Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Diagnostic Classification of Flash Drought Events Reveals Distinct Classes of Forcings and Impacts

Diagnostic Classification of Flash Drought Events Reveals Distinct Classes of Forcings and Impacts FEBRUARY 2022 OSM A N E T A L . 275 Diagnostic Classification of Flash Drought Events Reveals Distinct Classes of Forcings and Impacts a,b a c d d MAHMOUD OSMAN, BENJAMIN F. ZAITCHIK, HAMADA S. BADR, JASON OTKIN, YAFANG ZHONG, d e f,g f h DAVID LORENZ, MARTHA ANDERSON, TREVOR F. KEENAN, DAVID L. MILLER, CHRISTOPHER HAIN, AND THOMAS HOLMES Department of Earth and Planetary Sciences, The Johns Hopkins University, Baltimore, Maryland Irrigation and Hydraulics Department, Cairo University, Cairo, Egypt Department of Civil and Systems Engineering, The Johns Hopkins University, Baltimore, Maryland Space Science and Engineering Center, Cooperative Institute for Meteorological Satellite Studies, University of Wisconsin–Madison, Madison, Wisconsin Hydrology and Remote Sensing Laboratory, Agricultural Research Service, USDA, Beltsville, Maryland Department of Environmental Science, Policy, and Management, University of California, Berkeley, Berkeley, California Earth and Environmental Sciences Area, Lawrence Berkeley National Laboratory, Berkeley, California Earth Science Office, NASA Marshall Space Flight Center, Huntsville, Alabama Hydrological Sciences Laboratory, NASA Goddard Space Flight Center, Greenbelt, Maryland (Manuscript received 14 July 2021, in final form 15 December 2021) ABSTRACT: Recent years have seen growing appreciation that rapidly intensifying flash droughts are significant climate hazards with major economic and ecological impacts. This has motivated efforts to inventory, monitor, and forecast flash drought events. Here we consider the question of whether the term “flash drought” comprises multiple distinct classes of event, which would imply that understanding and forecasting flash droughts might require more than one framework. To do this, we first extend and evaluate a soil moisture volatility–based flash drought definition that we introduced in previous work and use it to inventory the onset dates and severity of flash droughts across the contiguous United States (CONUS) for the period 1979–2018. Using this inventory, we examine meteorological and land surface conditions associated with flash drought onset and recovery. These same meteorological and land surface conditions are then used to classify the flash droughts based on precursor conditions that may represent predictable drivers of the event. We find that distinct classes of flash drought can be diagnosed in the event inventory. Specifically, we describe three classes of flash drought: “dry and demanding” events for which antecedent evaporative demand is high and soil moisture is low, “evaporative” events with more modest antecedent evaporative demand and soil moisture anomalies, but positive antecedent evaporative anomalies, and “stealth” flash droughts, which are different from the other two classes in that precursor meteorological anomalies are modest relative to the other classes. The three classes exhibit somewhat different geographic and seasonal distributions. We conclude that soil moisture flash droughts are indeed a composite of distinct types of rapidly intensifying droughts, and that flash drought analyses and forecasts would benefit from approaches that recognize the existence of multiple phenome- nological pathways. KEYWORDS: Drought; Extreme events; Hydrometeorology; Soil moisture; Climate classification/regimes 1. Introduction losses, wildfires, and economic damages in the tens of billions of dollars. These droughts occurred at different times of the In recent years, a number of rapid-onset drought events year in different climate zones with different ecological char- have struck the contiguous United States (CONUS), with acteristics, yet they have all been described as flash droughts, severe consequences for ecological and agricultural systems. a term first coined by Peters et al. (2002) and Svoboda et al. For example, droughts in the Southern Plains in 2011, the cen- (2002) to reflect the fact that some droughts emerge rapidly tral United States in 2012, the Southeast in 2016, the Northern and quickly develop into high-impact extreme events. Plains in 2017, and Texas in 2019 led to widespread crop A challenging characteristic of flash droughts is that they appear suddenly}seemingly without warning}and therefore leave farmers, ranchers, and other vulnerable stakeholders lit- Denotes content that is immediately available upon publica- tle time to prepare mitigation responses (Otkin et al. 2015b, tion as open access. 2018a; Haigh et al. 2019). The 2012 flash drought, for exam- ple, received tremendous attention because of its impact on the nation’s corn crop. Yet there was virtually no sign of an Supplemental information related to this paper is available at the Journals Online website: https://doi.org/10.1175/10.1175/JHM- impending rapid intensification prior to the event in standard D-21-0134.s1. drought monitoring products at that time or in dynamically based seasonal forecasting systems (Hoerling et al. 2014). Postevent analyses concluded that the event was largely Corresponding author: Mahmoud Osman, mahmoud.osman@ jhu.com driven by random atmospheric variability, and perhaps was DOI: 10.1175/JHM-D-21-0134.1 Ó 2022 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses). 276 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 inherently unpredictable using conventional methods (Kumar severity when determining the intensity of a flash drought. et al. 2013). Poor model performance both in forecasting and Their study showed that there are important regional differ- reproducing these events presents an additional challenge in ences in flash drought severity when both of these components efforts to project flash drought impacts and feedbacks under are considered. nonstationary climate conditions (Wolf et al. 2016). Notwith- Most proposed definitions and intensity metrics for flash standing these challenges, there is evidence that flash droughts droughts have focused exclusively on capturing the phenome- are amenable to seasonal-to-subseasonal scale prediction on non rather than assessing whether it represents a coherent account of their sensitivity to initial conditions (Lorenz et al. class from the perspective of drought process. An exception is 2017a,b), the perceived importance of forecastable drivers of the work of Mo and Lettenmaier (2015, 2016), which explic- evaporative demand during flash drought intensification (Hob- itly distinguished between precipitation deficit flash droughts bins et al. 2016), and the potentially predictable role of vegeta- and heat wave flash droughts. The method used to define tion in flash drought processes (Wolf et al. 2016). these droughts has been debated, in large part because Mo Any such generalized statements on the predictability of flash and Lettenmaier consider duration of the heatwave event droughts, however, implicitly assume that the occurrence and rather than intensification rate, which is more typically under- severity of flash droughts can be diagnosed in a consistent and stood to be the defining characteristic of flash drought (Otkin process-relevant manner, and that the term “flash drought” et al. 2018b; Lisonbee et al. 2021), but their concept that flash refers to a single class of event. In recent years, many studies droughts might be the product of multiple different pathways have sought to describe and diagnose the occurrence of flash with distinct meteorological drivers is highly relevant to droughts by proposing a variety of definitions that can be used to understanding and prediction. While Mo and Lettenmaier inventory and map flash droughts. Otkin et al. (2013, 2014, made this distinction a priori by incorporating different varia- 2015a) identified flash droughts based on rapid changes in the bles and thresholds in their definitions, we are not aware of ratio between actual evapotranspiration (EVP) and potential any study that empirically classifies different flash drought evapotranspiration (PEVP). Other studies (Hunt et al. 2014; Mo types within an inventory generated using a common flash and Lettenmaier 2015)defined flash droughts as a function of drought definition. That is: if an inventory of flash drought the rapid drop in soil moisture with time. Chen et al. (2019) sug- events is generated using a definition based on flash drought gested the degradation of two categories in the U.S. Drought phenomenology alone, are there distinct classes within that Monitor (USDM) in a period of four weeks as a definition for inventory that can be identified due to different precursors in the onset of flash droughts. Christian et al. (2019) introduced the meteorology or surface conditions? If so, that implies that definition for flash droughts based on the rate of change in stan- understanding and predicting flash droughts may require that dardized ratio between EVP and PEVP over a six-pentad (6 3 5 we adopt different perspectives for each class. days) period. Another quantitative definition (Ford and Labosier Here, we apply our recently introduced SMVI flash drought 2017) identified flash droughts as the drop of the one pentad definition (Osman et al. 2021) to address this question. First, averaged soil moisture (SM) from the 40th to 20th percentiles in we extend the SMVI presented in Osman et al. (2021) to a period of four pentads or less. A subsequent study by Hoff- include estimates of drought severity, and we compare the mann et al. (2021) followed a similar methodology with adjust- SMVI to independent vegetation and crop datasets for semi- ments to reduce the number of identified events. In a recent nal flash drought events. Next, we apply a retrospective inven- study, (Osman et al. 2021) introduced a definition based on a soil tory of flash droughts, generated using SMVI, to derive moisture volatility index (SMVI), and also compared the SMVI composites of meteorological and surface conditions in the with six other definitions to highlight the fact that there are dif- predrought, onset, and recovery phases of all flash droughts. ferent pathways to identify flash drought onset. All of the listed Finally, we perform objective classification of the flash studies focused on CONUS, but the flash drought phenomenon drought inventory on the basis of meteorological and surface has been observed in many regions across the globe (Nguyen condition precursors to identify flash drought classes relevant et al. 2019; Zhang and Yuan 2020), with a number of studies to process understanding and prediction. focusing on China and India (Wang et al. 2016; Yuan et al. 2019; Mahto and Mishra 2020). These studies have yielded additional 2. Data and methods definitions. Indeed, the need to understand the implications of different definitions has become a research question in its own We generate an inventory of soil moisture flash droughts right (Lisonbee et al. 2021). for all of CONUS over the period 1979–2018 for spring Fewer studies have attempted to quantify the severity of the through fall (March–November). SMVI is calculated using flash droughts, but informative efforts do exist. Chen et al. root zone soil moisture (RZSM) from the SMERGE dataset. (2019) and Otkin et al. (2015a) both used USDM categories to SMERGE is a hybrid daily product at 0.1258 spatial resolution diagnose and assess severity of flash droughts. Christian et al. that combines satellite-derived soil moisture estimates from (2019) used standardized evaporative stress ratio (SESR) for the European Space Agency Climate Change Initiative and both purposes, Yuan et al. (2019) used soil moisture deficit, and NLDAS-2 Noah model output for RZSM averaged from 0- to Li et al. (2020) used evapotranspiration deficit. Basedonmod- 40-cm depth (Tobin et al. 2019). The SMERGE dataset has eled soil moisture, Otkin et al. (2021) developed a flash drought been evaluated against normalized difference vegetation intensity index (FDII) that explicitly accounts both for the mag- index (NDVI) products (Rouse et al. 1974) as well as in situ nitude of the rapid intensification and the resultant drought soil moisture observations, and it has been found to be a FEBRUARY 2022 OSM A N E T A L . 277 reliable dataset for agricultural and ecological applications method’s skill to capture changes in satellite-observed vegetation (Tobin et al. 2019). greenness due to flash drought. The cloud-free NDVI data were The SMVI is motivated by the fact that flash drought diagno- obtained from the 16-day MODIS composite product sis is concerned with capturing change that is more rapid than (MOD13C1) at 0.058 spatial resolution (Didan 2021)for the usual, so that it could be used to identify both rapid onset and years 2000 to present. NDVI grid points with anomalies below rapid intensification drought events. For SMVI, rapid changes 20.5 standard deviation from the mean are defined as are identified by the crossover of simple moving averages “negatively impacted” in comparisons with SMVI. This approxi- (SMAs) combined with duration and dryness thresholds. Onset mately corresponds to a probability of occurrence less than 30% is recorded when 1) the 5-day (1-pentad) RZSM SMA falls and for normally distributed conditions. Further, we evaluate the per- stays below the 20-day (4-pentad) SMA for at least a 20-day formance of the SMVI definition for the 2012 central United period or 2) both SMAs are below the 20th percentile of the States and 2017 Northern Plains flash droughts versus in situ 1979–2018 time-of-year RZSM climatology (Osman et al. reports of soil and crop conditions collected by the USDA 2021). If two sequential flash droughts are identified with a National Agricultural Statistics Service (NASS) observers. Data period of three pentads or less between them, then they are showing poor conditions are marked as negatively impacted. combined into a single event. We do this because a short rain- These data are collected at county scale, then spatially smoothed fall event may result in a temporary reduction in the severity of to reduce noise, and protect confidentiality (access to data at a flash drought but is often not sufficient to restore predrought county level was provided to the coauthors after signing a confi- conditions and end the drought event. dentiality agreement with the USDA NASS). The performance Severity is quantified basedonRZSMdeficit during the iden- analyses are carried out for the spring and summer seasonal aver- ages due to data availability and temporal resolution. tified flash drought event according to Eqs. (1) and (2) as illus- The performance of the SMVI is assessed with hit–miss confu- trated in the example in Fig. S1 in the online supplemental sion matrices that use NDVI and NASS data as observational material. This scale is based on the standardized distribution of reference datasets. True positive values represent grid points and the integrated RZSM deficit below the 20th percentile (and pentads depicted by SMVI as being in flash drought and also over the 5-day running average) during the drought event: marked as negatively impacted by the NASS or NDVI validation t5t datasets, while false positives are the events classified as flash SV 5() RZSM 2 RZSM (1) 20th 5d drought by SMVI where NASS or NDVI do not meet drought t5t impact criteria. True negative values represent grid points not marked as negatively impacted by the