Abstract

Flash drought refers to relatively short periods of warm surface temperature and anomalously low and rapid decreasing soil moisture (SM). Based on the physical mechanisms associated with flash droughts, these events are classified into two categories: heat wave and precipitation P deficit flash droughts. In previous work, the authors have defined heat wave flash droughts as resulting from the confluence of severe warm air temperature Tair, which increases evapotranspiration (ET), and anomalously low and decreasing SM. Here, a second type of flash drought caused by precipitation deficits is explored. The authors term these events P-deficit flash droughts, which they associate with lack of P. Precipitation deficits cause ET to decrease and temperature to increase. The P-deficit flash droughts are analyzed based on observations of P, Tair, and SM and ET reconstructed using land surface models for the period 1916–2013. The authors find that P-deficit flash droughts are more common than heat wave flash droughts. They are about twice as likely to occur as heat wave flash droughts over the conterminous United States. They are most prevalent over the southern United States with maxima over the southern Great Plains and the Southwest, in contrast to heat wave flash droughts that are mostly likely to occur over the Midwest and the Pacific Northwest, where the vegetation cover is dense.

1. Introduction

During spring 2012, high air temperatures Tair and severe depletion of soil moisture (SM) occurred suddenly over the agricultural heartland of the U.S. Midwest and withered recently planted crops in a matter of days. The relatively short period of intense warm air temperature and anomalously low and declining SM was termed a flash drought, following Senay et al. (2008) and Hunt et al. (2009): a term that had not previously been widely used and for which there was no accepted definition. This event was also detected by remote sensing using the evaporative stress index (ESI; Otkin et al. 2013, 2014; Anderson et al. 2011, 2013). The 2012 spring event was evidenced by rapid increases in evapotranspiration (ET). The early detection and severity of the 2012 event led to an increasing awareness of flash droughts.

In previous work (Mo and Lettenmaier 2015), we labeled flash droughts with characteristics similar to the 2012 event as heat wave flash droughts because such events are initialized by heat waves, which in turn lead to increased ET and reduced SM. Accordingly, we suggested a definition for heat wave flash droughts that includes anomalously high temperatures (Tair anomaly less than one standard deviation), increases of ET (ET anomaly >0), and soil moisture deficits (SM% < 40%). We also hypothesized (but did not explore) another kind of flash drought that is initialed by the lack of precipitation P.

Yang (2013), in a study of the 2011 Texas drought, referred to the sudden occurrence of heat waves and rapid reduction in SM in June 2011 as a flash drought. We argue that the 2011 Texas drought had characteristics that are fundamentally different from what we refer to as a heat wave flash drought. Myoung and Nielsen-Gammon (2010a,b) investigated the physical mechanisms responsible for the 2011 Texas drought. They found that the P deficits existed prior to the onset of the drought. We refer to this type of drought as P-deficit flash drought because the lack of SM was caused primarily by P deficits that were responsible for (rather than caused by) the onset of heat waves. There are other cases of P-deficit flash droughts. For example, Lyon and Dole (1995) studied the 1980 and 1988 heat waves over the central United States. They described these events as drought induced. Atmospheric circulation anomalies initialized the establishment of heat waves, but the lack of P and SM prolonged the events.

Both heat wave and P-deficit flash droughts are characterized by high temperature anomalies and SM deficits. However, they are caused and maintained by different physical mechanisms and have different characteristics. Most previous studies of heat waves and droughts have been regional in nature (Chang and Wallace 1987) and therefore are difficult to generalize. We attempt here to give a general definition and to examine characterization that should improve our ability to monitor and provide early warning of intense flash drought conditions and potentially to mitigate them. We first revisit the definition of heat wave flash drought and suggest a definition for P-deficit flash drought. We then examine the characteristics of the two types of flash droughts and their evolution by using reconstructed meteorological and soil moisture records over the conterminous United States (CONUS) for the period 1916–2013 using the model reconstruction approach described by Wood and Lettenmaier (2006).

2. Datasets and procedures

a. Datasets

We used daily gridded P and Tair data from a set of approximately 2400 index stations over the CONUS selected based on data quality and stability of the stations (Wood and Lettenmaier 2006). This is the same dataset used by Mo and Lettenmaier (2015) to study heat wave flash droughts. We then derived daily forcings for four land surface models (LSMs) from daily P, maximum temperature Tmax, minimum temperature Tmin, and surface wind speed. Surface wind speed was taken from the lowest level of the NCEP–NCAR reanalyses (Kalnay et al. 1996) as in Livneh et al. (2013) and Maurer et al. (2002). Prior to 1950, the surface wind speed was represented by the mean seasonal cycle. Livneh et al. (2013) have shown that this assumption has only a modest effect on the LSM output. The daily forcings drove the same four LSMs as in Mo and Lettenmaier (2015), which produced SM and ET. The models used were VIC, version 4.0.6 (Liang et al. 1994); Catchment (Koster et al. 2000; Ducharne et al. 2000); Noah, version 2.8 (Koren et al. 1999; Ek et al. 2003); and Sacramento Soil Moisture Accounting Model coupled with SNOW-17 model (SAC; Barnash et al. 1973; Anderson 1973). The datasets used here also form the foundation for the University of California, Los Angeles–University of Washington Experimental Surface Water Monitor for the Continental U.S. (http://www.hydro.washington.edu/forecast/monitor/; Wang et al. 2009). To eliminate initial condition effects, each of the four models was run for the period from 1915 to 2003, and the final soil moisture values at the end of 2003 were then used as the initial conditions at the beginning of 1915. We then iterated 10 times and took the final values at the end of 2015 as the initial conditions for the beginning of 1916.

