Abstract

Global and regional trends in drought for 1950–2000 are analyzed using a soil moisture–based drought index over global terrestrial areas, excluding Greenland and Antarctica. The soil moisture fields are derived from a simulation of the terrestrial hydrologic cycle driven by a hybrid reanalysis–observation forcing dataset. Drought is described in terms of various statistics that summarize drought duration, intensity, and severity. There is an overall small wetting trend in global soil moisture, forced by increasing precipitation, which is weighted by positive soil moisture trends over the Western Hemisphere and especially in North America. Regional variation is nevertheless apparent, and significant drying over West Africa, as driven by decreasing Sahel precipitation, stands out. Elsewhere, Europe appears to have not experienced significant changes in soil moisture, a trait shared by Southeast and southern Asia. Trends in drought duration, intensity, and severity are predominantly decreasing, but statistically significant changes are limited in areal extent, of the order of 1.0%–7.0% globally, depending on the variable and drought threshold, and are generally less than 10% of continental areas. Concurrent changes in drought spatial extent are evident, with a global decreasing trend of between −0.021% and −0.035% yr−1. Regionally, drought spatial extent over Africa has increased and is dominated by large increases over West Africa. Northern and East Asia show positive trends, and central Asia and the Tibetan Plateau show decreasing trends. In South Asia all trends are insignificant. Drought extent over Australia has decreased. Over the Americas, trends are uniformly negative and mostly significant.

Within the long-term trends there are considerable interannual and decadal variations in soil moisture and drought characteristics for most regions, which impact the robustness of the trends. Analysis of detrended and smoothed soil moisture time series reveals that the leading modes of variability are associated with sea surface temperatures, primarily in the equatorial Pacific and secondarily in the North Atlantic. Despite the overall wetting trend there is a switch since the 1970s to a drying trend, globally and in many regions, especially in high northern latitudes. This is shown to be caused, in part, by concurrent increasing temperatures. Although drought is driven primarily by variability in precipitation, projected continuation of temperature increases during the twenty-first century indicate the potential for enhanced drought occurrence.

1. Introduction

Drought can be regarded as one of the most damaging of natural disasters in human, environmental, and economic terms. It occurs as a result of extremes in climate that are driven by natural variability but may be exacerbated or dampened by anthropogenic influences. The variability of global climate is driven in the main by El Niño–Southern Oscillation (ENSO), which impacts the tropics and many regions in midlatitudes (Ropelewski and Halpert 1987). Other climate oscillations and modes of large-scale variability, such as the North Atlantic Oscillation (NAO), the Pacific decadal oscillation (PDO), and the Atlantic Multidecadal Oscillation (AMO), act on generally longer time scales and interact with ENSO or are the primary climate drivers elsewhere and more regional in their impacts. For example, the NAO is known to affect climate in eastern North America and Europe (Hurrell and VanLoon 1997) as well as North Africa (Wang 2003). The PDO (Mantua et al. 1997) is a primary driver of climate around the Pacific basin and interacts with ENSO, resulting in modifications of climate globally (Newman et al. 2003; Verdon and Franks 2006). The AMO (Kerr 2000) impacts on the North Atlantic and especially on North American (Enfield et al. 2001; McCabe et al. 2004) and European climate (Sutton and Hodson 2005), and is a potential modulating force of ENSO (Dong et al. 2006).

Of considerable interest is the change in variability and extremes under recent and future global warming and the potential acceleration of the water cycle, which may act to alter the occurrence and severity of drought. As temperatures rise, the capacity of the atmosphere to hold moisture would increase as governed by the Clausius–Clapeyron equation (Held and Soden 2000), with potential for increased evaporation and/or precipitation (Trenberth 1999), although these may be limited by other factors such as available energy and aerosol concentration. Climate model studies have shown that variability is likely to increase under plausible future climate scenarios (Wetherald and Manabe 2002), dependent upon climate sensitivity, with large regional changes in the water cycle. The potential for more droughts and of greater drought severity is a worrisome possibility (Wetherald and Manabe 1999; Wang 2005).

Huntington (2006) reviews the observational evidence so far for water cycle intensification to date and concludes that, despite some contradictions, the overall picture points toward intensification. For drought specifically, trends have been analyzed over the past 50–100 yr at regional (e.g., Lloyd-Hughes and Saunders 2002; Rouault and Richard 2005; Andreadis and Lettenmaier 2006) and global scales (Dai et al. 2004). When analyzing the Palmer Drought Severity Index (PDSI) and the Standardized Precipitation Index (SPI) over Europe, Lloyd-Hughes and Saunders (2002) found insignificant change in the proportion of land experiencing medium to extreme drought during the twentieth century. A drought analysis of South African SPI by Rouault and Richard (2005) found a substantial increase in 2-yr droughts since the 1970s. They also found interdecadal variability in the spatial extent of drought since the beginning of the century, the most severe of which is associated with ENSO. Andreadis and Lettenmaier (2006) analyzed a long-term (1915–2003) hydrological simulation over the United States and found a general increasing trend in soil moisture, with concurrent decrease in drought duration and extent, except for the Southwest and parts of the West. Dai et al. (2004) showed the global pattern of trends in annual PDSI and found that generally drier conditions have prevailed since the 1970s.

In this paper, we investigate variability and trends in soil moisture and drought characteristics, globally and regionally over the second half of the twentieth century. The analysis is based on a global soil moisture dataset derived from a model simulation of the terrestrial hydrologic cycle. The simulation is driven by a hybrid observation–reanalysis-based meteorological dataset and provides a globally consistent and physically based view of moisture availability. As drought can be described by any one or a combination of characteristics, and these are important to varying degrees depending on the situation, we are interested in changes in a number of aspects of drought. These include duration, intensity, and severity, which are dependent on the threshold for defining drought that is specific to the application. We focus on how soil moisture and drought characteristics vary at annual to decadal time scales, and whether there are any significant trends over the second half of the twentieth century. Intuitively, changes in precipitation will be the primary driver of variability in drought but will be modified by temperature changes, which is especially relevant given recent and potential future increases in surface air temperature. We therefore investigate the direct (precipitation, temperature) and indirect (large-scale climate oscillations) forcing mechanisms to understand what is driving changes and variability in soil moisture and drought occurrence.

2. Datasets and methods

To represent drought globally, we use soil moisture fields from a land surface hydrological model simulation driven by observation-based meteorological forcings (Sheffield and Wood 2007). Soil moisture balances the fluxes of precipitation, evapotranspiration and runoff and thus provides an aggregate measure of water availability and drought. In drought terminology, soil moisture falls somewhere in between meteorological and hydrological drought and may be representative of agricultural drought through its control on transpiration and thus vegetative vigor. We calculate an index of drought as the deficit of soil moisture relative to its seasonal climatology (Sheffield et al. 2004a). Wet spells can be calculated similarly but as the surplus of soil moisture. The simulation and the derivation of the drought index and related statistics are briefly described next. Further details can be found in Sheffield et al. (2004a).

a. Land surface hydrological simulation

The Variable Infiltration Capacity (VIC) land surface model (Liang et al. 1994; Cherkauer et al. 2002) was used to generate spatially and temporally consistent fields of soil moisture and other water budget flux and state variables. The VIC model simulates the terrestrial water and energy balances and distinguishes itself from other land surface schemes through the representation of subgrid variability in soil storage capacity as a spatial probability distribution, to which surface runoff is related, and by modeling base flow from a lower soil moisture zone as a nonlinear recession. The VIC model has been applied extensively at regional (e.g., Maurer et al. 2002) and global scales (Nijssen et al. 2001; Sheffield et al. 2004b).

For this study, the VIC model was run globally at 1.0° spatial resolution and 3-hourly time step for the period 1950–2000. This simulation was forced by a hybrid dataset of precipitation, near-surface meteorological and radiation data derived from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996) and a suite of global observation-based products. In effect, the subdaily variations in the reanalysis are used to downscale the monthly observations. These observations, which are generally available at higher spatial resolution, are concurrently used to downscale the reanalysis in space. Known biases in the reanalysis precipitation and near-surface meteorology were corrected at the monthly scale using observation-based datasets of precipitation, air temperature, and radiation. Corrections were also made to the rain day statistics of the reanalysis precipitation, which have been found to exhibit a spurious wavelike pattern in high-latitude wintertime. Other meteorological variables (downward short- and longwave, specific humidity, surface air pressure, and wind speed) were downscaled in space with account for changes in elevation. The forcing dataset is described in detail by Sheffield et al. (2006). The simulation has been validated against available observations of terrestrial hydrology (J. Sheffield and E. F. Wood 2007, unpublished manuscript), including in situ measurements of soil moisture, large basin streamflow, and remote sensing–based snow datasets.

Given recent and future potential increases in air temperature, we also carried out a second simulation to investigate the impact of trends in temperature on the drought trends. Higher temperatures will increase potential evapotranspiration and possibly result in increased drought occurrence, although actual changes will be controlled by available moisture from precipitation and be modified by temperature impacts on snow. Following Hamlet et al. (2007) and Dai et al. (2004), we forced the VIC model with climatological surface air temperature instead of annually varying values. In this way, any differences in the trends in soil moisture and drought characteristics between the two simulations would be attributable to trends in temperature. In the discussion in section 5b, the original simulation with annually varying air temperature forcing is referred to as TANN and the simulation with climatological air temperature as TCLIM.

b. Relationship with previous studies of drought using the VIC model

Previously, soil moisture fields from a retrospective simulation of the VIC model for the United States (Maurer et al. 2002) have been analyzed in terms of drought occurrence by Sheffield et al. (2004a), who found that the simulated soil moisture values were able to represent historic drought events, display coherency and sufficient detail at small space scales, and compare well with standard drought indices such as the PDSI. The PDSI is one of the most widely used drought indices both operationally and in climate research (Dai et al. 2004; Burke et al. 2006) and uses a generic two-layer soil model to describe the cumulative departure of moisture supply (Palmer 1965). In snow-dominated regions, Sheffield et al. (2004a) found that the VIC-based dataset and the PDSI dataset diverged, likely because of inadequate representation of cold season processes in the calculation of the PDSI.