NASS or NDVI validation datasets and not identified as flash drought grid points. False neg- SV SV 5 , (2) CAT atives represent grid points identified by SMVI as having no flash STD() SV drought while marked as negatively impacted by the NASS or where SV is the computed severity, and RZSM and RZSM NDVI validation datasets. Hit–miss statistics are calculated 20th 5d are the 20th percentile and 5-day moving average RZSM, according to Eqs. (3)–(10): respectively. Parameters t and t represent the times at which o f TP identified flash drought onset occurs and ends, respectively. The sensitivity() TPR 5 , (3) TP 1 FN standardized severity category is represented by SV with a CAT range from zero (no flash drought) up to 5 (maximum severity), and STD(SV ) is the severity standard deviation calcu- TN 1979–2018 () specificity TNR 5 , (4) lated from the flash drought inventory for all grid points, mea- TN 1 FP sured against the severity of all other identified flash drought events within the inventory. The use of categories to indicate FP drought severity is a common approach, as used in systems such () false discovery rate FDR 5 , (5) FP 1 TP as the USDM. In contrast to the USDM, the SMVI-based severity is intended to capture the severity of the rapid onset flash drought process. FN () false negative rate FNR 5 , (6) The end of the flash drought period (recovery period) date FN 1 TP is identified when the rate of drop in RZSM during an identi- fied flash drought event begins to recover (i.e., the 1-pentad FP () running average is no longer below 4-pentad running average) false positive rate FPR 5 , (7) FP 1 TN or the 1-pentad RZSM is no longer below the 20th percentile of the 1979–2018 time-of-year RZSM. TP SMVI performance was previously evaluated based on precision() PPV 5 , (8) TP 1 FP descriptions of reported major flash drought events (Osman et al. 2021). Influenced by the methodology followed by Peters et al. (2002) to detect drought using standardized NDVI, in this study TP 1 TN we use MODISNDVItime-of-yearanomaliestoassess the accuracy() ACC 5 , (9) TP 1 TN 1 FP 1 FN 278 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 FIG. 1. Flash drought maps as captured by SMVI definition during the active growing season (March–November): (left) 2012 and (right) 2017. (a),(b) Onset maps, where each color represents the month of flash drought onset. (c),(d) Estimated severity category maps. classes subjectively, but there are recommended diagnostics TP () critical success index CSI 5 , (10) for use in choosing the optimal number of classes. Here we TP 1 FN 1 FP apply the commonly used elbow method (Thorndike 1953) for this purpose. where TP, TN, FP, and FN represent true positive, true nega- tive, false positive, and false negative grid points, respectively. Values of Eqs. (3)–(10) range from 0 to 1, with 1 being the 3. Results and discussion perfect score for the TP or TN numerator-based ratios and a. The SMVI flash drought intensity metric the opposite for the FP and FN numerator-based ratios. Drawing on previous studies that have described meteoro- The United States was hit by several major flash drought logical and surface conditions associated with flash drought events over the past decade, resulting in excessive agricultural onset (Mo and Lettenmaier 2015, 2016; Ford and Labosier losses and livestock destruction. In 2012, the country experi- 2017; He et al. 2019; Osman et al. 2021), we select multiple enced one of the largest and most destructive flash droughts variables from the NLDAS-2 datasets (temperature, precipi- recorded to date, with more than $30 billion estimated dam- tation, RZSM, PEVP, EVP, and surface pressure) along with ages (Hoerling et al. 2013, 2014; Basara et al. 2019; Mallya the computed vapor pressure deficit (VPD) and total cloud et al. 2013; Fuchs et al. 2012; Otkin et al. 2016). A warm spring cover (TCC) from NCEP–NCAR reanalysis products (Kalnay followed by early summer heatwaves set the stage for a rap- et al. 1996), and analyze their progression through the pre- idly intensifying drought that struck much of the middle part drought, onset and end of the flash drought periods for all of the country in late spring and early summer and extended events included in the 40-yr (1979–2018) SMVI-derived flash to the north later in summer and in early fall (Fig. 1a). Nota- drought inventory. To focus on events with meaningful bly, though the occurrence of flash drought was very wide- impact, we analyze only SMVI-derived flash drought events spread (according to both SMVI and other definitions) with severity greater than 2. Unsupervised multivariate classi- (Osman et al. 2021), the central United States had the greatest fication is then performed as a function of these meteorologi- severity, as diagnosed by the SMVI (Fig. 1c). cal variables, using principal components transformation to Five years after the 2012 flash drought, the Northern Plains control for collinearity between variables. This classification is region was hit by another major flash drought, causing more used to characterize different types of flash droughts driven than $2.6 billion in agricultural losses and sparking wildfires. by different processes. The classes are determined using the The 2017 Northern Plains flash drought was focused on Mon- k-means partitioning unsupervised classification algorithm tana, North Dakota, South Dakota, and parts of Alberta and (Hartigan and Wong 1979; Lloyd 1982) as a heuristic cluster- Saskatchewan (Jencso et al. 2019). The event started in May ing method. We apply a sensitivity analysis to determine the over western Montana and swiftly intensified through high statistically optimal number of clusters. The anomalies are evaporative demand and precipitation deficits (Hoell et al. calculated as the in-time (predrought, onset, or recovery) pen- 2019a; Osman et al. 2021). The drought eventually spread tad anomaly relative to the 1979–2018 time-of-year average. over much of the Northern Plains region (Fig. 1b) causing The k-means algorithm allows the user to set the number of enormous economic losses (Gerken et al. 2018; Jencso et al. FEBRUARY 2022 OSM A N E T A L . 279 FIG.2.Maps of hit–miss analysis for the 2012 and the 2017 flash droughts during the actively growing season (March–November): (left) 2012 and (right) 2017. (a),(b) SMVI vs negative NDVI anomaly hit–miss map, in which lavender represents false positive (FP), orange rep- resents true positive (TP), white represents true negative (TN), green represents false negative (FN), and gray represents missing/unavail- able data. (c),(d) As in (a) and (b), but for NASS reported negative average crop conditions. (e),(f) As in (a) and (b), but for the observed topsoil moisture. 2019; He et al. 2019). Montana was the most impacted state With this caveat in mind, we compare the SMVI flash (Jencso et al. 2019), and this is evident in the SMVI-based drought index to MODIS NDVI anomalies and NASS crop severity analysis (Fig. 1d). The severity analysis is also consis- and topsoil condition anomalies. Using a simple hit/miss met- tent with the USDM reports that showed an exceptional (D4 ric in which negative anomalies in MODIS NDVI (more than category) drought over Montana (Jencso et al. 2019). It is 0.5 standard deviation below the mean) or the NASS condi- important to highlight that estimation of flash droughts’ sever- tion maps are interpreted as evidence of drought conditions, ity in this study is a method to relatively quantify soil moisture we find that there is broad agreement between the SMVI and deficit with a methodology similar to Yuan et al. (2019) study observed drought conditions for both the 2012 and 2017 flash given the different flash drought identification method. drought events (Figs. 1 and 2). We do see considerable false Independent, quantitative validation of drought indices is negatives on the margins of the drought-affected area, particu- notoriously difficult, since impacts of drought vary with cli- larly in 2012, but this is consistent with our liberal definition of mate context, land cover, and economic system. Since flash agricultural drought in the NDVI and NASS fields (i.e., flash drought is a subset of all droughts which is typically consid- drought identified area is smaller than NDVI and NASS nega- ered in agricultural and ecological contexts (Wang et al. 2016; tive anomalies). We also note a concentration of false positives Mo and Lettenmaier 2015; Christian et al. 2019; Otkin et al. along edge of drought regions, particularly in 2017, indicate 2018b), we consider vegetation health and crop status to be that the SMVI approach overestimated the extent of drought- two relevant indicators of drought impact that can verify the affected area relative to NASS estimates. utility of SMVI as a useful drought metric. In doing this, we Focusing on the central and northern High Plains regions recognize that the independent comparisons do not necessar- [as defined by Bukovsky (2011)] for the years 2012 and 2017, ily confirm the presence of flash drought; rather, they are respectively, we find that for flash droughts based on negative interpreted as indicators of whether an agricultural drought NDVI anomalies the accuracy was 0.68 in 2012 and 0.56 in may have occurred. 2017. Precision was higher in 2012 (0.74) than 2017 (0.50), 280 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 TABLE 1. SMVI–NASS and SMVI–NDVI summary hit–miss statistics for the 2012 central region flash drought showing the geographically dominant crops and observed soil moisture conditions. Corn Range Soybean Subsoil Topsoil Avg crop condition NDVI ACC 0.74 0.74 0.78 0.84 0.77 0.75 0.68 CSI 0.71 0.73 0.76 0.84 0.76 0.73 0.63 FDR 0.21 0.07 0.16 0.05 0.05 0.11 0.26 FNR 0.12 0.22 0.10 0.13 0.21 0.20 0.19 FPR 0.65 0.55 0.64 0.50 0.38 0.51 0.60 PPV 0.79 0.93 0.84 0.95 0.95 0.89 0.74 TNR 0.35 0.45 0.36 0.50 0.62 0.49 0.40 TPR 0.88 0.78 0.90 0.87 0.79 0.80 0.81 while the probability of detection (sensitivity) was higher in they are 0.84 and 0.95 for the 2017 event. We also note that 2017: 0.93, versus 0.81 in 2012 (Tables 1 and 2). The critical irrigation is a complicating factor that may affect comparison success index was significantly higher for the 2012 event (0.63) between datasets. While SMVI does include partial consider- compared to that observed in 2017 (0.48). These values of ation of irrigation, insomuch as SMERGE captures irrigation hit–miss statistics are consistent with moderate to strong per- signals, this representation is imperfect and might not align formance in event identification (Hoerling et al. 2013, 2014; with observed vegetation response to irrigation. Basara et al. 2019; Mallya et al. 2013; Fuchs et al. 2012; Otkin b. Proposed drivers of flash drought et al. 2016; Gerken et al. 2018; Jencso et al. 2019; He et al. 2019). It is important to note that this is an imperfect compari- Figure 3 presents composites of predrought (onset minus son. The SMVI approach is one pathway of identifying flash three pentads), onset, and recovery period conditions, using droughts, and a comparison with a vegetation index metric, composites of standardized anomalies of meteorological fields such as NDVI anomalies, is not exactly indicative of perfor- for all flash droughts of severity greater than 2 in the SMVI- mance in capturing a soil moisture flash drought. derived 1979–2018 inventory. Composites are calculated sepa- NASS-based evaluation, based on NASS identification of rately for each grid cell, such that the anomalies represent poor crop and soil conditions, led to comparable statistics for conditions when a flash drought occurred in that exact loca- each impacted region’s dominant crop (Figs. 2c–f). Tables 1 tion. Precipitation (PRCP) anomalies in the predrought and and 2 summarize SMVI–NASS statistics for both the 2012 onset periods are mostly negative, as one would expect, which and 2017 flash droughts. In the 2012 central U.S. flash is also associated with suppression of the convective available drought, SMVI shows an accuracy of 0.79, 0.75, and 0.74 for potential energy (CAPE) over most of CONUS (we include negatively impacted soybean, range, and corn, with a preci- CAPE in addition to precipitation in order to isolate local sion of 0.84, 0.79, and 0.89, respectively. The 2017 Northern convective potential as distinct from total realized rainfall). Plains flash drought captured by SMVI is similarly evaluated This is similar to the observed scenario before and during the and statistical evaluation was slightly higher than that seen for 2017 northern High Plains flash drought (Gerken et al. 2018). the NDVI analysis. Accuracy for detecting grids of flash The magnitude of these standardized anomalies, however, is drought in the Northern Plains compared to negatively generally small relative to the anomalies in RZSM and poten- impacted dominant crops (barley and spring wheat) are 0.8 tial evaporation (PEVP), particularly during the pentad of and 0.76, respectively, with precision values of 0.91 and 0.88, drought onset. and probability of detection greater than 0.84. Comparing These findings are consistent with previous studies (Otkin SMVI to the reported NASS topsoil moisture conditions et al. 2018b, 2013; Anderson et al. 2013), which have empha- shows a very similar pattern for the negatively reported condi- sized the importance of precursor soil moisture conditions and tions. The accuracy and precision of SMVI detection of the PEVP in the onset of a flash drought. Low RZSM, high PEVP reported negative NASS topsoil moisture conditions for the and high VPD conditions force the rapid transition from an 2012 flash drought event are 0.77 and 0.95, respectively, and energy limited environment to a water limited environment, TABLE 2. As in Table 1, but for the 2017 northern High Plains region flash drought. Barley Oats Spring wheat Winter wheat Subsoil Topsoil Avg crop condition NDVI ACC 0.80 0.73 0.76 0.78 0.84 0.84 0.72 0.56 CSI 0.79 0.72 0.75 0.77 0.83 0.83 0.70 0.48 FDR 0.09 0.15 0.12 0.08 0.03 0.02 0.18 0.50 FNR 0.15 0.18 0.16 0.17 0.15 0.16 0.17 0.07 FPR 0.55 0.73 0.65 0.61 0.29 0.25 0.72 0.72 PPV 0.91 0.85 0.88 0.92 0.97 0.98 0.82 0.50 TNR 0.45 0.27 0.35 0.39 0.71 0.75 0.28 0.28 TPR 0.85 0.82 0.84 0.83 0.85 0.84 0.83 0.93 FEBRUARY 2022 OSM A N E T A L . 