparatively modest amount of Great Plains soil moisture depletion in CCSM4 is 258
consistent with the multi-‐model results based on the land surface models of CMIP3 259
(e.g. Sheffield and Wood 2008). This relatively small signal of soil moisture 260
reduction is consistent with the limited detectability for a change in drought 261
frequency over the Great Plains until the latter portions of the 21st Century as noted 262
in other studies and assessments (Sheffield and Wood, 2008; Orlowsky and 263
Seneviratne 2011; IPCC 2012). Nonetheless, such soil moisture deficits sustained 264
over a long period could have appreciable consequences. 265
266
A physically based understanding for these dramatic differences in drought 267
projections is provided by comparing Palmer and CCSM4 projected 268
evapotranspiration (ET) changes. The ET in Palmer’s formulation (Fig. 3, yellow 269
curve) initially increases rapidly with the projected warming, with the moisture loss 270
determined largely by the increase in PET-‐driven demand, the latter being a close 271
proxy for the magnitude of projected temperature rise (compare red and orange 272
12
curves in Fig. 3). Since ET’s upper bound is dictated by precipitation and available 273
soil moisture, a decline in the rate of ET rise in Palmer’s model ensues in the latter 274
half of the 21st century as the bucket model’s soil moisture becomes substantially 275
depleted. At this point in Palmer’s projections, PDSI falls to unprecedented negative 276
values. By contrast, ET in CCSM4 (Fig. 3, purple curve) does not increase nearly as 277
much as ET in Palmer's model. The overall modest positive ET departure in CCSM4 278
projections is consistent with its projected modest soil moisture decline, whereas 279
higher frequency variability of ET in CCSM4 is mostly driven by fluctuations in 280
precipitation. 281
282
Finally, the simulations indicate that co-‐variability of annual departures in Great 283
Plains PDSI and soil moisture drastically changes in the warming 21st Century 284
climate. Figure 4 compares the relationship during the 20th Century (blue circles) 285
with that during the 21st Century (red circles) based on diagnosis of all three CCSM4 286
simulations. PDSI explains about 40% of the inter-‐annual variations in soil moisture 287
during the 20th Century, but explains only about 20% of the variance during the 21st 288
Century. A -‐1 PDSI value (an indication for mild drought) corresponds to about a -‐1 289
standardized departure of soil moisture during the 20th Century; however, a -‐6 PDSI 290
value (which is far beyond the range of the most extreme PDSI value during the 291
calibration period) also corresponds to a -‐1 standardized departure of soil moisture 292
during the 21th Century. The magnitude of the PDSI during the projection period of a 293
warming climate thus ceases to be a reliable indicator of land surface water 294
13
deficiencies and implied drought severity, at least within the simulations of CCSM4. 295
296
4. Conclusions and Discussion 297
Assessments of how climate change may affect the frequency and severity of 298
drought need to consider various drought indicators (e.g., IPCC 2012). Two such 299
indicators, PDSI and the soil moisture anomalies each present very different views 300
of how surface water balances may evolve over the Great Plains under global 301
warming. To understand these differences, we analyzed both drought indicators 302
using CCSM4 simulations. Our analysis reproduces disparate outlooks for Great 303
Plains drought—projections using the PDSI suggest that the Great Plains will be in a 304
semi-‐permanent state of severe Dust Bowl-‐like drought in coming decades, whereas 305
soil moisture projections reveal modest drying and comparatively low detectability 306
for changes in Great Plains drought frequency. Our analyses illustrate that the PDSI 307
