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