281 FIG. 3. Composite maps of standardized anomalies of climate conditions for selected atmospheric variables (TEMP: 2-m above ground temperature, PRCP: precipitation, RZSM: root-zone soil moisture, EVP: actual evapotranspiration, PEVP: potential evapotranspiration, SPRESS: surface pressure, TCC: total cloud cover, WS: 10-m above ground wind speed, CAPE: convective available potential energy, VPD: vapor pressure deficit) based on the full SMVI flash droughts inventory from 1979 to 2018 for severity higher than 2, during onset, recovery, and onset minus 3 pentads. leading to rapid drought onset and loss of green cover (Otkin more water limited environments the EVP anomalies are neg- et al. 2018b). This elevated PEVP only leads to an increase in ative in the predrought and onset periods, as elevated PEVP actual evapotranspiration (EVP) in regions with greater water cannot translate into an increase in EVP. As described later, variability}e.g., the Midwest and Great Lakes regions. In this distinction is important when considering process-based 282 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 FIG. 4. Boxplot of the standardized anomalies of atmospheric variables and root zone soil moisture averaged for the three pentads before drought onset for each class for the full SMVI inventory from 1979 to 2018. A separate figure for each of the fields’ variability across years is shown in Fig. S4. Maps of the anomalies averaged over the three pentads prior to onset are shown in Fig. S5. flash drought classification: the concept that elevated PEVP flash drought definition; analyses that use different definitions leads to elevated EVP, drying the soil column, is an important might lead to different conclusions. That said, Ford and Lab- aspect of some theories of vegetation-mediated flash drought osier (2017) examine some of the same variables and found intensification (Otkin et al. 2018b), but it is not a feature of all broadly similar patterns using a different flash drought defini- events in our inventory. tion formulation based on the drop in RZSM from the 40th to Other potential predictor variables show regionally variable the 20th percentile in a period that does not exceed four signals. Temperature (TEMP), often identified as a driver of pentads. flash drought, is generally elevated in the predrought period, c. Flash drought classification but the anomalies are weak, and the sign of anomaly is not entirely consistent. It is only during the onset pentad that ele- The composite analysis of conditions at different stages of vated temperatures are observed over most regions (though flash droughts shown in the previous section provides a useful even then the southeast is not particularly anomalously perspective on the flash drought development process; how- warm). Surface pressure (SPRES) might be expected to be ever, it does not consider the possibility that the inventoried anomalously high in the lead-up to a drought, but the anoma- flash droughts consist of distinct forms of drought develop- lies are weak and mixed over much of the country, as is the ment. It is therefore possible that the weak or mixed anoma- average near-surface wind speed (WS). TCC tends toward lies found for certain proposed drivers are simply an artifact negative anomalies in predrought and onset periods, matching of averaging across different types of events, blurring the expectation, but again there are weak or mixed anomalies for influence of hydrometeorological drivers in different drying a number of regions. scenarios. Considering the recovery pentad, which is defined as the To test this hypothesis, we perform K-means classification first pentad in which any of the onset conditions is violated, it on our SMVI-based flash drought inventory. We use onset is evident that the role of rainfall is significant in ending the pentad standardized anomalies for the nine variables applied flash drought. Both PRCP and TCC show strong positive in composite analysis (TMP, PRCP, RZSM, EVP, PEVP, anomalies in recovery, which stands in contrast to the modest SPRES, TCC, WS, CAPE, and VPD) as the basis for classifi- anomalies seen during the predrought and onset periods. cation, and first mask out unvegetated classes (bare soil and Rain breaks the flash drought cycle, quickly switching envi- urban classes) and potentially deep-rooted vegetation classes ronmental conditions to a non-water-limited status, provided (forest and woodland classes) according to the University of that the volume of rain is sufficient. TEMP, PEVP, EVP, Maryland (UMD) Land Cover Classification (Fig. S2). Only VPD, and SPRES anomalies are mixed during the recovery events with severity greater than 2 are included in the classifi- period. RZSM anomalies are still strongly negative, reflecting cation, and we perform principal component analysis on the fact that we have defined the recovery (end of flash meteorological variables prior to classification. Using the drought period) based on the change in rate of declination or elbow method (Thorndike 1953), we find that three classes if RZSM higher than the 20th percentile, which are still below normal conditions but no longer a flash drought. It is worth are optimal (Fig. S3). We emphasize that our classification is emphasizing that these composites are based on our SMVI intended to draw out indicative patterns and is not meant to FEBRUARY 2022 OSM A N E T A L . 283 FIG. 5. Daily time series plots of selected atmospheric variables and RZSM from four pentads prior to drought onset to one pentad after onset for (a) class 1, (b) class 2, and (c) class 3 events. For each class, the time series of each variable represents an average of 20 grid cells, each selected from the core area of a separate flash drought event. The y axis shows the standard devi- ation for the normalized variables’ values. imply that the three classes are entirely separable or indepen- its negative precipitation (PRCP) anomalies are modest relative dent phenomena. The use of a different dataset of meteoro- to the other two classes. logical variables, study region, or flash droughts identification These systematic differences between classes suggests that method may lead to a different number of classes. flash droughts can be triggered by a diversity of meteorologi- The character of each class with respect to precursor soil cal conditions. Class 2 bears the most classic signatures of moisture conditions and meteorology in the pentads leading up drought, with its dry antecedent conditions, high temperature to event onset is shown in Fig. 4. Notably, classes 2 and 3 are and evaporative demand conditions, low cloud cover, and characterized by elevated air temperature (TMP) and vapor reduced total evapotranspiration. From a flash drought per- pressure deficit (VPD) prior to flash drought onset, while class spective, these can be thought of as “dry and demanding” 1 is not. And while classes 2 and 3 have similar TMP anomalies, events, in which atmospheric evaporative demand combines class 2 exhibits substantially more severe antecedent VPD than with low rainfall and dry predrought conditions to allow for class 3, as well as stronger positive potential evapotranspiration rapid intensification of already dry conditions. Notably, (PEVP) anomalies and stronger negative root zone soil mois- PEVP anomalies for these events tend to be quite high, but ture (RZSM) and total cloud cover (TCC) anomalies. Class 3, EVP anomalies are strongly negative on account of the pre- meanwhile, is the only class that shows positive anomalies in vailing dry conditions prior to drought onset. It is important antecedent actual evapotranspiration (EVP) and in CAPE, and to emphasize that our interpretation of the different classes is 284 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 based on the mean value, which adds a margin of uncertainty in classifying an identified flash drought event. Figure 5b shows composite time series of key variables for 20 grid cells picked from the core of different class 2 drought events. As indicated in these time series, TMP, VPD, and PEVP are all elevated in the four pentads before flash drought onset while EVP anomalies are consistently negative over this period. PRCP anomalies are generally negative, with some noise evi- dent in this 20 grid cell sample, while NDVI and RZSM anomalies are strongly negative even four pentads before onset date. In contrast to the classic drought character of class 2, class 3 bears some surprising features. The fact that the events inten- sify rapidly even though, on average, the antecedent PRCP anomalies are modest and CAPE is enhanced, suggest that for these events rapid drying is largely driven by evaporative demand (positive VPD and PEVP anomalies) combined with sufficient moisture access to support elevated EVP. This com- bination makes class 3 the only class to exhibit anomalies con- sistent with the hypothesis that vegetation can contribute to flash drought onset by responding to elevated temperature and evaporative demand with increased evapotranspiration, accelerating depletion of root zone soil moisture. Based on these characteristics, we term class 3 events “evaporative” flash droughts. As shown in Fig. 5c for a random sample of points from different class 3 events, PRCP anomalies are mixed, with a negative signal only evident in the 2 pentads before onset, and positive anomalies seen at longer leads and even after flash drought onset. EVP is consistently elevated before and during onset, while strongly positive TMP, VPD, and PEVP anomalies emerge only in the two pentads before onset. Interestingly, RZSM and NDVI anomalies are, on average for this sample, positive until two pentads before onset, such that the rapid decline observed just before onset leads to negative anomalies that are substantially smaller than those observed for class 2 events at date of onset. Class 1, for its part, is noteworthy for the fact that air tem- perature and evaporative demand preceding flash drought FIG. 6. Frequency (% of years with an event) for each flash onset are unremarkable compared to average conditions. Pre- drought class at each grid point for the period 1979 to 2018, based cipitation is below average in the predrought period, skies are on the SMVI flash droughts definition. relatively clear (low TCC), and convective potential is low (negative CAPE anomaly). But anomalies in all other varia- bles commonly invoked to explain the rapidity of flash to predict with precision more than a few days in advance drought intensification are modest, i.e., there is a near-zero (Tian et al. 2017). The sample time series shown in Fig. 5a temperature, PEVP and VPD predrought anomalies. In this indicates that positive anomalies in VPD and PEVP are mod- sense, class 1 flash droughts appear to be dominated by pre- est and emerge only within two pentads of onset, and TMP cipitation deficit forcing rather than evaporative demand forc- ing, placing them at a far end of the PEVP versus PRCP anomalies are essentially neutral. Interestingly, the decline in balance of flash drought forcings described by Christian et al. NDVI is dramatic for this class, suggesting that these events (2021). As described later, class 1 events are, on average, strike vegetation that is particularly sensitive to drought stress slightly less severe than other classes, but they are not always on account of vegetation type or timing. The fact that NDVI low severity events. We will refer to these events as “stealth” anomalies are strongly positive at three and four pentad leads, flash droughts in that they have characteristics that would and that negative EVP signals are not evident at longer leads, make them difficult to forecast: where classes 2 and 3 show suggests that these events might be associated with favorable meteorological drivers that might be forecasted with skill at early season growing conditions leading to structural over- extended weather to subseasonal time scales, class 1 appears shoot in vegetation (Zhang et al. 2021). to be the product almost solely of moderately dry antecedent At the national scale, 45% of all flash drought events in our soil moisture and below average rainfall, which can be difficult inventory are class 1, 31% are class 2, and 22% are class 3. FEBRUARY 2022 OSM A N E T A L . 285 FIG.7.Classification maps of the 2011, 2012, and 2017 flash drought events. But there are distinct geographic patterns for each (Fig. 6). conditions are simply too dry. The prevalence of class 1 events Class 1 events are most common in the western High Plains, in the western High Plains is less easily explained, but it is class 2 are dominant in the southern Great Plains and Texas, consistent with experience in that the iconic 2017 flash and class 3 are the most common type in the upper Midwest. drought that affected Montana and North Dakota was a nota- This is not a deterministic split}all three classes are found in bly poorly predicted event (Jencso et al. 2019; Hoell et al. all regions}but the geographic distribution aligns with expec- 2019b). tation. In the relatively humid and cool upper Midwest, one Indeed, if we map the class associations of the 2017 flash might expect that high TMP and VPD can trigger elevated drought event, along with the seminal flash drought events of EVP even when soils are somewhat dry relative to their aver- 2011 and 2012 (Fig. 7), we see that 2017 was almost entirely age state, while in the warmer and drier southern Great Plains class 1. The 2011 event, focused on Texas and Oklahoma, is those conditions are less likely to be met with increased EVP: predominantly class 2. The widespread event of 2012 is a mix 286 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 FIG. 8. Average area in each flash drought class in each month included in this study. Average is calculated for the 1979–2018 period. of class 2 and class 3, consistent with the fact that this was a meanwhile, is the most widespread drought class in all months hot event affecting a broad swath of the Great Plains and except for June, when it is brieflyexceeded byclass 2. Thefact Midwest, including a diversity of climate zones and land cover that class 1 events continue to be relatively common in late types. summer is in part a reflection of geography, since these events Seasonally, all three flash drought classes can be observed in dominate in some of the coolest portions of the analysis any month included in our analysis (March–November; Fig. 8). domain. The drivers of flash drought risk, then, appear to vary Class 2 shows a dramatic peak in June, coincident with the by both region and season, a fact that is relevant for the devel- onset of summer heat and dryness over much of the drought- opment of flash drought risk monitoring and forecasting sys- susceptible United States. Class 3 shows a similar, albeit more tems. We note that these seasonal patterns are sensitive to our muted June peak. This is the least common flash drought class inventory method, which is subject to the previously discussed on average, but in the spring it does show slightly greater total assumptions, and clustering may vary accordingly. We note that area than class 2, and the drop in area after June is dramatic. our inventory method, which includes only the first instance of This is consistent with a drought process that includes sufficient flash drought in each grid cell in each year, may slightly under- available soil moisture to support elevated EVP. Class 1, represent late season flash droughts in general, since in cases of FIG. 9. Boxplots of the flash droughts average severity categories in the three classes after fil- tering out events of severity category less than 2 (box widths are proportional to the square root of the total number of grid points in each class). FEBRUARY 2022 OSM A N E T A L . 