To study the circulation anomalies associated with flash droughts, we used the daily 500-hPa heights from the Twentieth Century Reanalysis, version 2 (Compo et al. 2006, 2011). The horizontal resolution is 2° and the reanalysis data cover the period from 1871 to 2012. We used the 500-hPa heights for the Northern Hemisphere from 1916 to 2012.

Because flash drought events typically last only a few days, we used pentad data as in Mo and Lettenmaier (2015). We constructed the pentad data from the daily gridded data (in the case of leap years, the twelfth pentad has 6 days). The pentad mean climatologies were computed for the base period for each model and each variable. Anomalies are the departures from the climatology. The base period is 1916–2013 for the land surface variables and 1916–2012 for the Twentieth Century Reanalysis.

b. Procedures

Variations in evaporative demand are largest in the growing season; hence, we focused on pentads from April through September (36 per year). There are a total of 98 years in our record, and the record length Ntotal therefore is 3528 pentads. To determine the preferred regions for flash drought occurrence, we computed the frequency of occurrence (FOC) by using a threshold method. We processed each model separately. For a given pentad T and grid point x, we identified a flash drought event when a given definition of flash drought was met. That pentad was defined as the onset. For each grid point, we computed the total number of pentads N under flash drought of either type for the entire record for a given model. We defined the FOC as the percentage of pentads under heat wave or P-deficit flash droughts:

 
formula

We computed the FOC for each model separately and then took the ensemble mean of the FOC values over four models. To study the evolution of flash droughts, we made composites of P, Tair, ET, and SM anomalies from four pentads before to four pentads after the onset of both types of flash drought events. We assessed statistical significance using the Student’s t test. Areas where values are statistically significant at the 5% level are shaded. The field statistical significance for composites was tested using the method of Livezey and Chen (1983).

3. -deficit flash drought

a. Example

The 1980 summer drought over the central United States is a good example of a P-deficit flash drought event. The time–longitude plots of pentad mean Tair, ET, P, and SM anomalies averaged over the center of the heat waves from 34° to 40°N illustrate the life cycle of the event (Fig. 1). Figure 2 shows the corresponding 15-day mean 500-hPa height anomalies for June and July 1980. The SM and ET anomalies are ensemble means of four models.

Fig. 1.

Time–longitude diagram for (a) Tair anomalies (°C) averaged from 34° to 40°N from January to December 1980. Contours are given by the color bar. (b) As in (a), but for ET anomalies (mm day−1) averaged over four models. (c) As in (a), but for P anomalies (mm day−1). (d) As in (b), but for SM anomalies (mm).

Fig. 1.

Time–longitude diagram for (a) Tair anomalies (°C) averaged from 34° to 40°N from January to December 1980. Contours are given by the color bar. (b) As in (a), but for ET anomalies (mm day−1) averaged over four models. (c) As in (a), but for P anomalies (mm day−1). (d) As in (b), but for SM anomalies (mm).

Fig. 2.

The 500-hPa height anomalies (m) on (a) 1–15 Jun, (b) 16–30 Jun, (c) 1–15 Jul, and (d) 16–30 Jul 1980. Contours are given by the color bar.

Fig. 2.

The 500-hPa height anomalies (m) on (a) 1–15 Jun, (b) 16–30 Jun, (c) 1–15 Jul, and (d) 16–30 Jul 1980. Contours are given by the color bar.

Precipitation deficits (negative P anomalies) occurred in spring and early summer more than one month before the establishment of heat waves in July (Fig. 1). SM anomalies were already negative at the beginning of 1980 but recovered somewhat in early spring. In May, SM anomalies responded to the lack of P and decreased again. In June, an anticyclone moved into the central United States and intensified (Fig. 2). Precipitation deficits continued and reached a minimum in early July when the positive height anomalies reached a maximum (Fig. 2c). SM deficits (negative anomalies) increased as responses to the lack of P and persisted through the end of summer (Fig. 1d). The interesting point is that although SM and P deficits occurred in May and early June, ET did not respond to the depleted SM until early July (Fig. 1b), when SM anomalies were less than −40 mm (about one standard deviation). The Tair anomalies were already warm in June. Heat waves occurred only after ET started to decrease around 1 July. The Tair anomalies were above 5°C in mid-July when ET anomalies reached a minimum. The anticyclone moved out of the region after 16 July and temperatures declined.

The evolution of the 1980 event was very different from a heat wave flash drought even though both show rapid increases of temperature and decreases of SM. For heat wave flash droughts, high temperature is the major driver. ET anomalies are positive as they respond to high temperature (Mo and Lettenmaier 2015). In the 1980 case, P was a major driver. ET anomalies decreased in response to the decreases of SM caused by P deficits. The Bowen ratio increased and temperature increased as a response to (rather than cause of) the decreases in ET.

b. Definition and frequency of occurrence

During the dry season over the Southwest, there is little rain but temperatures can increase above one standard deviation before the onset of the monsoon season. Because this is a recurring climatological condition, it cannot be considered a flash drought. For the Southwest (25°–35°N, 110°–123°W), we only consider grid points where the pentad P climatology was greater than 0.2 mm day−1 to distinguish P-deficit flash droughts from monsoon onset conditions. Over the CONUS, this has the effect of screening out portions of Southern California in spring and the desert over the Southwest in spring and early summer. We imposed as our requirement for P deficiency that the P anomaly be <40%, that is, less than the 40th percentile for that pentad. We determined percentiles from data in the base period. In addition to the requirement of P anomalies, we required that the ET anomaly be negative in order to distinguish from heat wave flash droughts. We required the temperature anomalies to be above one standard deviation to assure that temperature is high. We tested four different scenarios for the possible definition of P-deficit flash droughts:

  • case 1: ET anomaly <0, P% < 40%;

  • case 2: ET anomaly <0, P% < 40%, and Tair anomaly greater than one standard deviation;

  • case 3: ET anomaly <0, SM% < 40%, and Tair anomaly greater than one standard deviation;

  • case 4: ET anomaly <0, P% < 20%, and Tair anomaly greater than one standard deviation.

