1. Introduction
The northeastern United States (NEUS) is well known not only for its dense population and thriving economy (U.S. Bureau of Economic Analysis 2016) but also for its humid and mild climate (Frumhoff et al. 2007). Unlike drier regions of the country that often suffer from drought, the abundant precipitation of the NEUS generally enables it to avoid prolonged and severe drought. This has been especially true over the relatively wet period since the unprecedented and infamous 1960s drought (Seager et al. 2012; Ford and Labosier 2017). Modern water resource planning in the NEUS is highly reliant on a model drought based off of this 1960s drought period, which was characterized by low temperatures and an anomalously long development period where local irrigation was insufficient to meet water demand (Barksdale 1968; Janes and Brumbach 1965; NIDIS/NOAA 2017, 2021; Xue and Ullrich 2021b). Although a general wetting trend is projected to continue over NEUS due to increases in precipitation, risk from extremely dry conditions does not disappear and short-term extreme droughts are even projected to become more frequent in the future (Frumhoff et al. 2007; Hayhoe et al. 2007; Krakauer et al. 2019). Evidence has also emerged that flash droughts, characterized by short-term lack of rainfall and abnormal evaporative demand associated with high temperature, are an increasing threat for this region (Moser et al. 2008; Trenberth et al. 2014; Otkin et al. 2018; Pendergrass et al. 2020; NIDIS/NOAA 2021). Recent examples of such rapid-onset droughts include the 2016 and 1999 droughts, which occurred during the growing season and led to severe impacts for local agriculture and related sectors (NIDIS/NOAA 2017; Lombard et al. 2020; NIDIS/NOAA 2021). Unlike traditional drought, which is common in dry regions (Kogan 1997; Mishra and Singh 2010), flash droughts can often occur in areas with dense vegetation, as plants can enhance evapotranspiration by drawing on water that is deeper in the soil (Mo and Lettenmaier 2016; Pendergrass et al. 2020; Chen et al. 2021). In examining hydroclimatic trends in this region, it is clear that water resource managers will be expected to deal with several impending challenges brought by the warming climate such as increased risk of wildfire, heatwaves, and drought (Frumhoff et al. 2007; Hayhoe et al. 2007). Historical data also may not be sufficiently reliable for future projections due to loss of stationarity, more intense extreme weather events, and enhanced climate variability (Milly et al. 2008; Armal et al. 2018; Yu et al. 2018; Stryker et al. 2018). Consequently, there remains a need for high-quality climate information to inform regional adaptation strategy (Milly et al. 2008; AghaKouchak et al. 2015).
Earth system models (ESMs) are commonly employed for projecting the impacts of climate change on water resources and the hydrologic cycle (Frumhoff et al. 2007; Kharin et al. 2007; Wagener et al. 2010). In particular, these models allow us to better understand the dynamic nature of drought under climate change (Strzepek et al. 2010; Joetzjer et al. 2012; Christian et al. 2019; Yuan et al. 2019). Although climate models have improved largely over the past several decades, in the process incorporating more complex representations of the atmosphere–land–ocean–sea ice system and its interactions (Rodgers et al. 2015; Bonan and Doney 2018), large and persistent uncertainties limit our ability to produce reliable projections (Xie et al. 2015; Xue and Ullrich 2021a). As one of the three main sources of such climate projection uncertainty, internal variability arises from the unforced natural variability of the climate system and is magnified by coupled model processes (Hawkins and Sutton 2009; Deser et al. 2012; Xie et al. 2015). It potentially accounts for half of the intermodel spread in near-surface variables such as precipitation and temperature over North America over the next 50 years (Deser et al. 2014, 2020). Due to drought’s low frequency and its relatively long duration, internal variability has meant that traditional studies attempting to characterize drought from only one or several future realizations have had difficulty drawing statistically significant conclusions (Hawkins and Sutton 2009; Deser et al. 2012; Xie et al. 2015; National Academies of Sciences Engineering and Medicine 2016). Consequently, projections of drought have largely remained ambiguous (Taylor et al. 2013; Trenberth et al. 2014; Van Loon 2015), suggesting a need for a large number of realizations to extract a signal from the noise (Taylor et al. 2013). Fortunately, this awareness has led to the development of datasets such as the Multi-Model Large Ensemble Archive (MMLEA), which presently includes seven CMIP5-class climate models with at least 16 (and up to 100) ensembles with slightly different initial conditions. This resource and analogous datasets have proven immensely useful for quantifying internal variability, especially at regional scales (Kay et al. 2015; Deser et al. 2020).
All seven large ensemble (LE) models presently available through the MMLEA are used in this study, so as to sample both internal variability and structural uncertainty. Meteorological droughts are identified using both the standardized precipitation index (SPI) and standardized net precipitation index (SNPI). SNPI is analogous to SPI but uses net precipitation (i.e., precipitation minus evapotranspiration) instead of precipitation, and so represents the net moisture input to the land surface. These indices are employed at multiple temporal scales over the historical (1950–2000) and future (2050–2100) periods, with the latter using the RCP8.5 emission scenario. Several recent studies have shown that plant response to increases in carbon dioxide need to be accounted for to properly understand future droughts, as vegetation plays an essential role in modulating evapotranspiration, surface hydrological conditions and the development of the flash droughts (Weiss et al. 2012; Swann et al. 2016; Bonfils et al. 2017; Knauer et al. 2017; Pendergrass et al. 2017; Dai et al. 2018). Consequently, indices such as SPI, SPEI, and the Palmer drought severity index (PDSI), which do not directly account for plant response, may not reliably capture the nature of future droughts (Swann et al. 2016; Dai et al. 2018; Yang et al. 2019; Mankin et al. 2019; Bonfils et al. 2017). Therefore, in this study, we principally rely on SNPI to project the change of water availability and its underlying factors in this region, whereas SPI is used to project dry conditions without considering water demand. Short-term, intermediate-term, and long-term droughts are analyzed using 3-month (SPI3/SNPI3), 6-month (SPI6/SNPI6), and 24-month (SPI24/SNPI24) drought indices (Wilhite 2005; Svoboda et al. 2012; Thomas et al. 2015; Svoboda and Fuchs 2016). Additionally, our investigation of flash droughts will be limited to those flash droughts identifiable with monthly data (here referred to as rapidly developing droughts) by using SNPI1 due to lacking weekly projections in LENS models.
With the MMLEA at our disposal, in this paper we address the following questions: How are droughts of different temporal scales changing over the NEUS? Will rapidly developing droughts become more intense and frequent over the NEUS? What are the underlying factors that affect rapidly developing drought in this region?
2. Data and methods
a. Data
In this study, our analysis uses data from the historical period (1950–2000) and future period (2050–2100) under the RCP8.5 emission scenario (note that MPI-ESM only has data through 2099, so its future period is 2050–99). All seven large ensemble (LE) CMIP5 models provided by the U.S. CLIVAR Working Group on Large Ensembles are employed (NCAR 2020). As shown in Table 1, the LE models have at least 16 and up to 100 ensemble members. Among the seven LE models, five of them capture plant response to increasing CO2, and six of them account for future increases in total vegetation (Rodgers et al. 2015; Hazeleger et al. 2010; Jeffrey et al. 2013; Kay et al. 2015; Kirchmeier-Young et al. 2017; Sun et al. 2018; Maher et al. 2019; Reick et al. 2021). We further use the aggregated set or the multiensemble mean of all ensembles of each model to make projections and estimate uncertainty for our projections. Note that without specific annotation, analysis is based on each model’s aggregated set of all ensembles. By using large ensembles from multiple models, we are able to incorporate structural uncertainties in our sample. Further, enough realizations are employed to compensate for the internal variability, while the 51-yr study period provides a long enough window to capture longer droughts. Drought and its impacts often occur at the watershed scale, and water management decisions are usually also made at this level as well (Diaz 1983; Wilhite and Glantz 1985; Wilhite 2000; Mishra and Singh 2010). Furthermore, surface water balance is primarily valid at the watershed scale (Gleick 1987; Arnell 1999; Wang and Dickinson 2012; Peel et al. 2010), and so our analysis focuses on regional monthly mean data over the New England region defined by the Watershed Boundary Dataset (WBD) (U.S. Geological Survey 2013). All datasets are first interpolated to 1° × 1° by conservative interpolation (Schulzweida 2019) before masking to ensure each dataset has the same boundary.
