department of civil and environmental engineering...
TRANSCRIPT
1
A multi-scale analysis of drought and pluvial mechanisms for the Southeastern United States 1
2
Jonghun Kam, Justin Sheffield, Eric F Wood 3
4
Department of Civil and Environmental Engineering, Princeton University, New Jersey 5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
2
Abstract 26
27
The southeast (SE) U.S. has experienced several severe droughts over the past 30 years, with 28
the most recent drought during 2006-2008 causing agricultural impacts of $1 billion. 29
However, the mechanisms that lead to droughts over the region and their persistence have 30
been poorly understood due to the region’s humid coastal environment and its complex 31
climate. In this study, we carry out a multi-scale analysis of drought mechanisms for the SE 32
U.S. over 1979-2008 using the North American Regional Reanalysis (NARR), to identify 33
conditions associated with drought and contrast with those associated with pluvials. These 34
conditions include land surface drought propagation, land-atmosphere feedbacks, regional 35
moisture sources, persistent atmospheric patterns and larger scale oceanic conditions. Typical 36
conditions for SE US droughts (pluvials) are identified as follows: 1) weaker (stronger) 37
southerly meridional fluxes and weaker (stronger) westerly zonal fluxes, 2) strong moisture 38
flux divergence (convergence) by transient eddies, and 3) strong (weak) coupling between the 39
land surface and atmosphere. The NARR demonstrates that historic SE droughts are mainly 40
derived from a combination of a strong North Atlantic Sub-tropical High (NASH) and 41
Icelandic Low (IL) during summer and winter, respectively, which peak one month earlier 42
than the onset of the drought. The land surface plays a moderate role in drought occurrence 43
over the SE via recycling of precipitation and the oceans show an asymmetric influence on 44
droughts and pluvials depending on the season. This study suggests that the NASH and IL 45
can be used as a predictor for SE droughts at one-month lead despite the overall that it 46
represents an atmospheric forcing. 47
48
49
50
3
51
1. Introduction 52
Drought over the U.S. has been studied extensively in terms of its causes and impacts, 53
with particular focus on the central U.S. because of its position as the bread-basket of the 54
country, and the southwestern U.S., because of the fragility of water availability and 55
increasing demands from a burgeoning population. The central U.S., in particular, has 56
experienced severe droughts during the 1930s, 1950s, late 1980s and 2000s. The 1930s Dust 57
Bowl drought was derived from persistent high pressure systems over the central U.S., 58
anomalous tropical sea surface temperatures (SSTs), and strong land surface-atmosphere 59
coupling (Schubert et al. 2004) and human-induced land degradation (Cook et al. 2009). The 60
1950s drought has been associated with cold phases of the Pacific Decadal Oscillation (PDO) 61
and the El-Niño and Southern Oscillation (ENSO) (Barlow et al. 2001), and the water phase 62
of the Atlantic Multidecadal Oscillation (AMO) (McCabe et al. 2004). The 1988 drought 63
showed similar atmospheric and oceanic conditions to the 1930s drought (Trenberth et al. 64
1988). 65
Droughts over the southeastern (SE) U.S. (box in Figure 1) have been less well 66
understood, which is partly due to the complexity of its climate but also its relative 67
abundance of water. Over the last three decades the region has experienced several severe 68
droughts with different characteristics due to high inter-annual and intra-seasonal 69
precipitation variability. In 1981, the SE experienced a severe “flash” drought (Figure 1 (b)), 70
which was driven by below-normal summertime precipitation, the least over the last three 71
decades, and above-normal temperature. The drought was alleviated by tropical storm Dennis 72
and above-normal winter precipitation. A multi-year moderately severe drought occurred 73
during 1984-1988 (Figure 1 (c)), which was initiated by below-normal precipitation in the 74
autumn of 1984. Precipitation during the summer of 1985 was not enough to recover this 75
4
drought, and this was exacerbated by the least wintertime precipitation in 1985 over the past 76
three decades. The drought of 1988 extended to the central U.S. and covered 58% of 77
contiguous U.S. (CONUS) during the summer (NCDC 2012). The most recent long-term 78
drought during 2006-2008 was initiated from below-normal precipitation during the autumn 79
and winter of 2005 (Figure 1 (d)). The severity of the drought peaked in 2007, with the least 80
annual precipitation over the last three decades. The expanding population over the region 81
has compounded the impacts of the droughts on water supply, agriculture and ecosystem 82
health (Keim 1997; Silliman et al. 2005). During 2006-2008 drought, agricultural damages of 83
$1 billion were incurred, and water storage and quality for municipal use deteriorated, 84
prompting multi-state water resources restrictions and legal action (Manuel, 2008; Campana 85
et al, 2012). This drought prompted a number of studies to investigate the causes of droughts 86
over the SE using observations and modeling (Seager et al. 2009; Ortegren et al. 2011). 87
Precipitation over the SE is derived from diverse moisture sources that vary with 88
driving mechanisms occurring at different temporal and spatial scales, including land-89
atmosphere feedbacks, variability in regional-scale circulation, and persistent oceanic 90
conditions over the north Atlantic, and the north and tropical Pacific. Moisture sources are 91
derived from 1) advected moisture from surrounding regions, and 2) recycled moisture from 92
within the region from local evapotranspiration. Variations in moisture transport play a 93
critical role in regional climate variability and have strong seasonality over the CONUS 94
(Shukla and Mintz, 1982; Ruane, 2010). For the southern U.S., the Great Plains Low Level 95
Jet (GPLLJ) plays a dominant role in moisture transport (Higgins et al. 1997; Mo et al. 2005; 96
Mestas-Nunez et al. 2007; Trenberth et al. 2011), bringing moisture from the Gulf of Mexico 97
into the Great Plains during the summer, but which may be deflected to the S.E. by high 98
pressure systems over the central U.S. (Berbery and Rasmusson 1999; Mo et al. 2005). Such 99
synoptic scale circulations plays an important role in driving variability in moisture transport 100
5
and precipitation over the SE (Seager et al. 2009; Wang et al. 2010; Diem, 2013), with 101
important features including the western ridge of the North Atlantic Subtropical High (NASH; 102
Li et al. 2011) and the upper-lever jet (Wang et al. 2010), and westerly and northwesterly 103
lower-troposphere flux over the SE region (Diem 2006). Seager et al. (2009) suggested that 104
interannual variability of summer precipitation over the SE is forced by “purely internal 105
atmospheric variability.” However, there are suggestions that this situation has been made 106
more complex in recent years, because of the potential effects of climate change. For example, 107
summer precipitation variability over the SE has intensified owing to changes in the NASH 108
that have been linked to recent warming (Wang et al. 2010; Li et al. 2011). 109
The contribution of the land surface to moisture availability and precipitation 110
depends on the regional hydroclimate and the strength of feedbacks with the atmosphere 111
(Dirmeyer 2011; Ferguson et al. 2012). The recycling ratio Budyko (1974) is the ratio of 112
recycled precipitation from a local source via evaporation to total precipitation and can be 113
used to quantify the strength of land-atmosphere coupling and its impact on precipitation 114
variability. Several recycling models have been proposed over the past three decades based 115
on various assumptions, such as a fully mixed atmosphere and negligible changes in 116
atmosphere moisture at monthly to seasonal time scales (Brubaker et al. 1993; Eltahir and 117
Bras, 1994; Burde and Zangvil, 2001b; Dominguez et al. 2006; Dirmeyer and Brubaker, 2007; 118
van der Ent et al. 2010). Using such a model, Dominguez and Kumar (2008) demonstrated 119
that during the summer, the eastern U.S. exhibits strong coupling between the land and 120
atmosphere with the strength of coupling depending on soil moisture availability. For the SE 121
in particular, Dominguez et al. (2006) reported that about 30% of summertime precipitation is 122
recycled on average. 123
At broader time and space scales, the oceans play one of the most important roles in 124
driving climate globally. Marshall et al. (2001) review the ocean teleconnections for the 125
6
North Atlantic and note that anomalous sea surface temperatures tend to force large scale 126