The simulations analyzed in Sheffield et al. (2004a) and in this paper were both generated by the VIC model (albeit slightly different versions) but differ substantially in terms of their domain (United States versus global), the meteorological forcings (gauge based versus a hybrid reanalysis–observation dataset), spatial resolution (0.125° versus 1.0°), and parameter data (different underlying datasets for the soil and vegetation distributions). Despite this, comparison of their representation of drought over the United States shows good agreement with respect to major drought events (not shown). Furthermore, the trends described in section 3 are consistent with Andreadis and Lettenmaier (2006) who used an extended version of the Maurer et al. (2002) U.S. dataset at 0.5° resolution. Work in progress has compared the 0.5° extended dataset with this global dataset in the framework of severity–area–duration curves (Andreadis et al. 2005) and shows close agreement that is encouraging given the differences in the simulations.

c. Soil moisture–based drought index

The drought index is calculated using the method of Sheffield et al. (2004a) and is briefly described here. Simulated soil moisture data at multiple model soil layers are aggregated over the total soil column, converted to volumetric values, and averaged to monthly values. For each model grid cell and month, a beta distribution is fitted to the 51 monthly values (1 value for each year in 1950–2000) using the method of moments. The current level of drought or wetness for a particular month and point in space can then be gauged relative to this fitted distribution or climatology. A drought is defined as a period of duration D months with a soil moisture quantile value, q(θ), less than an arbitrary threshold level, q0(θ), preceded and followed by a value above this level. The departure below this level at any particular time is the drought magnitude:

 
formula

and the mean magnitude over the drought duration is the intensity:

 
formula

The product of duration and intensity gives the drought severity:

 
formula

acknowledging that the impacts of drought are a balance between the length and the intensity of deficits. We also define classes of drought event based on their duration as follows:

 
formula

where the subscript to D indicates the range of drought duration in months. A climatological analysis of this dataset is given in Sheffield and Wood (2007).

3. Trends in soil moisture and drought

a. Trends in soil moisture

Trends are calculated using the nonparametric Mann–Kendall trend test (Mann 1945; Kendall 1975; Hirsch and Slack 1984), which is robust and distribution independent. We tested for serial correlation in the monthly data, which would invalidate the assumption of independent data. The areal extent of statistically significant serial correlation (0.01 level) is between 12% and 17% depending on the month, with about 50% of this area in drier regions and the majority of the remainder in very high latitudes. Therefore, the area that potentially invalidates the independence assumption is small and generally restricted to drier regions, such as the Sahara, which we ignore in the analysis. Figure 1 shows a map of the trends in annual volumetric soil moisture on a grid by grid basis. Results for the Sahara and other desert regions have been masked out based on a threshold of mean annual precipitation <0.5 mm day−1 to screen out serially correlated data and trend values that are essentially zero but are picked up by the ranked-based test. Table 1 summarizes trends of regional averaged time series. The regions are defined by Giorgi and Francisco (2000) and are shown in Fig. 2. For brevity, these regions may be referred to by acronyms that are defined in Table 1. The GRL region was originally defined as Greenland and northeastern Canada, but as the VIC model is not designed to simulate permanent ice sheets and glaciers we exclude the interior of Greenland from the definition of the GRL region and rename it northeastern Canada (NEC). We discuss the trend results in relation to precipitation and temperature trends in the forcing dataset (Fig. 3), which are calculated in the same manner as for the soil moisture quantiles.

Fig. 1.

Global distribution of linear trends in annual mean volumetric soil moisture, 1950–2000, calculated using the Mann–Kendall nonparametric trend test. Regions with mean annual precipitation less than 0.5 mm day−1 have been masked out because the VIC model simulates small drying trends in desert regions that, despite being essentially zero are identified by the nonparametric test. The trends in the bottom panel have been filtered for significance at the 0.05 level.

Fig. 1.

Global distribution of linear trends in annual mean volumetric soil moisture, 1950–2000, calculated using the Mann–Kendall nonparametric trend test. Regions with mean annual precipitation less than 0.5 mm day−1 have been masked out because the VIC model simulates small drying trends in desert regions that, despite being essentially zero are identified by the nonparametric test. The trends in the bottom panel have been filtered for significance at the 0.05 level.

Table 1.

Nonparametric trends in regional average soil moisture quantile, precipitation, and surface air temperature. Trends values in bold are significant at the 0.05 level. Statistical test values are given in parentheses.

Nonparametric trends in regional average soil moisture quantile, precipitation, and surface air temperature. Trends values in bold are significant at the 0.05 level. Statistical test values are given in parentheses.
Nonparametric trends in regional average soil moisture quantile, precipitation, and surface air temperature. Trends values in bold are significant at the 0.05 level. Statistical test values are given in parentheses.
Fig. 2.

Map of regions used in the analysis as defined by Giorgi and Francisco (2000). The GRL region has been modified from its original definition that covered Greenland and eastern Canada to exclude the interior of Greenland. This is because the VIC model is not designed to simulate permanent ice caps and glaciers. The region is renamed NEC.

Fig. 2.

Map of regions used in the analysis as defined by Giorgi and Francisco (2000). The GRL region has been modified from its original definition that covered Greenland and eastern Canada to exclude the interior of Greenland. This is because the VIC model is not designed to simulate permanent ice caps and glaciers. The region is renamed NEC.

Fig. 3.

Global distribution of linear trends, 1950–2000, in surface air temperature and precipitation as used to force the VIC model. The data are taken from the dataset of Sheffield et al. (2006), which are based on the CRU TS2.0 dataset of Mitchell and Jones (2005). The trends in the right-hand panels have been filtered for significance at the 0.05 level.

Fig. 3.

Global distribution of linear trends, 1950–2000, in surface air temperature and precipitation as used to force the VIC model. The data are taken from the dataset of Sheffield et al. (2006), which are based on the CRU TS2.0 dataset of Mitchell and Jones (2005). The trends in the right-hand panels have been filtered for significance at the 0.05 level.

At global scales, the trend in soil moisture is positive (wetting). Generally speaking, wetting trends occur in the Americas, Australia, Europe, and western Asia, and negative trends (drying) occur in Africa and parts of eastern Asia. The trends are generally collocated with equivalent trends in precipitation (Fig. 3). Statistically significant trends in soil moisture at the 0.05 level are, however, restricted to relatively small subareas of these continents. Regions of wetting trends are evident in the central Northern Territories of Canada (up to 0.2% vol yr−1), central USA (0.05%–0.2% vol yr−1) and northern Mexico (<0.1% vol yr−1). These are coincident with statistically significant increasing trends in precipitation (Fig. 3). For northern Canada, increased precipitation has also been noted by Zhang et al. (2000) and McBean et al. (2005). Nevertheless, the simulated increase in soil moisture (and increase in precipitation) appears contradictory to decreasing river discharge into the Arctic and North Atlantic from Canadian rivers since the mid-1960s (Déry and Wood 2005a). This could be explained by increasing evapotranspiration driven by higher temperatures (Zhang et al. 2000) that results in increased precipitation and decreased streamflow, although changes to snow will complicate this. However, calculation of soil moisture trends over a similar period (1964–2000) as used by Déry and Wood (2005a) shows widespread decreasing (but not always significant) trends over much of northern Canada that is consistent with decreasing streamflow. This indicates the large influence of increasing moisture during 1950–63 on the overall trend, which is further discussed in general in section 4b. The trends over the United States are also consistent, with increases in precipitation and soil wetness during the twentieth century reported by Groisman et al. (2004) and Andreadis and Lettenmaier (2006). Scattered regions in Brazil and Columbia and a large part of central Argentina show increasing trends up to 0.1% vol yr−1. These are generally collocated with increasing precipitation trends, and the trends are consistent with increased streamflow in large South American basins (Garcia and Mechoso 2005) and increased precipitation and streamflow in the La Plata basin (Berbery and Barros 2002). In the Western Hemisphere, parts of Scandinavia, eastern Europe, and western Russia show increasing trends (up to 0.2% vol yr−1). A few small regions of significant increasing trend of up to 0.4% vol yr−1 occur in western China in the northern Tibetan Plateau. In western Australia there are significant increasing trends up to 0.25% vol yr−1. These trends are generally consistent with increasing precipitation in these regions, also noted over the twentieth century by Dai et al. (1997).

Drying trends are most prominent in the Sahel (up to −0.6% vol yr−1 for individual grid cells), which has been well documented in terms of precipitation deficits during the 1970s and 1980s (Hulme 1992; L’Hôte et al. 2002). Also, significant decreasing trends occur in parts of central Africa and in southern Africa (up to −0.15% vol yr−1 through Angola and Zambia) that coincide with the southern extent of the ITCZ and are again collocated with decreasing trends in precipitation. Although the Arctic as a whole has likely experienced increased precipitation over the latter half of the twentieth century (McBean et al. 2005), several regions show significant drying trends, such as northern and southeastern Alaska. The majority of northeastern Asia shows drying trends (up to −0.2% vol yr−1) that are coincident with decreasing precipitation as shown in Fig. 3 and reported by McBean et al. (2005), although the area of statistically significant values is relatively small, being restricted to the central Yenesei and eastern Amur basins and far northeastern Siberia. The consistency of these trends with observed changes in related variables is unclear, as there is lack of consistency between increasing Arctic discharge from Siberian rivers (Peterson et al. 2002; Shiklomanov et al. 2006) and precipitation (Berezovskaya et al. 2004) and much debate over the impact of other processes such as changes in permafrost and fires (McClelland et al. 2006). Significant drying trends are also apparent in northern China and parts of Southeast Asia, up to −0.2% vol yr−1. This is consistent with Zou et al. (2005), who analyzed trends in PDSI data in China during 1951–2003 and found no significant changes except for northern regions.

To investigate whether the trends vary by season, which is more likely in monsoonal regions and continental interiors where interseasonal climate variability is relatively high, we also calculated trend values for each season separately (Fig. 4). Over the United States, the overall increasing trend is most prominent in winter months. In South America, the tendency is for higher trends in Argentina during the Austral summer [December–February (DJF)] and autumn [March–May (MAM)] whereas trends in Brazil and elsewhere in the north are greater in the drier seasons [June–July (JJA), September–November (SON)]. Over Africa, the largest trends tend to coincide with the peak or retreat of the ITCZ (JJA and SON over the Sahel; DJF and MAM in central and southern Africa). The few scattered regions of significant trends in Europe are mainly restricted to the winter months. Decreasing trends in far northeastern Siberia are dominant in the summer, whereas increasing trends east of the Urals are dominant in the spring. Decreasing trends in China are highest in the DJF–MAM, and decreasing trends in northern India and Southeast Asia are highest in the Monsoon season (JJA–SON). In Australia, increasing trends in the west are highest in the Austral summer (DJF).