Models of the multimodel large ensemble archive (MMLEA) and data repository from Deser et al. (2020).
b) Methods
1) Standardized precipitation index
As one of the most widely used drought indices, the SPI is a flexible drought indicator designed to capture the magnitude of meteorological drought conditions at different temporal scales (McKee et al. 1993; Guttman 1999; Svoboda and Fuchs 2016). Intuitively, the value of SPIn represents how much the accumulation of the past n months’ precipitation departs from average conditions. The value of the index is then used to classify droughts following Table 2. Here, we specifically use SPI3, SPI6, and SPI24 to quantify short-term, intermediate-term, and long-term droughts, following a number of past studies (Wilhite 2005; Svoboda et al. 2012; Svoboda et al. 2012; Thomas et al. 2015).
SPI must first be calibrated before use, which entails estimation of the arguments of the gamma distribution for a given region and realization. This procedure allows us to use the inverse CDF to transform precipitation data to be normally distributed. For each ensemble member, the historical period is used for this calibration. SPI values given here will then use the same calibration for the historical period (1950–2000) and the future period (2050–2100) to illustrate the impacts of climate change on drought conditions. If the future climate has the same drought statistics as the historical period, future SPI should follow the standard normal distribution, as in the historical period. On the other hand, any departure from normality is evidence of changing drought patterns. Note that for SPI, values are clipped to the range from −3.09 to 3.09 (corresponding to normal probabilities of 0.001 and 0.999) as suggested by Guttman (1999).
Although SPI is widely used in practice and research (Hayes et al. 2002; Wilhite 2005; Svoboda and Fuchs 2016), and is recommended by the World Meteorological Organization (WMO) (Svoboda et al. 2012; Svoboda et al. 2012) to assess meteorological droughts, it does have some known shortcomings (Vicente-Serrano et al. 2010; Dai 2011; Vicente-Serrano et al. 2012). For instance, SPI only relies on precipitation, under the argument that precipitation is the primary driver of drought and is more variable than other relevant quantities (McKee et al. 1993; McKee 1995; Hayes et al. 2002). As such, it does not account for intensification in evapotranspiration brought on by a warming climate (Seager et al. 2012; Lyon et al. 2005; Frumhoff et al. 2007) and associated enhancement in drought conditions (Vicente-Serrano et al. 2010; Dai 2011; Vicente-Serrano et al. 2012). As a result, SPI is unable to detect flash droughts primarily driven by enhanced evaporative demand (Otkin et al. 2013; Pendergrass et al. 2020). This has led to the use of alternate drought indices, such as the standardized precipitation evapotranspiration index (SPEI) (Vicente-Serrano et al. 2010) and PDSI (Palmer 1965), which rely on the potential evapotranspiration (PET) derived from offline empirical equations like the Penman–Monteith equation. However, as with SPI, the PET equations use only atmospheric variables and do not consider the change of physiological plant response to increasing CO2, a potentially essential factor in estimating future surface water balance. It is known that increases to atmospheric CO2 will enhance photosynthesis efficiency: by reducing their stomatal conductance, plants can decrease transpiration (water losses) per unit of carbon gain, then further mitigate plant water stress and increase soil moisture (Field et al. 1995; Swann et al. 2016; Milly and Dunne 2016; Swann 2018). Consequently, several studies have argued that capturing the transpiration response from plants to rising CO2 enables more reliable drought projections (Swann et al. 2016; Milly and Dunne 2016; Bonfils et al. 2017; Dai et al. 2018). This suggests that indices that employ PET will overestimate drought intensification under a warming climate (Sellers et al. 1996; Betts et al. 2007; Roderick et al. 2015; Milly and Dunne 2016; Yang et al. 2019) and so may not be appropriate for investigations of drought under climate change. Moreover, vegetation has been shown to be an essential factor for determining surface hydrology and estimating evapotranspiration, and so increases in evapotranspiration driven by more vegetation in the growing season can potentially trigger flash drought (Weiss et al. 2012; Zhang et al. 2015; Knauer et al. 2017; Pendergrass et al. 2017; Otkin et al. 2018; Pendergrass et al. 2020). This emphasizes the importance of a correct representation of vegetation in climate projections, which is otherwise absent in traditional PET calculations (Weiss et al. 2012, 2014; Frank et al. 2015; Zhang et al. 2015; Knauer et al. 2017).
2) Standardized net precipitation index
This adjustment is necessary since the gamma distribution cannot be used for negative quantities. As with the calculation of SPI (McKee et al. 1993; McKee 1995; Guttman 1999), the adjusted net precipitation is fitted to a gamma distribution and subsequently transformed to be normally distributed. Since in this study we will compare drought conditions defined by SPI/SNPI, fitting the data to the same distribution makes the comparison more straightforward. However, we have also confirmed that the fit of the gamma distribution to the adjusted net precipitation is as good as its fit to precipitation, and superior to the log-logistic distribution (see Text S4 in the online supplemental material).
Because the calibration of SPI is sensitive to the length of data employed (e.g., McKee et al. 1993; Guttman 1999), using more than 50 years of data is recommended. In our study, we calibrate both historical and future SNPI/SPI based on the historical period, which includes 51 years (1950–2000). That is, the gamma distribution parameters used in transforming the precipitation and adjusted net precipitation of both periods from a gamma distribution to a normal distribution are all calculated based on the historical data. Only the historical period is used as calibration to avoid future projections impacting the SPI/SNPI of the historical droughts, and so that differences in the future can be identified departures from normality.
3) Drought duration and magnitude
4) Defining rapidly developing drought
While the term “flash drought” is used widely in the literature, there remains some disagreement on the precise time scale of these events. In general, a flash drought is a subseasonal drought event (Christian et al. 2021; Pendergrass et al. 2020) and so can refer to rapid emergence of drought on submonthly or weekly time scales. Nonetheless, the most infamous flash drought events over CONUS are monthly flash droughts, including the 2012 U.S. Midwest flash drought and 2017 U.S. northern Great Plains flash drought (Hoell et al. 2020; Hoerling et al. 2012; Otkin et al. 2021). Additionally, as storage considerations have meant that climate models generally only provide monthly hydroclimate data like the LE models used in this study, analysis of submonthly events in the climate context remains difficult. For these reasons, our study focuses exclusively on rapidly developing droughts (a term here that includes flash droughts developing on monthly scales). Our use of monthly data means that we cannot fully capture all forms of flash drought, particularly those that develop on time scales shorter than a month. Submonthly flash drought development will be a target for future work, as data with daily time resolution comes available in the future.
Since flash drought is characterized by rapid development and extremely dry conditions after development (Otkin et al. 2018; Pendergrass et al. 2020), herein we define rapidly developing drought in month i as DSDi ≥ 2 and SNPI1i ≤ −2. Intuitively, this means that, at the very least, “extremely dry conditions” (SNPI1 ≤ −2) should develop from conditions that are at least neutral (SNPI ≥ 0) in only one month. We also define a notion of “extreme rapidly developing drought” since extreme events of this nature have the potential to result in enormous damage, particularly to agriculture (Otkin et al. 2018; Pendergrass et al. 2020). Herein, extreme rapidly developing drought is defined as drought with DSDi ≥ 4 and SNPI1i ≤ −2, indicating that extremely dry conditions develop from extremely wet conditions in only one month. Because flash drought is strongly associated with enhanced evaporative demand, previous studies of flash droughts usually exclude the winter season (Pendergrass et al. 2020; Ford and Labosier 2017; Otkin et al. 2018). Herein we will also ignore the winter season when we analyze trends for rapidly developing droughts. Moreover, we admit that such a definition based on SNPI1 might fail to capture flash droughts at submonthly scale and those occur in the middle of the months.