287 two flash droughts in the same location in the same year (which We emphasize that the classes defined here are representative are rare) the second event would not be captured by our of a continuum of processes associated with flash drought devel- method. opment. We choose to work with three classes because it proved Finally, we find that all three diagnosed classes of flash to be a stable, separable, and interpretable number of classes in drought include cases of severe drought [according to our cre- our analysis, but the result does not imply that there are only ated inventory of flash droughts severity from Eqs. (1) and three pathways that can lead to flash drought, or that an event (2)], but that there are statistical differences in severity cannot exhibit a mix of properties from two or three classes. The between classes, as estimated using the SMVI severity classes contrasting meteorological and surface process signatures of the defined in this study (Fig. 9). There is a slight tendency for three classes do, however, indicate that events identified as flash greater severity in class 2, the dry and demanding droughts, drought using a reasonable definition, including events that have and the most severe events in the record are dominated by been widely reported as seminal flash droughts, represent a class 2, followed by slightly decreased average severity for diversity of onset and intensification processes. Our results sug- class 3 and class 1. The differences in severity between classes gest that recognizing this diversity is critical to advance our are statistically significant, as evaluated using a Welch’s t test, understanding and ability to predict these events. for both raw and log transformed data, and confirmed with a nonparametric Wilcoxon signed-rank test. This result empha- Acknowledgments. We sincerely thank the journal editor sizes the potential severity of flash droughts that develop and the anonymous reviewers for their constructive comments. under the combined conditions of high evaporative demand, This work was funded by the National Science Foundation low precipitation, and dry antecedent conditions. Neverthe- (NSF) Grant 1854902. less, the distribution of event severities shown in Fig. 9 makes it clear that all three classes contain severe events. This is also REFERENCES clear from our analysis of seminal flash droughts (Fig. 1). We note that Fig. 9 shows results for events filtered for severity Anderson, M. C., C. Hain, J. Otkin, X. Zhan, K. Mo, M. Svoboda, greater than 2, but that the same general pattern is observed B. Wardlow, and A. Pimstein, 2013: An intercomparison of when we do not apply a severity threshold. drought indicators based on thermal remote sensing and NLDAS-2 simulations with U.S. Drought Monitor classifica- 4. Conclusions tions. J. Hydrometeor., 14, 1035–1056, https://doi.org/10.1175/ JHM-D-12-0140.1. Flash drought has proven to be a challenging phenomenon Basara, J. B., J. I. Christian, R. A. Wakefield, J. A. Otkin, E. H. for both monitoring and prediction. These challenges have Hunt, and D. P. Brown, 2019: The evolution, propagation, been associated with the rapidly evolving nature of the events andspreadof flash drought in the central United States dur- and, perhaps, with the fact that they depend on processes that ing 2012. Environ. Res. Lett., 14, 084025, https://doi.org/10. may not be explicitly resolved, or may be poorly predicted, in 1088/1748-9326/ab2cc0. standard subseasonal-to-seasonal forecast systems. But termi- Bukovsky, M. S., 2011: Masks for the Bukovsky regionalization of nology and definitions have also been challenging (Lisonbee North America. Regional Integrated Sciences Collective, et al. 2021), and the difficulty of establishing consistent and Institute for Mathematics Applied to Geosciences, National agreed-upon definitions is also a significant contributor to Center for Atmospheric Research, accessed 20 November associated challenges in prediction. If the physical interpreta- 2021, http://www.narccap.ucar.edu/contrib/bukovsky/. tion of a flash drought inventory is not sufficiently clear, then Chen, L. G., J. Gottschalck, A. Hartman, D. Miskus, R. Tinker, it is also not clear what one is predicting with a statistical and A. Artusa, 2019: Flash drought characteristics based on model trained using that inventory, or what one is evaluating U.S. Drought Monitor. Atmosphere, 10, 498, https://doi.org/ 10.3390/atmos10090498. when considering a dynamically based forecast of an event. Christian, J. I.,J.B. Basara, J. A. Otkin, E. D. Hunt,R.A.Wake- Here, we have examined meteorological drivers associated field, P. X. Flanagan, and X. Xiao, 2019: A methodology for with events inventoried using an SMVI-based definition of flash drought identification: Application of flash drought fre- flash drought events, and then classified all events in the quency across the United States. J. Hydrometeor., 20, 833–846, inventory on the basis of precursor meteorological and sur- https://doi.org/10.1175/JHM-D-18-0198.1. face conditions. We found three classes of flash droughts in }}, }},E.D. Hunt, J. A. Otkin,J.C.Furtado, V. Mishra, our inventory based on k-means clustering. We refer to these X. Xiao, and R. M. Randall, 2021: Global distribution, trends, classes as: dry and demanding droughts, with high evaporative and drivers of flash drought occurrence. Nat. Commun., 12, demand and antecedent low soil moisture levels; evaporative 6330, https://doi.org/10.1038/s41467-021-26692-z. droughts, which initiate under conditions of high demand and Didan, K., 2021: MODIS/Terra vegetation indices 16-Day L3 when elevated evapotranspiration accelerates soil drying; and global 0.05Deg CMG V061. NASA EOSDIS Land Processes stealth droughts, which may be hard to predict due to the lack DAAC, accessed 20 November 2021, https://doi.org/10.5067/ of a clear temperature or evaporative demand signal prior to MODIS/MOD13C1.061. initiation. These classes are associated with different meteoro- Ford, T. W., and C. F. Labosier, 2017: Meteorological conditions logical variables, regional distributions, seasonality, and cli- associated with the onset of flash drought in the Eastern matic and land surface risk factors, suggesting that there are United States. Agric. For. Meteor., 247, 414–423, https://doi. distinct forms of flash drought development. org/10.1016/j.agrformet.2017.08.031. 288 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 Fuchs, B., D. Wood, and D. Ebbeka, 2012: From too much to too Kumar, A., M. Chen, M. Hoerling, and J. Eischeid, 2013: Do little: How the Central US drought of 2012 evolved out of extreme climate events require extreme forcings? Geophys. Res. Lett., 40, 3440–3445, https://doi.org/10.1002/grl.50657. one of the most devastating floods on record in 2011. Li,J., Z. Wang,X.Wu, J. Chen,S. Guo, and Z. Zhang, 2020:A National Integrated Drought Information System, 105 pp. new framework for tracking flash drought events in space Gerken, T., G. T. Bromley, B. L. Ruddell, S. Williams, and P. C. and time. Catena, 194, 104763, https://doi.org/10.1016/j.catena. Stoy, 2018: Convective suppression before and during the 2020.104763. United States Northern Great Plains flash drought of 2017. Lisonbee, J., M. Woloszyn, and M. Skumanich, 2021: Making Hydrol. Earth Syst. Sci., 22, 4155–4163, https://doi.org/10. sense of flash drought: Definitions, indicators, and where we 5194/hess-22-4155-2018. go from here. J. Appl. Serv. Climatol., 2021,1–19, https://doi. Haigh, T. R., J. A. Otkin, A. Mucia, M. Hayes, and M. E. Bur- org/10.46275/JOASC.2021.02.001. bach, 2019: Drought early warning and the timing of range Lloyd, S., 1982: Least squares quantization in PCM. IEEE Trans. managers’ drought response. Adv. Meteor., 2019, 9461513, Inf. Theory, 28, 129–137, https://doi.org/10.1109/TIT.1982. https://doi.org/10.1155/2019/9461513. Hartigan, J. A., and M. A. Wong, 1979: A K-means clustering Lorenz, D. J., J. A. Otkin, M. Svoboda, C. R. Hain, M. C. Ander- algorithm. Appl. Stat., 28, 100, https://doi.org/10.2307/2346830. son, and Y. Zhong, 2017a: Predicting the U.S. drought moni- He, M., J. S. Kimball, Y. Yi, S. Running, K. Guan, K. Jensco, B. tor using precipitation, soil moisture, and evapotranspiration Maxwell, and M. Maneta, 2019: Impacts of the 2017 flash anomalies. Part II: Intraseasonal drought intensification fore- drought in the US northern plains informed by satellite-based casts. J. Hydrometeor., 18,1963–1982, https://doi.org/10.1175/ evapotranspiration and solar-induced fluorescence. Environ. JHM-D-16-0067.1. Res. Lett., 14, 074019, https://doi.org/10.1088/1748-9326/ab22c3. }}, }}, M. Svoboda, C. R. Hain, M. C. Anderson, and Y. Hobbins, M. T., A. Wood, D. J. McEvoy, J. L. Huntington, C. Zhong, 2017b: Predicting U.S. drought monitor states using Morton, M. Anderson, and C. Hain, 2016: The evaporative precipitation, soil moisture, and evapotranspiration anoma- demand drought index. Part I: Linking drought evolution to lies. Part I: Development of a nondiscrete USDM index. J. variations in evaporative demand. J. Hydrometeor., 17,1745– Hydrometeor., 18, 1943–1962, https://doi.org/10.1175/JHM-D- 1761, https://doi.org/10.1175/JHM-D-15-0121.1. 16-0066.1. Hoell, A., and Coauthors, 2019a: Anthropogenic contributions to Mahto, S. S., and V. Mishra, 2020: Dominance of summer mon- the intensity of the 2017 United States Northern Great Plains soon flash droughts in India. Environ. Res. Lett., 15, 104061, drought. Bull. Amer. Meteor. Soc., 100,S19–S24, https://doi. https://doi.org/10.1088/1748-9326/abaf1d. org/10.1175/BAMS-D-18-0127.1. Mallya, G., L. Zhao, X. C. Song, D. Niyogi, and R. S. Govindar- }}, J. Perlwitz, and J. Eischeid, 2019b: The causes, predictabil- aju, 2013: 2012 Midwest drought in the United States. J. ity, and historical context of the 2017 U.S. Northern Great Hydrol. Eng., 18,737–745, https://doi.org/10.1061/(ASCE)HE. Plains drought. National Integrated Drought Information 1943-5584.0000786. Mo, K. C., and D. P. Lettenmaier, 2015: Heat wave flash droughts System, 25 pp., https://www.drought.gov/documents/causes- in decline. Geophys. Res. Lett., 42, 2823–2829, https://doi.org/ predictability-and-historical-context-2017-us-northern-great- 10.1002/2015GL064018. plains-drought. }},and }}, 2016: Precipitation deficit flash droughts over the Hoerling, M., S. Schubert, and K. C. Mo, 2013: An interpretation United States. J. Hydrometeor., 17,1169–1184, https://doi.org/ of the origins of the 2012 Central Great Plains drought 10.1175/JHM-D-15-0158.1. assessment report. 50 pp., NOAA, https://psl.noaa.gov/csi/ Nguyen, H., M. C. Wheeler, J. A. Otkin, T. Cowan, A. Frost, and factsheets/pdf/noaa-gp-drought-assessment-report.pdf R. Stone, 2019: Using the evaporative stress index to monitor }}, J. Eischeid, A. Kumar, R. Leung, A. Mariotti, K. Mo, S. flash drought in Australia. Environ. Res. Lett., 14, 064016, Schubert, and R. Seager, 2014: Causes and predictability of https://doi.org/10.1088/1748-9326/ab2103. the 2012 Great Plains drought. Bull. Amer. Meteor. Soc., 95, Osman, M., B. F. Zaitchik, H. S. Badr, J. I. Christian, T. Tadesse, 269–282, https://doi.org/10.1175/BAMS-D-13-00055.1. J. A. Otkin, and M. C. Anderson, 2021: Flash drought onset Hoffmann, D., A. J. E. Gallant, and M. Hobbins, 2021: Flash over the contiguous United States: Sensitivity of inventories drought in CMIP5 models. J. Hydrometeor., 22, 1439–1454, and trends to quantitative definitions. Hydrol. Earth Syst. https://doi.org/10.1175/JHM-D-20-0262.1. Sci., 25, 565–581, https://doi.org/10.5194/hess-25-565-2021. Hunt, E. D., M. Svoboda, B. Wardlow, K. Hubbard, M. Hayes, Otkin, J. A.,M. C.Anderson, C. Hain,I. E.Mladenova, J. B. and T. Arkebauer, 2014: Monitoring the effects of rapid Basara, and M. Svoboda, 2013: Examining rapid onset onset of drought on non-irrigated maize with agronomic data drought development using the thermal infrared–based evap- and climate-based drought indices. Agric. For. Meteor., 191, orative stress index. J. Hydrometeor., 14,1057–1074, https:// 1–11, https://doi.org/10.1016/j.agrformet.2014.02.001. doi.org/10.1175/JHM-D-12-0144.1. Jencso, K., and Coauthors, 2019: Flash drought: Lessons learned }}, }}, }},and }}, 2014: Examining the relationship from the 2017 drought across the U.S. northern plains and between drought development and rapid changes in the evap- Canadian prairies. National Integrated Drought Information orative stress index. J. Hydrometeor., 15, 938–956, https://doi. System, 72 pp., https://www.drought.gov/documents/flash- org/10.1175/JHM-D-13-0110.1. drought-lessons-learned-2017-drought-across-us-northern- }}, }}, }},and }}, 2015a: Using temporal changes in plains-and-canadian. drought indices to generate probabilistic drought intensifica- Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year tion forecasts. J. Hydrometeor., 16,88–105, https://doi.org/10. Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471, 1175/JHM-D-14-0064.1. https://doi.org/10.1175/1520-0477(1996)077,0437:TNYRP.2. }}, M. Shafer, M. Svoboda, B. Wardlow, M. C. Anderson, C. 0.CO;2. Hain, and J. Basara, 2015b: Facilitating the use of drought FEBRUARY 2022 OSM A N E T A L . 289 early warning information through interactions with agricul- Thorndike, R. L., 1953: Who belongs in the family? Psychome- tural stakeholders. Bull. Amer. Meteor. Soc., 96, 1073–1078, trika, 18,267–276, https://doi.org/10.1007/BF02289263. https://doi.org/10.1175/BAMS-D-14-00219.1. Tian, D., E. F. Wood, and X. Yuan, 2017: CFSv2-based sub-sea- }}, and Coauthors, 2016: Assessing the evolution of soil mois- sonal precipitation and temperature forecast skill over the ture and vegetation conditions during the 2012 United States contiguous United States. Hydrol. Earth Syst. Sci., 21,1477– flash drought. Agric. For. Meteor., 218–219, 230–242, https:// 1490, https://doi.org/10.5194/hess-21-1477-2017. doi.org/10.1016/j.agrformet.2015.12.065. Tobin, K. J., W. T. Crow, J. Dong, and M. E. Bennett, 2019: Vali- }}, T. Haigh, A. Mucia, M. C. Anderson, and C. Hain, 2018a: dation of a new root-zone soil moisture product: Soil Comparison of agricultural stakeholder survey results and MERGE. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., drought monitoring datasets during the 2016 U.S. northern 12, 3351–3365, https://doi.org/10.1109/JSTARS.2019.2930946. plains flash drought. Wea. Climate Soc., 10, 867–883, https:// Wang,L., X. Yuan,Z. Xie,P.Wu, andY.Li, 2016:Increasing doi.org/10.1175/WCAS-D-18-0051.1. flash droughts over China during the recent global warming }}, M. Svoboda, E. D. Hunt, T. W. Ford, M. C. Anderson, C. hiatus. Sci. Rep., 6, 30571, https://doi.org/10.1038/srep30571. Hain, and J. B. Basara, 2018b: Flash droughts: A review and Wolf, S., and Coauthors, 2016: Warm spring reduced carbon cycle assessment of the challenges imposed by Rapid-onset impact of the 2012 US summer drought. Proc. Natl. Acad. droughts in the United States. Bull. Amer. Meteor. Soc., 99, Sci. USA, 113, 5880–5885, https://doi.org/10.1073/pnas. 911–919, https://doi.org/10.1175/BAMS-D-17-0149.1. }}, and Coauthors, 2021: Development of a flash drought inten- Yuan, X., L. Wang, P. Wu, P. Ji, J. Sheffield, and M. Zhang, 2019: sity index. Atmosphere, 12, 741, https://doi.org/10.3390/ Anthropogenic shift towards higher risk of flash drought over atmos12060741. China. Nat. Commun., 10, 4661, https://doi.org/10.1038/s41467- Peters, A., E. Walter-Shea, L. Ji, A. Vina,M.Hayes, and M. Svo- 019-12692-7. boda, 2002: Drought monitoring with NDVI-based standardized Zhang, M., and X. Yuan, 2020: Rapid reduction in ecosystem pro- vegetation index. Photogramm. Eng. Remote Sens., 68,71–75. ductivity caused by flash drought based on decade-long Rouse, J. W., Jr.,R.H. Haas, J. A. Schell, and D.W.Deering, FLUXNET observations. Hydrol. Earth Syst. Sci., 24,5579– 1974: Monitoring vegetation systems in the Great Plains with 5593, https://doi.org/10.5194/hess-24-5579-2020. ERTS. Goddard Space Flight Center 3d ERTS-1 Symp., Vol. Zhang, Y., T. F. Keenan, and S. Zhou, 2021: Exacerbated drought 1, Sect. A, NASA Paper A20, 9 pp. impacts on global ecosystems due to structural overshoot. Svoboda, M., and Coauthors, 2002: The Drought Monitor. Bull. Amer. Meteor. Soc., 83, 1181–1190, https://doi.org/10.1175/ Nat. Ecol. Evol., 5, 1490–1498, https://doi.org/10.1038/s41559- 1520-0477-83.8.1181. 021-01551-8. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Hydrometeorology American Meteorological Society

Diagnostic Classification of Flash Drought Events Reveals Distinct Classes of Forcings and Impacts

Loading next page...