exaggerates future drought severity over the Great Plains due to unrealistic 308
sensitivity to the projected warming of surface temperatures. 309
310
Scale analysis by Hu and Wilson (2000) shows the PDSI is about equally affected by 311
temperature and precipitation anomalies having similar magnitude, but that the 312
temperature sensitivity during the 20th Century is generally masked by the strong 313
negative correlation between temperature and rainfall variability. A GHG-‐forcing 314
change in surface temperature without a corresponding change in rainfall leads to 315
14
projected values of PDSI that are effectively proxies for the projected increases in 316
temperature, but are not suitable indicators for soil moisture. The simple 317
parameterization of potential evapotranspiration using a Thornthwaite formula is 318
not appreciably remedied using a more sophisticated Penman-‐Montieth formulation 319
(e.g. Dai 2011, 2012). Estimates of evapotranspiration based solely on temperature 320
have been regarded as problematic (e.g. Lockwood 1999; Burke et al. 2006; Burke 321
2011). Palmer’s formula for how PET relates to a specific location’s radiation 322
climate, as inferred from temperature alone, becomes violated in the presence of 323
appreciable warming and leads to unrealistic extreme drought indications 324
(Lockwood 1999). Additionally, the sensitivity of the PET calculation is strongly 325
affected by the choice of the calibration period (Karl 1986). 326
327
The PDSI was developed to define drought using a minimum number of widely 328
observed variables and provides a good snapshot of observed drought conditions in 329
a variety of climates. However Dai’s (2011) review of factors affecting drought 330
raised several limitations of PDSI under global warming. Dai noted that the Palmer 331
formula overestimates potential evapotranspiration under a warming climate, and, 332
“the fact that they [the PDSI index] may not work for the 21st century climate itself is 333
a troubling sign”. While past performance has been useful in appreciating the value 334
of the PDSI, it is no guarantee for future success, especially when applied to the non-‐335
stationary climate system currently unfolding. 336
337
15
Our result for severe drought indication over the Great Plains based on PDSI is 338
entirely consistent with a host of prior PDSI-‐based drought projections but derived 339
from an earlier generation of climate models (e.g., Dai 2011, Wehner et al. 2011). 340
The similar results are due to the robustness in projected surface warming 341
combined with the strong dependency of PDSI on temperature, rather than the 342
soundness of Palmer’s formulation of the complex surface water balance. Palmer’s 343
drought index avoids the complicating biological factors and the complex processes 344
of soil-‐vegetation–atmosphere transfer, which are essential in representing how 345
land surface moisture balances evolve in a warming climate. With the advent of 346
model-‐based reanalyses, and the need to examine the GHG-‐forced climate response 347
of drought, it is clear that drought indices need to be developed that incorporate all 348
relevant variables at the proper time resolution to consistently define drought 349
conditions across the full range of climates. 350
351
An extensive comparison of the Thornthwaite method with pan-‐based estimates of 352
evapotranspiration in China (Chen et al. 2005) illustrates severe deficiencies that 353
are only somewhat alleviated by properly calibrated Penman-‐Monteith (PM) 354
estimates. However, we stress the fact that while the use of the PM formulation to 355
estimate PET reduces the bias associated with the strong temperature dependence 356
in Palmer’s formulation (see e.g., Burke et al. 2006), the impact of other 357
simplifications remain. For instance, the PM formulation lacks a land surface model 358