3. Results
a. Result 1: A general wetting trend is undeniable in the mean
The significant wetting trend expected for the NEUS shows up clearly in Fig. 1 as a rightward shift of the mean of each model’s probability distribution. This trend is overwhelming in the SPI at all temporal scales, and for all LE models. However, the shift in SPI is unsurprising as this index only takes precipitation as input, and there is high confidence from both historical observations and climate models that annual mean precipitation is presently increasing (and will increase) over the NEUS (Huntington et al. 2004; Hayhoe et al. 2008; Frumhoff et al. 2007; Guilbert et al. 2015; Demaria et al. 2016). As for the SNPI, even when increases in evapotranspiration are taken into account, a long-term wetting trend remains in most models (5 out of 7), confirming a clear tendency toward wetting of the NEUS. Moreover, for both SPI and SNPI, this wetting trend is more obvious at longer temporal scales. Examining the ratio of the frequency density in Fig. 2, we see that long-term dry conditions (SPI24 < −1) are projected to become clearly less frequent (occurring ∼90% less often), while long-term pluvial conditions (SPI24 ≥ 2) are between 2 and 100 times more likely.
SPI/SNPI density of the future period (2050–2100) for all seven LE models. Each vertical line indicates the mean value from that model.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-21-0810.1
SPI/SNPI frequency density ratio for all seven LE models, defined as the frequency density of the future period divided by the frequency density of the historical period.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-21-0810.1
b. Result 2: More extremely wet conditions, with greater magnitude
The tendency toward wetter conditions is most obvious at the extreme right tail of Figs. 1 and 2 (i.e., “extremely wet” months), with both more frequent conditions with SPI/SNPI ≥ 2, and larger average SPI/SNPI for months in this category (as well as more months at the right boundary, 3.09). This accords with the continued intensification of extremely wet conditions over the NEUS (Melillo et al. 2014; Donat et al. 2016; Pfahl et al. 2017). Using SNPI, all LE models favor an increase in the frequency of extremely wet conditions and greater SNPI values within this category. For example, the multimodel mean probability of extremely wet months surges from 0.018 to 0.069 (for SPI1) and from 0.019 to 0.048 (for SNPI1). More significantly, exceptionally wet conditions (SPI1/SNPI1 equal to 3.09) will become 19 and 12 times more frequent, using SPI1 and SNPI1 respectively. These changes suggest greater challenges for water management in this region, as well as increased risk and intensity of flooding.
c. Result 3: More short-term extreme droughts and intensified evapotranspiration
While these two prior results may seem to suggest abundant water resources in the future, this general wetting trend does not imply that drought will disappear entirely. Although short-term extreme precipitation droughts (months with SPI3 ≤ −2) are projected to be less frequent in all LE models, nearly all models (6 out of 7) agree that short-term net precipitation droughts (SNPI3 ≤ −2) are projected to increase largely in frequency (Fig. 2). Further, short-term exceptional droughts (SNPI3 equal to −3.09) are expected to be more frequent in all models. This increase in short-term extreme drought appears to be driven by concordant increases in two factors: evapotranspiration and precipitation variability (Pendergrass et al. 2017). Because of evapotranspiration’s essential role in the water cycle (as the second-largest component in the water balance formula after precipitation; Melillo et al. 2014) and its significance in producing drying over land (Zhao and Dai 2015; Wilhite 2000; Dai et al. 2018), increases of evapotranspiration need to be accounted for when projecting the changing trends of droughts under a warming climate. Increasing precipitation variability, on the other hand, is associated with an increase in the frequency of both months with extremely low precipitation and high precipitation (and corresponding signals in streamflow), as has been documented in several past studies (Hayhoe et al. 2007; Frumhoff et al. 2007; Van Loon 2015; Demaria et al. 2016). As discussed later, increases in the frequency of extremely low values of SPI/SNPI in Figs. 1 and 2 are strongly related to a flattening of the probability density of precipitation/net precipitation intensity around the mean in Fig. 3 (Zhao and Dai 2015). This flattening suggests less frequent “normal conditions” and more frequent extremes. It is quantified by the departure from unity (the standard deviation of SPI/SNPI under a normal distribution) of the multimodel mean standard deviations of SPI1/SNPI1 to 1.16/1.20 in the future.
Each LE model’s density plot and density difference of precipitation, surface air temperature, evapotranspiration, and net precipitation.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-21-0810.1
d. Result 4: More rapidly developing drought with faster initiation and greater intensity
Because of the strong seasonality of flash drought (Otkin et al. 2018; Christian et al. 2019; Pendergrass et al. 2020), each LE model’s projections for the development speed of rapidly developing drought in each season except the winter season is shown in Fig. 4 (here MAM indicates the spring season, JJA indicates the summer season, and SON indicates the autumn season). Although there are discrepancies among seasons and models, several significant trends are apparent. The most concrete among these being that nearly all LE models agree that the frequency of rapidly developing drought is projected to increase significantly in all seasons (except for CanESM2 and CESM1-CAM5 in JJA). The average increase in the probability of rapidly developing drought is about 106%, indicating that rapidly developing drought could occur more than twice as frequently in the future period. Further, rapidly developing drought is also projected to initiate faster and with greater severity, as suggested by the smaller values of the median, third quartile, and lower whiskers of the DSD. Note that the p values in Fig. 4 indicate the statistical significance of the t test about whether the future and historical periods have the same average delta SNPI1 instead of if they have the same average rapidly developing drought frequency.
Development speed of rapidly developing drought for each season (except winter) and model. The black bold numbers above each boxplot indicate the average frequency within each period (51 years). The red bold numbers indicate the p values obtained from a t test of the delta SNPI1 mean between historical and future periods.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-21-0810.1
e. Result 5: Significant trends toward more extreme rapidly developing drought, particularly in the spring
Comparing Fig. 4 to the corresponding plot for extreme rapidly developing drought (Fig. 5), it is apparent that into the future extreme rapidly developing drought frequency tends increase even more than rapidly developing drought frequency. The average return period of extreme rapidly developing drought (defined as the multimodel mean years needed to occur once) drops from 65.9 years historically to 16.2 years in the future. By contrast, the return period for all rapidly developing droughts decreases from 6.3 to 2.9 years. This historical rapidly developing drought return period agrees with previous work by Ford and Labosier (2017), which is based on an examination of surface soil water content. Extreme rapidly developing drought also has a strong seasonality, with intensification being much more significant during the growing season, particularly in the spring (MAM). This poses substantial risk for agricultural productivity, and suggests the possibility of more frequent crop failure (Otkin et al. 2018; Pendergrass et al. 2020). The development speed of extreme rapidly developing drought is also trending higher, producing the possibility that extremely dry conditions may immediately follow extremely wet conditions (DSDi ≥ 4). Such rapid development of intense drought would be unprecedented and hard to predict, and contrary to the traditional view of drought as a slowly evolving extreme weather event.
Development speed of extreme rapidly developing drought for each season (except winter) and model. The black bold numbers above each boxplot indicate the average frequency within each period (51 years). The red bold numbers indicate the p values obtained from a t test of the delta SNPI1 mean between historical and future periods.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-21-0810.1
f. Result 6: No clear consensus on intermediate-term drought
As depicted by Fig. 1, for intermediate-term drought there are substantial disagreements of future change among LE models and between the two drought indexes (SPI6 and SNPI6). Although nearly all models indicate that the magnitude of exceptional intermediate droughts (SPI6/SNPI6 at the lowest boundary of −3.09) are projected to increase, SPI and SNPI show significant divergence at this time scale. All LE models agree that monthly intermediate-term dryness tends to become less frequent (months with SPI6 ≤ −1) and general wetness (the average SPI6) tends to increase; however, when looking at SNPI6, only four out of seven models project such increases. Therefore, to better describe the changing patterns of intermediate drought, we compare three essential drought features: drought duration, frequency, and magnitude.
From Fig. 6, we can see that, as a result of significant increases in precipitation, all LE models project decreases in SPI6-derived intermediate drought frequency, average duration, maximum duration, average magnitude, and maximum magnitude. On the other hand, for the SNPI6-derived intermediate drought, most LE models suggest that intermediate-term droughts tend to be more frequent (6 out of 7) but with shorter average duration (5 out of 7). However, the mean and maximum intermediate-term drought magnitude exhibits a wide spread between periods with little agreement between models, suggesting no statistically significant change.