 
/lp/american-meteorological-society/diagnostic-classification-of-flash-drought-events-reveals-distinct-6apSDCi3Nl
Publisher
American Meteorological Society
Copyright
Copyright © American Meteorological Society
ISSN
1525-7541
eISSN
1525-7541
DOI
10.1175/jhm-d-21-0134.1
Publisher site
See Article on Publisher Site

Abstract

FEBRUARY 2022 OSM A N E T A L . 275 Diagnostic Classification of Flash Drought Events Reveals Distinct Classes of Forcings and Impacts a,b a c d d MAHMOUD OSMAN, BENJAMIN F. ZAITCHIK, HAMADA S. BADR, JASON OTKIN, YAFANG ZHONG, d e f,g f h DAVID LORENZ, MARTHA ANDERSON, TREVOR F. KEENAN, DAVID L. MILLER, CHRISTOPHER HAIN, AND THOMAS HOLMES Department of Earth and Planetary Sciences, The Johns Hopkins University, Baltimore, Maryland Irrigation and Hydraulics Department, Cairo University, Cairo, Egypt Department of Civil and Systems Engineering, The Johns Hopkins University, Baltimore, Maryland Space Science and Engineering Center, Cooperative Institute for Meteorological Satellite Studies, University of Wisconsin–Madison, Madison, Wisconsin Hydrology and Remote Sensing Laboratory, Agricultural Research Service, USDA, Beltsville, Maryland Department of Environmental Science, Policy, and Management, University of California, Berkeley, Berkeley, California Earth and Environmental Sciences Area, Lawrence Berkeley National Laboratory, Berkeley, California Earth Science Office, NASA Marshall Space Flight Center, Huntsville, Alabama Hydrological Sciences Laboratory, NASA Goddard Space Flight Center, Greenbelt, Maryland (Manuscript received 14 July 2021, in final form 15 December 2021) ABSTRACT: Recent years have seen growing appreciation that rapidly intensifying flash droughts are significant climate hazards with major economic and ecological impacts. This has motivated efforts to inventory, monitor, and forecast flash drought events. Here we consider the question of whether the term “flash drought” comprises multiple distinct classes of event, which would imply that understanding and forecasting flash droughts might require more than one framework. To do this, we first extend and evaluate a soil moisture volatility–based flash drought definition that we introduced in previous work and use it to inventory the onset dates and severity of flash droughts across the contiguous United States (CONUS) for the period 1979–2018. Using this inventory, we examine meteorological and land surface conditions associated with flash drought onset and recovery. These same meteorological and land surface conditions are then used to classify the flash droughts based on precursor conditions that may represent predictable drivers of the event. We find that distinct classes of flash drought can be diagnosed in the event inventory. Specifically, we describe three classes of flash drought: “dry and demanding” events for which antecedent evaporative demand is high and soil moisture is low, “evaporative” events with more modest antecedent evaporative demand and soil moisture anomalies, but positive antecedent evaporative anomalies, and “stealth” flash droughts, which are different from the other two classes in that precursor meteorological anomalies are modest relative to the other classes. The three classes exhibit somewhat different geographic and seasonal distributions. We conclude that soil moisture flash droughts are indeed a composite of distinct types of rapidly intensifying droughts, and that flash drought analyses and forecasts would benefit from approaches that recognize the existence of multiple phenome- nological pathways. KEYWORDS: Drought; Extreme events; Hydrometeorology; Soil moisture; Climate classification/regimes 1. Introduction losses, wildfires, and economic damages in the tens of billions of dollars. These droughts occurred at different times of the In recent years, a number of rapid-onset drought events year in different climate zones with different ecological char- have struck the contiguous United States (CONUS), with acteristics, yet they have all been described as flash droughts, severe consequences for ecological and agricultural systems. a term first coined by Peters et al. (2002) and Svoboda et al. For example, droughts in the Southern Plains in 2011, the cen- (2002) to reflect the fact that some droughts emerge rapidly tral United States in 2012, the Southeast in 2016, the Northern and quickly develop into high-impact extreme events. Plains in 2017, and Texas in 2019 led to widespread crop A challenging characteristic of flash droughts is that they appear suddenly}seemingly without warning}and therefore leave farmers, ranchers, and other vulnerable stakeholders lit- Denotes content that is immediately available upon publica- tle time to prepare mitigation responses (Otkin et al. 2015b, tion as open access. 2018a; Haigh et al. 2019). The 2012 flash drought, for exam- ple, received tremendous attention because of its impact on the nation’s corn crop. Yet there was virtually no sign of an Supplemental information related to this paper is available at the Journals Online website: https://doi.org/10.1175/10.1175/JHM- impending rapid intensification prior to the event in standard D-21-0134.s1. drought monitoring products at that time or in dynamically based seasonal forecasting systems (Hoerling et al. 2014). Postevent analyses concluded that the event was largely Corresponding author: Mahmoud Osman, mahmoud.osman@ jhu.com driven by random atmospheric variability, and perhaps was DOI: 10.1175/JHM-D-21-0134.1 Ó 2022 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses). 276 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 inherently unpredictable using conventional methods (Kumar severity when determining the intensity of a flash drought. et al. 2013). Poor model performance both in forecasting and Their study showed that there are important regional differ- reproducing these events presents an additional challenge in ences in flash drought severity when both of these components efforts to project flash drought impacts and feedbacks under are considered. nonstationary climate conditions (Wolf et al. 2016). Notwith- Most proposed definitions and intensity metrics for flash standing these challenges, there is evidence that flash droughts droughts have focused exclusively on capturing the phenome- are amenable to seasonal-to-subseasonal scale prediction on non rather than assessing whether it represents a coherent account of their sensitivity to initial conditions (Lorenz et al. class from the perspective of drought process. An exception is 2017a,b), the perceived importance of forecastable drivers of the work of Mo and Lettenmaier (2015, 2016), which explic- evaporative demand during flash drought intensification (Hob- itly distinguished between precipitation deficit flash droughts bins et al. 2016), and the potentially predictable role of vegeta- and heat wave flash droughts. The method used to define tion in flash drought processes (Wolf et al. 2016). these droughts has been debated, in large part because Mo Any such generalized statements on the predictability of flash and Lettenmaier consider duration of the heatwave event droughts, however, implicitly assume that the occurrence and rather than intensification rate, which is more typically under- severity of flash droughts can be diagnosed in a consistent and stood to be the defining characteristic of flash drought (Otkin process-relevant manner, and that the term “flash drought” et al. 2018b; Lisonbee et al. 2021), but their concept that flash refers to a single class of event. In recent years, many studies droughts might be the product of multiple different pathways have sought to describe and diagnose the occurrence of flash with distinct meteorological drivers is highly relevant to droughts by proposing a variety of definitions that can be used to understanding and prediction. While Mo and Lettenmaier inventory and map flash droughts. Otkin et al. (2013, 2014, made this distinction a priori by incorporating different varia- 2015a) identified flash droughts based on rapid changes in the bles and thresholds in their definitions, we are not aware of ratio between actual evapotranspiration (EVP) and potential any study that empirically classifies different flash drought evapotranspiration (PEVP). Other studies (Hunt et al. 2014; Mo types within an inventory generated using a common flash and Lettenmaier 2015)defined flash droughts as a function of drought definition. That is: if an inventory of flash drought the rapid drop in soil moisture with time. Chen et al. (2019) sug- events is generated using a definition based on flash drought gested the degradation of two categories in the U.S. Drought phenomenology alone, are there distinct classes within that Monitor (USDM) in a period of four weeks as a definition for inventory that can be identified due to different precursors in the onset of flash droughts. Christian et al. (2019) introduced the meteorology or surface conditions? If so, that implies that definition for flash droughts based on the rate of change in stan- understanding and predicting flash droughts may require that dardized ratio between EVP and PEVP over a six-pentad (6 3 5 we adopt different perspectives for each class. days) period. Another quantitative definition (Ford and Labosier Here, we apply our recently introduced SMVI flash drought 2017) identified flash droughts as the drop of the one pentad definition (Osman et al. 2021) to address this question. First, averaged soil moisture (SM) from the 40th to 20th percentiles in we extend the SMVI presented in Osman et al. (2021) to a period of four pentads or less. A subsequent study by Hoff- include estimates of drought severity, and we compare the mann et al. (2021) followed a similar methodology with adjust- SMVI to independent vegetation and crop datasets for semi- ments to reduce the number of identified events. In a recent nal flash drought events. Next, we apply a retrospective inven- study, (Osman et al. 2021) introduced a definition based on a soil tory of flash droughts, generated using SMVI, to derive moisture volatility index (SMVI), and also compared the SMVI composites of meteorological and surface conditions in the with six other definitions to highlight the fact that there are dif- predrought, onset, and recovery phases of all flash droughts. ferent pathways to identify flash drought onset. All of the listed Finally, we perform objective classification of the flash studies focused on CONUS, but the flash drought phenomenon drought inventory on the basis of meteorological and surface has been observed in many regions across the globe (Nguyen condition precursors to identify flash drought classes relevant et al. 2019; Zhang and Yuan 2020), with a number of studies to process understanding and prediction. focusing on China and India (Wang et al. 2016; Yuan et al. 2019; Mahto and Mishra 2020). These studies have yielded additional 2. Data and methods definitions. Indeed, the need to understand the implications of different definitions has become a research question in its own We generate an inventory of soil moisture flash droughts right (Lisonbee et al. 2021). for all of CONUS over the period 1979–2018 for spring Fewer studies have attempted to quantify the severity of the through fall (March–November). SMVI is calculated using flash droughts, but informative efforts do exist. Chen et al. root zone soil moisture (RZSM) from the SMERGE dataset. (2019) and Otkin et al. (2015a) both used USDM categories to SMERGE is a hybrid daily product at 0.1258 spatial resolution diagnose and assess severity of flash droughts. Christian et al. that combines satellite-derived soil moisture estimates from (2019) used standardized evaporative stress ratio (SESR) for the European Space Agency Climate Change Initiative and both purposes, Yuan et al. (2019) used soil moisture deficit, and NLDAS-2 Noah model output for RZSM averaged from 0- to Li et al. (2020) used evapotranspiration deficit. Basedonmod- 40-cm depth (Tobin et al. 2019). The SMERGE dataset has eled soil moisture, Otkin et al. (2021) developed a flash drought been evaluated against normalized difference vegetation intensity index (FDII) that explicitly accounts both for the mag- index (NDVI) products (Rouse et al. 1974) as well as in situ nitude of the rapid intensification and the resultant drought soil moisture observations, and it has been found to be a FEBRUARY 2022 OSM A N E T A L . 277 reliable dataset for agricultural and ecological applications method’s skill to capture changes in satellite-observed vegetation (Tobin et al. 2019). greenness due to flash drought. The cloud-free NDVI data were The SMVI is motivated by the fact that flash drought diagno- obtained from the 16-day MODIS composite product sis is concerned with capturing change that is more rapid than (MOD13C1) at 0.058 spatial resolution (Didan 2021)for the usual, so that it could be used to identify both rapid onset and years 2000 to present. NDVI grid points with anomalies below rapid intensification drought events. For SMVI, rapid changes 20.5 standard deviation from the mean are defined as are identified by the crossover of simple moving averages “negatively impacted” in comparisons with SMVI. This approxi- (SMAs) combined with duration and dryness thresholds. Onset mately corresponds to a probability of occurrence less than 30% is recorded when 1) the 5-day (1-pentad) RZSM SMA falls and for normally distributed conditions. Further, we evaluate the per- stays below the 20-day (4-pentad) SMA for at least a 20-day formance of the SMVI definition for the 2012 central United period or 2) both SMAs are below the 20th percentile of the States and 2017 Northern Plains flash droughts versus in situ 1979–2018 time-of-year RZSM climatology (Osman et al. reports of soil and crop conditions collected by the USDA 2021). If two sequential flash droughts are identified with a National Agricultural Statistics Service (NASS) observers. Data period of three pentads or less between them, then they are showing poor conditions are marked as negatively impacted. combined into a single event. We do this because a short rain- These data are collected at county scale, then spatially smoothed fall event may result in a temporary reduction in the severity of to reduce noise, and protect confidentiality (access to data at a flash drought but is often not sufficient to restore predrought county level was provided to the coauthors after signing a confi- conditions and end the drought event. dentiality agreement with the USDA NASS). The performance Severity is quantified basedonRZSMdeficit during the iden- analyses are carried out for the spring and summer seasonal aver- ages due to data availability and temporal resolution. tified flash drought event according to Eqs. (1) and (2) as illus- The performance of the SMVI is assessed with hit–miss confu- trated in the example in Fig. S1 in the online supplemental sion matrices that use NDVI and NASS data as observational material. This scale is based on the standardized distribution of reference datasets. True positive values represent grid