We computed both pentad standard deviations and percentiles for the base period from 1916 to 2013. For each case and each model, we selected pentads that met the criteria listed above and computed FOC. We then composited standardized Tair, P anomalies, and SM percentiles for all pentads under drought (Fig. 3). Modest changes in the criteria change the number of events but not the general space–time patterns.

Fig. 3.

(a) The FOC for case 1 (P anomaly <40%, ET anomaly <0) and (b) the composite of standardized Tair anomalies (°C) for pentads under flash drought for case 1. Values are given by the color bar. (c) As in (b), but for P anomalies (mm day−1) under flash drought for case 1. (d) As in (b), but for SM percentiles (%) under flash drought for case 1. (e)–(h) As in (a)–(d), but for case 2 (P anomaly <40%, ET anomaly <0, and Tair anomaly greater than one std dev). (i)–(l) As in (a)–(d), but for case 3 (SM percentile <40%, ET anomaly <0, and Tair anomaly greater than one std dev).

Fig. 3.

(a) The FOC for case 1 (P anomaly <40%, ET anomaly <0) and (b) the composite of standardized Tair anomalies (°C) for pentads under flash drought for case 1. Values are given by the color bar. (c) As in (b), but for P anomalies (mm day−1) under flash drought for case 1. (d) As in (b), but for SM percentiles (%) under flash drought for case 1. (e)–(h) As in (a)–(d), but for case 2 (P anomaly <40%, ET anomaly <0, and Tair anomaly greater than one std dev). (i)–(l) As in (a)–(d), but for case 3 (SM percentile <40%, ET anomaly <0, and Tair anomaly greater than one std dev).

Case 1 tests whether the lack of P and negative ET anomalies alone are sufficient to increase Tair above one standard deviation so these events can be qualified as P-deficit flash droughts. Case 2 is a subset of events in case 1 for which Tair anomalies are specified to be above one standard deviation. The requirements for case 3 are similar to the definition of heat wave flash droughts except ET anomalies are negative. Comparison of case 2 and case 4 tests the sensitivity of P anomaly requirements.

The case 1 composites indicate that, on average, the lack of P and negative ET anomalies alone usually decrease SM to below 40% (Fig. 3d). However, the SM deficits are not strong enough to increase Tair anomalies above one standard deviation (Fig. 3b). Therefore, P deficits and negative ET anomalies are necessary but not sufficient conditions for P-deficit flash droughts.

Case 2 is a subset of case 1. The composite shows that Tair is above 1.4 standard deviations. There is no explicit requirement of SM%, but the composite indicates that SM% is below 30%, except in the western dry region where SM is below 40%. The SM minimum is located in the Great Plains. Because of the SM connection, flash droughts as defined by the case 2 criteria are agricultural droughts. The comparison between case 1 and case 2 indicates that P deficits alone do not always cause flash drought to occur. We found that SM needs to be below about 30% to create favorable conditions for flash drought to occur.

Case 3 is similar to the definition of heat wave flash droughts except ET anomalies are negative. The FOC pattern is similar to case 2 but there are more events over the Great Plains and southern states. The composites of P, Tair, and SM anomalies are also similar to case 2, but magnitudes are weaker. There is no requirement for P, but the composite indicates that P is below normal but anomalies are slightly weaker than case 2. When P deficits cause SM% to drop below 40%, case 3 events will occur even if P deficits are greater than 40%. Case 3 has more relaxed requirements than case 2. Because these are P deficit droughts, we decided to use the P anomaly as an indicator, and in particular, we adopted case 2 as our definition of P-deficit flash drought. Case 4 tests the sensitivity of the FOC to P deficits. When the P percentile requirements are below 20%, the FOC pattern is similar to case 2 but values are lower. The maximum of the FOC is still located over Texas, but the magnitudes are 1% less (Fig. 4c).

Fig. 4.

Ensemble mean frequency of occurrence of pentads under (a) P-deficit flash drought, (b) heat wave flash drought, and (c) P-deficit flash drought for case 4 (P anomaly <20%, ET anomaly <0, and Tair anomaly greater than one std dev) averaged over four models. Contours are given by the color bar. (d) Ratio between the FOC for the P deficit and the FOC for the heat wave flash droughts.

Fig. 4.

Ensemble mean frequency of occurrence of pentads under (a) P-deficit flash drought, (b) heat wave flash drought, and (c) P-deficit flash drought for case 4 (P anomaly <20%, ET anomaly <0, and Tair anomaly greater than one std dev) averaged over four models. Contours are given by the color bar. (d) Ratio between the FOC for the P deficit and the FOC for the heat wave flash droughts.