with plant phenology from which transpiration is physically derived, and still uses 359
16
simple accounting assumptions of how ET balances either precipitation or PET in 360
Palmer’s water balance calculations. 361
362
In the context of our study, it is important to consider the implications that a non-‐363
stationary, warming climate has for misinterpreting drought severity based on 364
model projections of PDSI. The limitations are: 1) the absolute value of PDSI should 365
be viewed as an unreliable measure of drought severity in a warming climate (and 366
for these same reasons in an extreme heat wave event like over the Great Plains in 367
2012), and 2) inferences of possible amplifying effects of temperature alone 368
(separate from precipitation effects) on drought severity derived from analysis of 369
Palmer’s model should likewise be viewed as unreliable. For real-‐time drought 370
monitoring the first limitation could be mitigated by annually updating the 371
calibration of the Palmer equations and their coefficients using the latest data. By 372
updating to include the data through 2011–12, one can derive a more representative 373
estimate of the "climatological water balance", given the known appreciable 374
variability and emergent trends; however, the problem in using PDSI for assessing 375
future drought conditions in climate model projections remains. 376
377
So is a transition to semi-‐permanent drought imminent for the Great Plains? While 378
our results cannot rule out the possibility of a Dust Bowl like-‐period in the near 379
future, the physical processes for such an occurrence are likely to be related to 380
prolonged deficiencies in Great Plains precipitation, rather than to the local 381
temperature effect of projected surface warming alone.382
17
383
References 384
385
Burke, E. J., 2011: Understanding the sensitivity of different drought metrics 386
to the drivers of drought under increased atmospheric CO2. J. Hydromet., 12, 387
1378–1394. 388
389
Burke, E. J., and S. J. Brown, 2008: Evaluating uncertainties in the projection 390
of future drought. J. Hydrometeor., 9, 292–299. 391
392
Burke, E. J., S. J. Brown, and N. Christidis, 2006: Modeling the recent evolution 393
of global drought and projections for the 21st century with the Hadley Centre 394
climate model. J. Hydromet., 7, 1113–1125. 395
396
Chen, D., G. Gao, C.-‐Y. Xu, J. Guo, and G. Ren, 2005: Comparison of the 397
Thornthwaite method and pan data with the standard Penman-‐Monteith 398
estimates of reference evapotranspiration in China. Climate Research, 28, 399
123–132. 400
401 402
Dai, A., 2011: Drought under global warming: A review. Wiley 403
Interdisciplinary Reviews: Climate Change, DOI: 10.1002/wcc.81. 404
405
18
Dai, A., 2012: Increasing drought under global warming in observations and 406
models. Nature Climate Change, DOI: 10.1038/NCLIMATE1633. 407
408
Fennessey, N.M. and P. H. Kirshen, 1994. Evaporation and evapotranspiration 409
under climate change in New England. ASCE J. Water Resour. Plann. Mgmt., 410
120, 48–69. 411
412
Gent, Peter R., and Coauthors, 2011: The Community Climate System Model 413
Version 4. J. Climate, 24, 4973–4991. 414
415
Hoerling, M., X. Quan, and J. Eischeid (2009), Distinct causes for two principal 416
U.S. droughts of the 20th century, Geophys. Res. Lett., 36, L19708, 417
doi:10.1029/2009GL039860. 418
419
Hu, Q., and G. D. Willson, 2000: Effects of temperature anomalies on the 420
Palmer Drought Severity Index in The Central United States. Intl. J. Climatol., 421
20, 1899–1911. 422
423 IPCC, 2012: Managing the Risks of Extreme Events and Disasters to Advance 424
Climate Change Adaptation. A Special Report of Working Groups I and II of 425
the Intergovernmental Panel on Climate Change [Field, C.B., V. Barros, T.F. 426
Stocker, D. Qin, D.J. Dokken, K.L. Ebi, M.D. Mastrandrea, K.J. Mach, G.-‐K. 427