Fig. 4.

Global distribution of linear trends in seasonal mean volumetric soil moisture, 1950–2000, calculated using the Mann–Kendall nonparametric trend test. Regions with mean annual precipitation less than 0.5 mm day−1 have been masked out in the same way as for Fig. 1. The bottom panel shows the seasonal range in trends calculated as the maximum minus the minimum trend of the four seasonal values at each grid cell.

Fig. 4.

Global distribution of linear trends in seasonal mean volumetric soil moisture, 1950–2000, calculated using the Mann–Kendall nonparametric trend test. Regions with mean annual precipitation less than 0.5 mm day−1 have been masked out in the same way as for Fig. 1. The bottom panel shows the seasonal range in trends calculated as the maximum minus the minimum trend of the four seasonal values at each grid cell.

b. Global trends in drought characteristics

Next we investigate trends in droughts characteristics (duration, magnitude, and severity) over the 50-yr period. Trends are again calculated using the Mann–Kendall nonparametric test. Individual drought events are assumed to be independent and the period between events is calculated from the beginning of an event to the beginning of the next. We tested for serial correlation in the characteristics of events and found that about 5% or less of grid cells had significant values (0.05 level) and, similar to the results for soil moisture, that most of these were located in the Sahara region which we ignore in the analysis. The geographic distribution of trends in drought duration, intensity, and severity is shown in Fig. 5 for statistically significant trends only, at the 0.05 level for soil moisture quantile threshold q0(θ) = (10.0, 50.0), corresponding to severe and mild drought, respectively. Table 2 shows the percent area of each continent that has statistically significant trends in drought characteristics. In general, hydrological and meteorological variables exhibit spatial correlation, which reduces the number of independent data points over a region. To determine the field significance of the area of significant trends, we estimated the distribution of trend areas using a bootstrap approach in which we generate 1000 time series of soil moisture fields by resampling from the original dataset. Tests over a single region showed that 1000 samples were sufficient to give stable results. The area of significant trends was then calculated for each resampled series and the 95th percentile calculated from the total sample. If the original trend area is greater than this percentile value it is field significant.

Fig. 5.

Nonparametric trend in drought duration D, intensity I, and severity S for q0(θ) = {10.0%, 50.0%} threshold values for 1950–2000. Note that the units have been scaled by 10 000, 1000, and 100, respectively. Regions with average precipitation <0.5 mm day−1 have been masked out.

Fig. 5.

Nonparametric trend in drought duration D, intensity I, and severity S for q0(θ) = {10.0%, 50.0%} threshold values for 1950–2000. Note that the units have been scaled by 10 000, 1000, and 100, respectively. Regions with average precipitation <0.5 mm day−1 have been masked out.

Table 2.

Percent area with statistically significant positive or negative trends in drought duration (D), intensity (I), and severity (S) for q0(θ) = (10.0, 50.0%). The values in parentheses are field significance calculated using a bootstrap resampling approach.

Percent area with statistically significant positive or negative trends in drought duration (D), intensity (I), and severity (S) for q0(θ) = (10.0, 50.0%). The values in parentheses are field significance calculated using a bootstrap resampling approach.
Percent area with statistically significant positive or negative trends in drought duration (D), intensity (I), and severity (S) for q0(θ) = (10.0, 50.0%). The values in parentheses are field significance calculated using a bootstrap resampling approach.

Overall, the area that has undergone statistically significant changes in drought characteristics is small (Table 2). In general, the area of negative trends is greater than that of positive trends, a result of the global wetting trend in soil moisture. Significant trends are more spatially extensive for q0(θ) = 50%, a result influenced by the greater number of droughts at this threshold value. Also, the area of significant trends in drought severity is generally greater than that for drought intensity, which is in turn greater than that for duration. Globally, only 0.6%–4.1% of the land has experienced increasing trends in duration, intensity and severity, and 1.8%–6.8% has experienced decreasing trends. At continental scales, Africa is dominated by significant increasing trends in drought severity [area = 10.4% for q0(θ) = 50%]. Over Asia, the areas of significant decreasing trends at q0(θ) = 50% are highest for duration (4.5%), intensity (4.0%), and severity (4.0%). Negative trends tend to dominate over Europe, especially for drought duration [area = 9.6% for q0(θ) = 50%]. Elsewhere, decreasing trends are more prevalent in North America (e.g., area of decreasing trends in duration = 8.3%) and Oceania especially for drought intensity [area = 10.3% for q0(θ) = 50%]. In South America, the area of decreasing trends is dominant, especially for q0(θ) = 50% (8.1%, 8.1%, and 9.7% for duration, intensity, and severity, respectively).

c. Regional trends in drought spatial extent

Table 3 gives trends in the spatial extent of drought for the world and regionally, for various threshold values. Globally, there is an overall decreasing trend in drought extent of −0.021 to −0.035% yr−1, although only for a threshold q0(θ) ≤ 20% are the trends statistically significant at the 0.05 level. Regionally there are distinct differences in the sign of the trends as well as the magnitude and statistical significance that are consistent with the trends in soil moisture. Over northern Europe the tendency is for a decrease in spatial extent, although only trends for q0(θ) ≤ 20.0 are significant, and in the Mediterranean region all trends are positive but essentially statistically zero. Lloyd-Hughes and Saunders (2002) analyzed PDSI and SPI over Europe, and similarly found generally insignificant change in the proportion of land experiencing medium to extreme drought during the twentieth century. West Africa is dominated by events in the Sahel, which result in large increases in spatial extent that are approximately proportional to the threshold. For example, for q0(θ) = 50% the trend is 0.527% yr−1, which translates into about 28% increase over the full period. Although eastern Africa shows consistently increasing trends also (up to 0.15% yr−1), they are only significant for q0(θ) ≥ 40%. In southern Africa (SAF), positive trends of 0.038%–0.234% yr−1 are only significant at the 0.1 level.

Table 3.

Trends in the spatial extent of drought for various q0(θ) values. The trends are calculated using the Mann–Kendall nonparametric test. Trend values in bold are significant at the 0.05 level.

Trends in the spatial extent of drought for various q0(θ) values. The trends are calculated using the Mann–Kendall nonparametric test. Trend values in bold are significant at the 0.05 level.
Trends in the spatial extent of drought for various q0(θ) values. The trends are calculated using the Mann–Kendall nonparametric test. Trend values in bold are significant at the 0.05 level.

Over the northern part of Asia (regions NAS, CAS, TIB, and EAS) the picture is mixed, with positive and significant trends over northern Asia, positive but insignificant trends over eastern Asia and negative trends over central Asia and the Tibetan Plateau, although only over TIB are the majority of the trends significant. All trends in southern Asia are insignificant with negative trends for SEA and positive for SAS. The Australian region trends are negative and all significant ranging from −0.08% to −0.32% yr−1. For North America, the trends in spatial extent are uniformly negative and almost always significant [the exceptions are in WNA for q0(θ) = (10, 20), ENA for q0(θ) = (10, 20, 30) threshold, and ALA for q0(θ) = (40, 50) threshold]. The largest trends are in CNA for q0(θ) = 50% (approximately −0.4% yr−1 or 19% decrease in spatial extent over the full period) and NEC for q0(θ) = 50% (approximately −0.5% yr−1 or 26% decrease). Over south America, all trends are negative, with all AMZ trends significant, but only trends for q0(θ) ≤ 30% significant in CAM and for q0(θ) ≥ 30% in SSA.

d. Epochal changes in drought frequency

We next calculated the change in drought statistics between the first (1950–75) and second (1976–99) halves of the simulation period (Fig. 6), which shows that the number of droughts has tended to decrease over most parts of the world. For short-term droughts, D3–6, midlatitudes are dominated by decreases, most notably in central and eastern Europe, Australia, southern South America, and central North America. Increases are evident across southern Canada, southwest Europe, and across northern Russia and most of Siberia, although these are localized. The pattern for medium-term (D6–12) droughts is more organized, with large and spatially coherent decreases across most of Alaska and northern Canada, eastern Europe and western Russia, subtropical Asia, and central Australia. Conversely, a large expanse of increased frequency traverses Siberia. The number of long-term droughts (D12) in both epochs is limited to a few regions (northern Canada, Tibetan Plateau) and the changes are all decreasing. Mean drought duration has decreased in northern midlatitudes and northern Canada, but has increased over the northwest United States and large regions of Siberia. In Africa, mean drought duration increased in the Sahel and southern Africa. For drought intensity the changes are generally small and localized, and where they are more prominent they tend to collocate with regions of large changes in mean drought length. For drought severity, the distribution of changes is similar to that for mean drought duration, and given that changes in mean drought intensity are relatively small, the indication is that changes in drought severity are driven mainly by changes in drought duration.

Fig. 6.

Percent change in frequency of short- (D3–6), medium- (D6–12), and long-term (D12+) duration droughts and mean drought duration (D), intensity (I), and severity (S) for q0(θ) = 10.0%, between 1950–75 and 1976–99.

Fig. 6.

Percent change in frequency of short- (D3–6), medium- (D6–12), and long-term (D12+) duration droughts and mean drought duration (D), intensity (I), and severity (S) for q0(θ) = 10.0%, between 1950–75 and 1976–99.

4. Temporal variability of soil moisture and drought

a. Regional temporal variability

Within the long-term linear trends identified there is considerable variability at interannual to decadal time scales. Figure 7 shows the temporal variation of several drought characteristics {mean drought duration, D for [q0(θ) = 50.0%]; number of short, high-intensity droughts [q0(θ) = 10.0%, D = D1–3]; number of long, low-intensity droughts [q0(θ) = 50.0%, D = D12+]; spatial extent of drought [q0(θ) = 50.0%, D = D1] calculated for an 11-yr moving window. The drought characteristics are averaged over each region. The time series for long, low-intensity droughts is multiplied by a factor of 3 to aid visualization. At global scales, there is little variation over the time period in drought duration and frequency, although the spatial extent of drought has swung from the high of the 1950s through a low in the 1960–70s and back up again in recent years.

Fig. 7.

Regional average time series of various drought statistics: mean duration of drought [q0(θ) = 50.0%], number of droughts [q0(θ) = 10.0%, D = D1–3], number of droughts [q0(θ) = 50.0%, D = D12+], and spatial extent of drought [% area, q0(θ) = 50.0%, D = D1]. The statistics are calculated over an 11-yr moving window and are plotted at the center of the window. The data series for the number of long-term droughts [q0(θ) = 50.0%, D = D12+] are multiplied by 3 for ease of comparison.