Duration and magnitude of intermediate-term droughts derived from (left) SPI6 and (right) SNPI6. For duration, the black bold numbers above each boxplot indicate the average frequency (the number of dry months per 51-yr period). The red bold numbers indicate the p values obtained from a t test of the drought duration mean between historical and future periods. For drought magnitude (the accumulated SPI6/SNPI6 during each drought event), the black bold numbers above each boxplot indicate the mean drought magnitude. The red bold numbers indicate the p values obtained from a t test of the drought magnitude mean between historical and future periods.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-21-0810.1
4. Discussion
Rapidly developing drought and unprecedented rapidly developing extreme droughts are projected to be a major challenge into the future, with faster initiation and greater intensity; this result holds especially true in the spring season, and is certainly projected to threaten local agricultural production. Meanwhile, the underlying causes of flash drought intensification is still ambiguous, so we will further explore the potential drivers of increasing flash droughts/rapidly developing droughts. In this study, we have used SNPI1 and its change over subsequent months to quantify the frequency and intensity of rapidly developing droughts, that is, rapidly developing water shortage related to precipitation, temperature, and evaporative demand (Taylor et al. 2013). By construction, changes to the frequency and character of rapidly developing drought are projected to be brought about through changes to net precipitation (i.e., precipitation inputs minus evapotranspiration outputs). In this section, we turn our attention to explaining future trends in drought and rapidly developing drought through a detailed examination of precipitation, temperature, evapotranspiration, leaf-area index, and their associated relationships. In general, we note that increasing monthly precipitation variability affects the inputs to the net precipitation, while P/E anticorrelation modifies the outputs. Consequently, these modifications to net precipitation drive a clear intensification of flash droughts/rapidly developing droughts over the NEUS, particularly in the spring season.
a. More precipitation and more variable precipitation in the NEUS
In Fig. 3 (left column), a clear positive shift in mean precipitation intensity is observed in all models, along with a flattening of the frequency density and subsequent increase in the frequency in both tails. This change is indicative of the well-known shift to increased precipitation variability under global warming, driven by increasing moisture and mitigated by weaker atmosphere circulation (Pendergrass et al. 2017). With that said, precipitation change over the NEUS has a significant seasonal pattern (Frumhoff et al. 2007; Hayhoe et al. 2007) that necessitates a deeper examination of each season separately. To that end, Fig. 7 shows the precipitation mean and standard deviation of each LE model. From here it is quickly apparent that nearly all LE models project that precipitation exhibit a significant increase in its mean and variance over all seasons, in agreement with historical observations of increasing annual mean precipitation and intensified extreme precipitation events (Frumhoff et al. 2007; Demaria et al. 2016; Pendergrass et al. 2017). Among the seven LE models, the relative change between historical (1950–2000) and future periods (2050–2100) of the mean ranges from 7.57% to 15.70%, while the increase in the standard deviation ranges from 14.10% to 35.90%.
Mean and standard deviation of each LE model’s precipitation during historical and future periods. The bold numbers above each bar indicate the value of the standard deviation.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-21-0810.1
Increasing regional mean precipitation is certainly the main driver behind the general wetting trend in this region; however, the increase is not uniform over four seasons. As depicted in Fig. 7, winter (DJF) and spring (MAM) produce a greater precipitation increase, with multimodel mean relative change of 20.36% and 16.75%, significantly exceeding increases in the summer (4.38%) and autumn (3.31%). This seasonal dependency agrees with previous studies examining GCM projections (Frumhoff et al. 2007; Hayhoe et al. 2008; Xue and Ullrich 2021b). Similarly, increases in precipitation variability in the winter and spring seasons are also larger than corresponding increases in summer and autumn. Notably, each LE model’s mean and standard deviation are strongly correlated in almost all seasons: the correlation in winter, spring, and autumn is 0.95, 0.96, and 0.74; however, this correlation is not significant in summer (0.04), suggesting this season is dominated by different processes. This synchronization is a topic for potential future exploration. In conjunction with warmer temperatures shifting snowfall to rainfall, particularly in the spring, it is clear that these two seasons may experience more extreme precipitation and flooding (Xue and Ullrich 2021b). On the other hand, as shown in Fig. 3, six out of seven models agree that months with nearly zero monthly precipitation (monthly precipitation less than 0.5 mm day−1) tend to be more likely in the future, thus supporting an increase in the occurrence of rapidly developing droughts.
b. Evapotranspiration is energy limited within the humid NEUS
Increased precipitation variability is one of the primary drivers behind possibility of emerging rapidly developing drought; however, this single factor is insufficient to explain why spring is at the greatest risk for increased rapidly developing drought, while it does not have the greatest increase in variability. To understand this discrepancy, it is necessary to additionally consider evapotranspiration’s role in the development of rapidly developing droughts. Evapotranspiration depends on the water and energy availability, and so it is important to ascertain whether particular regions are energy-limited or water-limited (Roderick et al. 2009a,b; Teuling et al. 2009). Put simply, the amount of available energy can limit evapotranspiration in regions where water is abundant, whereas water supply limits it in regions of abundant energy. As not all LE models provide radiation data, here we employ surface air temperature and precipitation to quantify the supply of energy and water, respectively. While other factors are important for estimating evapotranspiration, such as surface wind and relative humidity, changes in these factors under global warming are modest compared with changes in precipitation and temperature (Held and Soden 2006; Laîné et al. 2014; Trenberth et al. 2014; Ma et al. 2016). If this probability is lower than the conventional 5% ( p < 0.05) the correlation coefficient is called statistically significant.
In the NEUS, all LE models exhibit very strong correlations between the multiensemble mean surface air temperature and evapotranspiration. Correlations range from 0.92 to 0.99 (with p values much less than 0.05, which indicates that the correlation is statistically significant at 95% confidence level), with the mean of 0.95 during the historical period (1950–2000). This trend is even more significant in the spring season with a mean correlation of 0.97 (with p values much less than 0.05). Consequently, it is safe to say that evapotranspiration over the NEUS is primarily energy-limited. On the other hand, correlations between evapotranspiration and precipitation vary among models, with a mean correlation of 0.04 (with all p values less than 0.05). To further confirm that surface air temperature increases are the primary driver of evapotranspiration differences, we calculate the correlation of each ensemble’s average deltas of precipitation, evapotranspiration, and surface air temperature between the historical and future periods over all LE members. As in the historical data, the correlations between the deltas of evapotranspiration and surface air temperature are obvious, with the multiensemble mean ranging from 0.15 to 0.83 (5 out of 7 models have p values less than 0.05). However, examining the correlations between deltas of evapotranspiration and precipitation does not produce robust relationship, with the multiensemble mean correlations varying from −0.24 to 0.65 with a multimodel mean of 0.07 (only one model has a p value less than 0.05).
c. Differences in the trends of precipitation and evapotranspiration variability
From Fig. 3, we see that the density plots of surface air temperature and evapotranspiration essentially all shift to the right while maintaining the same “shape” of the distribution. This is in contrast to precipitation, which exhibits both an increase in its mean and a widening of its distribution. Although evapotranspiration and precipitation show increases in the multimodel mean of 13.76% and 11.43% over the NEUS, the relative change of the standard deviation of surface air temperature (0.77%) and evapotranspiration (−0.96%) are essentially negligible compared when compared with that of precipitation (22.86%). Note that the small decrease in evapotranspiration variability appears to primarily emerge from a decrease in the frequency of low evapotranspiration months. Nonetheless, net precipitation is subject to both a shift in its mean (with a multimodel mean increase of 7.52%) and an increase to its variability (with a multimodel mean increase of 20.16%). Namely, while the regional climate becomes wetter, there will be an increase in the frequency of both the extremely dry and wet conditions that is clearly apparent in the fourth column of Fig. 3. Looking more closely into each season, the variability increase is much more significant in the spring as a result of greater precipitation variability in these seasons; however, MAM and SON evapotranspiration variability actually decreases slightly (with multimodel mean relative change of −3.11% and −2.43%). MAM is projected to produce the most discrepant variability trends, with reduced evapotranspiration variability and simultaneously greater precipitation variability. Consequently, the spring tends to have the largest increases in the variability of net precipitation. The disparate behavior of precipitation and evapotranspiration leads to increasing anticorrelation of these processes, which we investigate next.