points and the integrated RZSM deficit below the 20th percentile (and pentads depicted by SMVI as being in flash drought and also over the 5-day running average) during the drought event: marked as negatively impacted by the NASS or NDVI validation t5t datasets, while false positives are the events classified as flash SV 5() RZSM 2 RZSM (1) 20th 5d drought by SMVI where NASS or NDVI do not meet drought t5t impact criteria. True negative values represent grid points not marked as negatively impacted by the NASS or NDVI validation datasets and not identified as flash drought grid points. False neg- SV SV 5 , (2) CAT atives represent grid points identified by SMVI as having no flash STD() SV drought while marked as negatively impacted by the NASS or where SV is the computed severity, and RZSM and RZSM NDVI validation datasets. Hit–miss statistics are calculated 20th 5d are the 20th percentile and 5-day moving average RZSM, according to Eqs. (3)–(10): respectively. Parameters t and t represent the times at which o f TP identified flash drought onset occurs and ends, respectively. The sensitivity() TPR 5 , (3) TP 1 FN standardized severity category is represented by SV with a CAT range from zero (no flash drought) up to 5 (maximum severity), and STD(SV ) is the severity standard deviation calcu- TN 1979–2018 () specificity TNR 5 , (4) lated from the flash drought inventory for all grid points, mea- TN 1 FP sured against the severity of all other identified flash drought events within the inventory. The use of categories to indicate FP drought severity is a common approach, as used in systems such () false discovery rate FDR 5 , (5) FP 1 TP as the USDM. In contrast to the USDM, the SMVI-based severity is intended to capture the severity of the rapid onset flash drought process. FN () false negative rate FNR 5 , (6) The end of the flash drought period (recovery period) date FN 1 TP is identified when the rate of drop in RZSM during an identi- fied flash drought event begins to recover (i.e., the 1-pentad FP () running average is no longer below 4-pentad running average) false positive rate FPR 5 , (7) FP 1 TN or the 1-pentad RZSM is no longer below the 20th percentile of the 1979–2018 time-of-year RZSM. TP SMVI performance was previously evaluated based on precision() PPV 5 , (8) TP 1 FP descriptions of reported major flash drought events (Osman et al. 2021). Influenced by the methodology followed by Peters et al. (2002) to detect drought using standardized NDVI, in this study TP 1 TN we use MODISNDVItime-of-yearanomaliestoassess the accuracy() ACC 5 , (9) TP 1 TN 1 FP 1 FN 278 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 FIG. 1. Flash drought maps as captured by SMVI definition during the active growing season (March–November): (left) 2012 and (right) 2017. (a),(b) Onset maps, where each color represents the month of flash drought onset. (c),(d) Estimated severity category maps. classes subjectively, but there are recommended diagnostics TP () critical success index CSI 5 , (10) for use in choosing the optimal number of classes. Here we TP 1 FN 1 FP apply the commonly used elbow method (Thorndike 1953) for this purpose. where TP, TN, FP, and FN represent true positive, true nega- tive, false positive, and false negative grid points, respectively. Values of Eqs. (3)–(10) range from 0 to 1, with 1 being the 3. Results and discussion perfect score for the TP or TN numerator-based ratios and a. The SMVI flash drought intensity metric the opposite for the FP and FN numerator-based ratios. Drawing on previous studies that have described meteoro- The United States was hit by several major flash drought logical and surface conditions associated with flash drought events over the past decade, resulting in excessive agricultural onset (Mo and Lettenmaier 2015, 2016; Ford and Labosier losses and livestock destruction. In 2012, the country experi- 2017; He et al. 2019; Osman et al. 2021), we select multiple enced one of the largest and most destructive flash droughts variables from the NLDAS-2 datasets (temperature, precipi- recorded to date, with more than $30 billion estimated dam- tation, RZSM, PEVP, EVP, and surface pressure) along with ages (Hoerling et al. 2013, 2014; Basara et al. 2019; Mallya the computed vapor pressure deficit (VPD) and total cloud et al. 2013; Fuchs et al. 2012; Otkin et al. 2016). A warm spring cover (TCC) from NCEP–NCAR reanalysis products (Kalnay followed by early summer heatwaves set the stage for a rap- et al. 1996), and analyze their progression through the pre- idly intensifying drought that struck much of the middle part drought, onset and end of the flash drought periods for all of the country in late spring and early summer and extended events included in the 40-yr (1979–2018) SMVI-derived flash to the north later in summer and in early fall (Fig. 1a). Nota- drought inventory. To focus on events with meaningful bly, though the occurrence of flash drought was very wide- impact, we analyze only SMVI-derived flash drought events spread (according to both SMVI and other definitions) with severity greater than 2. Unsupervised multivariate classi- (Osman et al. 2021), the central United States had the greatest fication is then performed as a function of these meteorologi- severity, as diagnosed by the SMVI (Fig. 1c). cal variables, using principal components transformation to Five years after the 2012 flash drought, the Northern Plains control for collinearity between variables. This classification is region was hit by another major flash drought, causing more used to characterize different types of flash droughts driven than $2.6 billion in agricultural losses and sparking wildfires. by different processes. The classes are determined using the The 2017 Northern Plains flash drought was focused on Mon- k-means partitioning unsupervised classification algorithm tana, North Dakota, South Dakota, and parts of Alberta and (Hartigan and Wong 1979; Lloyd 1982) as a heuristic cluster- Saskatchewan (Jencso et al. 2019). The event started in May ing method. We apply a sensitivity analysis to determine the over western Montana and swiftly intensified through high statistically optimal number of clusters. The anomalies are evaporative demand and precipitation deficits (Hoell et al. calculated as the in-time (predrought, onset, or recovery) pen- 2019a; Osman et al. 2021). The drought eventually spread tad anomaly relative to the 1979–2018 time-of-year average. over much of the Northern Plains region (Fig. 1b) causing The k-means algorithm allows the user to set the number of enormous economic losses (Gerken et al. 2018; Jencso et al. FEBRUARY 2022 OSM A N E T A L . 279 FIG.2.Maps of hit–miss analysis for the 2012 and the 2017 flash droughts during the actively growing season (March–November): (left) 2012 and (right) 2017. (a),(b) SMVI vs negative NDVI anomaly hit–miss map, in which lavender represents false positive (FP), orange rep- resents true positive (TP), white represents true negative (TN), green represents false negative (FN), and gray represents missing/unavail- able data. (c),(d) As in (a) and (b), but for NASS reported negative average crop conditions. (e),(f) As in (a) and (b), but for the observed topsoil moisture. 2019; He et al. 2019). Montana was the most impacted state With this caveat in mind, we compare the SMVI flash (Jencso et al. 2019), and this is evident in the SMVI-based drought index to MODIS NDVI anomalies and NASS crop severity analysis (Fig. 1d). The severity analysis is also consis- and topsoil condition anomalies. Using a simple hit/miss met- tent with the USDM reports that showed an exceptional (D4 ric in which negative anomalies in MODIS NDVI (more than category) drought over Montana (Jencso et al. 2019). It is 0.5 standard deviation below the mean) or the NASS condi- important to highlight that estimation of flash droughts’ sever- tion maps are interpreted as evidence of drought conditions, ity in this study is a method to relatively quantify soil moisture we find that there is broad agreement between the SMVI and deficit with a methodology similar to Yuan et al. (2019) study observed drought conditions for both the 2012 and 2017 flash given the different flash drought identification method. drought events (Figs. 1 and 2). We do see considerable false Independent, quantitative validation of drought indices is negatives on the margins of the drought-affected area, particu- notoriously difficult, since impacts of drought vary with cli- larly in 2012, but this is consistent with our liberal definition of mate context, land cover, and economic system. Since flash agricultural drought in the NDVI and NASS fields (i.e., flash drought is a subset of all droughts which is typically consid- drought identified area is smaller than NDVI and NASS nega- ered in agricultural and ecological contexts (Wang et al. 2016; tive anomalies). We also note a concentration of false positives Mo and Lettenmaier 2015; Christian et al. 2019; Otkin et al. along edge of drought regions, particularly in 2017, indicate 2018b), we consider vegetation health and crop status to be that the SMVI approach overestimated the extent of drought- two relevant indicators of drought impact that can verify the affected area relative to NASS estimates. utility of SMVI as a useful drought metric. In doing this, we Focusing on the central and northern High Plains regions recognize that the independent comparisons do not necessar- [as defined by Bukovsky (2011)] for the years 2012 and 2017, ily confirm the presence of flash drought; rather, they are respectively, we find that for flash droughts based on negative interpreted as indicators of whether an agricultural drought NDVI anomalies the accuracy was 0.68 in 2012 and 0.56 in may have occurred. 2017. Precision was higher in 2012 (0.74) than 2017 (0.50), 280 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 TABLE 1. SMVI–NASS and SMVI–NDVI summary hit–miss statistics for the 2012 central region flash drought showing the geographically dominant crops and observed soil moisture conditions. Corn Range Soybean Subsoil Topsoil Avg crop condition NDVI ACC 0.74 0.74 0.78 0.84 0.77 0.75 0.68 CSI 0.71 0.73 0.76 0.84 0.76 0.73 0.63 FDR 0.21 0.07 0.16 0.05 0.05 0.11 0.26 FNR 0.12 0.22 0.10 0.13 0.21 0.20 0.19 FPR 0.65 0.55 0.64 0.50 0.38 0.51 0.60 PPV 0.79 0.93 0.84 0.95 0.95 0.89 0.74 TNR 0.35 0.45 0.36 0.50 0.62 0.49 0.40 TPR 0.88 0.78 0.90 0.87 0.79 0.80 0.81 while the probability of detection (sensitivity) was higher in they are 0.84 and 0.95 for the 2017 event. We also note that 2017: 0.93, versus 0.81 in 2012 (Tables 1 and 2). The critical irrigation is a complicating factor that may affect comparison success index was significantly higher for the 2012 event (0.63) between datasets. While SMVI does include partial consider- compared to that observed in 2017 (0.48). These values of ation of irrigation, insomuch as SMERGE captures irrigation hit–miss statistics are consistent with moderate to strong per- signals, this representation is imperfect and might not align formance in event identification (Hoerling et al. 2013, 2014; with observed vegetation response to irrigation. Basara et al. 2019; Mallya et al. 2013; Fuchs et al. 2012; Otkin b. Proposed drivers of flash drought et al. 2016; Gerken et al. 2018; Jencso et al. 2019; He et al. 2019). It is important to note that this is an imperfect compari- Figure 3 presents composites of predrought (onset minus son. The SMVI approach is one pathway of identifying flash three pentads), onset, and recovery period conditions, using droughts, and a comparison with a vegetation index metric, composites of standardized anomalies of meteorological fields such as NDVI anomalies, is not exactly indicative of perfor- for all flash droughts of severity greater than 2 in the SMVI- mance in capturing a soil moisture flash drought. derived 1979–2018 inventory. Composites are calculated sepa- NASS-based evaluation, based on NASS identification of rately for each grid cell, such that the anomalies represent poor crop and soil conditions, led to comparable statistics for conditions when a flash drought occurred in that exact loca- each impacted region’s dominant crop (Figs. 2c–f). Tables 1 tion. Precipitation (PRCP) anomalies in the predrought and and 2 summarize SMVI–NASS statistics for both the 2012 onset periods are mostly negative, as one would expect, which and 2017 flash droughts. In the 2012 central U.S. flash is also associated with suppression of the convective available drought, SMVI shows an accuracy of 0.79, 0.75, and 0.74 for potential energy (CAPE) over most of CONUS (we include negatively impacted soybean, range, and corn, with a preci- CAPE in addition to precipitation in order to isolate local sion of 0.84, 0.79, and 0.89, respectively. The 2017 Northern convective potential as distinct from total realized rainfall). Plains flash drought captured by SMVI is similarly evaluated This is similar to the observed scenario before and during the and statistical evaluation was slightly higher than that seen for 2017 northern High Plains flash drought (Gerken et al. 2018). the NDVI analysis. Accuracy for detecting grids of flash The magnitude of these standardized anomalies, however, is drought in the Northern Plains compared to negatively generally small relative to the anomalies in RZSM and poten- impacted dominant crops (barley and spring wheat) are 0.8 tial evaporation (PEVP), particularly during the pentad of and 0.76, respectively, with precision values of 0.91 and 0.88, drought onset. and probability of detection greater than 0.84. Comparing These findings are consistent with previous studies (Otkin SMVI to the reported NASS topsoil moisture conditions et al. 2018b, 2013; Anderson et al. 2013), which have empha- shows a very similar pattern for the negatively reported condi- sized the importance of precursor soil moisture conditions and tions. The accuracy and precision of SMVI detection of the PEVP in the onset of a flash drought. Low RZSM, high PEVP reported negative NASS topsoil moisture conditions for the and high VPD conditions force the rapid transition from an 2012 flash drought event are 0.77 and 0.95, respectively, and energy limited environment to a water limited environment, TABLE 2. As in Table 1, but for the 2017 northern High Plains region flash drought. Barley Oats Spring wheat Winter wheat Subsoil Topsoil Avg crop condition NDVI ACC 0.80 0.73 0.76 0.78 0.84 0.84 0.72 0.56 CSI 0.79 0.72 0.75 0.77 0.83 0.83 0.70 0.48 FDR 0.09 0.15 0.12 0.08 0.03 0.02 0.18 0.50 FNR 0.15 0.18 0.16 0.17 0.15 0.16 0.17 0.07 FPR 0.55 0.73 0.65 0.61 0.29 0.25 0.72 0.72 PPV 0.91 0.85 0.88 0.92 0.97 0.98 0.82 0.50 TNR 0.45 0.27 0.35 0.39 0.71 0.75 0.28 0.28 TPR 0.85 0.82 0.84 0.83 0.85 0.84 0.83 0.93 FEBRUARY 2022 OSM A N E T A L . 