Fig. 7.

Regional average time series of various drought statistics: mean duration of drought [q0(θ) = 50.0%], number of droughts [q0(θ) = 10.0%, D = D1–3], number of droughts [q0(θ) = 50.0%, D = D12+], and spatial extent of drought [% area, q0(θ) = 50.0%, D = D1]. The statistics are calculated over an 11-yr moving window and are plotted at the center of the window. The data series for the number of long-term droughts [q0(θ) = 50.0%, D = D12+] are multiplied by 3 for ease of comparison.

Figure 8 shows the spatial loadings of the first three principal components of monthly soil moisture quantiles and represents the major modes of variability in global drought and wet spells. Figure 9 shows smoothed time series of these principal components. Although the variance explained by the first three components is small, the value decreases rapidly for higher order components, and so we show only the first three components for brevity. PC1 explains 8.1% of the global variability in soil moisture and is negative in the Amazon, the Sahel, southern Africa, northeastern Siberia, Southeast Asia, and Australia among other places. It is positive in southern South America, central and southern United States, northern Canada, and central and eastern Asia. The distribution of loadings for PC1 reminisce ESNO impacts. In fact the time series in Fig. 9 is correlated with Niño-3.4 SST variability (r = 0.50) and so, notwithstanding the variability associated with the overall trend, tropical Pacific temperatures appear to be the primary driver of global variability in soil moisture, a result also shown by Dai et al. (2004) for PDSI data. PC2 explains 6.4% of the variance and shows strong positive loadings over northern Canada, the Amazon, southern Africa, and central Australia. Negative loadings are highest in Alaska, northern Europe, and far eastern Asia. The time series of PC2 shows a low-frequency multidecadal oscillation that covaries well with the AMO index (r = 0.67) and is consistent with the PDSI analysis of McCabe and Palecki (2006). The third component, PC3, explains 6.0% of the variance and shows strong positive loadings over central Europe through Russia, Australia, South America and Alaska, and strong negative loadings over northern Siberia, east Africa, and northeast Canada. The decadal variability follows the NAO somewhat (Fig. 9), however the correlation is insignificant (r = 0.21) and especially weak during the mid-1980s–mid-1990s.

Fig. 8.

Spatial loadings of the first three principal components of annual soil moisture quantile for 1950–2000. The amount of variance explained by each component is also given.

Fig. 8.

Spatial loadings of the first three principal components of annual soil moisture quantile for 1950–2000. The amount of variance explained by each component is also given.

Fig. 9.

Smoothed time series of the first three principal components of monthly soil moisture quantile for 1950–2000 compared to climate indices. Niño-3.4 SSTs are defined as over 5.0°S–5.0°N, 120.0°–170.0°W. AMO: Atlantic multidecadal oscillation defined as detrended area weighted average SSTs over the North Atlantic, 0°–70°N. NAO: DJFM wintertime North Atlantic Oscillation defined as the difference of normalized sea level pressure between stations in Portugal and Iceland.

Fig. 9.

Smoothed time series of the first three principal components of monthly soil moisture quantile for 1950–2000 compared to climate indices. Niño-3.4 SSTs are defined as over 5.0°S–5.0°N, 120.0°–170.0°W. AMO: Atlantic multidecadal oscillation defined as detrended area weighted average SSTs over the North Atlantic, 0°–70°N. NAO: DJFM wintertime North Atlantic Oscillation defined as the difference of normalized sea level pressure between stations in Portugal and Iceland.

Regionally, considerable variation in drought characteristics is evident (Fig. 7), and we can also relate the variations in soil moisture to large-scale climate oscillations that act at interannual to decadal time scales. Table 4 shows the correlation between regional principal components of smoothed monthly soil moisture quantiles and various climate indices. The soil moisture data are detrended to avoid spurious correlations at multidecadal time scales and the PCs are smoothed using a 13-month moving window. It should be noted that the large extent of some regions may hide any strong connectivity at smaller scales.

Table 4.

Correlation between regional principal components of detrended annual soil moisture quantile and various climate indices. The climate indices are also annual values and are compared year for year (51 values) with no lag relative to the soil moisture data. The correlations are the maximum values for the first three PCs and those in bold are statistically significant at the 0.01 level. Niño-3.4: SST anomalies in the Niño-3.4 region, 5.0°S–5.0°N, 120.0°–170.0°W. PDO: Pacific decadal oscillation defined as the leading PC of monthly SST anomalies in the North Pacific Ocean, poleward of 20°N. AMO: Atlantic multidecadal oscillation defined as detrended area weighted average SSTs over the North Atlantic, 0°–70°N. NAO: DJFM wintertime North Atlantic Oscillation defined as the difference of normalized sea level pressure between Portugal and Iceland.

Correlation between regional principal components of detrended annual soil moisture quantile and various climate indices. The climate indices are also annual values and are compared year for year (51 values) with no lag relative to the soil moisture data. The correlations are the maximum values for the first three PCs and those in bold are statistically significant at the 0.01 level. Niño-3.4: SST anomalies in the Niño-3.4 region, 5.0°S–5.0°N, 120.0°–170.0°W. PDO: Pacific decadal oscillation defined as the leading PC of monthly SST anomalies in the North Pacific Ocean, poleward of 20°N. AMO: Atlantic multidecadal oscillation defined as detrended area weighted average SSTs over the North Atlantic, 0°–70°N. NAO: DJFM wintertime North Atlantic Oscillation defined as the difference of normalized sea level pressure between Portugal and Iceland.
Correlation between regional principal components of detrended annual soil moisture quantile and various climate indices. The climate indices are also annual values and are compared year for year (51 values) with no lag relative to the soil moisture data. The correlations are the maximum values for the first three PCs and those in bold are statistically significant at the 0.01 level. Niño-3.4: SST anomalies in the Niño-3.4 region, 5.0°S–5.0°N, 120.0°–170.0°W. PDO: Pacific decadal oscillation defined as the leading PC of monthly SST anomalies in the North Pacific Ocean, poleward of 20°N. AMO: Atlantic multidecadal oscillation defined as detrended area weighted average SSTs over the North Atlantic, 0°–70°N. NAO: DJFM wintertime North Atlantic Oscillation defined as the difference of normalized sea level pressure between Portugal and Iceland.

In northern Europe, a decadal oscillation, most evident in the spatial extent, underlies the small decreasing trend in drought (section 3c), with peaks in the 1950s, 1970s, and 1990s (Fig. 7). Lloyd-Hughes and Saunders (2002) and van der Schrier et al. (2006) also found the 1950s and 1990s to be the most drought-prone periods in terms of PDSI and 3- and 12-month SPI. For the Mediterranean, although overall trends are insignificant, there is a slight decreasing trend until the early 1980s when there is a sharp increase in frequencies, especially for drought spatial extent. The underlying soil moisture variability in the Mediterranean appears to be weakly correlated with the NAO (r = 0.56), as documented previously (e.g., Rodo et al. 1997). The relationship in northern Europe is insignificant (r = −0.19, Table 4), which may be a result of the variability in strength of connectivity across the region and seasonally (Uvo 2003).

The change in the number of droughts over West Africa is dominated by the increasing trend up to the mid-1980s but there is a large reversal in trend in the following years. Correlations between the detrended soil moisture time series and the AMO are modest (r = −0.46 for PC1) but are consistent with observational and model-based studies (Zhang and Delworth 2006). Previous studies have shown that NAO has some influence over the climate in this region (Oba et al. 2001), although there is contradictory evidence (Wang 2003) and the correlation here is weakly significant (r = −0.43 for PC2), which reflects this uncertainty. A similar, but less pronounced, picture is apparent in East Africa, with a noticeable decrease in drought frequency in the 1950–60s. The AMO and Niño-3.4 SSTs provide the highest, but weak, correlations (r = 0.43 and 0.48, respectively). In southern Africa, peaks occur in the late 1960s and early 1990s, which overlays the positive trend in drought characteristics and decreasing soil moisture trend. Rouault and Richard (2005) analyzed South African SPI and found a substantial increase in 2-yr droughts since the 1970s. They surmised that the change was likely driven by stronger connections with ENSO, although correlations here with the Niño-3.4 index are low, likely due to the large size of the SAF region.

A decadal oscillation in north Asia (NAS) overlays the general increasing trend for all characteristics that is consistent with decreasing soil moisture. Of note is the increase in the number of long, low-intensity droughts during the 1980s and 1990s, which may be related to the switch to a positive NAO phase (Visbeck et al. 2001), although correlations of soil moisture with climate indices are insignificant. Central Asian droughts also follow a decadal cycle, peaking in the 1970s and ending with an upward trend in the 1990s. Correlations with climate indices are weak but greatest with the Niño-3.4 and AMO indices. Over the Tibetan Plateau, the series are dominated by a peak in all variables around 1960, and a decreasing trend thereafter until the early 1990s, when there are slight increases again, which has continued into recent years (Barlow et al. 2002). The AMO provides the only significant, although weak, correlation (r = 0.42). Drought in East Asia shows little variation over the 50 yr as seen before and is most closely tied to Niño-3.4 variability.

For Southeast and South Asia, there are small but increasing trends over much of the period with a slight decreasing trend in the 1990s. Note the high and expected correlation (r = −0.87, PC1) with the Niño-3.4 index. The decreasing trends over Australia are dominated by a large amplitude decadal variation that peaks in the 1960s and late 1980s. The correlation with Niño-3.4 is expected (r = −0.49, PC1) but is weak, likely owing to the size of the AUS region.

Over North America, the overall wetting trend is reflected in decreasing trends in all drought variables, yet there is large variability within this. Alaska and Northeastern Canada show large decreases since the 1950s but an upturn in the 1990s with weak correlation between soil moisture and the AMO (r = 0.46 and 0.49, respectively, PC1). The number and spatial extent of droughts in western North America decreases until the early 1970s, at which time they increase to the end of the record. In central North America there is an overall decreasing trend, although both the western and central regions exhibit an upward jump in longer duration drought frequencies during the 1970s. In eastern North America the series are dominated by decadal variability overlaid by a decreasing trend. Low values occur during the 1970s and for a brief period around the early 1990s. The changes across these three latter regions are generally consistent with the overall decreasing drought trends found by Andreadis and Lettenmaier (2006) for the United States.