d. Increasing anticorrelation of evapotranspiration and precipitation
In the NEUS, where evapotranspiration is energy limited, there is a very strong positive correlation between monthly mean surface air temperature and evapotranspiration that occurs in all seven LE models over the historical period; however, correlations between precipitation and evapotranspiration vary largely across the seven LE models, ranging from −0.29 to 0.65 (with p values less than 0.05). Under future global warming, the positive correlations between surface air temperature and evapotranspiration continue to be evident, with a multimodel mean correlation of 0.92 (with p values less than 0.05). In the spring season, these correlations even reach a multimodel mean value of 0.98 (with p values less than 0.05). On the other hand, the correlations between precipitation and evapotranspiration become increasingly negative into the future over all seasons (with an average decrease of 0.24). This trend is even more significant in the spring (with an average decrease of 0.46). Annually, six out of seven LE models project that these correlations will be far more negative in the future, compared with the historical period. For the only model that shows an increased correlation, CSIRO-Mk3.6.0, precipitation and evapotranspiration in the future spring season remains negative (−0.34) (with p value less than 0.05). Across all models, multimodel mean historical correlations between precipitation and evapotranspiration are roughly zero annually (0.04) and in the spring season (−0.04); in the future they decrease to −0.21 and −0.50 respectively. Such a negative shift in correlation indicates that, during future periods of low precipitation, evapotranspiration will be more likely to be greater than under analogous historical circumstances, in turn magnifying the local moisture deficit. Similarly, when precipitation is high, evapotranspiration is not expected to be obviously larger than is mean state as what it would have been historically. These two effects, in turn, exacerbate extreme events such as rapidly developing drought and the extremely wet conditions, as apparent in both tails of density plots of Fig. 8.
Frequency plots of net precipitation in each season for (top) the historical period and (middle) the future period, along with (bottom) their difference.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-21-0810.1
Model-ensemble average normalized regional monthly mean precipitation and evapotranspiration in the spring during historical and future periods. Data are normalized to their percentiles within historical and future period respectively.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-21-0810.1
The points of anticorrelated low moisture condition we are interested in are located at the upper left corner of the figure. Indeed, we see that six out of seven models project a large increase of such anticorrelated low moisture conditions in the spring with a multimodel mean relative increase of 54.80%, with little doubt that this shift is attributed to P/E anticorrelation. As a result of increasing P/E anticorrelation, anticorrelated high moisture conditions (precipitation larger than 50th percentile and evapotranspiration no larger than 50th percentile) will also increase, providing favorable moisture conditions for floods.
In the next section we explore the processes underlying these shifts and argue that an earlier onset (spring) of growing seasons is a likely culprit (Backlund et al. 2008; Christiansen et al. 2011).
e. Modified evapotranspiration partitioning brought by the extension of the growing season
Several previous studies have shown that changes in vegetation are a key factor behind trends in evapotranspiration (Zhang et al. 2001, 2015; Peel et al. 2010), while intensified water demand from plants in the growing season also plays an essential role in the development of flash drought (Otkin et al. 2018; Pendergrass et al. 2017, 2020). These processes have certain implications for the regional hydroclimate of the NEUS, as this region is predicted to experience a longer growing season with denser vegetation under a warming climate (Frumhoff et al. 2007; Christiansen et al. 2011; Xue and Ullrich 2021b). Indeed, we now argue that this is the primary factor driving the increasing P/E anticorrelation in the spring (and to a lesser degree in the autumn). Specifically, we argue that the modified partitioning of evapotranspiration brought on by a prolonged growing season directly drives this anticorrelation and, consequently, flash droughts/rapidly developing drought.
Total evapotranspiration is defined as the sum of evaporation from soil and surface water (Esoil), the evaporation from water intercepted by plants (Eplant), and the plant transpiration (Tplant) (Ferguson and Veizer 2007; Wang and Dickinson 2012; Lee et al. 2010). Of these, transpiration is the largest contributor, although all three are captured in modern ESMs (Dirmeyer et al. 2005; Lawrence et al. 2007). A number of past studies have used observations and models to demonstrate the impacts of interannual shifts in vegetation on evapotranspiration that occur through modification of its partitioning into these categories (Lawrence et al. 2007; Lawrence and Chase 2009; Gong et al. 2007; Jung et al. 2010; Wang et al. 2010; Wang and Dickinson 2012). It has been noted that soil evaporation can only occur if it acquires water at the surface level, because of the disconnect between surface and deep soil layers, especially during the dry season and over bare soil (Heitman et al. 2008; Wang and Dickinson 2012); however, transpiration from plants can extract water from deeper soil through their rooting systems. As a result, most soil evaporation occurs along with or shortly following precipitation, while transpiration can exhibit a lagged response to precipitation and is associated more with biological processes and energy available (solar radiation) (Williams et al. 2004; Wang and Dickinson 2012). As a direct consequence of increased vegetation in the NEUS, particularly in the spring, there is good reason to believe vegetation plays a greater role in driving surface evapotranspiration. Namely, evapotranspiration tends to draw more water from the rooting zones instead of surface soil, and so will be more weakly associated with recent or concurrent precipitation. In turn, there is an increasing anticorrelation of evapotranspiration and precipitation.
To further confirm this hypothesis, we employ the large ensemble CESM1 model, the only LE model that provides three evapotranspiration components (Esoil, Eplant, and Tplant) and total leaf area index (TLAI), to examine changing trends in vegetation and evapotranspiration partitioning. We also introduce the plant transpiration component ratio (PTCR), which is defined as the ratio of plant transpiration to the soil evaporation. PTCR is employed as a metric to quantify the increasingly noteworthy role of plants in driving surface evapotranspiration. From Fig. 10, we observe a clear increase in ensemble-mean TLAI over the next century, which is particularly significant in the spring season: compared with the historical period (1950–2000), the future period (2050–2100) TLAI is projected to be 33.42% larger over all seasons and 45.24% larger over the spring season only. Annual mean PTCR is highly associated with more vegetation (TLAI) (with a correlation of 0.75 (with a p value less than 0.05) over all seasons and 0.91 (with a p value less than 0.05) during the spring season). It has risen continuously since the 1950s, indicative of how transpiration from plants is becoming a greater component of surface total evapotranspiration. The relative change of PTCR in the spring (39.54%) is much larger than the relative change during all season (10.35%) and other individual seasons, supporting our conclusion that the extension of growing season is the main driver of P/E anticorrelation in the spring. Note that the TLAI has a faster increase than PTCR, which is potentially caused by the increasing water use efficiency, as has been illustrated by the observations (Keenan et al. 2013; Swann et al. 2016; Keeling et al. 2017).
Ensemble mean annually averaged monthly plant evapotranspiration component ratio and leaf area index from CESM1-CAM5 during all seasons and the spring season.
Citation: Journal of Climate 35, 22; 10.1175/JCLI-D-21-0810.1
Although changes to evapotranspiration partitioning in the growing season and its downstream effect on the increasing P/E anticorrelation is well supported by past studies and CESM1-CAM5 data, one could hypothesize that the increasing P/E anticorrelation is instead related to the increasing soil moisture brought by earlier snowmelt in the growing season (Frumhoff et al. 2007; Xue and Ullrich 2021b). To demonstrate this is not the case, we employ all LE models that provide soil moisture data (CESM1-CAM5, CSIRO-Mk3.6.0, GFDL CM3, and MPI-ESM) and examine if soil moisture is higher in the spring season into the future. While this would, in turn, provide more water for evapotranspiration (see Text S5 in the online supplemental material), it is notable that evapotranspiration is primarily energy limited in this region (as discussed earlier). However, even if this is the case, these models actually exhibit soil drying in the spring season: although there are some discrepancies among the four models, they all agree that the topsoil moisture (defined as the average soil moisture of the first 0.1-m soil) will decrease into the future. Indeed, three out of four models agree that such a decrease also occur within all soil layers. This indeed supports the claim that snowmelt’s influence on springtime soil moisture is not the primary driver of the observed P/E anticorrelation. Importantly, we find that all models agree that the correlation between the ensemble mean soil moisture and net precipitation is much larger than the correlation of ensemble mean soil moisture and precipitation, again illustrating the role of net precipitation in driving soil moisture.