circulation over the U.S. and Europe, contributing to anomalous precipitation. McCabe et al. 127
(2004) found that the positive phase of the AMO generally introduces more frequent drought 128
events over the CONUS and that the cold phase of the PDO coupled with the warm phase of 129
the Atlantic Ocean influences the spatial extent of the droughts over the CONUS. Climate 130
model studies (Feng et al. 2011; Nigam et al. 2011) show that the warm phase of the AMO 131
decreases summer and autumn rainfall over the US, especially for the south and SE regions, 132
but increases precipitation over the west African monsoon region. Furthermore, atmosphere-133
ocean interactions are also critical for precipitation variability over the SE. During the boreal 134
winter (December-February; DJF), air-sea flux variability can drive anomalous SST patterns 135
over the North Atlantic (Gulev et al. 2013). Anomalous SSTs over the Pacific can modulate 136
the westerly jet stream in the mid-troposphere leading to anomalous large circulation across 137
the CONUS (e.g. 1988 drought) (Trenberth et al. 1988). North Atlantic tropical cyclones 138
(TCs) play a moderate role in impeding and alleviating droughts over the SE depending on 139
their frequency and rainfall intensity (Kam et al. 2013). However, despite that the influences 140
of the Pacific and Atlantic oceans on continental scale droughts over the CONUS have been 141
well studied, their roles in regional scale drought over the SE are less well understood. 142
Of course, these individual mechanisms do not act alone but interact at different time 143
and space scales. For example, inter-annual variability of the north Atlantic is linked to 144
development of TCs that play a role at intra-seasonal time scales in modulating drought 145
occurrence and recovery (Kam et al., 2013). Therefore, a thorough understanding of drought 146
mechanisms for the SE requires a multi-scale study of land-atmosphere-ocean interactions. 147
Here, we use the North American Regional Reanalysis (NARR) to examine the land-148
atmosphere-ocean conditions and feedbacks that are associated with drought (and pluvial) 149
conditions over the SE, how they vary seasonally and how consistent these conditions are 150
7
between different events. The SE is defined here as 30°-35°N, 91°-82°W and is shown in 151
Figure 1 (b)-(d) as outlined by the black box. This area primarily includes the states of 152
Mississippi, Alabama, and Georgia with small portions of Arkansas and South Carolina. We 153
focus on meteorological droughts and pluvials based on precipitation anomalies during 1979-154
2008, but also examine anomalies in other parts of the terrestrial water budget. Section 2 155
describes the data and methods used. In Section 3, we present the hydroclimate over the SE 156
and compare its variation between the drought and pluvial years at regional scale. Section 4 157
investigates the typical conditions of the atmosphere and oceans during the development of 158
droughts and their potential driving mechanisms, at ocean basin to planetray scales, and then 159
Section 5 focuses on how these conditions evolved for the recent multi-year drought (2006-160
2008) across the various land, atmosphere, and ocean scales. In Section 6, we summarize our 161
findings and discuss their implications for the predictability of SE droughts and its 162
application to seasonal forecasting. 163
164
2. Data and Methods 165
166
2.1. Datasets 167
The NARR (Mesinger et al. 2006) is a state-of-the-art regional reanalysis for North 168
America and adjoining oceans that was proposed to better capture weather and climate 169
variability at regional and continental scales and associated hydrometerological extremes at 170
high resolution. The NARR has numerous improvements compared to previous global 171
reanalyses. It is based on the NCEP Eta atmospheric model (Black, 1988), its data 172
assimilation system (Zupanski and Mesinger 1995), and the Noah land-surface model (Ek et 173
al. 2003). It uses lateral boundary conditions from the NCEP-DOE Global Reanalysis 174
(Kanamitsu et al. 2002). The NARR includes land surface, atmospheric, and oceanic 175
8
variables at high temporal (3-hour) and spatial resolution (32 km). It has 29 vertical layers 176
and covers from 1979 to the present. Notable improvements in the NARR over previous 177
global reanalyses include finer resolution and additional/improved observational datasets, 178
assimilation of precipitation and introduction of a comprehensive land surface model 179
(Mesinger et al. 2006). 180
Past studies have shown that the NARR is a useful data resource for exploring 181
atmospheric and terrestrial water cycles at basin and regional scales (Mo et al. 2005; Ruiz-182
Barradas and Nigam 2006; Dominguez and Kumar 2008; Li et al. 2010; Ruane, 2010; Kumar 183
and Merwade 2011), for examining extreme hydrologic events such as pluvials and droughts 184
(Mo and Chelliah 2006; Weaver et al. 2009; Neiman et al. 2011; Sheffield et al. 2012), and as 185
an observational estimate for validation of other reanalyses and coupled models (Xiaoming 186
and Barros, 2010; Mei and Wang, 2012; Ferguson et al. 2012). We use the NARR 187
precipitation and soil moisture data at monthly scale to calculate anomalies to represent the 188
timing and severity of droughts and pluvials. 189
Additional datasets of geopotential heights (GHs) and SSTs are used to analyze 190
teleconnections with SE hydroclimate beyond the domain of the NARR. SSTs are taken from 191
the monthly Extended Reconstructed SST version 3b (ERSSTv3b) dataset (Smith et al., 2008) 192
and monthly GHs at 500 mb and 850 mb from the NCEP/NCAR global reanalysis (Kalnay et 193
al., 1996). 194
195
2.2. Analysis of moisture fluxes 196
2.2.1. COMPUTATION OF VERTICALLY INTEGRATED MOISTURE FLUXES 197
We compute the seasonal mean vertically integrated moisture fluxes (the surface to 198
250 mb; ignoring the negligible amount of water vapor above 250 mb) from the 3-hourly data 199
over 1979-2008 as follows: 200
9
Qλ1g qu Eqn. 1
201
QΦ1g qv Eqn. 2
The overbars represent the seasonal means over 1979-2008. 202
In this study, other forms of liquid and frozen water in the atmosphere are ignored. 203
The overbar represents a seasonal average derived from the three hourly data. We partition 204
the total column moisture transport into the lower (1000 mb - 850 mb) and upper 205
tropospheric levels (825 mb - 250 mb) in order to assess the contribution of low/high level 206
jets to total moisture transport over the SE. We partition the moisture fluxes into their mean 207
component and transient eddies (Peixoto and Oort 1992; Eqn.3 and Eqn. 4) 208
Qλ1g qu 1g q′u′ Eqn. 3
209
QΦ1g qv 1g q′v′ Eqn. 4
210
The transient eddy terms, q′u′ and q′v′, are the cross-covariance between the wind and 211
humidity variations computed from the difference between the seasonal means of the product 212
of the 3-hourly specific humidity and wind, and the product of the seasonal means of specific 213
humidity and wind (qv - qv and qu - qu). For the zonal eddy moisture fluxes, a positive 214
(negative) value indicates higher moisture flux than the average in the westerly (easterly) 215
direction (Brubaker et al. 1994). 216
10
217
2.2.2. COMPUTATION OF REGIONAL FLUX CONVERGENCE 218
Sub-integrals for each boundary of the rectangular region were computed following 219
the method of Simmonds et al. (1999). Seasonal fluxes (mm/day) along the four boundaries 220
were computed by aggregating the 3-hourly moisture fluxes (ks/m/s) along the respective 221
boundary (kg/s) and dividing by the area of the region. Then, the regional moisture flux 222
convergence was computed by summing the representative fluxes along the four boundaries. 223
Positive values represent convergence, that is, the region is a moisture sink, and negative 224
values represent divergence or a moisture source. 225
Despite the recommendation of a broader study area (> 106 km2) (Rasmusson 1977), 226
our study area for the SE is 0.4 * 106 km2, which is supported by the high-resolution data of 227
the NARR. To validate this, the moisture flux convergence averaged over the long term 228
(1979-2008) was compared with the long-term average of the difference between regional 229
precipitation and evapotranspiration. Under the assumption that the precipitable water 230
tendency is negligible, these two quantities should be equal. 231
232
2.3. Precipitation Recycling 233
We use the concept of precipitation recycling as a measure of the strength of land-234
atmosphere coupling and its impact on drought development and persistence. Recycled 235
precipitation is defined as precipitation originated from local sources via evaportranspiration 236
and recycling ratio is a ratio of recycled precipitation to total precipitation over a region 237