We replotted the FOC for P-deficit flash drought in Fig. 4a and reproduced the FOC for heat wave flash drought in Fig. 4b from Mo and Lettenmaier (2015) for comparison. It is apparent that there are more P-deficit flash drought events than heat wave flash drought events. Furthermore, P-deficit flash drought events can occur everywhere but are more likely to occur over the South, with maxima extending from the southern Great Plains to the Southwest, where heat wave droughts are infrequent. There are fewer events over the north-central region and the Ohio Valley, where heat wave flash droughts are most likely to occur. This contrasts with a maximum in the FOC of heat wave flash droughts in the north-central region of the United States, with a secondary maximum in the Pacific Northwest. Figure 4d shows the ratio between the FOC for P-deficit flash droughts and FOC for the heat wave flash droughts. The ratio is near 1 over the north-central region and the Ohio Valley, where both types of flash droughts can occur. The ratio is greater than 4 where the P-deficit flash drought is more prominent over the southern plains and the Southwest.

One feature that distinguishes flash droughts from longer meteorological and agricultural droughts is that flash droughts generally do not persist because Tair anomalies tend not to be persistent. For heat wave flash droughts, most events only last for one to two pentads (Mo and Lettenmaier 2015). We evaluated persistence for P-deficit flash drought as the number of events that persisted for one pentad to three or more pentads after the onset. We did so for each event and for each LSM separately. Then, we averaged over all events and averaged over all four models (Fig. 5). Most P-deficit flash drought events over the CONUS only persist for one pentad. Events over Texas tend to be more persistent; 20%–30% of events over Texas persist for three pentads or longer. Even though flash droughts tend not to persist, they initiate the depletion of SM, and the persisting soil moisture deficits can causes large damages to the agricultural community.

Fig. 5.

Percentages of P-deficit flash droughts that persist for (a) one pentad, (b) two pentads, and (c) three pentads after onset. Contours are given by the color bar.

Fig. 5.

Percentages of P-deficit flash droughts that persist for (a) one pentad, (b) two pentads, and (c) three pentads after onset. Contours are given by the color bar.

4. Physical mechanisms for flash droughts

Heat wave flash droughts are temperature driven. Figure 6c shows the vegetation coverage averaged from April to September. From the FOC (Fig. 4b), it is evident that heat wave flash droughts tend to be located over areas where vegetation coverage is dense. Over the interior of the West where vegetation cover is sparse, there are few events. As the term implies, high temperatures cause ET to increase and SM to decrease because of vegetation moisture extraction (Mo and Lettenmaier 2015).

Fig. 6.

(a) Correlation between pentad Tair and SM anomalies for pentads from April to September from 1916 to 2013 averaged over four models. Contours are given by the color bar. (b) As in (a), but for correlation between pentad ET and SM anomalies. (c) Vegetation cover (%) averaged from April to September.

Fig. 6.

(a) Correlation between pentad Tair and SM anomalies for pentads from April to September from 1916 to 2013 averaged over four models. Contours are given by the color bar. (b) As in (a), but for correlation between pentad ET and SM anomalies. (c) Vegetation cover (%) averaged from April to September.

In contrast, P-deficit flash droughts are P driven. From the 1980 case, we noticed that Tair was already warm at the onset of the events but had not reached one standard deviation before the event’s onset. From this example, we posit that for a P-deficit flash drought to occur, the sequence is that 1) the lack of P drives down SM and negative SM anomalies cause Tair to increase and 2) Tair anomalies are positive by the onset of the event.

a. SM and relationship

Huang and van den Dool (1993) examined the relationship between P and Tair and found that the strongest relationships occur in the Great Plains. They also quantified the role played by SM. One measure of the relationship between Tair and SM anomalies is the correlation between pentad SM and Tair anomalies over the growing season from April to September. We computed correlations for each LSM separately. The ensemble mean over the four LSMs is given in Fig. 6a. Assuming the degrees of freedom to be the years of record minus 2, correlations needed to be greater than 0.148 (or less than −0.148) in order to be statistically significant at the 5% level. Figure 6a suggests that (positive) negative SM anomalies are related to cool (warm) Tair over the Great Plains and the southern United States with a minimum over the southern Great Plains. Koster et al. (2009) indicated that these are areas where temperature is sensitive to wetness and meteorological drought can lead to warmer Tair. The processes go through ET because in these regions, ET responds to SM almost linearly (Koster et al. 2009). Figure 6b shows the correlation between pentad ET and SM anomalies for April–September from 1916 to 2013. They are positively correlated over the areas where P-deficit flash droughts are most likely to occur (Fig. 4a). Hence, ET decreases in response to declining SM, and the reduced ET results in warming Tair. The P-deficit flash droughts occur less often over the north-central region and the West. The north-central region has relatively dense vegetation coverage (Fig. 6c); hence, ET tends to respond positively to high temperatures. Therefore, conditions in these regions are also favorable to heat wave flash droughts. The West, in general, is a dry region. Accordingly, Koster et al. (2009) argued that SM variability there is too weak to cause strong ET responses.

b. before onset

In the 1980 case, Tair was already warm before the onset of the flash drought. In this section, we explore why this is a necessary condition for P-deficit flash droughts to occur. We have shown that P deficits will lead to increases of Tair. The lack of P is likely to be associated with clear sky and an increase of net radiation to reach the surface. The net radiation will increase Tair. The question is whether the increases are large enough to push Tair above one standard deviation, which is one of our requirements for occurrence of P-deficit flash droughts. The composite of Tair anomalies when P percentiles are less than 40% and ET is negative (case 1) averaged over four models indicates that the changes of Tair anomalies are less than one standard deviation (Fig. 3b). These increments are not large enough to be qualified as P-deficit flash droughts unless Tair anomalies are already positive before the onset of the events. This is consistent with the 1980 case. This is also similar to the heat wave drought case. High temperatures will cause SM to decrease, but increments are not large enough to be qualified as heat wave flash drought unless SM is already negative at the onset. Over the Southwest monsoon region, there is intense heat just before monsoon rainfall begins. Once the monsoon starts, temperatures cool. This is a part of seasonal cycle so we do not consider it to constitute flash drought.