We note that CSIRO-Mk3.6.0 is the only model among the LE models that does not produce an increased P/E anticorrelation into the future. This is also the only model among those analyzed that does not account for changes in vegetation into the future. This lends credence to our theory of vegetation being the primary driver of increasing anticorrelation. Also, we need to clarify that although some models project that there exists a negative correlation between precipitation and evapotranspiration in the future MAM, it does not mean that relatively high precipitation directly determines the relatively low evapotranspiration and also does not suggest that the region is water limited. Instead, this is caused by different monthly changing trends brought by various factors. For example, the precipitation will increase more in March (0.81 mm day−1) instead of May (0.26 mm day−1) in the future, which we believe is mainly driven by more warming in cold months (also shown above) due to the snow-albedo feedback (Xue and Ullrich 2021b). Meanwhile, compared with March, May will experience a stronger growing season extension because the length of the growing season is usually defined by the threshold based on daily minimum temperature (Christiansen et al. 2011). Therefore, a stronger growing season extension makes May have a larger transpiration increase compared with March; we can see the CESM1 multiensemble plant transpiration increase is much larger in May (0.28 mm day−1) compared with March (0.06 mm day−1). More warming in March and a larger growing season extension in May make evapotranspiration have a fairly even increase in March, April, and May with multimodel means of 0.44, 0.39, and 0.35 mm day−1. Therefore, evapotranspiration will not have the same monthly changing trend as the precipitation, which partially induces their negative correlations in the future.
Note that we rely on the SNPI1 to project rapidly developing drought, including some slower-developing flash droughts, in this study. However, because the data are only available at monthly frequency, such an analysis will not capture the most rapidly emerging flash droughts, such as those that emerge and resolve at submonthly time scales. It is nonetheless apparent that SNPI1 performs well for a number of real-world flash drought case studies, conducted over the NEUS and U.S. Midwest (see the online supplemental material). A comprehensive examination of SNPI across a larger selection of regions and at finer temporal scales will be considered in future work and is needed before SNPI can be employed more widely as a flash drought metric. Also, this study only analyzes those projections under the RCP8.5 emission scenario. Therefore, it lacks consideration of other emission scenarios and their corresponding projections. Moreover, the large ensemble models used are imperfect and uncertainties in the relevant processes, especially those related to the land surface, necessitate further study of projected changes in actual soil moisture.
5. Conclusions
To better understand trends in drought character and frequency over the NEUS, we have applied both the standardized precipitation index (SPI) and standardized net precipitation index (SNPI) to seven large ensemble model datasets. These two indices have subsequently enabled insight into shifts in this region’s precipitation characteristics and atmospheric water demand characteristics, respectively. Short-term, intermediate-term, and long-term droughts are explored with SPI3/SNPI3, SPI6/SNPI6, and SPI24/SNPI24, while our study of rapidly developing drought employs SNPI1, which captures drought events that are characterized by rapid development. A clear rightward shift in the SPI probability distributions indicates that the NEUS will experience significant wetting as a result of precipitation increases, particularly at longer time scales. Even in light of similarly increasing mean evapotranspiration, most models (5 out of 7) still project a positive trend in overall water availability. Flattening of the frequency distribution of SPI/SNPI among all LE models indicates more frequent extremes, particularly wet extremes, at the expense of more moderate periods. For example, the multimodel mean frequency of exceptionally wet months (SPI1/SNPI1 equal to 3.09) is predicted to increase by 19 times (using SPI1) and 12 times (using SNPI1). Consequently, more frequent and intense flooding is expected to become a growing concern for local water management. This wetting trend does not imply that the drought is a purely historical concern, however. In light of surging evapotranspiration and increasing precipitation variability, 6 out of 7 models suggest that short-term extreme droughts (SNPI3 ≤ −2) are also projected to be much more frequent in the future. Moreover, rapidly developing droughts (as indicated by SNPI1) will see a 106% increase in frequency, dropping the average return time from 6.3 to 2.9 years, with faster development. Extreme rapidly developing drought, where SNPI1 drops by at least 4 points to below −2, is projected to exhibit a drop in return period from 65.9 to 16.2 years. This increase in frequency for both rapidly developing droughts and rapidly developing extreme droughts is most pronounced in the spring season, and will likely drive significant challenges to the agriculture and ecosystem. Although the LE models do not provide a clear consensus on intermediate-term drought, most models agree that intermediate-term drought tends to be more frequent, with a lower average duration. The projected changes in rapidly developing drought, particularly during the spring season, are attributed to an increase in precipitation variability that is not matched by evapotranspiration; consequently, these two fields become increasingly anticorrelated. Increasing precipitation variability has been well studied, and is generally attributed to increased atmospheric moisture counteracted by a weakening circulation. We confirm that the increasing anticorrelation, which is strongest in the spring, is largely a result of the extension of the growing season and attributed to increase in leaf area index (i.e., more vegetation). Namely, direct soil evaporation is typically highly correlated with precipitation because it draws from moisture in the surface soil; however, rooting enables plants to derive soil moisture from deeper in the soil. As a result, we argue that increased vegetation in the NEUS, which is associated with the shift to a warmer, moister climate is expected to increase the risk of the sudden drying episodes explored in this work.
Acknowledgments.
This research was supported by the RGMA program area(s) in the U.S. Department of Energy’s Office of Biological and Environmental Research as part of the multi-program, collaborative Integrated Coastal Modeling (ICoM) project. This project is also supported by the National Institute of Food and Agriculture, U.S. Department of Agriculture, hatch project under California Agricultural Experiment Station project accession No. 1016611. We thank the U.S. CLIVAR Working Group on Large Ensembles for making available the Multi-Model Large Ensemble Archive via the NCAR Climate Data Gateway. The authors certify that they have no affiliations with or involvement in any organization or entity with any financial interest or non-financial interest in the subject matter or materials discussed in this manuscript.
Data availability statement.
All large ensemble models used in this study are openly available from Multi-Model Large Ensemble Archive provided by the U.S. CLIVAR Working Group on Large Ensembles at https://www.cesm.ucar.edu/projects/community-projects/MMLEA/ as cited in Deser et al. (2020).
REFERENCES
AghaKouchak, A., D. Feldman, M. Hoerling, T. Huxman, and J. Lund, 2015: Water and climate: Recognize anthropogenic drought. Nature, 524, 409–411, https://doi.org/10.1038/524409a.
Armal, S., N. Devineni, and R. Khanbilvardi, 2018: Trends in extreme rainfall frequency in the contiguous United States: Attribution to climate change and climate variability modes. J. Climate, 31, 369–385, https://doi.org/10.1175/JCLI-D-17-0106.1.
Arnell, N. W., 1999: A simple water balance model for the simulation of streamflow over a large geographic domain. J. Hydrol., 217, 314–335, https://doi.org/10.1016/S0022-1694(99)00023-2.
Backlund, P., A. Janetos, and D. Schimel, 2008: The Effects of Climate Change on Agriculture, Land Resources, Water Resources, and Biodiversity in the United States. DIANE Publishing, 240 pp.
Barksdale, H. C., 1968: The Northeast Water Supply Crisis of the 1960’s. U.S. Department of the Interior Geological Survey, 14 pp., https://doi.org/10.3133/70039594.
Betts, R. A., and Coauthors, 2007: Projected increase in continental runoff due to plant responses to increasing carbon dioxide. Nature, 448, 1037–1041, https://doi.org/10.1038/nature06045.
Bonan, G. B., and S. C. Doney, 2018: Climate, ecosystems, and planetary futures: The challenge to predict life in Earth system models. Science, 359, eaam8328, https://doi.org/10.1126/science.aam8328.
Bonfils, C., and Coauthors, 2017: Competing influences of anthropogenic warming, ENSO, and plant physiology on future terrestrial aridity. J. Climate, 30, 6883–6904, https://doi.org/10.1175/JCLI-D-17-0005.1.
Cassano, J. J., P. Uotila, A. H. Lynch, and E. N. Cassano, 2007: Predicted changes in synoptic forcing of net precipitation in large Arctic river basins during the 21st century. J. Geophys. Res., 112, G04S49, https://doi.org/10.1029/2006JG000332.
Chen, L., T. W. Ford, and P. Yadav, 2021: The role of vegetation in flash drought occurrence: A sensitivity study using Community Earth System Model, version 2. J. Hydrometeor., 22, 845–857, https://doi.org/10.1175/JHM-D-20-0214.1.
Christian, J. I., J. B. Basara, J. A. Otkin, E. D. Hunt, R. A. Wakefield, P. X. Flanagan, and X. Xiao, 2019: A methodology for flash drought identification: Application of flash drought frequency across the United States. J. Hydrometeor., 20, 833–846, https://doi.org/10.1175/JHM-D-18-0198.1.