during a given time. We use a regional recycling ratio (r) as an indicator for the strength of 238
land-atmosphere coupling based on the Brubaker (1993) recycling model. It can be applied at 239
monthly and seasonal time scales, although it has been found to underestimate the strength of 240
recycling (Burde and Zangvil, 2001b). Our estimates of the recycling ratio for the SE from 241
11
the NARR are consistent with the estimates across the CONUS from different recycling 242
models (Dirmeyer and Brubaker 2007; Dominguez and Kumar, 2008). 243
244
2.4. Water budget non-closure in the NARR 245
Uncertainties exist in estimates of regional moisture flux convergence from 246
reanalysis products because of the assimilation of observations (Higgins et al. 1997 and 247
Ruane 2010). The errors in an individual product can be estimated by the analysis increments, 248
which are defined as the difference between the analysis field and the model first guess and in 249
general represents the impact of observations on the model forecast (Higgins et al. 1997). In 250
the NARR, assimilation of precipitation observations introduces moisture increments in the 251
atmospheric column, which include water vapor and water condensate increments, to reduce 252
errors in moisture tendencies. Here, we consider only the water vapor increment. For example, 253
over the SE, overestimation in simulated precipitation introduces negative moisture 254
increments into the atmospheric column without adjustment of the overestimates in 255
evaporation and/or moisture convergence. The maximum value of the increments occurs in 256
the summer with a magnitude comparable to the total convergence itself (Ruane, 2010). This 257
means that there are large uncertainties in the NARR data especially during the summer. The 258
increments are larger for pluvial years but relatively small for drought years due to 259
overestimation of precipitation (Ruane, 2010) and evaporation (Sheffiled et al. 2012) by the 260
NARR atmospheric and land models and the imbalance in the water budget because of the 261
lack of feedback from the land surface (Dominguez et al. 2006; Luo et al. 2007; Bisselink and 262
Dolman 2008; Weaver et al. 2009; van der Ent et al. 2010; Ruane, 2010). Over our study 263
region, the moisture increments are -1.29 mm/day on annual average, which contributes to 264
most of the residuals from the atmospheric water budget (1.5 mm/day). 265
266
12
2.5 Selection of drought and pluvial years 267
To select pluvial and drought years, we focus on meteorological drought defined by 268
anomalous precipitation relative to the 1979-2008 climatology, which is a precursor to 269
agricultural and hydrologic drought. We choose the six wettest and driest years based on the 270
ranks of seasonal precipitation anomalies (Figure 2) and then composite hydroclimate 271
variables for the pluvial and drought years. The winter period is defined as from December in 272
the previous year to February of the current year. For example, the winter of 2006, which was 273
the fourth driest winter of the time period, is from December 2006 to February 2007. Table 1 274
shows the statistics for summer and winter precipitation over the SE for 1979-2008, and 275
indicates that the driest periods were in 1980 and 2006. 276
277
3. Moisture Transport and Water Budgets over the SE U.S. 278
We show results for different aspects of the water budgets over the SE: the transport 279
of moisture into the region, the land and atmospheric water budgets, and the recycling of 280
precipitation. We compare the climatologies of the moisture transport on a seasonal basis and 281
the water budget and recycling ratio on a monthly basis with the composites for the six driest 282
and wettest years. 283
284
3.1. Moisture Transport and Convergence: Climatology, Drought and Pluvial Years 285
The mean and transient eddy components from the NARR are presented in Figure 3. 286
Across the eastern US, the mean moisture flux component dominates the total moisture 287
transport via the Great Plain Low Level Jet (GPLLJ) during summer while the transient 288
eddies play a minor role. During winter, the mean component dominates the total moisture 289
transport in the zonal direction due to large scale meanders of the jet stream, that is, Rossby 290
waves along mid-latitude regions (Schubert et al. 2011). In the meridional direction, however, 291
13
the mean and transient eddy components are comparable (Brubaker et al. 1994). Surface 292
temperature maps for the summer and winter climatology show large gradients along the 293
boundary between moisture convergence and divergence by the transient eddy component 294
(Figure 3 (e) and (f)). This indicates that transient eddy moisture fluxes are associated with 295
heat exchange between the land and ocean and also via storm tracks, especially during the 296
winter (Chen, 1985). 297
The summertime total-column moisture flux convergence over the SE is weakly 298
divergent (very close to zero) in the climatology (1979-2008), due to higher 299
evaportranspiration from the land surface during summer, while it is strongly convergence 300
(2.5 mm/day) during the winter (Figure 4 (a)). The weak summer divergence is derived from 301
weak zonal moisture transport deficit (surplus) between the west (north) side of the region 302
and the east (south) side of the region. During winter, moisture convergence is dominated by 303
the surplus of moisture transport in the meridional direction. During the six driest years, weak 304
meridional moisture fluxes along the Gulf of Mexico drive relatively strong divergence 305
around our study region during the summer and relatively weak convergence during winter. 306
This indicates that the moisture transport in the meridional direction plays an important role 307
in moisture convergence over the SE and its interannual variability can influence moisture 308
availability over the region. The total column mean and transient eddy components are both 309
convergent over the SE during winter, indicating that advected moisture into the region 310
modulates the variation of wintertime precipitation over the SE, and especially in relation to 311
the negligible recycling of moisture (see below). Interestingly, strong (weak) high-level 312
meridional moisture transport along the coastline is associated with summertime and 313
wintertime pluvials (droughts) over the SE, while consistent moisture is transported by low 314
level moisture flux from the Gulf of Mexico regardless of the season. This suggests that 315
extreme summer and winter precipitation is related to fluctuations of high-level meridional 316
14
moisture transport over the south side of region. 317
Total column moisture flux is partitioned into the mean (Figure 5 (a), (b), and (c)) 318
and transient eddy component (Figure 5 (d), (e), and (f)). Over 1979-2008, the moisture flux 319
convergence by the mean component compensates moisture divergence by the transient 320
eddies, resulting in weak moisture flux convergence during the summer. Based on the 321
climatology, the transient eddy component plays a critical role in moisture flux convergence 322
during the winter since most of the moisture flux convergence is derived from the transient 323
eddy component. During the six driest summers and winters, the low-level mean component 324
(from surface to 850mb) in the meridional direction is a major portion of moisture transport 325
over the study region, and the direction of the mean component at low level and high level is 326
opposite to each other. This indicates that there is heat exchange from land to ocean via 327
transient eddies, which suggested warm SSTs in the Gulf of Mexico. During the six driest 328
winters, transient eddies also play a dominant role in the regional moisture transport. 329
Interestingly, strong (weak) high-level meridional moisture transport along the coastline is 330
associated with summertime and wintertime pluvials (droughts) over the SE, while consistent 331
moisture is transported by low level moisture flux from the Gulf of Mexico regardless of the 332
season. This suggests that extreme summer and winter precipitation is related to fluctuations 333
of high-level meridional moisture transport over the south side of region. This indicates that 334
for the six driest (wettest) summers and winters, moisture flux divergence (convergence) is 335
mainly derived from transient eddies, which are related to sub-seasonal meteorological 336
phenomenon driven by atmosphere variability, as noted by Seager et al. (2009), rather than 337
more persistent oceanic forcing. This suggests that the potential predictability of droughts at 338
seasonal time scale over the SE might be limited. However, the results demonstrate that there 339
exists a consistent contrast of meridional moisture transport along the coastline of the Gulf of 340