c. Differences between -deficit flash droughts and conventional droughts

Flash droughts are the confluence of heat waves and dryness. Both heat wave and P-deficit flash droughts require Tair to be higher than one standard deviation in addition to dryness. One might question how P-deficit flash droughts differ from conventional droughts, given that both are related to precipitation deficits. The main difference, aside from the fact that conventional droughts have much longer length scales, is the role of temperature anomalies. Conventional droughts are not necessarily associated with warm temperature anomalies, whereas heat waves are typically a result of P-deficit flash droughts.

5. Trends in the occurrence of -deficit flash droughts

Figures 7e and 7f show the number of pentads under P-deficit flash drought per warm season (April–September) averaged over a box over the southern Great Plains (28°–36°N, 95°–105°W) and a box over the Southwest (32°–36°N, 110°–116°W) for each LSM along with the respective ensemble means. Modest changes in the location of the boxes do not change the conclusions. Unlike the heat wave flash drought, there are only modest differences among LSMs because the forcing terms P and Tair are the same for all; only ET is different from model to model. There are no statistically significant trends in the occurrence of P-deficit flash drought events in the southern Great Plains. The 2011 event, however, was exceptionally strong (Fig. 7e). Over the Southwest, there were slight upward trends in the number of P-deficit events from 1920 to 1960. However, after 1990, the occurrence of the flash drought events increased dramatically. The trends correspond roughly to increasing trends in Tair anomalies averaged over the Southwest (Fig. 7g); Tair warming accelerated after 1990. The P trends (Fig. 7h) are also mostly decreasing from 1920 to 1960 in this region, but there is no dramatic decrease post-1990.

Fig. 7.

(a) Trends in the number of pentads per year under P-deficit flash drought. Trends that are statistically significant at the 5% significance level as determined by the Mann–Kendall test for all four models (VIC, SAC, Noah, and Catchment) are shaded. Green shading indicates upward trends and orange shading indicates downward trends. (b) As in (a), but for mean Tair anomalies averaged from April to September. (c) As in (a), but for P anomalies. (d) As in (a), but for heat wave flash drought events. (e) Number of pentads under P-deficit flash drought per year averaged over a rectangle in the southern plains (28°–36°N, 95°–105°W) for four models and the ensemble mean. (f) As in (e), but for a rectangle over the Southwest (32°–36°N, 105°–116°W). (g) As in (f), but for Tair anomalies averaged from April to September. (h) As in (g), but for P anomalies.

Fig. 7.

(a) Trends in the number of pentads per year under P-deficit flash drought. Trends that are statistically significant at the 5% significance level as determined by the Mann–Kendall test for all four models (VIC, SAC, Noah, and Catchment) are shaded. Green shading indicates upward trends and orange shading indicates downward trends. (b) As in (a), but for mean Tair anomalies averaged from April to September. (c) As in (a), but for P anomalies. (d) As in (a), but for heat wave flash drought events. (e) Number of pentads under P-deficit flash drought per year averaged over a rectangle in the southern plains (28°–36°N, 95°–105°W) for four models and the ensemble mean. (f) As in (e), but for a rectangle over the Southwest (32°–36°N, 105°–116°W). (g) As in (f), but for Tair anomalies averaged from April to September. (h) As in (g), but for P anomalies.

To determine trends in the occurrence of flash droughts, we applied the Mann–Kendall test to the time series of the total number of pentads under drought each year on a gridcell-by-gridcell basis. The test was performed for each model separately. Only trends that were present in all four models were plotted. We reproduce the Mann–Kendall test for heat wave flash drought events in Fig. 7d from Mo and Lettenmaier (2015) for comparison. Figure 7a, which passes the field significance test of Livezey and Chen (1983), indicates that there were upward trends over the Southwest and Four Corners regions for P-deficit flash droughts while heat wave flash droughts had downward trends over the north-central region (Fig. 7d). For meteorological drought, there were downward trends in the occurrence over the CONUS due to the upward trends in P and SM (Andreadis and Lettenmaier 2006; Andreadis et al. 2005).

For P, there were upward trends in the north-central region where heat wave flash droughts had downward trends. But P-deficit flash droughts are relatively infrequent in that region, so the influence of P trends is in the larger context. Overall, trends in the number of P-deficit flash droughts tend to be associated mostly with trends in Tair. The lack of P will increase Tair, but the increments as indicated by Fig. 3b are not large. If a particular decade is anomalously warm and Tair is already positive before the onset of P-deficit flash drought events, then they are more likely to occur. A P-deficit flash drought is P driven, but the trends in occurrence are associated with trends in Tair, while a heat wave flash drought is temperature driven, but trends in occurrence are associated with trends in P.