Plattner, S.K. Allen, M. Tignor, and P.M. Midgley (eds.)]. Cambridge University 428
Press, Cambridge, UK, and New York, NY, USA, 582 pp. 429
19
Jones, P., M. Hulme, K. Briffa, and C. Jones, 1996: Summer moisture 430
availability over Europe in the Hadley Centre general circulation model 431
based on the Palmer Drought Severity Index. Int. J. Climatol., 16, 155–172. 432
433
Karl. T., 1986: The sensitivity of the Palmer Drought Severity Index and 434
Palmer's Z-‐Index to their calibration coefficients including potential 435
evapotranspiration. J. Climate Appl. Meteor., 25, 77–86. 436
437
Lawrence, D., and Coauthors, 2011: Parameterization improvements and 438
functional and structural advances in version 4 of the Community Land 439
Model. J. Adv. Model. Earth Syst., 3, doi:10.1029/2011MS000045. 440
441
Lawrence, D., and Coauthors, 2012: The CCSM4 Land Simulation, 1850-‐2005: 442
Assessment of surface climate and new capabilities. J. Climate, 25, 2240– 443
2260. 444
445
Lockwood, J. G., 1999: Is potential evapotranspiration and its relationship 446
with actual evapotranspiration sensitive to elevated atmospheric CO2 levels? 447
Climatic Change, 41, 193–212. 448
449
Madden, R., and J. Williams, 1978: The correlation between temperature and 450
precipitation in the United States and Europe. Mon. Wea. Rev., 106, 142–147. 451
20
452
Milly, P., K. Dunne, and A. Vecchia, 2005: Global pattern of trends in 453
streamflow and water availability in a changing climate. Nature, 438, 347–50. 454
455
Moss, R. H., J. A. Edmonds, K. A. Hibbard, M. R. Manning, S. K. Rose, D. P. van 456
Vuuren, T. R. Carter, S. Emori, M. Kainuma, T. Kram, G. Meehl, J. F. B. Mitchell, 457
N. Nakicenovic, K. Riahi, S. J. Smith, R. J. Stouffer, A. M. Thomson, J. P. Weyant, 458
T. J. Wilbanks, 2010: The next generation of scenarios for climate change 459
research and assessment. Nature, 463, 747–756. 460
461
Oleson and Coauthors, 2010: Technical Description of version 4.0 of the 462
Community Land Model (CLM). NCAR Technical Note TN-‐478+STR, 238 pp. 463
464
Orlowsky, B., and S. Seneviratne, 2011: Global changes in extreme events: 465
regional and seasonal dimensions. Clim. Change. DOI: 10.1007/s10584-‐011-‐466
0122-‐9. 467
468
Palmer, W. C., 1965: Meteorological drought. Research Paper 45, U.S. Dept. of 469
Commerce, 58 pp. 470
471
Rind, D., R. Goldberg, J. Hansen, C. Rosenzweig, and R. Ruedy, 1990: Potential 472
evapotranspiration and the likelihood of future drought. J. Geophys. Res., 95, 473
21
9983-‐10,004. 474
475
Schubert, S., M. Suarez, P. Pegion, R. Koster, and J. Bacmeister, 2004: On the 476
cause of the 1930s dust bowl. Science, 303, 1855–1859 DOI: 477
10.1126/science.1095048 478
479 480
Sheffield, J., and E. Wood, 2008: Projected changes in drought occurrence 481
under future global warming from multi-‐model, multi-‐scenario, IPCC AR4 482
simulations. Clim. Dyn., 31, 79-‐105. 483
484
Shuttleworth, W., 1993: Evaporation. Handbook of Hydrology. D. R. Maidment, 485
Ed., McGraw-‐Hill, 4.1–4.53. 486
487
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: The CMIP5 Experiment 488
Design. Bull. Amer. Meteorol. Soc., in review. (A summary of the CMIP5 489
Experimental Design is available at http://cmip-‐pcmdi.llnl.gov/cmip5/ 490
docs/Taylor_CMIP5_design.pdf). 491
492
Thornthwaite, C., 1948: An approach toward a rational classification of 493
climate. Geogr. Rev., 38, 55-‐94. 494
495
22
Wehner, M., D. Easterling, J. Lawrimore, R. Heim, R. Vose, and B. Santer, 2011: 496
Projections of future drought in the continental Unites States and Mexico. 497
J. Hydromet., 12, 1359-‐1377. 498
499
Wells, N., S. Goddard, and M. Hayes, 2004: A Self-‐Calibrating Palmer Drought 500
Severity Index. J. Climate, 17, 2335-‐2351. 501
502
Winter, J.M., and E.A.B. Eltahir, 2012: Modeling the hydroclimatology of the 503
midwestern United States. Part 2: Future climate. Clim. Dyn., 38, 595-‐611, 504