the Mexico between drought and pluvial years, and better understanding of its driving 341
15
mechanisms can lead to improved predictability as discussed in section 4. 342
343
3. 2. The Regional Water Budget: Climatology, Drought and Pluvial Years 344
Monthly anomalies of moisture flux convergence and precipitation have large intra-345
seasonal variability (standard deviation = 3 and 2.5 mm/day, respectively; Figure 6 (a) and 346
(b)) and are highly correlated (correlation coefficient > 0.7) (Table 2). The terrestrial water 347
budget components (evaporation, runoff and soil moisture) have progressively lower short-348
term variability (Figure 6 (c), (d), and (e)) indicating that they have longer persistence 349
compared to components of the atmospheric budget (precipitation and moisture flux 350
convergence, MFC). Soil moisture at one-month lag shows the highest temporal correlation 351
with precipitation (Table 1), indicating that it takes about one month for the land surface to 352
respond to atmospheric anomalies. 353
Figure 7 shows the climatology of components of the terrestrial and atmospheric 354
water budgets for the SE and compares it with those during the six driest and wettest 355
summers. We focus on the warm season (JJA) to examine the maximum interaction of the 356
land and atmospheric components. To quantify the impact of precipitation assimilation in the 357
NARR on the water budgets, the moisture increments are also represented in Figure 7. Over 358
the SE, the overestimation of precipitation by the NARR atmospheric model is reduced by 359
adding negative moisture increments in the atmosphere while there is no adjustment to the 360
overestimates of evaporation and moisture convergence. Especially during the wetter seasons, 361
the moisture increments are over -2.0 mm/day while they are less than -1.0 mm/day during 362
the drier seasons. This indicates that the NARR overestimates evaporation and/or moisture 363
flux convergence over the SE and therefore can overestimate the recycling of rainfall as 364
analysed in the next sub section. The residuals from the atmospheric water budget (Figure 7; 365
P-E-MFC) are 1.5 mm/day averaged over 1979-2008, which are derived mostly from the 366
16
precipitation data assimilation (the increments: -1.2 mm/day). 367
There is a large correlation between MFC and precipitation (Table 1). Evaporation 368
has a marked seasonal cycle, which peaks in June and is at a minimum in December causing 369
a transition from an overall moisture source (evaporation exceeding precipitation) during the 370
summer to a moisture sink during winter. Runoff, which includes surface runoff and 371
subsurface baseflow, peaks in the spring as driven by the peak in precipitation and low 372
evaporation from winter to spring. The inferred change in terrestrial water storage (Figure 7; 373
P-E-Q) is positive in the winter and early spring, and negative in the warmer months. 374
During the summer drought years, the MFC decreases from spring onwards driving 375
less precipitation over the SE with no peak in June. On the other hand, the summer pluvial 376
years show a higher peak of MFC in June, which is related to a concurrent increase in 377
precipitation. The NARR indicates that there is a significant increase in the atmospheric 378
water budget during summer pluvial years and a decrease during summer drought years as 379
expected. In drought years, higher solar radiation due to reduced cloud cover, and higher 380
surface temperatures (not shown), causes evaporation to exceed precipitation from May 381
through August and enhances the role of soil moisture in providing moisture to the 382
atmosphere to compensate the high evaporative demand. In pluvial years, the opposite 383
changes occur. Thus, drought years tend to have stronger coupling between the land surface 384
and atmosphere than normal years, which can be quantified by the recycling ratio as shown in 385
the next section in terms of the inter-annual variability of recycling. 386
387
3. 3. Recycling Ratio and Recycled Rainfall: Climatology, Drought Years and Pluvial Years 388
The recycling ratio and recycled precipitation over the SE are shown in Figure 8. 389
Over the course of the year, the recycling ratio increases from near zero in winter to a 390
maximum in August, when the recycling ratio is over 0.1, meaning that more than 10% of 391
17
monthly precipitation is derived from local evaporation. The variance of the recycling ratio 392
also increases during the warm season (Figure 8 (a)). The recycling ratio and its variance are 393
higher during summer droughts (mean of about 0.15) than during summer pluvials (mean of 394
about 0.06). To assess the actual impact of recycling on precipitation, the recycled 395
precipitation is computed from the monthly precipitation multiplied by the recycling ratio. 396
Interestingly, there is more recycled precipitation on average (0.5 mm/day) than during 397
extreme years. During summer pluvials, precipitation is recycled a little bit more (0.4 398
mm/day) than during drought years (0.3 mm/day), despite a lower recycling ratio, since the 399
absolute size of precipitation is much higher. These estimates are likely overestimated, 400
however, because of the overestimates in evaporation and/or moisture flux convergence 401
discussed above. The low average recycling ratio indicates that the land surface plays a 402
moderate role in summer precipitation over the SE during normal years, mainly due to its 403
proximity to ocean sources of moisture. 404
405
4. Multi-scale Atmospheric and Oceanic Drivers: Warm and Cold Seasons 406
To identify typical conditions of large-scale circulation during drought and pluvial 407
years, we plot the difference in moisture fluxes (vectors) and 500mb GHs (shading) over the 408
CONUS between drought and pluvial years (Figure 9 (a) and (b), respectively). During 409
summer drought years, moisture transport from the Gulf of Mexico into the SE is generally 410
blocked by anomalous anticyclonic motion in the mid-troposphere induced by positive GH 411
anomalies over the Central U.S. During winter drought years, a dipole (high-low) anomalous 412
pressure system brings dry air from Canada through the Great Plains down to the SE. Dong et 413
al. (2011) also identified this pattern in their study of the winter 2006 Great Plains drought. 414
These anomalous patterns suggest that different driving mechanisms for extreme events exist 415
over the SE region at seasonal scale. We explore next the driving mechanisms of these 416
18
patterns at different spatial scales, from regional up to planetary scales, first during warm 417
seasons, and then during cold seasons. 418
419
4.1. Warm season variability 420
During dry summers, a strong high pressure system at 850 mb (green contour in 421
Figure 10) extends to the southern U.S. and thus its western ridge is shifted northward, which 422
is derived from strong positive GH anomalies over the north Atlantic and the central U.S. 423
This drives less precipitation along the Gulf of Mexico coastal states. This situation is 424
reversed during wet summers (Figure 10 (b)). This high pressure system is well known as the 425
NASH, which is a major driver of summer climate variability for the SE (Wang et al. 2010). 426
The strength of the NASH is generally measured by the extent of the 1560 geopotential meter 427
(gpm) lines near Bermuda, and quantified by the Bermuda High Index (BHI) and Western 428
Bermuda High Index (WBHI). Here, the definition of the BHI and the WHBI are from the 429
study of Diem (2013), where the BHI (WBHI) index is computed from the normalized 430
difference of sea level pressures (850-hPa geopotential heights) between two regions marked 431
as circle (cross) in Figure 10. During dry summers, the NARR show positive values of the 432
BHI and WBHI, which derive from positive and negative surface pressure anomalies over the 433
land and oceans, respectively. The positive values of the BHI and WBHI are related to the 434
blocking effect of moisture transport from the Gulf of Mexico in the meridional direction and 435
thus it results in below-normal precipitation. Also, the shift of the NASH westward reduces 436
the number of landfalling TCs in the SE by deflecting their paths to the southern U.S. 437
(Rosenberg 1978). 438
Summer droughts are also associated with warmer SSTs in the north Atlantic (55-439
25W, 25-55N) and cooler SSTs over the north Pacific (160-130W, 30-60N) while positive 440
anomalies of SSTs over the Niño 3.4 region are weakly associated in terms of the magnitude 441
19
(Figure 10 (c)). We quantify the consistency of the anomalous patterns by counting the 442
number of the years that show the same sign as the composited patterns with a threshold of 443