6. Evolution of flash droughts

We explore the evolution of both types of flash droughts through examination of P, Tair, SM, and ET anomalies from four pentads before to four pentads after the onset. To show the atmospheric circulation anomalies associated with P-deficit flash drought, we formed composites of the 500-hPa height pentad anomalies from the Twentieth Century Reanalysis, version 2 (Compo et al. 2006, 2011), using an index method. The index we used in Mo and Lettenmaier (2015) for heat wave flash droughts was based on the average number of events in the rectangle 38°–45°N, 90°–95°W. For P-deficit flash drought, we took the box as 28°–36°N, 95°–105°W. The selection of the boxes in both cases was intended to locate them over the area of the maximum FOCs (Fig. 4). We processed each LSM separately. For each pentad, we calculated the number of flash drought events averaged over the rectangle. We then averaged over the four LSMs. We computed the composite of 500-hPa height anomalies for pentads for which the index was greater than 0.5.

a. Heat wave flash droughts

The life cycle of “typical” heat wave droughts is given by composites in Fig. 8. Heat wave flash drought is temperature driven, but P deficits are needed to set up desirable conditions for flash drought to occur. Precipitation deficits start to appear over the north-central region two pentads before the onset, and that causes SM to decrease to below 40% one pentad before onset (Fig. 8l). Temperature anomalies are about 1°–2°C one pentad before onset. Temperature increases rapidly over the Midwest and the north-central during onset. When Tair anomalies are above 3°C, ET starts to increase. That drives SM below 30%. SM anomalies persist more than two pentads after onset. The interesting point is that P deficits only appear before onset. As indicated by Mo and Lettenmaier (2015), the important role of P is to set up favorable conditions for heat wave flash drought to occur. It does not play an active role in raising temperature or increasing ET. Heat waves in general are not very persistent. At one pentad after onset, temperature already starts to cool down on average. Heat wave flash drought has atmospheric circulation support. The composite of 500-hPa heights based on the number of heat wave flash drought events in the box (38°–45°N, 90°–95°W) indicates that there are anticyclones located in the north-central region near the maxima of the FOC of heat wave flash droughts (Fig. 9a).

Fig. 8.

Composite of Tair anomalies (°C) for heat wave flash droughts for (a) two pentads before onset, (b) one pentad before onset, (c) onset, (d) one pentad after onset, and (e) two pentads after onset. Contours are given by the color bar. (f)–(j) As in (a)–(e), but for ET anomalies (mm day−1) averaged over four models. (k)–(o) As in (f)–(j), but for SM percentiles (%). (p)–(t) As in (a)–(e), but for P anomalies (mm day−1).

Fig. 8.

Composite of Tair anomalies (°C) for heat wave flash droughts for (a) two pentads before onset, (b) one pentad before onset, (c) onset, (d) one pentad after onset, and (e) two pentads after onset. Contours are given by the color bar. (f)–(j) As in (a)–(e), but for ET anomalies (mm day−1) averaged over four models. (k)–(o) As in (f)–(j), but for SM percentiles (%). (p)–(t) As in (a)–(e), but for P anomalies (mm day−1).

Fig. 9.

(a) Composite of 500-hPa height anomalies (m) based on the number of heat wave flash drought events in rectangle (38°–45°N, 90°–95°W). (b) As in (a), but for P-deficit flash drought events in a rectangle (28°–36°N, 95°–105°W) during onset. Contours are given by the color bar.

Fig. 9.

(a) Composite of 500-hPa height anomalies (m) based on the number of heat wave flash drought events in rectangle (38°–45°N, 90°–95°W). (b) As in (a), but for P-deficit flash drought events in a rectangle (28°–36°N, 95°–105°W) during onset. Contours are given by the color bar.

b. -deficit flash droughts

The evolution of typical P-deficit flash droughts is shown in Fig. 10. Precipitation deficits start to appear about two pentads before onset and gain strength toward onset. At pentad −2, there are already −2 mm day−1 negative P anomalies over the Great Plains, which is about half a standard deviation. SM deficits develop faster over the Great Plains than the other regions because P deficits over the Great Plains increase faster (Figs. 10a–c). ET decreases to respond to the SM deficits. The Tair anomalies are already above 1°–2°C before onset. During onset, the composite of 500-hPa height anomalies shows an anticyclone strengthening over the southern Great Plains, and both P and SM anomalies reach a minimum (Fig. 9b). SM anomalies persist through +2 pentads and beyond. The Tair anomalies reach 2°–3°C during the event onset, which is over one standard deviation of Tair. There is a strong correspondence between SM and ET and also good correspondence between the decreases of ET and the increases of Tair.

Fig. 10.

Composite of P anomalies (mm day−1) for P-deficit flash droughts for (a) two pentads before onset, (b) one pentad before onset, (c) onset, (d) one pentad after onset, and (e) two pentads after onset. Contours are given by the color bar. (f)–(j) As in (a)–(e), but for SM percentiles (%). (k)–(o) As in (a)–(e), but for ET anomalies (mm day−1) averaged over four models. (p)–(t) As in (a)–(e), but for Tair anomalies (°C).

Fig. 10.

Composite of P anomalies (mm day−1) for P-deficit flash droughts for (a) two pentads before onset, (b) one pentad before onset, (c) onset, (d) one pentad after onset, and (e) two pentads after onset. Contours are given by the color bar. (f)–(j) As in (a)–(e), but for SM percentiles (%). (k)–(o) As in (a)–(e), but for ET anomalies (mm day−1) averaged over four models. (p)–(t) As in (a)–(e), but for Tair anomalies (°C).