doi:10.1007/s00382-‐011-‐1183-‐1. 505
506
507
508
509
510
511
512
513
514
515
516
517
518
23
Figure Captions 519
520
Figure 1. Annual time series of PDSI averaged over the Great Plains (30°N–50°N, 521
105°W–95°W) for 1851–2100 (the calibration period is 1901–2000). Individual 522
traces correspond to each of the three CCSM4 model integrations (gray curves) 523
while the 3-‐run ensemble average is shown by the yellow curve. The time series are 524
smooth with a 9 point Gaussian filter. 525
526
Figure 2. Spatial patterns of decadally averaged land surface conditions for PDSI 527
(top row), 10cm soil moisture (second row), precipitation (third row), and 528
temperature (bottom row) for 1865–1875 (left panel) and 2045–2055 (right panel) 529
for the CCSM4 model integrations. Anomalies are computed from the 1901–2000 530
reference period. The 1865–1875 results are based on a single model run, and the 531
results for 2045–2055 are based on the 3-‐run ensemble average. Box in lower left 532
panel denotes the Great Plains domain over which spatial averages are computed 533
for time series analysis in Figs. 1 and 3, and scatter plot relationships in Fig. 4. 534
535
Figure 3. Standardized annual time series of Great Plains surface temperature (red 536
curve), evapotranspiration (purple), precipitation (green), soil moisture (blue), and 537
PDSI (yellow) as simulated in the ensemble of CCSM4 forced integrations. Also 538
shown are the Palmer-‐model derived annual time series of evapotranspiration 539
(yellow) and potential evapotranspiration (orange). The reference period for the 540
24
standardization is 1901–2000 and each of the curves is smoothed with a 9 point 541
Gaussian filter. 542
543
Figure 4. Comparison of the relationship between annual 10-‐cm soil moisture and 544
PDSI for the 1901–2000 period (blue circles, R=+0.61) with that of the 2001–2100 545
period (R=+0.48) based on the diagnosis of all three CCSM4 simulations. 546
547
548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569
25
570 571 Figure 1. Annual time series of PDSI averaged over the Great Plains (30°N–50°N, 572 105°W–95°W) for 1850–2100 (the calibration period is 1901-‐2000). Individual 573 traces correspond to each of the three CCSM4 model integrations (gray curves) 574 while the 3-‐run ensemble average is shown by the yellow curve. The time series are 575 smooth with a 9 point Gaussian filter. 576 577
26
578 579 Figure 2. Spatial patterns of decadally averaged land surface conditions for PDSI 580 (top row), 10-‐cm soil moisture (second row), precipitation (third row), and 581 temperature (bottom row) for 1865–1875 (left panel) and 2045–2055 (right panel) 582 for the CCSM4 model integrations. Anomalies are computed from the 1901–2000 583 reference period. The 1865–1875 results are based on a single model run, and the 584 results for 2045–2055 are based on the 3-‐run ensemble average. Box in lower left 585 panel denotes the Great Plains domain over which spatial averages are computed 586 for time series analysis in Figs. 1 and 3, and scatter plot relationships in Fig. 4. 587 588 589
27
590 591 Figure 3. Standardized annual time series of Great Plains surface temperature (red 592 curve), evapotranspiration (purple), precipitation (green), soil moisture (blue), and 593 PDSI (yellow) as simulated in the ensemble of CCSM4 forced integrations. Also 594 shown are the Palmer-‐model-‐derived annual time series of evapotranspiration 595 (yellow) and potential evapotranspiration (orange). The reference period for the 596 standardization is 1901–2000 and each of the curves is smoothed with a 9 point 597 Gaussian filter. 598 599
28
600 601 Figure 4. Comparison of the relationship between annual 10cm soil moisture and 602 PDSI for the 1901-‐2000 period (blue circles, R=+0.61) with that of the 2001-‐2100 603 period (R=+0.48) based on the diagnosis of all three CCSM4 simulations. 604