Christian, J. I., J. B. Basara, E. D. Hunt, J. A. Otkin, J. C. Furtado, V. Mishra, X. Xiao, and R. M. Randall, 2021: Global distribution, trends, and drivers of flash drought occurrence. Nat. Commun., 12, 6330, https://doi.org/10.1038/s41467-021-26692-z.
Christiansen, D. E., S. L. Markstrom, and L. E. Hay, 2011: Impacts of climate change on the growing season in the United States. Earth Interact., 15, https://doi.org/10.1175/2011EI376.1.
Crowley, J. W., J. X. Mitrovica, R. C. Bailey, M. E. Tamisiea, and J. L. Davis, 2006: Land water storage within the Congo Basin inferred from GRACE satellite gravity data. Geophys. Res. Lett., 33, L19402, https://doi.org/10.1029/2006GL027070.
Dai, A., 2011: Drought under global warming: A review. Wiley Interdiscip. Rev.: Climate Change, 2, 45–65, https://doi.org/10.1002/wcc.81.
Dai, A., T. Zhao, and J. Chen, 2018: Climate change and drought: A precipitation and evaporation perspective. Curr. Climate Change Rep., 4, 301–312, https://doi.org/10.1007/s40641-018-0101-6.
Demaria, E. M. C., R. N. Palmer, and J. K. Roundy, 2016: Regional climate change projections of streamflow characteristics in the Northeast and Midwest US. J. Hydrol., 5, 309–323, https://doi.org/10.1016/j.ejrh.2015.11.007.
Deser, C., R. Knutti, S. Solomon, and A. S. Phillips, 2012: Communication of the role of natural variability in future North American climate. Nat. Climate Change, 2, 775–779, https://doi.org/10.1038/nclimate1562.
Deser, C., A. S. Phillips, M. A. Alexander, and B. V. Smoliak, 2014: Projecting North American climate over the next 50 years: Uncertainty due to internal variability. J. Climate, 27, 2271–2296, https://doi.org/10.1175/JCLI-D-13-00451.1.
Deser, C., and Coauthors, 2020: Insights from Earth system model initial-condition large ensembles and future prospects. Nat. Climate Change, 10, 277–286, https://doi.org/10.1038/s41558-020-0731-2.
Diaz, H. F., 1983: Drought in the United States. J. Appl. Meteor. Climatol., 22, 3–16, https://doi.org/10.1175/1520-0450(1983)022<0003:DITUS>2.0.CO;2.
Dirmeyer, P. A., X. Gao, M. Zhao, Z. Guo, T. Oki, and N. Hanasaki, 2005: The second Global Soil Wetness Project (GSWP-2): Multi-model analysis and implications for our perception of the land surface. COLA Tech. Rep. 185, 44 pp.
Donat, M. G., A. L. Lowry, L. V. Alexander, P. A. O’Gorman, and N. Maher, 2016: More extreme precipitation in the world’s dry and wet regions. Nat. Climate Change, 6, 508–513, https://doi.org/10.1038/nclimate2941.
Ferguson, P. R., and J. Veizer, 2007: Coupling of water and carbon fluxes via the terrestrial biosphere and its significance to the Earth’s climate system. J. Geophys. Res., 112, D24S06, https://doi.org/10.1029/2007JD008431.
Field, C. B., R. B. Jackson, and H. A. Mooney, 1995: Stomatal responses to increased CO2: Implications from the plant to the global scale. Plant Cell Environ., 18, 1214–1225, https://doi.org/10.1111/j.1365-3040.1995.tb00630.x.
Ford, T. W., and C. F. Labosier, 2017: Meteorological conditions associated with the onset of flash drought in the eastern United States. Agric. For. Meteor., 247, 414–423, https://doi.org/10.1016/j.agrformet.2017.08.031.
Frank, D. C., and Coauthors, 2015: Water-use efficiency and transpiration across European forests during the Anthropocene. Nat. Climate Change, 5, 579–583, https://doi.org/10.1038/nclimate2614.
Frumhoff, P. C., J. J. McCarthy, J. M. Melillo, S. C. Moser, and D. J. Wuebbles, 2007: Confronting climate change in the U.S. Northeast: Science, impacts, and solutions. UCS Publications, 146 pp., https://www.nrc.gov/docs/ML1434/ML14345A554.pdf.
Gleick, P. H., 1987: The development and testing of a water balance model for climate impact assessment: Modeling the Sacramento basin. Water Resour. Res., 23, 1049–1061, https://doi.org/10.1029/WR023i006p01049.
Gong, D., S. Kang, L. Yao, and L. Zhang, 2007: Estimation of evapotranspiration and its components from an apple orchard in northwest China using sap flow and water balance methods. Hydrol. Proc., 21, 931–938, https://doi.org/10.1002/hyp.6284.
Guilbert, J., A. K. Betts, D. M. Rizzo, B. Beckage, and A. Bomblies, 2015: Characterization of increased persistence and intensity of precipitation in the northeastern United States. Geophys. Res. Lett., 42, 1888–1893, https://doi.org/10.1002/2015GL063124.
Guttman, N. B., 1999: Accepting the standardized precipitation index: A calculation algorithm. J. Amer. Water Resour. Assoc., 35, 311–322, https://doi.org/10.1111/j.1752-1688.1999.tb03592.x.
Hawkins, E., and R. Sutton, 2009: The potential to narrow uncertainty in regional climate predictions. Bull. Amer. Meteor. Soc., 90, 1095–1108, https://doi.org/10.1175/2009BAMS2607.1.
Hayes, M. J., C. Alvord, and J. Lowrey, 2002: Drought indices. National Drought Mitigation Center, University of Nebraska, 18 pp., http://danida.vnu.edu.vn/cpis/files/Refs/Drought/Drought%20Indices.pdf.
Hayhoe, K., and Coauthors, 2007: Past and future changes in climate and hydrological indicators in the US Northeast. Climate Dyn., 28, 381–407, https://doi.org/10.1007/s00382-006-0187-8.
Hayhoe, K., and Coauthors, 2008: Regional climate change projections for the Northeast USA. Mitigation. Adapt. Strategies Global. Change, 13, 425–436, https://doi.org/10.1007/s11027-007-9133-2.
Hazeleger, W., and Coauthors, 2010: EC-Earth: A seamless Earth-system prediction approach in action. Bull. Amer. Meteor. Soc., 91, 1357–1364, https://doi.org/10.1175/2010BAMS2877.1.
Heitman, J. L., R. Horton, T. J. Sauer, and T. M. DeSutter, 2008: Sensible heat observations reveal soil–water evaporation dynamics. J. Hydrometeor., 9, 165–171, https://doi.org/10.1175/2007JHM963.1.
Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19, 5686–5699, https://doi.org/10.1175/JCLI3990.1.
Hoell, A., and Coauthors, 2020: Lessons learned from the 2017 flash drought across the U.S northern Great Plains and Canadian prairies. Bull. Amer. Meteor. Soc., 101, E2171–E2185, https://doi.org/10.1175/BAMS-D-19-0272.1.
Hoerling, M. P., J. K. Eischeid, X.-W. Quan, H. F. Diaz, R. S. Webb, R. M. Dole, and D. R. Easterling, 2012: Is a transition to semipermanent drought conditions imminent in the U.S. Great Plains? J. Climate, 25, 8380–8386, https://doi.org/10.1175/JCLI-D-12-00449.1.
Huntington, T. G., G. A. Hodgkins, B. D. Keim, and R. W. Dudley, 2004: Changes in the proportion of precipitation occurring as snow in New England (1949–2000). J. Climate, 17, 2626–2636, https://doi.org/10.1175/1520-0442(2004)017<2626:CITPOP>2.0.CO;2.
Janes, B. E., and J. J. Brumbach, 1965: The 1964 agriculture drought in Connecticut. Agricultural Experiment Station University of Connecticut, 22 pp.
Jeffrey, S., L. Rotstayn, M. Collier, S. Dravitzki, C. Hamalainen, C. Moeseneder, K. Wong, and J. Syktus, 2013: Australia’s CMIP5 submission using the CSIRO-Mk3.6 model. Aust. Meteor. Oceanogr. J., 63, 1–13, https://doi.org/10.22499/2.6301.001.