281 FIG. 3. Composite maps of standardized anomalies of climate conditions for selected atmospheric variables (TEMP: 2-m above ground temperature, PRCP: precipitation, RZSM: root-zone soil moisture, EVP: actual evapotranspiration, PEVP: potential evapotranspiration, SPRESS: surface pressure, TCC: total cloud cover, WS: 10-m above ground wind speed, CAPE: convective available potential energy, VPD: vapor pressure deficit) based on the full SMVI flash droughts inventory from 1979 to 2018 for severity higher than 2, during onset, recovery, and onset minus 3 pentads. leading to rapid drought onset and loss of green cover (Otkin more water limited environments the EVP anomalies are neg- et al. 2018b). This elevated PEVP only leads to an increase in ative in the predrought and onset periods, as elevated PEVP actual evapotranspiration (EVP) in regions with greater water cannot translate into an increase in EVP. As described later, variability}e.g., the Midwest and Great Lakes regions. In this distinction is important when considering process-based 282 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 FIG. 4. Boxplot of the standardized anomalies of atmospheric variables and root zone soil moisture averaged for the three pentads before drought onset for each class for the full SMVI inventory from 1979 to 2018. A separate figure for each of the fields’ variability across years is shown in Fig. S4. Maps of the anomalies averaged over the three pentads prior to onset are shown in Fig. S5. flash drought classification: the concept that elevated PEVP flash drought definition; analyses that use different definitions leads to elevated EVP, drying the soil column, is an important might lead to different conclusions. That said, Ford and Lab- aspect of some theories of vegetation-mediated flash drought osier (2017) examine some of the same variables and found intensification (Otkin et al. 2018b), but it is not a feature of all broadly similar patterns using a different flash drought defini- events in our inventory. tion formulation based on the drop in RZSM from the 40th to Other potential predictor variables show regionally variable the 20th percentile in a period that does not exceed four signals. Temperature (TEMP), often identified as a driver of pentads. flash drought, is generally elevated in the predrought period, c. Flash drought classification but the anomalies are weak, and the sign of anomaly is not entirely consistent. It is only during the onset pentad that ele- The composite analysis of conditions at different stages of vated temperatures are observed over most regions (though flash droughts shown in the previous section provides a useful even then the southeast is not particularly anomalously perspective on the flash drought development process; how- warm). Surface pressure (SPRES) might be expected to be ever, it does not consider the possibility that the inventoried anomalously high in the lead-up to a drought, but the anoma- flash droughts consist of distinct forms of drought develop- lies are weak and mixed over much of the country, as is the ment. It is therefore possible that the weak or mixed anoma- average near-surface wind speed (WS). TCC tends toward lies found for certain proposed drivers are simply an artifact negative anomalies in predrought and onset periods, matching of averaging across different types of events, blurring the expectation, but again there are weak or mixed anomalies for influence of hydrometeorological drivers in different drying a number of regions. scenarios. Considering the recovery pentad, which is defined as the To test this hypothesis, we perform K-means classification first pentad in which any of the onset conditions is violated, it on our SMVI-based flash drought inventory. We use onset is evident that the role of rainfall is significant in ending the pentad standardized anomalies for the nine variables applied flash drought. Both PRCP and TCC show strong positive in composite analysis (TMP, PRCP, RZSM, EVP, PEVP, anomalies in recovery, which stands in contrast to the modest SPRES, TCC, WS, CAPE, and VPD) as the basis for classifi- anomalies seen during the predrought and onset periods. cation, and first mask out unvegetated classes (bare soil and Rain breaks the flash drought cycle, quickly switching envi- urban classes) and potentially deep-rooted vegetation classes ronmental conditions to a non-water-limited status, provided (forest and woodland classes) according to the University of that the volume of rain is sufficient. TEMP, PEVP, EVP, Maryland (UMD) Land Cover Classification (Fig. S2). Only VPD, and SPRES anomalies are mixed during the recovery events with severity greater than 2 are included in the classifi- period. RZSM anomalies are still strongly negative, reflecting cation, and we perform principal component analysis on the fact that we have defined the recovery (end of flash meteorological variables prior to classification. Using the drought period) based on the change in rate of declination or elbow method (Thorndike 1953), we find that three classes if RZSM higher than the 20th percentile, which are still below normal conditions but no longer a flash drought. It is worth are optimal (Fig. S3). We emphasize that our classification is emphasizing that these composites are based on our SMVI intended to draw out indicative patterns and is not meant to FEBRUARY 2022 OSM A N E T A L . 283 FIG. 5. Daily time series plots of selected atmospheric variables and RZSM from four pentads prior to drought onset to one pentad after onset for (a) class 1, (b) class 2, and (c) class 3 events. For each class, the time series of each variable represents an average of 20 grid cells, each selected from the core area of a separate flash drought event. The y axis shows the standard devi- ation for the normalized variables’ values. imply that the three classes are entirely separable or indepen- its negative precipitation (PRCP) anomalies are modest relative dent phenomena. The use of a different dataset of meteoro- to the other two classes. logical variables, study region, or flash droughts identification These systematic differences between classes suggests that method may lead to a different number of classes. flash droughts can be triggered by a diversity of meteorologi- The character of each class with respect to precursor soil cal conditions. Class 2 bears the most classic signatures of moisture conditions and meteorology in the pentads leading up drought, with its dry antecedent conditions, high temperature to event onset is shown in Fig. 4. Notably, classes 2 and 3 are and evaporative demand conditions, low cloud cover, and characterized by elevated air temperature (TMP) and vapor reduced total evapotranspiration. From a flash drought per- pressure deficit (VPD) prior to flash drought onset, while class spective, these can be thought of as “dry and demanding” 1 is not. And while classes 2 and 3 have similar TMP anomalies, events, in which atmospheric evaporative demand combines class 2 exhibits substantially more severe antecedent VPD than with low rainfall and dry predrought conditions to allow for class 3, as well as stronger positive potential evapotranspiration rapid intensification of already dry conditions. Notably, (PEVP) anomalies and stronger negative root zone soil mois- PEVP anomalies for these events tend to be quite high, but ture (RZSM) and total cloud cover (TCC) anomalies. Class 3, EVP anomalies are strongly negative on account of the pre- meanwhile, is the only class that shows positive anomalies in vailing dry conditions prior to drought onset. It is important antecedent actual evapotranspiration (EVP) and in CAPE, and to emphasize that our interpretation of the different classes is 284 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 based on the mean value, which adds a margin of uncertainty in classifying an identified flash drought event. Figure 5b shows composite time series of key variables for 20 grid cells picked from the core of different class 2 drought events. As indicated in these time series, TMP, VPD, and PEVP are all elevated in the four pentads before flash drought onset while EVP anomalies are consistently negative over this period. PRCP anomalies are generally negative, with some noise evi- dent in this 20 grid cell sample, while NDVI and RZSM anomalies are strongly negative even four pentads before onset date. In contrast to the classic drought character of class 2, class 3 bears some surprising features. The fact that the events inten- sify rapidly even though, on average, the antecedent PRCP anomalies are modest and CAPE is enhanced, suggest that for these events rapid drying is largely driven by evaporative demand (positive VPD and PEVP anomalies) combined with sufficient moisture access to support elevated EVP. This com- bination makes class 3 the only class to exhibit anomalies con- sistent with the hypothesis that vegetation can contribute to flash drought onset by responding to elevated temperature and evaporative demand with increased evapotranspiration, accelerating depletion of root zone soil moisture. Based on these characteristics, we term class 3 events “evaporative” flash droughts. As shown in Fig. 5c for a random sample of points from different class 3 events, PRCP anomalies are mixed, with a negative signal only evident in the 2 pentads before onset, and positive anomalies seen at longer leads and even after flash drought onset. EVP is consistently elevated before and during onset, while strongly positive TMP, VPD, and PEVP anomalies emerge only in the two pentads before onset. Interestingly, RZSM and NDVI anomalies are, on average for this sample, positive until two pentads before onset, such that the rapid decline observed just before onset leads to negative anomalies that are substantially smaller than those observed for class 2 events at date of onset. Class 1, for its part, is noteworthy for the fact that air tem- perature and evaporative demand preceding flash drought FIG. 6. Frequency (% of years with an event) for each flash onset are unremarkable compared to average conditions. Pre- drought class at each grid point for the period 1979 to 2018, based cipitation is below average in the predrought period, skies are on the SMVI flash droughts definition. relatively clear (low TCC), and convective potential is low (negative CAPE anomaly). But anomalies in all other varia- bles commonly invoked to explain the rapidity of flash to predict with precision more than a few days in advance drought intensification are modest, i.e., there is a near-zero (Tian et al. 2017). The sample time series shown in Fig. 5a temperature, PEVP and VPD predrought anomalies. In this indicates that positive anomalies in VPD and PEVP are mod- sense, class 1 flash droughts appear to be dominated by pre- est and emerge only within two pentads of onset, and TMP cipitation deficit forcing rather than evaporative demand forc- ing, placing them at a far end of the PEVP versus PRCP anomalies are essentially neutral. Interestingly, the decline in balance of flash drought forcings described by Christian et al. NDVI is dramatic for this class, suggesting that these events (2021). As described later, class 1 events are, on average, strike vegetation that is particularly sensitive to drought stress slightly less severe than other classes, but they are not always on account of vegetation type or timing. The fact that NDVI low severity events. We will refer to these events as “stealth” anomalies are strongly positive at three and four pentad leads, flash droughts in that they have characteristics that would and that negative EVP signals are not evident at longer leads, make them difficult to forecast: where classes 2 and 3 show suggests that these events might be associated with favorable meteorological drivers that might be forecasted with skill at early season growing conditions leading to structural over- extended weather to subseasonal time scales, class 1 appears shoot in vegetation (Zhang et al. 2021). to be the product almost solely of moderately dry antecedent At the national scale, 45% of all flash drought events in our soil moisture and below average rainfall, which can be difficult inventory are class 1, 31% are class 2, and 22% are class 3. FEBRUARY 2022 OSM A N E T A L . 285 FIG.7.Classification maps of the 2011, 2012, and 2017 flash drought events. But there are distinct geographic patterns for each (Fig. 6). conditions are simply too dry. The prevalence of class 1 events Class 1 events are most common in the western High Plains, in the western High Plains is less easily explained, but it is class 2 are dominant in the southern Great Plains and Texas, consistent with experience in that the iconic 2017 flash and class 3 are the most common type in the upper Midwest. drought that affected Montana and North Dakota was a nota- This is not a deterministic split}all three classes are found in bly poorly predicted event (Jencso et al. 2019; Hoell et al. all regions}but the geographic distribution aligns with expec- 2019b). tation. In the relatively humid and cool upper Midwest, one Indeed, if we map the class associations of the 2017 flash might expect that high TMP and VPD can trigger elevated drought event, along with the seminal flash drought events of EVP even when soils are somewhat dry relative to their aver- 2011 and 2012 (Fig. 7), we see that 2017 was almost entirely age state, while in the warmer and drier southern Great Plains class 1. The 2011 event, focused on Texas and Oklahoma, is those conditions are less likely to be met with increased EVP: predominantly class 2. The widespread event of 2012 is a mix 286 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 FIG. 8. Average area in each flash drought class in each month included in this study. Average is calculated for the 1979–2018 period. of class 2 and class 3, consistent with the fact that this was a meanwhile, is the most widespread drought class in all months hot event affecting a broad swath of the Great Plains and except for June, when it is brieflyexceeded byclass 2. Thefact Midwest, including a diversity of climate zones and land cover that class 1 events continue to be relatively common in late types. summer is in part a reflection of geography, since these events Seasonally, all three flash drought classes can be observed in dominate in some of the coolest portions of the analysis any month included in our analysis (March–November; Fig. 8). domain. The drivers of flash drought risk, then, appear to vary Class 2 shows a dramatic peak in June, coincident with the by both region and season, a fact that is relevant for the devel- onset of summer heat and dryness over much of the drought- opment of flash drought risk monitoring and forecasting sys- susceptible United States. Class 3 shows a similar, albeit more tems. We note that these seasonal patterns are sensitive to our muted June peak. This is the least common flash drought class inventory method, which is subject to the previously discussed on average, but in the spring it does show slightly greater total assumptions, and clustering may vary accordingly. We note that area than class 2, and the drop in area after June is dramatic. our inventory method, which includes only the first instance of This is consistent with a drought process that includes sufficient flash drought in each grid cell in each year, may slightly under- available soil moisture to support elevated EVP. Class 1, represent late season flash droughts in general, since in cases of FIG. 9. Boxplots of the flash droughts average severity categories in the three classes after fil- tering out events of severity category less than 2 (box widths are proportional to the square root of the total number of grid points in each class). FEBRUARY 2022 OSM A N E T A L . 