more than four years in agreement (that is 5 or 6 years). During dry summers (Figure 10 (c)), 444
these anomalous SST patterns over the north Pacific and north Atlantic are consistent across 445
individual dry summers (hatched areas in Figure 10 (a)), however a warmer phase of the 446
tropical Pacific is consistent only during wet summers. The contrast in summertime SST 447
anomalies between the north Pacific and north Atlantic is notable and is reminiscent of the 448
pattern identified at decadal or longer scales (the cold phase of PDO-the warm phase of AMO) 449
by McCabe et al. (2004) and Schubert et al. (2007) as favorable for drought conditions for the 450
U.S. in general. 451
The center of the anomalous SST pattern over the north Atlantic is well matched with 452
the center of the NASH, suggesting that either the ocean or atmosphere could be a predictor 453
over the Atlantic for the other. To examine the mechanisms, the composited maps for 500mb 454
GH are calculated from two month lead time (-2 month) to two month lagged time (+2 month) 455
during summer droughts and pluvials (Figure 11). Due to the small sample size, we chose 456
Barnard’s exact test to estimate the statistical significance of the signs of the composite 457
geopotential height anomalies during droughts and pluvials, as compared to during non-458
drought and non-pluvial years. In general, Barnard’s exact test is used to test the 459
independence of rows and columns in a contingency table (Barnard, 1945). Here, the two 460
rows of the table are: droughts (pluvials) and non-droughts (non-pluvials); and the two 461
columns are: the signs of the composited 500mb geopotential height anomalies and their 462
opposite sign. The null hypothesis is that the signs of the composite 500mb geopotential 463
height anomalies are independent of droughts or pluvials conditions at the 0.1 significance 464
level (hatched areas in Figure 11 and 13). During summer droughts, consistent and positive 465
anomalies of 500 mb GH over the north Atlantic are greatest at two month lead time (AMJ; 466
20
Figure 12 (a)) and persist until zero lead time, after which it disappears at one month lag time 467
(Figure 12 (g)), and vice versa. This indicates that a strong NASH during late spring forces 468
positive north Atlantic SST anomalies and thus leads to reduced precipitation over the SE. 469
Therefore, variability in the NASH is one of main drivers of summer droughts over the SE. 470
During summer pluvials, consistent 500 mb GH anomalies over the Gulf of Alaska and north 471
Atlantic persist from two month lead time until zero time, and their locations are well 472
matched with the summertime SST anomalies over the north Pacific and north Atlantic 473
(Figure 11 (b)). This implies that the atmosphere forces the anomalous SSTs over the north 474
Pacific and Atlantic during late spring (AMJ) via Rossby waves at mid-latitudes and then 475
theses anomalous GHs and SSTs result in the summertime pluvials. Overall, the results 476
suggest that SE hydroclimate has a symmetric connection with conditions over the Atlantic 477
but an asymmetric connection with the Pacific. Zhao et al. (2011) suggested that during the 478
summer, these subtropical anticyclones over the Pacific and Atlantic and anomalous SSTs 479
might be associated with the Asian-Pacific Oscillation at interdecadal scale, which is 480
predictable by some climate forecast systems (Chen et al. 2013). 481
482
4.2. Cold Season Variability 483
Winter droughts (pluvials) show a westward (eastward) shift of a low pressure system 484
(green line in Figure 12 (a)), which is located between Iceland and southern Greenland for the 485
climatology (the contour black line of -5 gpm in Figure 13 (a) and (b)). This low pressure 486
system is the Icelandic Low, which is downstream of the mid-tropospheric wave trough 487
towards Europe (Blackmon et al. 1984). Over the north Atlantic, the IL teleconnection with 488
the pressure system over Bermuda was introduced by the study of Blackmon et al. (1984) and 489
its interannual variability is linked with interannual Arctic sea ice extent variations and the 490
North Atlantic Oscillation (NAO; Deser et al. 2000). For example, eastern US winter 491
21
droughts relate to higher values of the IL index, which resembles the negative phase of the 492
NAO. The IL index is calculated from 500mb GH anomalies on a regional basis (55°-70°N, 493
60°W-10°E). Our results demonstrate that the westward shift of the IL results in a higher 494
value of the IL and contributes to a reduction (increase) in precipitation over the SE. 495
The winter drought years are associated with weak but consistent cold phases of SSTs 496
in the Niño 3.4 region and the north Pacific (Figure 12 (c)). The anomalous SST patterns 497
during dry winters are close to the SSTs pattern of “the perfect ocean for drought” during 498
1998-2002 as identified by Hoerling and Kumar (2003). During winter pluvials, the El Niño 499
phase over the tropical Pacific is strong but inconsistent across individual wet winters in the 500
composite SST maps (Figure 12 (d)) since it is derived from two strong El Niño in 1982 and 501
1997. There is a horse-shoe like SST anomaly pattern over the central Pacific, similar to the 502
negative phase of the PDO. This is consistent across years during winter pluvials (hatched 503
areas in Figure 12 (d)). This suggests that the north Pacific is related to winter pluvials over 504
the SE and the signal from the North Atlantic is weak and inconsistent during dry winters. 505
During the winter droughts, the composited maps for mid-troposhperic GHs indicate 506
that the IL (box in Figure 13) peaks at one month lead time (NDJ) due to a strong high-low 507
dipole pattern of mid-tropospheric GH anomalies, which then weakens gradually and 508
disappears during the early spring (FMA). At one month lead time there is a strong 509
atmospheric forcing of SST anomalies over the Atlantic and winter droughts over the SE, 510
with the opposite mechanism in place during the winter pluvials. Negative 500mb GH 511
anomalies over the central Pacific appear abruptly at zero lead time and persist until early 512
spring (FMA), indicating that during winter pluvials, SSTs over the central Pacific force the 513
atmosphere, and induce more precipitation over the CONUS, especially the SE. Furthermore, 514
we constructed composite maps for 500mb geopotential height anomalies at monthly time 515
step (October-February; not shown), The monthly composite maps demonstrate the north-516
22
south dipole pattern of anomalous 500mb geopotential heights (low-high) from November 517
through February of the next year, which is representative of the typical pattern of a positive 518
IL. Our results demonstrate that the IL at one month lead time step can be used as a predictor 519
of winter droughts over the SE while a negative phase of PDO forces the atmosphere and 520
then induces winter pluvials over the SE. 521
522
5. A Case Study of the 2006-2008 SE Drought 523
As a case-study, we examine the roles of the above-discussed climate drivers in the 524
recent 2006-2008 SE drought by examining of the evolution of hydroclimate variables and 525
atmospheric and oceanic drivers. Figure 14 shows monthly time series of hydroclimate 526
variables normalized by their standard deviations for 1979-2008. It also shows related 527
atmospheric and oceanic drivers: monthly time series of the IL index, the BHI and the WBHI, 528
SSTs over the Niño 3.4 region, the north Atlantic SSTs averaged over the region 55º-25ºW 529
and 30º-60ºN, and the difference between the regional averaged SSTs over the north Atlantic 530
(55º-25ºW, 25º-55ºN) and north Pacific (160º-130ºW, 30º-60ºN), each normalized by its 531
standard deviation. Positive values of the northwestern Atlantic/northeastern Pacific 532
difference indicates a stronger SST contrast which is associated with less precipitation during 533
summer as shown above. The IL index is computed from the regional average (60ºW-10ºE, 534
55ºN-70ºN, the dashed box in Figure 9) of the normalized 500mb GHs. Figure 14 also shows 535
time series of TC related rainfall anomalies again normalized by the standard deviation to 536
indicate the strength of TC activity (Kam et al. 2012). 537
During the late summer of 2005, below-normal moisture flux convergence led to less 538
precipitation (including close to normal TC-related rainfall). This period is coincident with a 539
strong positive phase of the IL index. This induced negative soil moisture anomalies 540
beginning in the fall of 2005, reaching a minimum during the late spring of 2007. During the 541
23
spring and early summer of 2006, high potential evaporation increased evapotranspiration, 542
but this turned to negative anomalies in subsequent years because of the lack of available soil 543