Joetzjer, E., H. Douville, C. Delire, P. Ciais, B. Decharme, and S. Tyteca, 2012: Evaluation of drought indices at interannual to climate change timescales: A case study over the Amazon and Mississippi River basins. Hydrol. Earth Syst. Sci., 9, 13231–13249, https://doi.org/10.5194/hessd-9-13231-2012.
Jung, M., and Coauthors, 2010: Recent decline in the global land evapotranspiration trend due to limited moisture supply. Nature, 467, 951–954, https://doi.org/10.1038/nature09396.
Kay, J. E., and Coauthors, 2015: The Community Earth System Model (CESM) large ensemble project: A community resource for studying climate change in the presence of internal climate variability. Bull. Amer. Meteor. Soc., 96, 1333–1349, https://doi.org/10.1175/BAMS-D-13-00255.1.
Keeling, R. F., and Coauthors, 2017: Atmospheric evidence for a global secular increase in carbon isotopic discrimination of land photosynthesis. Proc. Natl. Acad. Sci. USA, 114, 10 361–10 366, https://doi.org/10.1073/pnas.1619240114.
Keenan, T. F., D. Y. Hollinger, G. Bohrer, D. Dragoni, J. W. Munger, H. P. Schmid, and A. D. Richardson, 2013: Increase in forest water-use efficiency as atmospheric carbon dioxide concentrations rise. Nature, 499, 324–327, https://doi.org/10.1038/nature12291.
Kharin, V. V., F. W. Zwiers, X. Zhang, and G. C. Hegerl, 2007: Changes in temperature and precipitation extremes in the IPCC ensemble of global coupled model simulations. J. Climate, 20, 1419–1444, https://doi.org/10.1175/JCLI4066.1.
Kirchmeier-Young, M. C., F. W. Zwiers, and N. P. Gillett, 2017: Attribution of extreme events in Arctic sea ice extent. J. Climate, 30, 553–571, https://doi.org/10.1175/JCLI-D-16-0412.1.
Knauer, J., S. Zaehle, M. Reichstein, B. E. Medlyn, M. Forkel, S. Hagemann, and C. Werner, 2017: The response of ecosystem water-use efficiency to rising atmospheric CO2 concentrations: Sensitivity and large-scale biogeochemical implications. New Phytol., 213, 1654–1666, https://doi.org/10.1111/nph.14288.
Kogan, F. N., 1997: Global drought watch from space. Bull. Amer. Meteor. Soc., 78, 621–636, https://doi.org/10.1175/1520-0477(1997)078<0621:GDWFS>2.0.CO;2.
Krakauer, N. Y., T. Lakhankar, and D. Hudson, 2019: Trends in drought over the Northeast United States. Water, 11, 1834, https://doi.org/10.3390/w11091834.
Laîné, A., H. Nakamura, K. Nishii, and T. Miyasaka, 2014: A diagnostic study of future evaporation changes projected in CMIP5 climate models. Climate Dyn., 42, 2745–2761, https://doi.org/10.1007/s00382-014-2087-7.
Lawrence, D. M., P. E. Thornton, K. W. Oleson, and G. B. Bonan, 2007: The partitioning of evapotranspiration into transpiration, soil evaporation, and canopy evaporation in a GCM: Impacts on land–atmosphere interaction. J. Hydrometeor., 8, 862–880, https://doi.org/10.1175/JHM596.1.
Lawrence, P. J., and T. N. Chase, 2009: Climate impacts of making evapotranspiration in the community land model (CLM3) consistent with the simple biosphere model (SiB). J. Hydrometeor., 10, 374–394, https://doi.org/10.1175/2008JHM987.1.
Lee, D., J. Kim, K.-S. Lee, and S. Kim, 2010: Partitioning of catchment water budget and its implications for ecosystem carbon exchange. Biogeosciences, 7, 1903–1914, https://doi.org/10.5194/bg-7-1903-2010.
Lombard, P. J., J. R. Barclay, and D.-A. E. McCarthy, 2020: 2020 drought in New England. Version 1.1, U.S. Geological Survey Tech. Rep. 2020–1148, 12 pp.
Lyon, B., N. Christie-Blick, and Y. Gluzberg, 2005: Water shortages, development, and drought in Rockland County, New York. J. Amer. Water Resour. Assoc., 41, 1457–1469, https://doi.org/10.1111/j.1752-1688.2005.tb03812.x.
Ma, J., G. R. Foltz, B. J. Soden, G. Huang, J. He, and C. Dong, 2016: Will surface winds weaken in response to global warming? Environ. Res. Lett., 11, 124012, https://doi.org/10.1088/1748-9326/11/12/124012.
Maher, N., and Coauthors, 2019: The Max Planck Institute Grand Ensemble: Enabling the exploration of climate system variability. J. Adv. Model. Earth Syst., 11, 2050–2069, https://doi.org/10.1029/2019MS001639.
Mankin, J. S., R. Seager, J. E. Smerdon, B. I. Cook, and A. P. Williams, 2019: Mid-latitude freshwater availability reduced by projected vegetation responses to climate change. Nat. Geosci., 12, 983–988, https://doi.org/10.1038/s41561-019-0480-x.
McKee, T. B., 1995: Drought monitoring with multiple time scales. Proc. Ninth Conf. on Applied Climatology, Dallas, TX, Amer. Meteor. Soc., 233–236.
McKee, T. B., N. J. Doesken, and J. Kleist, 1993: The relationship of drought frequency and duration to time scales. Proc. Eighth Conf. on Applied Climatology, Anaheim, CA, 17–22.
Melillo, J. M., T. C. Richmond, and G. W. Yohe, 2014: Climate change impacts in the United States. Third National Climate Assessment, U.S. Global Change Research Program, 841 pp.
Milly, P. C. D., and K. A. Dunne, 2016: Potential evapotranspiration and continental drying. Nat. Climate Change, 6, 946–949, https://doi.org/10.1038/nclimate3046.
Milly, P. C. D., J. Betancourt, M. Falkenmark, R. M. Hirsch, Z. W. Kundzewicz, D. P. Lettenmaier, and R. J. Stouffer, 2008: Stationarity is dead: Whither water management? Science, 319, 573–574, https://doi.org/10.1126/science.1151915.
Mintz, Y., and Y. V. Serafini, 1992: A global monthly climatology of soil moisture and water balance. Climate Dyn., 8, 13–27, https://doi.org/10.1007/BF00209340.
Mishra, A. K., and V. P. Singh, 2010: A review of drought concepts. J. Hydrol., 391, 202–216, https://doi.org/10.1016/j.jhydrol.2010.07.012.
Mo, K. C., and D. P. Lettenmaier, 2016: Precipitation deficit flash droughts over the United States. J. Hydrometeor., 17, 1169–1184, https://doi.org/10.1175/JHM-D-15-0158.1.
Moser, S. C., R. E. Kasperson, G. Yohe, and J. Agyeman, 2008: Adaptation to climate change in the northeast United States: Opportunities, processes, constraints. Mitigation. Adapt. Strategies Global Change, 13, 643–659, https://doi.org/10.1007/s11027-007-9132-3.
National Academies of Sciences, Engineering, and Medicine, 2016: Attribution of Extreme Weather Events in the Context of Climate Change. National Academies Press, 186 pp.
NCAR, 2020: Multi-Model Large Ensemble Archive. National Center for Atmospheric Research, https://www.cesm.ucar.edu/projects/community-projects/MMLEA/.
NIDIS/NOAA, 2017: Coping with drought and its aftermath in the Northeast. National Integrated Drought Information System, 5 pp., https://www.drought.gov/documents/coping-drought-and-its-aftermath-northeast.
NIDIS/NOAA, 2021: 2018–2019 Northeast DEWS Strategic Plan. National Integrated Drought Information System, 29 pp., https://www.drought.gov/documents/2018-2019-northeast-dews-strategic-plan.
Osman, M., B. F. Zaitchik, H. S. Badr, J. I. Christian, T. Tadesse, J. A. Otkin, and M. C. Anderson, 2021: Flash drought onset over the contiguous United States: Sensitivity of inventories and trends to quantitative definitions. Hydrol. Earth Syst. Sci., 25, 565–581, https://doi.org/10.5194/hess-25-565-2021.