A decreasing trend is apparent in Central America, most prominently at the beginning of the period. The Amazon is dominated by a decadal cycle, peaking in the early 1960s and mid-1980s that relates to Niño- 3.4 SSTs (r = −0.72). In southern South America there are similar oscillations with peaks in frequencies and extent in the mid-1960s followed by a drop until the early 1980s and then increasing conditions thereafter.

b. Variation and robustness of trends

The regional time series of soil moisture and drought statistics show large variability within the long-term trends identified in section 3. Of interest is how these variations affect the robustness of the global wetting trend, especially as we move into the twenty-first century and the potential impacts of global warming. Recent increases in global temperatures (e.g., Jones et al. 1999; Jones and Moberg 2003; Hansen et al. 1999; Brohan et al. 2006) may have already caused an acceleration of the water cycle (Huntington 2006) and intensification of drought. For example, many regions show decadal variations that switch during the 1970s, which has been reported previously (Dai et al. 2004; Rouault and Richard 2005) and may be indicative of temperature impacts on drought, either directly or indirectly through intensification of climate drivers such as ENSO (Hunt 1999; Herbert and Dixon 2003). Nevertheless, evidence of increasing summertime soil moisture across Asia despite increasing temperatures (Robock et al. 2000) and forcing of increased drought by large-scale climate anomalies [e.g., decreased late-spring precipitation in China driven by a shift to positive phase of the NAO (Xin et al. 2006)] add to the uncertainty of current and future drought response to changing temperatures.

Figure 10 shows a time series of trends in regional mean soil moisture quantile calculated over an 11-yr moving window. The trends are color coded according to the sign of the trend and the statistical significance at the 0.05 level. At the global scale the trends oscillate from wetting to drying around the mid-1970s with peak drying trends in the mid-1980s and subsequent reduction in magnitude toward the end of the century. Over northern Europe, mostly insignificant wetting trends are separated by drying trends at the beginning and end of the series. In the Mediterranean, initially increasing soil moisture is overwhelmed by decreasing trends from the mid-1960s onward. For Africa, generally decreasing trends dominate (with a spate of increasing trends centered on 1970 in southern Africa), although all regions begin to experience increasing trends at the end of the time period. Over northern Asian regions, mostly increasing trends are juxtaposed with decreasing trends in the last 20 yr, although the north Asia region shows drying trends since the mid-1960s. The decadal oscillation in trends over Southeast and southern Asia, Australia, and the Amazon indicates a consistency over most of the tropics, which mirrors the variation at the global scale as well. Over the Americas, all regions (except the Amazon) end the period with decreasing trends, although there is considerable variability previously in some regions (e.g., central North America). Of particular note are the large trends in Alaska and northern Canada in the last 10–20 yr that are concurrent with increasing temperatures (not shown).

Fig. 10.

Trends in regional average soil moisture for a 21-yr moving window. The x axis indicates the middle date over which the trend is calculated. Trends that are significant at the 0.05 level are shaded in darker colors. Positive (negative) trend values are shaded in warm (cool) colors. The values for “world” have been multiplied by 3 for ease of visualization.

Fig. 10.

Trends in regional average soil moisture for a 21-yr moving window. The x axis indicates the middle date over which the trend is calculated. Trends that are significant at the 0.05 level are shaded in darker colors. Positive (negative) trend values are shaded in warm (cool) colors. The values for “world” have been multiplied by 3 for ease of visualization.

These results indicate a switch to drying trends in soil moisture in many regions and more generally at global scales, despite a long-term wetting trend. Although there is considerable variability over the whole period, we hypothesize that this is caused in part by warming temperatures that act to increase evapotranspiration and/or early snowmelt and therefore the occurrence of drought. This is despite the possibility of concurrent increases in precipitation, although this is unlikely in some regions because of the anticorrelation of precipitation and temperature (Déry and Wood 2005b). We explore the relationship between soil moisture/drought and precipitation and temperature variability next and then address the impact of warming trends.

5. Relationships with meteorological forcings

a. Relationship with precipitation and temperature variability

Drought is driven primarily by lack of precipitation, but this is accentuated or diminished by associated changes in temperature and other meteorological processes. Anomalously high temperatures will tend to increase evapotranspiration, while low precipitation will obviously reduce recharge of the soil column. These processes may interact in complex and nonlinear ways such that drought can, for example, be induced by many months of below-normal rainfall, be prolonged by high temperatures, and then be alleviated by a single storm. These relationships may also have implications for the occurrence of drought under future climates that are likely to be warmer but with associated changes in precipitation that are regionally dependent and may show increase or decreases. Drought development may also lag anomalies in precipitation and other meteorological forcings, but this relationship is not well understood. For example, months of anomalously low precipitation may not result in drought conditions until some time later. A subsequent return to normal conditions may similarly be delayed as moisture takes time to filter through the hydrologic system and replenish depleted stores. These processes are complicated by seasonal variations where precipitation may dominate in cool seasons and be modified in warm seasons by temperature effects.

Figure 11 shows scatterplots of the trends in precipitation and temperature stratified by trends in soil moisture quantile for several regions chosen to represent a diversity of climates. Wetting (drying) trends that are significant at the 0.05 level in soil moisture are associated with positive (negative) trends in precipitation in all regions. Of note is the small spread in the distribution of precipitation trends and clear delineation between wetting and drying soil moisture trends in high-latitude regions (ALA and NAS). Tropical and Southern Hemisphere regions (e.g., AMZ, SAF, SAS) show larger spread in precipitation trends and some overlap in the range of positive soil moisture trends with negative precipitation trends. The relationship between soil moisture and temperature is somewhat unclear, however (in part, because decreasing temperature trends are uncommon). As significant trends in soil moisture (wetting and drying) are generally associated with positive temperature trends, indicating that increasing temperatures do not necessarily hinder increasing soil moisture, and, conversely, that they may enhance decreasing soil moisture.

Fig. 11.

Scatterplot of trends in precipitation and surface air temperature, stratified by trends in soil moisture for selected regions. Blue (red) symbols represent positive (negative), significant trends in soil moisture at the 0.05 level.

Fig. 11.

Scatterplot of trends in precipitation and surface air temperature, stratified by trends in soil moisture for selected regions. Blue (red) symbols represent positive (negative), significant trends in soil moisture at the 0.05 level.

At seasonal scales, the relationship between soil moisture, precipitation, and temperature is more complex, especially for cooler regions where snowpack storage and seasonally frozen soil water play an important role. As an example, the results for the NAS region (Fig. 12) show very different relationships between trends in soil moisture and those in the forcing variables. During summer (JJA) and autumn (SON), drying (wetting) soil moisture trends are generally associated with decreasing (increasing) precipitation. However, in the cooler months (DJF and MAM) this distinction is not apparent and the sign of trends in soil moisture is independent of the sign of the precipitation trend. Springtime soil moisture is dominated by snowmelt, which, although driven in part by springtime temperatures and precipitation, is a function of the snowpack accumulated over the preceding winter. As a comparison, humid regions such as the Amazon (not shown) show little seasonal variability in the precipitation–temperature–soil moisture relationships.

Fig. 12.

Scatterplot of seasonal trends in precipitation and surface air temperature, stratified by trends in soil moisture for the NAS region. Blue (red) symbols represent positive (negative), significant trends in soil moisture at the 0.05 level.

Fig. 12.

Scatterplot of seasonal trends in precipitation and surface air temperature, stratified by trends in soil moisture for the NAS region. Blue (red) symbols represent positive (negative), significant trends in soil moisture at the 0.05 level.

b. Sensitivity of drought to temperature trends

To further investigate the impact of temperature on soil moisture and drought trends we compare the results of the TANN and the TCLIM simulations. Figure 13 shows the difference between the two simulations in terms of mean soil moisture and trends in soil moisture, and indicates where changes are attributable to trends in temperature. Regions of maximum differences in mean soil moisture tend to occur in the Northern Hemisphere, in mid–high latitudes, with TANN driving increases in soil moisture in eastern Europe, northern Eurasia, southern Alaska, and southeastern Canada, and decreases in the eastern United States, Pacific Northwest, eastern Canada, and small parts of central Europe, central Asia, and eastern Siberia. Globally, the tendency is for equal areas of increases and decreases in the Northern Hemisphere and a ratio of 3:2 in favor of increases in the Southern Hemisphere. The differences in trends between the two simulations are small (of the order of 0.002% yr−1 or less), although this is the trend over the full 50 yr (≈0.1% 50 yr−1). The largest differences are found in high northern latitudes, with increased trend magnitude for TCLIM over northern Canada and northern Europe and decreased trend magnitude for the area from eastern Europe through central Siberia. Eastern Siberia also shows decreased magnitudes. Elsewhere, the northern half of the Andes shows lower magnitudes in Peru and higher magnitudes in Columbia.

Fig. 13.

Comparison of soil moisture between the original simulation with annually varying air temperature forcing (TANN) and that with time invariant or climatological air temperature (TCLIM). (top) Difference in mean soil moisture, 1950–2000. (bottom) Difference in soil moisture trend, 1950–2000.

Fig. 13.

Comparison of soil moisture between the original simulation with annually varying air temperature forcing (TANN) and that with time invariant or climatological air temperature (TCLIM). (top) Difference in mean soil moisture, 1950–2000. (bottom) Difference in soil moisture trend, 1950–2000.

In Fig. 14, the differences between the two simulations are shown for 11-yr moving averages in trends in regional mean soil moisture quantile. For the majority of regions the differences are generally less than 1% yr−1. However, outstanding are a few regions with differences up to 6% yr−1 (ALA, NEC) and 2%–3% yr−1 (NEU, NAS). Note that these are differences in trends over 11-yr windows and will generally be much higher than the 50-yr trends. Of particular interest are the large differences in regions ALA and NEC during the last 20 yr of the record that indicate that decreasing soil moisture is exaggerated by the increasing trend in air temperature. The temperature trend is particularly pronounced during 1990–2000 (not shown). Similar behavior in the latter years of the record exists for other regions, such as EAS, WNA, ENA, and SSA, but the magnitude of the differences are considerably smaller. Even at the global scale a difference is noticeable in the last 20 yr. Despite some long term variability over the full period in most regions, the evidence points toward a temperature effect in recent years that tends to exaggerate or force decreasing trends in soil moisture.

Fig. 14.

Difference in 11-yr moving window trends in soil moisture quantile between the TANN and TCLIM simulations. The TANN simulation was forced with annually varying surface air temperature. The TCLIM simulation was forced with climatological surface air temperature.

Fig. 14.