moisture, despite persistently positive potential evaporation anomalies. These combined to 544
force positive recycling ratio anomalies during the summer, of over 0.3. During the winters, 545
the positive phase of the IL was derived from the southward shift of the minimum in GH 546
anomalies, which brought a reduction in winter precipitation over the SE. The north Atlantic 547
was in a warm phase during the entire period and the wintertime SSTs over the Niño 3.4 548
region (5°S-5°N, 170°W-120°W) showed the La-Niña phase (a cold phase) during the 549
subsequent winters, which were also associated with a reduction in SE precipitation. During 550
the summers of 2006-2009, differences between SSTs in the north Atlantic and north Pacific 551
were positive and provided favorable oceanic conditions for less summertime precipitation 552
over the SE (McCabe et al. 2004). Also, the WBHI index indicates that there existed a strong 553
NASH during the summers of 2006 and 2007. From August 2008, the precipitation deficit 554
over the SE region weakened gradually and more moisture transport and TC-related rainfall 555
produced above-normal summertime precipitation, leading to drought recovery in the early 556
spring of 2009. This was linked to negative phases of the IL, the BHI, and the WBHI through 557
the winter of 2008. In summary, this drought was mainly forced by the atmosphere, 558
especially a strong NASH and IL, and persisted by a combination of favorable conditions in 559
different aspects of the land and ocean. 560
561
6. Summary and Conclusions 562
This study has evaluated the mechanisms associated with drought over the SE U.S. 563
from the point of view of the land, atmosphere and ocean, as well as the conditions associated 564
with pluvial events. The evaluation is based on the NARR, which while providing a 565
consistent and continuous dataset of relevant variables for exploring extreme hydrologic 566
24
events over the CONUS, does have limited spatial and temporal coverage and is subject to 567
uncertainties at all scales. Of importance is the nonclosure of the water budget as derived 568
from its data assimilation method that despite being novel in its indirect assimilation of 569
observed precipitation results in significant closure residuals. This will impact the strength of 570
the calculated anomalies, for example, the overestimate in moisture flux and recycling ratio 571
computation, but is unlikely to affect the sign of the anomalies or the overall conclusions of 572
this study. 573
The hydroclimate of SE U.S. is subject to different dominant moisture sources and 574
different climate drivers depending on the season. Typical conditions for SE droughts 575
(pluvials) are: 1) weaker (stronger) southerly meridional moisture fluxes and weaker 576
(stronger) westerly zonal moisture fluxes, 2) strong moisture flux divergence (convergence) 577
by transient eddies, that is, a strong (weak) heat and moisture exchange between tropical and 578
extra-tropical regions, and 3) strong (weak) coupling between the land surface and 579
atmosphere. For summer, a strong NASH at two month (AMJ) until zero month lead time 580
(JJA) forces a warmer phase of the north Atlantic and SE drought conditions. For winter, a 581
peak in the IL index in NDJ propagates a wave train in the mid-troposphere, which resembles 582
the negative phase of NAO. This leads to less moisture convergence and less precipitation 583
over the SE. For oceanic influences, the roles of the Pacific and the Atlantic are different 584
during droughts and pluvials depending on the season. The tropical Pacific induces more 585
interannual variability in wintertime precipitation over the SE, the central Pacific forces the 586
atmosphere during winter pluvials, and the Atlantic is forced by the interannual variability of 587
the NASH during the summer. During summer droughts and pluvials, different Niño regions 588
also play different roles in the interannual variability of SE precipitation. 589
Overall, the mechanisms for variability in SE hydroclimate are complex, especially 590
during summer. Seager et al. (2009) suggested that SE precipitation variability is purely 591
25
controlled by internal atmospheric processes and thus is unpredictable by a SST-forced global 592
circulation model. This study also indicates that the symmetry of SE droughts and pluvials is 593
mainly driven from the symmetrical atmospheric responses at one to two month lead time 594
step despite asymmetric terrestrial and oceanic responses. Recent studies indicate that 595
summer precipitation variability relates to the NASH (Wang et al., 2010) and to low 596
frequency variability of the North Atlantic Ocean (Ortegren et al. 2011). This study has 597
provided a better understanding of the land, atmosphere, and ocean conditions associated 598
with historical SE droughts, which, despite the overall preponderance of atmospheric controls, 599
shows potential for enhanced seasonal forecast skill via one to two month lead time of 600
anomalies in the NASH and the IL. Further study is necessary to understand whether these 601
associations are stationary. For example, variability in SE precipitation has increased in 602
recent years as linked to warming of the north Atlantic (Wang et al., 2010), and Pacific SST 603
teleconnections with CONUS precipitation and vary on multi-decadal time scales (Kam et al., 604
2014). 605
606
Acknowledgments 607
We acknowledge the support of the Princeton TIGRESS high performance computer center. 608
This work was supported by the NOAA Climate Program Office (NA08OAR4310579) and 609
the USGS (G11AP20215). The authors are pleased to acknowledge that the work reported on 610
in this paper was substantially performed at the TIGRESS high performance computer center 611
at Princeton University which is jointly supported by the Princeton Institute for 612
Computational Science and Engineering and the Princeton University Office of Information 613
Technology's Research Computing department.614
26
References 615
Barlow, M., S. Nigam, and E. H. Berbery, 2001: ENSO, Pacific decadal variability, and US 616
summertime precipitation, drought, and stream flow. J. Climate, 14, 2105-2128. 617
Barnard, G. A., 1945: A new test for 2x2 tables, Nature, 156, 177. 618
Berbery, E. H., and E. M. Rasmusson, 1999: Mississippi moisture budgets on regional 619
scales. Mon. Weather Rev., 127, 2654-2673. 620
Black, T. L., 1988: The step-mountain eta coordinate regional model: A documentation. 621
NOAA/NWS National Meteorological Center, 47 pp. 622
Bisselink, B., and A. J. Dolman, 2008: Precipitation recycling: moisture sources over Europe 623
using ERA-40 Data. J. Hydrometeorol, 9, 1073-1083. 624
Brubaker, K. L., D. Entekhabi, and P. S. Eagleson, 1993: Estimation of continental 625
precipitation recycling. J. Climate, 6, 1077-1089. 626
——, 1994: Atmospheric water-vapor transport and continental hydrology over the 627
America. J. Hydrology, 155, 407-428. 628
Burde, G. I., and A. Zangvil, 2001a: The estimation of regional precipitation recycling. Part I: 629
Review of recycling models. J. Climate, 14, 2497-2508. 630
——, 2001b: The estimation of regional precipitation recycling. Part II: A new recycling 631
model. J. Climate, 14, 2509-2527. 632
Campana, P., J. Knox, A. Grundstein, and J. Dowd, 2012: The 2007-2009 drought in Athens, 633
Georgia, United States: a climatological analysis and an assessment of future water 634
availability. J. Am. Water Resour. As., 48, 379-390. 635
Chen, J., P. Zhao, S. Yang, G. Liu, and X. Zhou, 2013: Simulation and Dynamical Prediction 636
of the Summer Asian–Pacific Oscillation and Associated Climate Anomalies by the NCEP 637
CFSv2, J. Climate, 26, 3644-3656. 638
Chen, T. C., 1985: Global water-vapor flux and maintenance during FGGE. Mon. Weather 639
27
Rev., 113, 1801-1819. 640
Cook, B. I., R. L. Miller and R. Seager, 2009: Amplification of the North American "Dust 641
Bowl" drought through human-induced land degradation. Proc. Natl. Acad. Sci. U.S.A., 642
106, 4997-5001. 643
Dai, A. G., 2011: Drought under global warming: a review. Wiley Interdisciplinary Reviews-644
Climate Change, 2, 45-65. 645
Deser, C., J. E. Walsh, and M. S. Timlin, 2000: Arctic sea ice variability in the context of 646
recent atmospheric circulation trends. J. Climate, 13, 617-633. 647
Diem, J. E., 2006: Synoptic-scale controls of summer precipitation in the Southeastern United 648
States. J. Climate, 19, 613-621. 649
Diem, J. E., 2013: Influences of the Bermuda High and atmospheric moistening on changes 650
in summer rainfall in the Atlanta, Georgia region, USA. Int. J. Climatol., 33, 160-172. 651
Dirmeyer, P. A., 2011: The terrestrial segment of soil moisture-climate coupling. Geophys. 652
Res. Lett., 38, L16702, doi:10.1029/2011GL048268. 653