7. Discussion and conclusions

We define flash droughts as short periods of warm temperature and anomalously low and rapidly declining SM. Based on physical mechanisms associated with flash droughts, we have classified them into two types: heat wave flash droughts and P-deficit flash droughts, the characteristics of which are compared in Table 1. Both are manifested by SM deficits that cause damage to crops. In that sense, both are agricultural droughts. We have shown that, following our definitions, P-deficit flash droughts are more common than heat wave flash droughts. They can occur in most areas of the United States, but they are most prevalent in the Great Plains and the southern United States while heat wave flash droughts are most likely to occur in the north-central region and the Pacific Northwest. The P-deficit flash droughts are precipitation-driven events. The lack of rain prior to the onset of an event reduces SM and decreases ET, which in turn leads to high temperatures. It is very different from heat wave flash droughts, which are temperature driven and tend to occur in more densely vegetated areas.

Table 1.

Features of heat wave and P-deficit flash droughts.

Features of heat wave and P-deficit flash droughts.
Features of heat wave and P-deficit flash droughts.

Are flash droughts forced by sea surface temperature anomalies (SSTAs)? We performed an exploratory investigation of the relationships between SSTAs and the occurrence of flash droughts. For heat wave flash droughts, we plotted the number of events per year over a box (36°–42°N, 80°–100°W) where the FOC has a maximum. We then selected 10 years that had the most P-deficit flash drought events and composited the SSTAs over the season (April–September) for these years. Most anomalies in the composite were not statistically significant, and the composite did not pass the field significant test of Livezey and Chen (1983).

The same procedures can be applied to the relationships between SSTAs and the number of P-deficit flash drought events per season. We selected a box located over the southern Great Plains (28°–36°N, 93°–107°W) where the FOC has a maximum. We composited nine events; the composite SSTAs in the North Atlantic are warm and there is a weak cold ENSO signal in the central Pacific. Results are suggestive rather than conclusive. More detailed study is needed to determine whether flash droughts are forced by local or remote conditions, and if so, whether there are precursors that could be used to provide advance warning of flash droughts.

Flash droughts have a large impact on crops, and early warming may help to mitigate associated agricultural damages. Because heat waves do not persist, most flash droughts only last one or two pentads. However, the depletion of SM can last for a long time and the persisting SM depletion causes severe damages to crops. P-deficit flash droughts occur during meteorological drought, so conventional drought indices such as the standardized precipitation index and runoff and SM percentiles have some ability to detect them (Svoboda et al. 2002). Heat wave flash droughts occur when SM is already in deficit. Given the short duration of flash droughts relative to conventional droughts, shorter time steps (e.g., pentad or weekly vs monthly) need to be used to monitor them. More importantly, the shorter time scales relative to conventional drought (which are close to weather forecast time scales) offer some hope of forecasting both the onset and termination of these events.

Acknowledgments

This project was supported by NOAA Grant GC14-168A to the NOAA/Climate Prediction Center and by NOAA Grant NA14OAR4310293 to the University of California, Los Angeles.