Difference in 11-yr moving window trends in soil moisture quantile between the TANN and TCLIM simulations. The TANN simulation was forced with annually varying surface air temperature. The TCLIM simulation was forced with climatological surface air temperature.

6. Discussion and conclusions

a. Uncertainties in the meteorological forcings and hydrologic modeling

The trends discussed are only as robust as the meteorological data that forces the simulation and the land surface model that is used to derive the soil moisture data. It has been shown that modeled land surface hydrology is sensitive to the forcing dataset that drives it, and especially precipitation (Ngo-Duc et al. 2005; Fekete et al. 2004; Berg et al. 2003; Sheffield et al. 2004b; Guo et al. 2006). Fekete et al. (2004) showed that the uncertainty in precipitation datasets was of the order of interannual variability, and that the impact of precipitation uncertainties on the terrestrial water budget was of at least the same magnitude, especially in semiarid regions where the hydrologic response is highly nonlinear. In the second Global Soil Wetness Project (GSWP-2) multimodel comparison, Guo et al. (2006) found that uncertainties in the forcings were as large as differences between land surface models. The first-order drivers of drought, monthly precipitation, and temperature are derived in this study from the Climatic Research Unit (CRU) TS2.0 gauge-based dataset (Mitchell and Jones 2005). For time periods when station observations are limited or nonexistent the CRU dataset relies on climatological values, or so-called relaxation to climatology (Mitchell and Jones 2005). Additionally, the gauge density that contributes to a grid cell may force errors in the simulated hydrology for densities of the order of less than 30 gauges per 106 km2 (Oki et al. 1999).

It is therefore likely that errors in the forcing dataset used here will result in errors in the representation of drought, and that the reliability of the time series at the grid scale (1.0°) may be reduced (Patz et al. 2002), although this can be alleviated through spatial and temporal averaging (Giorgi 2002). Nevertheless, at larger scales, and especially in data-poor regions, we argue that this is our best estimate. In data-rich regions, such as the United States and Europe, our estimates may compare less favorably against that obtained when using data from dense station networks. For example, Decharme and Douville (2006) showed that the GSWP-2 forcing dataset drastically overestimated precipitation compared to data from a dense network in France with resulting impacts on modeled river discharge, and that this overestimation was systematic globally. However, comparisons of drought characteristics derived from the dataset in this paper and datasets based on higher spatial resolution modeling forced with gauge-based observations (Sheffield et al. 2004a; Andreadis et al. 2005; Andreadis and Lettenmaier 2006) are encouraging (section 2b). Furthermore, biases in the modeling of the land surface budgets, through simplifications in the modeling and uncertainties in the parameter data, may result in errors in the trends (Sheffield et al. 2004a). These biases are generally unquantifiable but most importantly are systematic and therefore uniform over time. Therefore it is likely they will not impact the sign or strength of the calculated trends appreciably. We similarly argue that comparable results would be obtained if we used a different model. Intercomparison of land surface models driven by the same forcings has been carried out regionally (Wood et al. 1998; Mitchell et al. 2004) and globally (Guo and Dirmeyer 2006) and have concluded that although the models do poorly at reproducing the absolute values of observed soil moisture they do reasonably well at mimicking the anomalies and interannual variability. They may, therefore, provide useful information on drought occurrence and trends when viewed with respect to their own climatologies.

The PDSI can be considered as another type of model but designed specifically to monitor drought (although much simpler in its treatment of physical processes when compared to hydrologic land surface models) and has been used to assess trends in global drought previously (Dai et al. 2004; Burke et al. 2006). Burke et al. (2006) looked at trends in PDSI over the second half of the twentieth century as calculated from i) the observation-driven PDSI dataset of Dai et al. (2004) and ii) a PDSI dataset driven by precipitation and temperature from coupled and uncoupled runs of the Hadley climate model. They found that the two datasets show a global drying trend of between −0.2 and −0.3 decade−1 in PDSI units. This is at odds with the results of this paper, which show a small wetting trend globally, although the pattern of regional trends is similar (cf. Fig. 1 with their Fig. 3). The difference is mainly because of the larger drying trend in the PDSI datasets for the last 20 yr as shown by Burke et al. (2006) and Sheffield and Wood (2007) and is systematic of the PDSI, whether driven by the Burke et al. (2006), the Dai et al. (2004), or the VIC forcings (Sheffield and Wood 2007). This may be due to a number of differences in the PDSI and VIC modeling approaches, such as the model time step (PDSI is monthly; VIC is 3 hourly), the sensitivity of the model to precipitation and temperature changes [e.g., the PDSI uses the Thornthwaite method to calculate PE that may be biased for higher temperatures (Burke et al. 2006)], and the fundamental physical processes that are modeled (e.g., the PDSI does not include snow processes) and requires further investigation.

There are also a number of nonmeteorological boundary conditions, such as land cover, that are assumed time invariant (although vegetation parameters such as leaf area index do vary seasonally) but may actually have an appreciable impact on the trends. For example, anthropogenic factors—including irrigation, water withdrawals, and land use change—and natural processes—such as vegetation dynamics and wildfire—are not modeled explicitly, and the impact of these may vary in time also, thus affecting the trends. Estimates of current day irrigation are that 16.3% of cultivated regions are equipped for irrigation (Siebert et al. 2005), which can have a significant impact on the water cycle (Haddeland et al. 2007), although historically this may have been offset by changes in land use. Land cover has changed dramatically over the past 300 yr (Ramankutty and Foley 1999) and more so in tropical/developing regions over the twentieth century (Klein Goldewijk 2001). The impact that this has had on the water cycle may be substantial (Zhang and Schilling 2006; Scanlon et al. 2007), likely reducing evapotranspiration and increasing runoff with possible implications for the results presented here. Furthermore, elevated levels of CO2 and increased growing season length may be responsible for recent increases in net primary productivity (NPP) and thus transpiration (Friend et al. 2007), although stomatal closure response to elevated CO2 levels may have had the opposite effect (Gedney et al. 2006).

b. Summary and conclusions

Global and regional trends in drought over the past 50 yr are analyzed using a soil moisture–based drought index over global terrestrial areas, excluding Greenland and Antarctica. Drought is described in terms of various statistics that summarize drought duration, intensity, and severity. Trends in soil moisture and drought characteristics were calculated using a nonparametric trend test on a grid cell basis and for regional averages. Despite some uncertainties in the forcings and the modeling process, we have confidence in the results as derived from a validated dataset, especially at larger scales and when put in context of other studies.

An overall increasing trend in global soil moisture, driven by increasing precipitation, underlies the whole analysis, which is reflected most obviously over the western hemisphere and especially in North America. Regional variation is nevertheless apparent and significant drying over West Africa, as driven by decreasing Sahel precipitation, stands out. Elsewhere, Europe appears to have not experienced significant changes in soil moisture, a trait shared by Southeast and southern Asia. Trends in drought characteristics are predominantly decreasing, but statistically significant changes are limited in areal extent, of the order of 1.0%–7.0% globally, depending on the drought threshold and variable and being generally less than 10% of continental areas. Concurrent changes in drought spatial extent are evident, with a global decreasing trend of −0.021% to −0.035% yr−1. Regionally, drought extent over Africa has increased and is dominated by large increases over West Africa. Northern and East Asia show positive trends and central Asia and the Tibetan Plateau show decreasing trends. In South Asia all trends are insignificant. Drought extent over Australia has decreased. Over the Americas, trends are uniformly negative and mostly significant.

Within the long-term trends there are interannual and decadal variations in soil moisture and drought characteristics that are apparent in many regions. Globally, variations are driven mainly by ENSO variability, although the AMO appears to play an important role globally and in many regions, such as West and East Africa, central Asia, and the high latitudes of North America. However, the short length of record relative to the scale of the AMO precludes any definite conclusions. High correlation values are found between the Mediterranean and the NAO, and Southeast Asia and the Amazon basin and Niño-3.4 SSTs. Stronger connection are likely at scales smaller than the regions examined here and by using seasonal and lagged correlations. The decadal variations in soil moisture and drought characteristics impact the robustness of the long-term trends. In general, they are responsible for diminishing the long-term trends. In fact, despite the overall wetting trend, there is a switch in later years to a drying trend, globally and in many regions, which is concurrent with increasing temperatures. Although drought is driven primarily by variability in precipitation, temperature has an effect that appears to be exaggerated in the last decade or so especially in high northern latitudes. This is most pertinent within the context of potential continued temperature increases during the twenty-first century.

Future climate projections from coupled models predict increases in global temperatures and generally increasing temperatures over land regions for most emission scenarios (e.g., Giorgi and Bi 2005). The range in predictions varies among scenarios but is generally increasing. If temperature is a secondary forcing of drought (precipitation being the primary forcing) in most regions, then the implication is that droughts will increase in the future, especially given the magnitude of predicted temperature increases. On the other hand, predicted changes in precipitation are highly variable in space and are scenario dependent to the extent that precipitation is predicted to increase in some regions and decrease in others (Giorgi and Bi 2005). Temperature-driven changes in drought will be modified by the changes to precipitation, although the fact that precipitation and temperature are anticorrelated in many regions (Trenberth and Shea 2005; Déry and Wood 2005b) may lead to enhanced drought occurrence.

Acknowledgments

This work has been supported by NOAA Grants NA86GP0248 and NA0303AR4310001 and NASA Grant NAG5-9486. We thank three anonymous reviewers for their useful comments and suggestions.