Dirmeyer, P. A., and K. L. Brubaker, 2007: Characterization of the global hydrologic cycle 654
from a back-trajectory analysis of atmospheric water vapor. J. Hydrometeorol, 8, 20-37. 655
Dominguez, F., and P. Kumar, 2008: Precipitation recycling variability and ecoclimatological 656
stability - A study using NARR data. Part I: Central U. S. Plains ecoregion. J. Climate, 21, 657
5165-5185. 658
Dominguez, F., P. Kumar, X. Z. Liang, and M. F. Ting, 2006: Impact of atmospheric moisture 659
storage on precipitation recycling. J. Climate, 19, 1513-1530. 660
Dong, X. Q., and Coauthors, 2011: Investigation of the 2006 drought and 2007 flood 661
extremes at the Southern Great Plains through an integrative analysis of observations. J. 662
Geophys. Res., 116, D03204, doi:10.1029/2010JD01477. 663
Ek, M. B., and Coauthors, 2003: Implementation of Noah land surface model advances in the 664
28
National Centers for Environmental Prediction operational mesoscale Eta model. J. 665
Geophys. Res., 108, D03204, doi:10.1029/2002JD003296. 666
Eltahir, E. A. B., 1998: A soil moisture rainfall feedback mechanism 1. Theory and 667
observations. Water Resour. Res., 34, 765-776, doi:10.1029/97WR03499. 668
Eltahir, E. A. B., and R. L. Bras, 1994: Precipitation recycling in the Amazon basin. Q. J. Roy. 669
Meteor. So., 120, 861-880. 670
Feng, S., Q. Hu, and R. J. Oglesby, 2011: Influence of Atlantic sea surface temperatures on 671
persistent drought in North America. Clim. Dyn., 37, 569-586. 672
Ferguson, C. R., and E. F. Wood, 2011: Observed land-atmosphere coupling from satellite 673
remote sensing and reanalysis. J. Hydrometeorol, 12, 1221-1254. 674
Ferguson, C. R., E. F. Wood, and R. K. Vinukollu, 2012: A Global intercomparison of 675
modeled and observed land-atmosphere coupling. J. Hydrometeorol, 13, 749-784. 676
Gulev, S. K., Latif, M., Keenlyside, N., Park, W. and Koltermann, K. P., 2013: North Atlantic 677
Ocean control on surface heat flux on multidecal timescales. Nature, 499, 464-467. 678
Held, I. M., and B. J. Soden, 2000: Water vapor feedback and global warming. Annu. Rev. 679
Energ. Env. , 25, 441-475. 680
Higgins, R. W., Y. Yao, E. S. Yarosh, J. E. Janowiak, and K. C. Mo, 1997: Influence of the 681
Great Plains low-level jet on summertime precipitation and moisture transport over the 682
central United States. J. Climate, 10, 481-507. 683
Hoerling, M., and A. Kumar, 2003: The perfect ocean for drought. Science, 299, 691-694. 684
Huntington, T. G., 2006: Evidence for intensification of the global water cycle: Review and 685
synthesis. J. Hydrology, 319, 83-95. 686
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-year reanalysis project. Bull. Amer. 687
Meteor. Soc., 77, 437-471. 688
Kam, J., J. Sheffield, X. Yuan, and E. F. Wood, 2013: Influence of Atlantic tropical cyclones 689
29
on drought over the eastern US (1980-2007). J. Clim., 26, doi:10.1175/JCLI-D-12-690
00244.1. 691
Kam, J., J. Sheffield, X. Yuan, and E. F. Wood, 2014: Did a skillful prediction of sea surface 692
temperatures help or hinder forecasting of the 2012 Midwestern US drought?, Environ. 693
Res. Lett., 9, 034005, doi:10.1088/1748-9326/9/3/034005. 694
Kanamitsu, M., W. Ebisuzaki, J. Woollen, S. K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter, 695
2002: NCEP-DOE AMIP-II reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, 1631-1643. 696
Kumar, S., and V. Merwade, 2011: Evaluation of NARR and CLM3.5 outputs for surface 697
water and energy budgets in the Mississippi River basin. J. Geophys. Res., 116, D08115, 698
doi:10.1029/2010JD014909. 699
Li, W., L. Li, R. Fu, Y. Deng, and H. Wang, 2011: Changes to the North Atlantic subtropical 700
high and its role in the intensification of summer rainfall variability in the southeastern 701
United States. J. Climate, 24, 1499-1506. 702
Li, X. P., S. Y. Zhong, X. D. Bian, W. E. Heilman, Y. Luo, and W. J. Dong, 2010: 703
Hydroclimate and variability in the Great Lakes region as derived from the North 704
American Regional Reanalysis. J. Geophys. 705
Res., 115, D12104, doi: 10.1029/2009JD012756 . 706
Luo, Y., E. H. Berbery, K. E. Mitchell, and A. K. Betts, 2007: Relationships between land 707
surface and near-surface atmospheric variables in the NCEP north American regional 708
reanalysis. J. Hydrometeorol, 8, 1184-1203. 709
Manuel, J., 2008: Drought in the southeast: Lessons for water management. Environ. Health 710
Persp., 116, A168-A171. 711
Marshall, J., and Coauthors, 2001: North Atlantic climate variability: Phenomena, impacts 712
and mechanisms. Iny. J. Climatol., 21, 1863-1898. 713
McCabe, G. J., M. A. Palecki, and J. L. Betancourt, 2004: Pacific and Atlantic Ocean 714
30
influences on multidecadal drought frequency in the United States. P. Natl. Acad. Sci. 715
USA., 101, 4136-4141. 716
McCabe, G. J., J. L. Betancourt, S. T. Gray, M. A. Palecki, and H. G. Hidalgo, 2008: 717
Associations of multi-decadal sea-surface temperature variability with US 718
drought. Quatern. Int., 188, 31-40. 719
Mei, R., and G. L. Wang, 2012: Summer land-atmosphere coupling strength in the United 720
States: Comparison among observations, reanalysis data, and numerical models. J. 721
Hydrometeorol,13, 1010-1022. 722
Mesinger, F., and Coauthors, 2006: North American regional reanalysis. Bull. Amer. Meteor. 723
Soc., 87, 343-360. 724
Mestas-Nunez, A. M., and D. B. Enfield, 2007: Water vapor fluxes over the Intra-Americas 725
Sea: Seasonal and interannual variability and associations with rainfall. J. 726
Climate, 20, 1910-1922. 727
Mo, K. C., and M. Chelliah, 2006: The modified Palmer drought severity index based on the 728
NCEP North American Regional Reanalysis. J. Appl. Meteorol. Clim., 45, 1362-1375. 729
Mo, K. C., M. Chelliah, M. L. Carrera, R. W. Higgins, and W. Ebisuzaki, 2005: Atmospheric 730
moisture transport over the United States and Mexico as evaluated in the NCEP regional 731
reanalysis. J. Hydrometeorol, 6, 710-728. 732
Neiman, P. J., L. J. Schick, F. M. Ralph, M. Hughes, and G. A. Wick, 2011: Flooding in 733
western Washington: the connection to atmospheric rivers. J. Hydrometeorol, 12, 1337-734
1358. 735
Nigam, S., B. Guan, and A. Ruiz-Barradas, 2011: Key role of the Atlantic Multidecadal 736
Oscillation in 20th century drought and wet periods over the Great Plains. Geophys. Res. 737
Lett., 38, L16713, doi:10.1029/2011GL048650. 738
Ortegren, J. T., P. A. Knapp, J. T. Maxwell, W. P. Tyminski, and P. T. Soule, 2011: Ocean-739
31
atmosphere influences on low-frequency warm-season drought variability in the Gulf 740
coast and southeastern United States. J. Appl. Meteorol. Clim., 50, 1177-1186. 741
Peixoto, J. P. and A. H. Oort, 1992: Physics of climate. American Institute of Physics, New 742
York, 520 pp. 743
Rasmusson, E. M., 1977: Hydrological application of atmospheric vapor-flux analyses. 744
Operational Hydrology Rep. 11. WMO, Geneva, 50 pp 745
Ralph, F. M., P. J. Neiman, G. A. Wick, S. I. Gutman, M. D. Dettinger, D. R. Cayan, and A. B. 746
White, 2006: Flooding on California's Russian River: Role of atmospheric 747
rivers. Geophys. Res. Lett., 33, L13801, doi: 10.1029/2006GL026689. 748
Ruane, A. C., 2010: NARR's atmospheric water cycle components. Part I: 20-year mean and 749
annual interactions. J. Hydrometeorol, 11, 1205-1219. 750
Ruiz-Barradas, A., and S. Nigam, 2006: Great plains hydroclimate variability: The view from 751
North American regional reanalysis. J. Climate, 19, 3004-3010. 752
Sahsamanoglou, H. S., 1990: A contribution to the study of action centers in the North-753
Atlantic. Iny. J. Climatol., 10, 247-261. 754
Schubert, S. D., M. J. Suarez, P. J. Pegion, R. D. Koster, and J. T. Bacmeister, 2004: On the 755
cause of the 1930s Dust Bowl. Science, 303, 1855-1859. 756
Seager, R., A. Tzanova, and J. Nakamura, 2009: Drought in the Southeastern United States: 757
causes, variability over the last millennium, and the potential for future hydroclimate 758
change. J. Climate, 22, 5021-5045. 759
Sheffield, J., and E. F. Wood, 2008: Projected changes in drought occurrence under future 760
global warming from multi-model, multi-scenario, IPCC AR4 simulations. Clim. 761
Dyn., 31, 79-105. 762
Sheffield, J., B. Livneh, and E. F. Wood, 2012: Representation of terrestrial hydrology and 763
large-scale drought of the continental United States from the North American Regional 764
32
Reanalysis. J. Hydrometeorol, 13, 856-876. 765
Shukla, J., and Y. Mintz, 1982: Influence of land-surface evapo-transpiration on the Earth’s 766
climate. Science, 215, 1498-1501. 767
Simmonds, I., D. H. Bi, and P. Hope, 1999: Atmospheric water vapor flux and its association 768
with rainfall over China in summer. J. Climate, 12, 1353-1367. 769
Smith, T. M., R. W. Reynolds, T. C. Peterson, and J. Lawrimore, 2008: Improvements to 770
NOAA's historical merged land-ocean surface temperature analysis (1880-2006). J. 771
Climate, 21, 2283-2296. 772