REFERENCES

REFERENCES
Anderson
,
E. A.
,
1973
: National Weather Service River Forecast System—Snow accumulation and ablation model. NOAA Tech. Memo. NWS HYDRO-17, 87 pp. [Available online at ftp://ftp.wcc.nrcs.usda.gov/wntsc/H&H/snow/AndersonHYDRO17.pdf.]
Anderson
,
M. C.
,
C.
Hain
,
B.
Wardlow
,
A.
Prinstein
,
J. R.
Mecikalski
, and
W. P.
Kustas
,
2011
:
Evaluation of drought indices based on thermal remote sensing and evapotranspiration over the continental United States
.
J. Climate
,
24
,
2025
2044
, doi:.
Anderson
,
M. C.
,
C.
Hain
,
J.
Otkin
,
X.
Zhan
,
K. C.
Mo
,
M.
Svoboda
,
B.
Waedlow
, and
A.
Pimstein
,
2013
:
Intercomparison of drought indicators based on thermal remote sensing and NLDAS-2 simulations with U.S. Drought Monitor classifications
.
J. Hydrometeor.
,
14
,
1035
1056
, doi:.
Andreadis
,
K. M.
, and
D. P.
Lettenmaier
,
2006
:
Trends in 20th century drought over the continental United States
.
Geophys. Res. Lett.
,
33
,
L10403
, doi:.
Andreadis
,
K. M.
,
E. A.
Clark
,
A. W.
Wood
,
A.
Hamlet
, and
D. P.
Lettenmaier
,
2005
:
Twentieth-century drought in the conterminous United States
.
J. Hydrometeor.
,
6
,
985
1001
, doi:.
Barnash
,
R. J. C.
,
R. L.
Ferral
, and
R. A.
McGuire
,
1973
: A generalized streamflow simulation system: Conceptual models for digital computers. Joint Federal and State River Forecast Center, U.S. National Weather Service, and California Department of Water Resources Tech. Rep., 204 pp.
Chang
,
F. C.
, and
J. M.
Wallace
,
1987
:
Meteorological conditions during heat waves and droughts in the United States Great Plains
.
Mon. Wea. Rev.
,
115
,
1253
1269
, doi:.
Compo
,
G. P.
,
J. S.
Whitaker
, and
P. D.
Sardeshmukh
,
2006
:
Feasibility of a 100-year reanalysis using only surface pressure data
.
Bull. Amer. Meteor. Soc.
,
87
,
175
190
, doi:.
Compo
,
G. P.
, and Coauthors
,
2011
:
The Twentieth Century Reanalysis Project
.
Quart. J. Roy. Meteor. Soc.
,
137
,
1
28
, doi:.
Ducharne
,
A.
,
R. D.
Koster
,
M. J.
Suarez
,
M.
Stieglitz
, and
P.
Kumar
,
2000
:
A catchment-based approach to modeling land surface processes in a general circulation model
.
J. Geophys. Res.
,
105
,
24 823
24 838
, doi:.
Ek
,
M. B.
,
K. E.
Mitchell
,
Y.
Lin
,
E.
Rogers
,
P.
Grunmann
,
V.
Koren
,
G.
Gayno
, and
J. D.
Tarpley
,
2003
:
Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model
.
J. Geophys. Res.
,
108
,
8851
, doi:.
Huang
,
J.
, and
H. M.
van den Dool
,
1993
:
Monthly precipitation–temperature relation and temperature prediction over the United States
.
J. Climate
,
6
,
1111
1132
, doi:.
Hunt
,
E. D.
,
K. G.
Hubbard
,
D. A.
Wilhite
,
T.
Arkebauer
, and
A. L.
Dutcher
,
2009
:
The development and evaluation of a soil moisture index
.
Int. J. Climatol.
,
29
,
747
759
, doi:.
Kalnay
,
E.
, and Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project
.
Bull. Amer. Meteor. Soc.
,
77
,
437
471
, doi:.
Koren
,
V.
,
J.
Schaake
,
K.
Mitchell
,
Q.
Duan
,
F.
Chen
, and
J.
Baker
,
1999
:
A parameterization of snowpack and frozen ground intended for NCEP weather and climate models
.
J. Geophys. Res.
,
104
,
19 569
19 585
, doi:.
Koster
,
R. D.
,
M. J.
Suarez
,
A.
Ducharne
,
M.
Stieglitz
, and
P.
Kumar
,
2000
:
A catchment-based approach to modeling land surface processes in a general circulation model: 1. Model structure
.
J. Geophys. Res.
,
105
,
24 809
24 822
, doi:.
Koster
,
R. D.
,
S. D.
Schubert
, and
M. J.
Suarez
,
2009
:
Analyzing the concurrence of meteorological droughts and warm periods, with implications for the determination of evaporative regime
.
J. Climate
,
22
,
3331
3341
, doi:.
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.
,
99
,
14 415
14 428
, doi:.
Livezey
,
R. E.
, and
W. Y.
Chen
,
1983
:
Statistical field significant test and its determination by Monte Carlo techniques
.
Mon. Wea. Rev.
,
111
,
46
59
, doi:.
Livneh
,
B.
,
E. A.
Rosenberg
,
C.
Lin
,
B.
Nijssen
,
V.
Mishra
,
K.
Andreadis
,
E. P.
Maurer
, and
D. P.
Lettenmaier
,
2013
:
A long-term hydrologically based dataset of land surface fluxes and states for the conterminous United States: Update and extensions
.
J. Climate
,
26
,
9384
9392
, doi:.
Lyon
,
B.
, and
R. M.
Dole
,
1995
:
A diagnostic comparison of the 1980 and 1988 U.S. summer heat wave–droughts
.
J. Climate
,
8
,
1658
1675
, doi:.
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
, doi:.
Mo
,
K. C.
, and
D. P.
Lettenmaier
,
2015
:
Heat wave flash droughts in decline
.
Geophys. Res. Lett.
,
42
,
2823
2829
, doi:.
Myoung
,
B.
, and
J. W.
Nielsen-Gammon
,
2010a
:
The convective instability pathway to warm season drought in Texas. Part I: The role of convective inhibition and its modulation by soil moisture
.
J. Climate
,
23
,
4461
4470
, doi:.
Myoung
,
B.
, and
J. W.
Nielsen-Gammon
,
2010b
:
The convective instability pathway to warm season drought in Texas. Part II: Free tropospheric modulation of convective inhibition
.
J. Climate
,
23
,
4474
4488
, doi:.
Otkin
,
J. A.
,
M. C.
Anderson
,
C.
Hain
,
I. E.
Mladenova
,
J. B.
Basara
, and
M.
Svoboda
,
2013
:
Examining rapid onset drought development using the thermal infrared based evaporative stress index
.
J. Hydrometeor.
,
14
,
1057
1074
, doi:.
Otkin
,
J. A.
,
M. C.
Anderson
,
C.
Hain
, and
M.
Svoboda
,
2014
:
Examining the relationship between drought development and rapid changes in the evaporative stress index
.
J. Hydrometeor.
,
15
,
938
956
, doi:.
Senay
,
G. B.
,
M. B.
Budde
,
T. F.
Brown
, and
J. P.
Verdin
,
2008
: Mapping drought in the Southern Great Plains. 22nd Conf. on Hydrology, New Orleans, LA, Amer. Meteor. Soc., 6.8. [Available online at https://ams.confex.com/ams/88Annual/techprogram/paper_134349.htm.]
Svoboda
,
M.
, and Coauthors
,
2002
:
The Drought Monitor
.
Bull. Amer. Meteor. Soc.
,
83
,
1181
1190
, doi:.
Wang
,
A.
,
T. J.
Bohn
,
S. P.
Mahanama
,
R. D.
Koster
, and
D. P.
Lettenmaier
,
2009
:
Multimodel reconstruction of drought over the continental United States
.
J. Climate
,
22
,
2684
2712
, doi:.
Wood
,
A. W.
, and
D. P.
Lettenmaier
,
2006
:
A test bed for new seasonal hydrologic forecasting approaches in the western United States
.
Bull. Amer. Meteor. Soc.
,
87
,
1699
1712
, doi:.
Yang
,
Z.
,
2013
: Developing a flash drought indicator for the U.S. Great Plains. M.S. thesis, Dept. of Geological Sciences, University of Texas at Austin, 24 pp. [Available online at http://hdl.handle.net/2152/21828.]