REFERENCES

REFERENCES
Andreadis
,
K. M.
, and
D. P.
Lettenmaier
,
2006
:
Trends in 20th century drought over the continental United States.
Geophys. Res. Lett.
,
33
.
L10403, doi:10.1029/2006GL025711
.
Andreadis
,
K. M.
,
E. A.
Clark
,
A. W.
Wood
,
A. F.
Hamlet
, and
D. P.
Lettenmaier
,
2005
:
Twentieth-century drought in the conterminous United States.
J. Hydrometeor.
,
6
,
985
1001
.
Barlow
,
M.
,
H.
Cullen
, and
B.
Lyon
,
2002
:
Drought in central and southwest Asia: La Niña, the warm pool, and Indian Ocean precipitation.
J. Climate
,
15
,
697
700
.
Berbery
,
E. H.
, and
V. R.
Barros
,
2002
:
The hydrologic cycle of the La Plata basin in South America.
J. Hydrometeor.
,
3
,
630
645
.
Berezovskaya
,
S.
,
D.
Yang
, and
D. L.
Kane
,
2004
:
Compatibility analysis of precipitation and runoff trends over the large Siberian watersheds.
Geophys. Res. Lett.
,
31
.
L21502, doi:10.1029/2004GL021277
.
Berg
,
A. A.
,
J. S.
Famiglietti
,
J. P.
Walker
, and
P. R.
Houser
,
2003
:
Impact of bias correction to reanalysis products on simulations of North American soil moisture and hydrological fluxes.
J. Geophys. Res.
,
108D
.
4490, doi:10.1029/2002JD003334
.
Brohan
,
P.
,
J. J.
Kennedy
,
I.
Harris
,
S. F. B.
Tett
, and
P. D.
Jones
,
2006
:
Uncertainty estimates in regional and global observed temperature changes: A new dataset from 1850.
J. Geophys. Res.
,
111
.
D12106, doi:10.1029/2005JD006548
.
Burke
,
E. J.
,
S. J.
Brown
, and
N.
Christidis
,
2006
:
Modeling the recent evolution of global drought and projections for the twenty-first century with the Hadley Centre climate model.
J. Hydrometeor.
,
7
,
1113
1125
.
Cherkauer
,
K. A.
,
L. C.
Bowling
, and
D. P.
Lettenmaier
,
2002
:
Variable infiltration capacity (VIC) cold land process model updates.
Global Planet. Change
,
38
,
151
159
.
Dai
,
A. G.
,
I. Y.
Fung
, and
A. D.
Del Genio
,
1997
:
Surface observed global land precipitation variations during 1900–88.
J. Climate
,
10
,
2943
2962
.
Dai
,
A. G.
,
K. E.
Trenberth
, and
T.
Qian
,
2004
:
A global dataset of Palmer Drought Severity Index for 1870–2002: Relationship with soil moisture and effects of surface warming.
J. Hydrometeor.
,
5
,
1117
1130
.
Decharme
,
B.
, and
H.
Douville
,
2006
:
Uncertainties in the GSWP-2 precipitation forcing and their impacts on regional and global hydrological simulations.
Climate Dyn.
,
27
,
695
713
.
Déry
,
S. J.
, and
E. F.
Wood
,
2005a
:
Decreasing river discharge in northern Canada.
Geophys. Res. Lett.
,
32
.
L10401, doi:10.1029/ 2005GL022845
.
Déry
,
S. J.
, and
E. F.
Wood
,
2005b
:
Observed twentieth century land surface air temperature and precipitation covariability.
Geophys. Res. Lett.
,
32
.
L21414, doi:10.1029/2005GL024234
.
Dong
,
B.
,
R. T.
Sutton
, and
A. A.
Scaife
,
2006
:
Multidecadal modulation of El Niño–Southern Oscillation (ENSO) variance by Atlantic Ocean sea surface temperatures.
Geophys. Res. Lett.
,
33
.
L08705, doi:10.1029/2006GL025766
.
Enfield
,
D. B.
,
A. M.
Mestas-Nuñez
, and
P. J.
Trimble
,
2001
:
The Atlantic multidecadal oscillation and its relation to rainfall and river flows in the continental U.S.
Geophys. Res. Lett.
,
28
,
2077
2080
.
Fekete
,
B. M.
,
C. J.
Vörösmarty
,
J. O.
Roads
, and
C. J.
Willmott
,
2004
:
Uncertainties in precipitation and their impacts on runoff estimates.
J. Climate
,
17
,
294
304
.
Friend
,
A. D.
, and
Coauthors
,
2007
:
FLUXNET and modelling the global carbon cycle.
Global Change Biol.
,
13
,
610
633
.
Garcia
,
N. O.
, and
C. R.
Mechoso
,
2005
:
Variability in the discharge of South American rivers and in climate.
Hydrol. Sci. J.
,
50
,
459
478
.
Gedney
,
N.
,
P. M.
Cox
,
R. A.
Betts
,
O.
Boucher
,
C.
Huntingford
, and
P. A.
Stott
,
2006
:
Detection of a direct carbon dioxide effect in continental river runoff records.
Nature
,
439
,
835
838
.
Giorgi
,
F.
,
2002
:
Variability and trends of sub-continental scale surface climate in the twentieth century. Part 1: Observations.
Climate Dyn.
,
18
,
675
691
.
Giorgi
,
F.
, and
R.
Francisco
,
2000
:
Uncertainties in regional climate change predictions: A regional analysis of ensemble simulations with the HADCM2 GCM.
Climate Dyn.
,
16
,
169
182
.
Giorgi
,
F.
, and
X.
Bi
,
2005
:
Updated regional precipitation and temperature changes for the 21st century from ensembles of recent AOGCM simulations.
Geophys. Res. Lett.
,
32
.
L21715, doi:10.1029/2005GL024288
.
Groisman
,
P. Ya
,
R. W.
Knight
,
T. R.
Karl
,
D. R.
Easterling
,
B.
Sun
, and
J. H.
Lawrimore
,
2004
:
Contemporary changes of the hydrological cycle over the contiguous United States: Trends derived from in situ observations.
J. Hydrometeor.
,
5
,
64
85
.
Guo
,
Z.
, and
P. A.
Dirmeyer
,
2006
:
Evaluation of the Second Global Soil Wetness Project soil moisture simulations: 1. Intermodel comparison.
J. Geophys. Res.
,
111D
.
D22S02, doi:10.1029/2006JD007233
.
Guo
,
Z.
,
P. A.
Dirmeyer
,
Z-Z.
Hu
,
X.
Gao
, and
M.
Zhao
,
2006
:
Evaluation of the Second Global Soil Wetness Project soil moisture simulations: 2. Sensitivity to external meteorological forcing.
J. Geophys. Res.
,
111D
.
D22S03, doi:10.1029/2006JD007845
.
Haddeland
,
I.
,
T.
Skaugen
, and
D. P.
Lettenmaier
,
2007
:
Hydrologic effects of land and water management in North America and Asia: 1700–1992.
Hydrol. Earth Syst. Sci.
,
11
,
1035
1045
.
Hamlet
,
A. F.
,
P. W.
Mote
,
M. P.
Clark
, and
D. P.
Lettenmaier
,
2007
:
Twentieth-century trends in runoff, evapotranspiration, and soil moisture in the western United States.
J. Climate
,
20
,
1468
1486
.
Hansen
,
J.
,
R.
Ruedy
,
J.
Glascoe
, and
M.
Sato
,
1999
:
GISS analysis of surface temperature change.
J. Geophys. Res.
,
104
,
D24
.
30997
31022
.
Held
,
I. M.
, and
B. J.
Soden
,
2000
:
Water vapor feedback and global warming.
Annu. Rev. Energy Environ.
,
25
,
441
475
.
Herbert
,
J. M.
, and
R. W.
Dixon
,
2002
:
Is the ENSO phenomenon changing as a result of global warming.
Phys. Geogr.
,
23
,
196
211
.
Hirsch
,
R. M.
, and
J. R.
Slack
,
1984
:
A nonparametric trend test for seasonal data with serial dependence.
Water Resour. Res.
,
20
,
727
732
.
Hulme
,
M.
,
1992
:
Rainfall changes in Africa: 1931–1960 to 1961–1990.
Int. J. Climatol.
,
12
,
685
699
.
Hunt
,
A. G.
,
1999
:
Understanding a possible correlation between El Niño occurrence frequency and global warming.
Bull. Amer. Meteor. Soc.
,
80
,
297
300
.
Huntington
,
T. G.
,
2006
:
Evidence for intensification of the global water cycle: Review and synthesis.
J. Hydrol.
,
319
,
83
95
.
Hurrell
,
J. W.
, and
H.
Van Loon
,
1997
:
Decadal variations in climate associated with the North Atlantic oscillation.
Climatic Change
,
36
,
301
326
.
Jones
,
P. D.
, and
A.
Moberg
,
2003
:
Hemispheric and large-scale surface air temperature variations: An extensive revision and an update to 2001.
J. Climate
,
16
,
206
223
.
Jones
,
P. D.
,
M.
New
,
D. E.
Parker
,
S.
Martin
, and
I. G.
Rigor
,
1999
:
Surface air temperature and its changes over the last 150 years.
Rev. Geophys.
,
37
,
173
199
.
Kalnay
,
E.
, and
Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project.
Bull. Amer. Meteor. Soc.
,
77
,
437
471
.
Kendall
,
M. G.
,
1975
:
Rank Correlation Methods.
Charles Griffin, 202 pp
.
Kerr
,
R. A.
,
2000
:
A North Atlantic climate pacemaker for the centuries.
Science
,
288
,
1984
1985
.
Klein Goldewijk
,
K.
,
2001
:
Estimating global land use change over the past 300 years: The HYDE Database.
Global Biogeochem. Cycles
,
15
,
417
433
.
L’Hôte
,
Y.
,
G.
Mahé
,
B.
Somé
, and
J. P.
Triboulet
,
2002
:
Analysis of a Sahelian annual rainfall index from 1896 to 2000; the drought continues.
Hydrol. Sci. J.
,
47
,
563
572
.
Liang
,
X.
,
D. P.
Lettenmaier
,
E. F.
Wood
, and
S. J.
Burges
,
1994
:
A simple hydrologically based model of land surface water and energy fluxes for general circulation models.
J. Geophys. Res.
,
99D
,
14415
14428
.
Lloyd-Hughes
,
B.
, and
M. A.
Saunders
,
2002
:
A drought climatology for Europe.
Int. J. Climatol.
,
22
,
1571
1592
.
Mann
,
H. B.
,
1945
:
Non-parametric test against trend.
Econometrika
,
13
,
245
259
.
Mantua
,
N. J.
,
S. R.
Hare
,
Y.
Zhang
,
J. M.
Wallace
, and
R. C.
Francis
,
1997
:
A Pacific interdecadal climate oscillation with impacts on salmon production.
Bull. Amer. Meteor. Soc.
,
78
,
1069
1079
.
Maurer
,
E. P.
,
A. W.
Wood
,
J. C.
Adam
,
D. P.
Lettenmaier