Sun, X., and A. P. Barros, 2010: An evaluation of the statistics of rainfall extremes in rain 773
gauge observations, and satellite-based and reanalysis products using universal 774
multifractals. J. Hydrometeorol, 11, 388-404. 775
Trenberth, K. E., and C. J. Guillemot, 1996: Physical processes involved in the 1988 drought 776
and 1993 floods in North America. J. Climate, 9, 1288-1298. 777
——, 1998: Evaluation of the atmospheric moisture and hydrological cycle in the 778
NCEP/NCAR reanalyses. Clim. Dyn., 14, 213-231. 779
Trenberth, K. E., G. W. Branstator, and P. A. Arkin, 1988: Origins of the 1988 North-780
American drought. Science, 242, 1640-1645. 781
Trenberth, K. E., J. T. Fasullo, and J. Mackaro, 2011: Atmospheric moisture transports from 782
ocean to land and global energy flows in reanalyses. J. Climate, 24, 4907-4924. 783
van der Ent, R. J., H. H. G. Savenije, B. Schaefli, and S. C. Steele-Dunne, 2010: Origin and 784
fate of atmospheric moisture over continents. Water Resour. Res., 46, W09525, 785
doi:10.1029/2010WR009127. 786
Wang, H., R. Fu, A. Kumar, and W. Li, 2010: Intensification of summer rainfall variability in 787
the southeastern United States during recent decades. J. Hydrometeorol, 11, 1007-1018. 788
Weaver, S. J., A. Ruiz-Barradas, and S. Nigam, 2009: Pentad evolution of the 1988 drought 789
33
and 1993 flood over the Great Plains: An NARR perspective on the Atmospheric and 790
terrestrial water balance. J. Climate, 22, 5366-5384. 791
Zhao, P., S. Yang, H. Wang, and Q. Zhang, 2011: Interdecadal relationships between the 792
Asian–Pacific Oscillation and summer climate anomalies over Asia, North Pacific, and 793
North America during a recent 100 years, J. Climate, 24, 4793-4799. 794
Zhu, Y., and R. E. Newell, 1998: A proposed algorithm for moisture fluxes from atmospheric 795
rivers. Mon. Weather Rev., 126, 725-735. 796
Zupanski, D., and F. Mesinger, 1995: 4-dimensional variation assimilation of precipitation 797
data. Mon. Weather Rev., 123, 1112-1127.798
34
Table captions 799
Table 1: Lag correlation coefficients among the monthly averages of the components of land 800
surface and atmospheric water budgets during 1979-2008. Negative (positive) indicates 801
that precipitation (first five rows) and soil moisture (last four rows) lead (lag) other 802
current components by the indicated month. Correlation coefficients in bold have p-803
values less than 0.01. 804
Figure captions 805
Figure 1: (a): Time series of annual precipitation anomalies over 1979-2008. (b), (c), and (d): 806
Spatial distribution of annual precipitation anomalies over the eastern US in 1981, 1986, 807
and 2007. The anomalies are based on the climatology period of 1979-2008. The study 808
region is shown by the box. 809
Figure 2: Time series of precipitation anomalies over the southeastern US during 1979-2008 810
for (a) summer, and (b) winter. 811
Figure 3: (a) and (b): Mean component of moisture fluxes during summer and winter. The 812
box represents the study region. (c) and (d): Climatology of transient eddy component of 813
moisture fluxes during summer and winter. The background shading is the moisture 814
convergence by each component. (e) and (f): Spatial distributions of summertime and 815
wintertime surface temperature averages over the eastern US and the adjacent oceans 816
during 1979-2008. To aid visualization, the vectors are shown at a coarser spatial 817
resolution (~150km) than the original data (32km). 818
Figure 4: The magnitude of moisture transport for the SE in terms of regional moisture 819
convergence and meridional and zonal moisture fluxes, for (a) the climatology (1979-820
2008), (b) six driest years, and (c) six wettest years, during summer (JJA) and winter 821
(DJF). Total column moisture flux convergence into the region is represented by the 822
middle box (positive: convergence and negative: divergence) and moisture fluxes at each 823
35
of the regions four boundaries are represented by the boxes in each cardinal direction. 824
Red (blue) bars are total column moisture transport in the zonal (meridional) direction. 825
Positive values in the zonal (meridional) direction represent westerly (southerly) fluxes. 826
Total column (dark shading) fluxes are separated into lower level (medium shading) and 827
upper level (light shading). 828
Figure 5: Same as Figure 4 except for the mean component (top) and the transient eddy 829
component (bottom) of total moisture fluxes over the study region. 830
Figure 6. Time series of anomalies of the components of the land surface and atmosphere 831
water budgets based on the climatology period of 1979-2007. Blue (red) represents 832
positive (negative) anomalies of each component. 833
Figure 7: Terrestrial and atmospheric water budget components over the SE region on a 834
monthly basis for the climatology (1979-2008), six driest years, and six wettest years ((a), 835
(b), and (c), respectively). The six driest and wettest years are selected based on summer 836
precipitation. Residuals are shown from the terrestrial water budget components (open 837
circle) and atmospheric water budget components (closed circle) and including the 838
moisture increments (open rectangle) over the climatology (1979-2008), six driest years, 839
and six wettest years ((d), (e), and (f), respectively). Components are precipitation (P), 840
evapotranspiration (E), runoff (Q), moisture flux convergence (MFC), and analysis 841
increment (INC). 842
Figure 8: Recycling ratio and recycled precipitation over the SE on a monthly basis over the 843
climatology (1979-2008; (a) and (c)), six driest years ((b) and (e)), and six wettest years 844
((c) and (f)). (a) and (d): the boxes represent a range of one standard deviation of the 845
recycling ratio and recycled precipitation. (b), (c), (e), and (f): the error bars represent the 846
minimum and maximum of the recycling ratio and the recycled precipitation on a 847
monthly basis during the six driest and wettest years. The circles represent their monthly 848
36
means. 849
Figure 9: The differences of total moisture fluxes (vector; mean + transient eddies) and the 850
geopotential height at 500 mb (background) between the drought and pluvial years ((a): 851
summer and (b): winter). 852
Figure 10. Geopotential heights anomalies at 850mb ((a) and (b)) and SST anomalies ((c) and 853
(d)) during summer drought and pluvial years. The black contours represent 854
climatological geopotential heights and the green contour, 1560 gpm, represents the 855
drought or pluvial year anomalies. The Bermuda High Index (Western Bermuda High 856
Index) is computed from the differences of sea level pressures (geopotential heights at 857
850mb) between the two circles (crosses). The north Pacific and Atlantic are represented 858
as a box in Figure (c). Hatched areas represent areas with at least four of the six driest or 859
wettest years having the same sign as their average. The climatology is computed based 860
on 1979-2008. 861
Figure 11: Leading and lagged composited maps for geopotential height (GH) anomalies at 862
500 mb relative to climatology (1979-2008) for drought summers and pluvial summers at 863
two month lead time ((a) and (b): AMJ), one month lead time ((c) and (d): MJJ), zero 864
month lead time ((e) and (f): JJA), one month lagged time ((g) and (h):JAS), and two 865
month lagged time ((i) and (j): ASO). Hatched areas represent areas with statistically 866
significant positive (negative) anomalies at 10% level based on Barnard’s exact test. 867
Figure 12: Same as Figure 10 except for winter. 868
Figure 13: Same as Figure 11 except for winter ((a) and (b): OND; (c) and (d): NDJ; (e) and 869
(f): DJF; (g) and (h): JFM; (i) and (j): FMA). 870
Figure 14: Time series of standardized anomalies of (left) atmospheric and land surface 871
conditions over the SE, and (right) atmospheric and oceanic conditions over surrounding 872
oceans, for the multi-year SE drought of 2005-2008.873
37
Table 1: Lag correlation coefficients among the monthly averages of the components of land 874 surface and atmospheric water budgets during 1979-2008. Negative (positive) indicates that 875 precipitation (first five rows) and soil moisture (last four rows) lead (lag) other current 876 components by the indicated month. Correlation coefficients in bold have p-values less than 877 0.01. 878
Lag Precipitation
-3 month -2 month -1 month 0 +1 month +2 month +3 month
MFC 0.062 0.084 0.060 0.850 0.127 0.121 0.082
SM 0.457 0.480 0.531 0.369 0.178 0.168 0.114
Evap. 0.305 0.354 0.285 0.032 0.089 0.053 -0.061
Runoff 0.185 0.268 0.385 0.592 0.238 0.130 0.045
Lag Soil moisture (0-200cm)
-3 month -2 month -1 month 0 +1 month +2 month +3 month
MFC 0.017 0.076 0.799 0.232 0.386 0.366 0.355
Evap. 0.548 0.592 0.634 0.614 0.529 0.477 0.409
Runoff 0.381 0.432 0.476 0.558 0.565 0.516 0.470
879