, and
B.
Nijssen
,
2002
:
A long-term hydrologically based dataset of land surface fluxes and states for the conterminous United States.
J. Climate
,
15
,
3237
3251
.
McBean
,
G.
, and
Coauthors
,
2005
:
Arctic climate—Past and present.
Arctic Climate Impact Assessment, C. Symon, L. Arris, and B. Heal, Eds., Cambridge University Press, 22–55
.
McCabe
,
G. J.
, and
M. A.
Palecki
,
2006
:
Multidecadal climate variability of global lands and oceans.
Int. J. Climatol.
,
26
,
849
865
.
McCabe
,
G. J.
,
M. A.
Palecki
, and
J. L.
Betancourt
,
2004
:
Pacific and Atlantic Ocean influences on multidecadal drought frequency in the United States.
Proc. Natl. Acad. Sci.
,
101
,
4136
4141
.
McClelland
,
J. W.
,
S. J.
Déry
,
B. J.
Peterson
,
R. M.
Holmes
, and
E. F.
Wood
,
2006
:
A pan-arctic evaluation of changes in river discharge during the latter half of the 20th century.
Geophys. Res. Lett.
,
33
.
L06715, doi:10.1029/2006GL025753
.
Mitchell
,
K. E.
, and
Coauthors
,
2004
:
The multi-institution North American Land Data Assimilation System (NLDAS): Utilizing multiple GCIP products and partners in a continental distributed hydrological modeling system.
J. Geophys. Res.
,
109D
.
D07S90, doi:10.1029/2003JD003823
.
Mitchell
,
T. D.
, and
P. D.
Jones
,
2005
:
An improved method of constructing a database of monthly climate observations and associated high-resolution grids.
Int. J. Climatol.
,
25
,
693
712
.
Newman
,
M.
,
G. P.
Compo
, and
M. A.
Alexander
,
2003
:
ENSO-forced variability of the Pacific decadal oscillation.
J. Climate
,
16
,
3853
3857
.
Ngo-Duc
,
T.
,
J.
Polcher
, and
K.
Laval
,
2005
:
A 53-year forcing data set for land surface models.
J. Geophys. Res.
,
110D
.
D06116, doi:10.1029/2004JD005434
.
Nijssen
,
B.
,
R.
Schnur
, and
D. P.
Lettenmaier
,
2001
:
Global retrospective estimation of soil moisture using the Variable Infiltration Capacity land surface model, 1980–93.
J. Climate
,
14
,
1790
1808
.
Oba
,
G.
,
E.
Post
, and
N. C.
Stenseth
,
2001
:
Sub-Saharan desertification and productivity are linked to hemispheric climate variability.
Global Change Biol.
,
7
,
241
246
.
Oki
,
T.
,
T.
Nishimura
, and
P.
Dirmeyer
,
1999
:
Assessment of annual runoff from land surface models using Total Runoff Integrating Pathways (TRIP).
J. Meteor. Soc. Japan
,
77
,
235
255
.
Palmer
,
W.
,
1965
:
Meteorological drought. U.S. Dept. of Commerce Weather Bureau Research Paper 45, 58 pp
.
Patz
,
J. A.
, and
Coauthors
,
2002
:
Climate change: Regional warming and malaria resurgence.
Nature
,
420
,
627
628
.
Peterson
,
B. J.
,
R. M.
Holmes
,
J. W.
McClelland
,
C. J.
Vörösmarty
,
R. B.
Lammers
,
A. I.
Shiklomanov
,
I. A.
Shiklomanov
, and
S.
Rahmstorf
,
2002
:
Increasing river discharge to the Arctic Ocean.
Science
,
298
,
2171
2173
.
Ramankutty
,
N.
, and
J. A.
Foley
,
1999
:
Estimating historical changes in global land cover: Croplands from 1700 to 1992.
Global Biogeochem. Cycles
,
13
,
997
1027
.
Robock
,
A.
,
K. Y.
Vinnikov
,
G.
Srinivasan
,
J. K.
Entin
,
S. E.
Hollinger
,
N. A.
Speranskaya
,
S.
Liu
, and
A.
Namkhai
,
2000
:
The Global Soil Moisture Data Bank.
Bull. Amer. Meteor. Soc.
,
81
,
1281
1299
.
Rodó
,
X.
,
E.
Baert
, and
F. A.
Comin
,
1997
:
Variations in seasonal rainfall in southern Europe during present century: Relationships with the North Atlantic Oscillation and the El Niño–Southern Oscillation.
Climate Dyn.
,
13
,
275
284
.
Ropelewski
,
C. F.
, and
M. S.
Halpert
,
1987
:
Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation.
Mon. Wea. Rev.
,
115
,
1606
1626
.
Rouault
,
M.
, and
Y.
Richard
,
2005
:
Intensity and spatial extent of droughts in southern Africa.
Geophys. Res. Lett.
,
32
.
L15702, doi:10.1029/2005GL022436
.
Scanlon
,
B. R.
,
I.
Jolly
,
M.
Sophocleous
, and
L.
Zhang
,
2007
:
Global impacts of conversions from natural to agricultural ecosystems on water resources: Quantity versus quality.
Water Resour. Res.
,
43
.
W03437, doi:10.1029/2006WR005486
.
Sheffield
,
J.
, and
E. F.
Wood
,
2007
:
Characteristics of global and regional drought, 1950–2000: Analysis of soil moisture data from off-line simulation of the terrestrial hydrologic cycle.
J. Geophys. Res.
,
112
.
D17115, doi:10.1029/2006JD008288
.
Sheffield
,
J.
,
G.
Goteti
,
F.
Wen
, and
E. F.
Wood
,
2004a
:
A simulated soil moisture based drought analysis for the United States.
J. Geophys. Res.
,
109D
.
D24108, doi:10.1029/2004JD005182
.
Sheffield
,
J.
,
A. D.
Ziegler
,
E. F.
Wood
, and
Y.
Chen
,
2004b
:
Correction of the high-latitude rain day anomaly in the NCEP–NCAR reanalysis for land surface hydrological modeling.
J. Climate
,
17
,
3814
3828
.
Sheffield
,
J.
,
G.
Goteti
, and
E. F.
Wood
,
2006
:
Development of a 50-year high-resolution global dataset of meteorological forcings for land surface modeling.
J. Climate
,
19
,
3088
3111
.
Shiklomanov
,
A. I.
,
T. I.
Yakovleva
,
R. B.
Lammers
,
I. Ph
Karasev
,
C. J.
Vörösmarty
, and
E.
Linder
,
2006
:
Cold region river discharge uncertainty—estimates from large Russian rivers.
J. Hydrol.
,
326
,
231
256
.
Siebert
,
S.
,
P.
Döll
,
J.
Hoogeveen
,
J. M.
Faures
,
K.
Frenken
, and
S.
Feick
,
2005
:
Development and validation of the global map of irrigation areas.
Hydrol. Earth Syst. Sci.
,
9
,
535
547
.
Sutton
,
R. T.
, and
D. L. R.
Hodson
,
2005
:
Atlantic Ocean forcing of North American and European summer climate.
Science
,
309
,
115
118
.
Trenberth
,
K. E.
,
1999
:
Atmospheric moisture recycling: Role of advection and local evaporation.
J. Climate
,
12
,
1368
1381
.
Trenberth
,
K. E.
, and
D. J.
Shea
,
2005
:
Relationships between precipitation and surface temperature.
Geophys. Res. Lett.
,
32
.
L14703, doi:10.1029/2005GL022760
.
Uvo
,
C. B.
,
2003
:
Analysis and regionalization of northern European winter precipitation based on its relationship with the North Atlantic oscillation.
Int. J. Climatol.
,
23
,
1185
1194
.
van der Schrier
,
G.
,
K. R.
Briffa
,
P. D.
Jones
, and
T. J.
Osborn
,
2006
:
Summer moisture variability across Europe.
J. Climate
,
19
,
2818
2834
.
Verdon
,
D. C.
, and
S. W.
Franks
,
2006
:
Long-term behaviour of ENSO: Interactions with the PDO over the past 400 years inferred from paleoclimate records.
Geophys. Res. Lett.
,
33
.
L06712, doi:10.1029/2005GL025052
.
Visbeck
,
M. H.
,
J. W.
Hurrell
,
L.
Polvani
, and
H. M.
Cullen
,
2001
:
The North Atlantic Oscillation: Past, present, and future.
Proc. Natl. Acad. Sci. USA
,
98
,
12876
12877
.
Wang
,
G. L.
,
2003
:
Reassessing the impact of North Atlantic Oscillation on the sub-Saharan vegetation productivity.
Global Change Biol.
,
9
,
493
499
.
Wang
,
G. L.
,
2005
:
Agricultural drought in a future climate: Results from 15 global climate models participating in the IPCC 4th assessment.
Climate Dyn.
,
25
,
739
753
.
Wetherald
,
R. T.
, and
S.
Manabe
,
1999
:
Detectability of summer dryness caused by greenhouse warming.
Climatic Change
,
43
,
495
511
.
Wetherald
,
R. T.
, and
S.
Manabe
,
2002
:
Simulation of hydrologic changes associated with global warming.
J. Geophys. Res.
,
107D
.
4379, doi:10.1029/ 2001JD001195
.
Wood
,
E. F.
, and
Coauthors
,
1998
:
The Project for Intercomparison of Land-Surface Parameterization Schemes (PILPS) Phase 2(c) Red-Arkansas River basin experiment: 1. Experiment description and summary intercomparisons.
Global Planet. Change
,
19
,
115
135
.
Xin
,
X.
,
R.
Yu
,
T.
Zhou
, and
B.
Wang
,
2006
:
Drought in late spring of South China in recent decades.
J. Climate
,
19
,
3197
3206
.
Zhang
,
R.
, and
T. L.
Delworth
,
2006
:
Impact of Atlantic multidecadal oscillations on India/Sahel rainfall and Atlantic hurricanes.
Geophys. Res. Lett.
,
33
.
L17712, doi:10.1029/2006GL026267
.
Zhang
,
X. B.
,
L. A.
Vincent
,
W. D.
Hogg
, and
A.
Niitsoo
,
2000
:
Temperature and precipitation trends in Canada during the 20th century.
Atmos.–Ocean
,
38
,
395
429
.
Zhang
,
Y-K.
, and
K. E.
Schilling
,
2006
:
Increasing streamflow and baseflow in Mississippi River since the 1940s: Effect of land use change.
J. Hydrol.
,
324
,
412
422
.
Zou
,
X.
,
P.
Zhai
, and
Q.
Zhang
,
2005
:
Variations in droughts over China: 1951–2003.
Geophys. Res. Lett.
,
32
.
L04707, doi:10.1029/2004GL021853
.

Footnotes

Corresponding author address: Justin Sheffield, Department of Civil Engineering, Princeton University, Princeton, NJ 08544. Email: justin@princeton.edu