287 two flash droughts in the same location in the same year (which We emphasize that the classes defined here are representative are rare) the second event would not be captured by our of a continuum of processes associated with flash drought devel- method. opment. We choose to work with three classes because it proved Finally, we find that all three diagnosed classes of flash to be a stable, separable, and interpretable number of classes in drought include cases of severe drought [according to our cre- our analysis, but the result does not imply that there are only ated inventory of flash droughts severity from Eqs. (1) and three pathways that can lead to flash drought, or that an event (2)], but that there are statistical differences in severity cannot exhibit a mix of properties from two or three classes. The between classes, as estimated using the SMVI severity classes contrasting meteorological and surface process signatures of the defined in this study (Fig. 9). There is a slight tendency for three classes do, however, indicate that events identified as flash greater severity in class 2, the dry and demanding droughts, drought using a reasonable definition, including events that have and the most severe events in the record are dominated by been widely reported as seminal flash droughts, represent a class 2, followed by slightly decreased average severity for diversity of onset and intensification processes. Our results sug- class 3 and class 1. The differences in severity between classes gest that recognizing this diversity is critical to advance our are statistically significant, as evaluated using a Welch’s t test, understanding and ability to predict these events. for both raw and log transformed data, and confirmed with a nonparametric Wilcoxon signed-rank test. This result empha- Acknowledgments. We sincerely thank the journal editor sizes the potential severity of flash droughts that develop and the anonymous reviewers for their constructive comments. under the combined conditions of high evaporative demand, This work was funded by the National Science Foundation low precipitation, and dry antecedent conditions. Neverthe- (NSF) Grant 1854902. less, the distribution of event severities shown in Fig. 9 makes it clear that all three classes contain severe events. This is also REFERENCES clear from our analysis of seminal flash droughts (Fig. 1). We note that Fig. 9 shows results for events filtered for severity Anderson, M. C., C. Hain, J. Otkin, X. Zhan, K. Mo, M. Svoboda, greater than 2, but that the same general pattern is observed B. Wardlow, and A. Pimstein, 2013: An intercomparison of when we do not apply a severity threshold. drought indicators based on thermal remote sensing and NLDAS-2 simulations with U.S. Drought Monitor classifica- 4. Conclusions tions. J. Hydrometeor., 14, 1035–1056, https://doi.org/10.1175/ JHM-D-12-0140.1. Flash drought has proven to be a challenging phenomenon Basara, J. B., J. I. Christian, R. A. Wakefield, J. A. Otkin, E. H. for both monitoring and prediction. These challenges have Hunt, and D. P. Brown, 2019: The evolution, propagation, been associated with the rapidly evolving nature of the events andspreadof flash drought in the central United States dur- and, perhaps, with the fact that they depend on processes that ing 2012. Environ. Res. Lett., 14, 084025, https://doi.org/10. may not be explicitly resolved, or may be poorly predicted, in 1088/1748-9326/ab2cc0. standard subseasonal-to-seasonal forecast systems. But termi- Bukovsky, M. S., 2011: Masks for the Bukovsky regionalization of nology and definitions have also been challenging (Lisonbee North America. Regional Integrated Sciences Collective, et al. 2021), and the difficulty of establishing consistent and Institute for Mathematics Applied to Geosciences, National agreed-upon definitions is also a significant contributor to Center for Atmospheric Research, accessed 20 November associated challenges in prediction. If the physical interpreta- 2021, http://www.narccap.ucar.edu/contrib/bukovsky/. tion of a flash drought inventory is not sufficiently clear, then Chen, L. G., J. Gottschalck, A. Hartman, D. Miskus, R. Tinker, it is also not clear what one is predicting with a statistical and A. Artusa, 2019: Flash drought characteristics based on model trained using that inventory, or what one is evaluating U.S. Drought Monitor. Atmosphere, 10, 498, https://doi.org/ 10.3390/atmos10090498. when considering a dynamically based forecast of an event. Christian, J. I.,J.B. Basara, J. A. Otkin, E. D. Hunt,R.A.Wake- Here, we have examined meteorological drivers associated field, P. X. Flanagan, and X. Xiao, 2019: A methodology for with events inventoried using an SMVI-based definition of flash drought identification: Application of flash drought fre- flash drought events, and then classified all events in the quency across the United States. J. Hydrometeor., 20, 833–846, inventory on the basis of precursor meteorological and sur- https://doi.org/10.1175/JHM-D-18-0198.1. face conditions. We found three classes of flash droughts in }}, }},E.D. Hunt, J. A. Otkin,J.C.Furtado, V. Mishra, our inventory based on k-means clustering. We refer to these X. Xiao, and R. M. Randall, 2021: Global distribution, trends, classes as: dry and demanding droughts, with high evaporative and drivers of flash drought occurrence. Nat. Commun., 12, demand and antecedent low soil moisture levels; evaporative 6330, https://doi.org/10.1038/s41467-021-26692-z. droughts, which initiate under conditions of high demand and Didan, K., 2021: MODIS/Terra vegetation indices 16-Day L3 when elevated evapotranspiration accelerates soil drying; and global 0.05Deg CMG V061. NASA EOSDIS Land Processes stealth droughts, which may be hard to predict due to the lack DAAC, accessed 20 November 2021, https://doi.org/10.5067/ of a clear temperature or evaporative demand signal prior to MODIS/MOD13C1.061. initiation. These classes are associated with different meteoro- Ford, T. W., and C. F. Labosier, 2017: Meteorological conditions logical variables, regional distributions, seasonality, and cli- associated with the onset of flash drought in the Eastern matic and land surface risk factors, suggesting that there are United States. Agric. For. Meteor., 247, 414–423, https://doi. distinct forms of flash drought development. org/10.1016/j.agrformet.2017.08.031. 288 J OUR N A L O F H Y D R O M E T E OR O L OGY VOLUME 23 Fuchs, B., D. Wood, and D. Ebbeka, 2012: From too much to too Kumar, A., M. Chen, M. Hoerling, and J. Eischeid, 2013: Do little: How the Central US drought of 2012 evolved out of extreme climate events require extreme forcings? Geophys. Res. Lett., 40, 3440–3445, https://doi.org/10.1002/grl.50657. one of the most devastating floods on record in 2011. Li,J., Z. Wang,X.Wu, J. Chen,S. Guo, and Z. Zhang, 2020:A National Integrated Drought Information System, 105 pp. new framework for tracking flash drought events in space Gerken, T., G. T. Bromley, B. L. Ruddell, S. Williams, and P. C. and time. Catena, 194, 104763, https://doi.org/10.1016/j.catena. Stoy, 2018: Convective suppression before and during the 2020.104763. United States Northern Great Plains flash drought of 2017. Lisonbee, J., M. Woloszyn, and M. Skumanich, 2021: Making Hydrol. Earth Syst. Sci., 22, 4155–4163, https://doi.org/10. sense of flash drought: Definitions, indicators, and where we 5194/hess-22-4155-2018. go from here. J. Appl. Serv. Climatol., 2021,1–19, https://doi. Haigh, T. R., J. A. Otkin, A. Mucia, M. Hayes, and M. E. Bur- org/10.46275/JOASC.2021.02.001. bach, 2019: Drought early warning and the timing of range Lloyd, S., 1982: Least squares quantization in PCM. IEEE Trans. managers’ drought response. Adv. Meteor., 2019, 9461513, Inf. Theory, 28, 129–137, https://doi.org/10.1109/TIT.1982. https://doi.org/10.1155/2019/9461513. Hartigan, J. A., and M. A. Wong, 1979: A K-means clustering Lorenz, D. J., J. A. Otkin, M. Svoboda, C. R. Hain, M. C. Ander- algorithm. Appl. Stat., 28, 100, https://doi.org/10.2307/2346830. son, and Y. Zhong, 2017a: Predicting the U.S. drought moni- He, M., J. S. Kimball, Y. Yi, S. Running, K. Guan, K. Jensco, B. tor using precipitation, soil moisture, and evapotranspiration Maxwell, and M. Maneta, 2019: Impacts of the 2017 flash anomalies. Part II: Intraseasonal drought intensification fore- drought in the US northern plains informed by satellite-based casts. J. Hydrometeor., 18,1963–1982, https://doi.org/10.1175/ evapotranspiration and solar-induced fluorescence. Environ. JHM-D-16-0067.1. Res. Lett., 14, 074019, https://doi.org/10.1088/1748-9326/ab22c3. }}, }}, M. Svoboda, C. R. Hain, M. C. Anderson, and Y. Hobbins, M. T., A. Wood, D. J. McEvoy, J. L. Huntington, C. Zhong, 2017b: Predicting U.S. drought monitor states using Morton, M. Anderson, and C. Hain, 2016: The evaporative precipitation, soil moisture, and evapotranspiration anoma- demand drought index. Part I: Linking drought evolution to lies. Part I: Development of a nondiscrete USDM index. J. variations in evaporative demand. J. Hydrometeor., 17,1745– Hydrometeor., 18, 1943–1962, https://doi.org/10.1175/JHM-D- 1761, https://doi.org/10.1175/JHM-D-15-0121.1. 16-0066.1. Hoell, A., and Coauthors, 2019a: Anthropogenic contributions to Mahto, S. S., and V. Mishra, 2020: Dominance of summer mon- the intensity of the 2017 United States Northern Great Plains soon flash droughts in India. Environ. Res. Lett., 15, 104061, drought. Bull. Amer. Meteor. Soc., 100,S19–S24, https://doi. https://doi.org/10.1088/1748-9326/abaf1d. org/10.1175/BAMS-D-18-0127.1. Mallya, G., L. Zhao, X. C. Song, D. Niyogi, and R. S. Govindar- }}, J. Perlwitz, and J. Eischeid, 2019b: The causes, predictabil- aju, 2013: 2012 Midwest drought in the United States. J. ity, and historical context of the 2017 U.S. Northern Great Hydrol. Eng., 18,737–745, https://doi.org/10.1061/(ASCE)HE. Plains drought. National Integrated Drought Information 1943-5584.0000786. Mo, K. C., and D. P. Lettenmaier, 2015: Heat wave flash droughts System, 25 pp., https://www.drought.gov/documents/causes- in decline. Geophys. Res. Lett., 42, 2823–2829, https://doi.org/ predictability-and-historical-context-2017-us-northern-great- 10.1002/2015GL064018. plains-drought. }},and }}, 2016: Precipitation deficit flash droughts over the Hoerling, M., S. Schubert, and K. C. Mo, 2013: An interpretation United States. J. Hydrometeor., 17,1169–1184, https://doi.org/ of the origins of the 2012 Central Great Plains drought 10.1175/JHM-D-15-0158.1. assessment report. 50 pp., NOAA, https://psl.noaa.gov/csi/ Nguyen, H., M. C. Wheeler, J. A. Otkin, T. Cowan, A. Frost, and factsheets/pdf/noaa-gp-drought-assessment-report.pdf R. Stone, 2019: Using the evaporative stress index to monitor }}, J. Eischeid, A. Kumar, R. Leung, A. Mariotti, K. Mo, S. flash drought in Australia. Environ. Res. Lett., 14, 064016, Schubert, and R. Seager, 2014: Causes and predictability of https://doi.org/10.1088/1748-9326/ab2103. the 2012 Great Plains drought. Bull. Amer. Meteor. Soc., 95, Osman, M., B. F. Zaitchik, H. S. Badr, J. I. Christian, T. Tadesse, 269–282, https://doi.org/10.1175/BAMS-D-13-00055.1. J. A. Otkin, and M. C. Anderson, 2021: Flash drought onset Hoffmann, D., A. J. E. Gallant, and M. Hobbins, 2021: Flash over the contiguous United States: Sensitivity of inventories drought in CMIP5 models. J. Hydrometeor., 22, 1439–1454, and trends to quantitative definitions. Hydrol. Earth Syst. https://doi.org/10.1175/JHM-D-20-0262.1. Sci., 25, 565–581, https://doi.org/10.5194/hess-25-565-2021. Hunt, E. D., M. Svoboda, B. Wardlow, K. Hubbard, M. Hayes, Otkin, J. A.,M. C.Anderson, C. Hain,I. E.Mladenova, J. B. and T. Arkebauer, 2014: Monitoring the effects of rapid Basara, and M. Svoboda, 2013: Examining rapid onset onset of drought on non-irrigated maize with agronomic data drought development using the thermal infrared–based evap- and climate-based drought indices. Agric. For. Meteor., 191, orative stress index. J. Hydrometeor., 14,1057–1074, https:// 1–11, https://doi.org/10.1016/j.agrformet.2014.02.001. doi.org/10.1175/JHM-D-12-0144.1. Jencso, K., and Coauthors, 2019: Flash drought: Lessons learned }}, }}, }},and }}, 2014: Examining the relationship from the 2017 drought across the U.S. northern plains and between drought development and rapid changes in the evap- Canadian prairies. National Integrated Drought Information orative stress index. J. Hydrometeor., 15, 938–956, https://doi. System, 72 pp., https://www.drought.gov/documents/flash- org/10.1175/JHM-D-13-0110.1. drought-lessons-learned-2017-drought-across-us-northern- }}, }}, }},and }}, 2015a: Using temporal changes in plains-and-canadian. drought indices to generate probabilistic drought intensifica- Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year tion forecasts. J. Hydrometeor., 16,88–105, https://doi.org/10. Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471, 1175/JHM-D-14-0064.1. https://doi.org/10.1175/1520-0477(1996)077,0437:TNYRP.2. }}, M. Shafer, M. Svoboda, B. Wardlow, M. C. Anderson, C. 0.CO;2. Hain, and J. Basara, 2015b: Facilitating the use of drought FEBRUARY 2022 OSM A N E T A L . 289 early warning information through interactions with agricul- Thorndike, R. L., 1953: Who belongs in the family? Psychome- tural stakeholders. Bull. Amer. Meteor. Soc., 96, 1073–1078, trika, 18,267–276, https://doi.org/10.1007/BF02289263. https://doi.org/10.1175/BAMS-D-14-00219.1. Tian, D., E. F. Wood, and X. Yuan, 2017: CFSv2-based sub-sea- }}, and Coauthors, 2016: Assessing the evolution of soil mois- sonal precipitation and temperature forecast skill over the ture and vegetation conditions during the 2012 United States contiguous United States. Hydrol. Earth Syst. Sci., 21,1477– flash drought. Agric. For. Meteor., 218–219, 230–242, https:// 1490, https://doi.org/10.5194/hess-21-1477-2017. doi.org/10.1016/j.agrformet.2015.12.065. Tobin, K. J., W. T. Crow, J. Dong, and M. E. Bennett, 2019: Vali- }}, T. Haigh, A. Mucia, M. C. Anderson, and C. Hain, 2018a: dation of a new root-zone soil moisture product: Soil Comparison of agricultural stakeholder survey results and MERGE. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., drought monitoring datasets during the 2016 U.S. northern 12, 3351–3365, https://doi.org/10.1109/JSTARS.2019.2930946. plains flash drought. Wea. Climate Soc., 10, 867–883, https:// Wang,L., X. Yuan,Z. Xie,P.Wu, andY.Li, 2016:Increasing doi.org/10.1175/WCAS-D-18-0051.1. flash droughts over China during the recent global warming }}, M. Svoboda, E. D. Hunt, T. W. Ford, M. C. Anderson, C. hiatus. Sci. Rep., 6, 30571, https://doi.org/10.1038/srep30571. Hain, and J. B. Basara, 2018b: Flash droughts: A review and Wolf, S., and Coauthors, 2016: Warm spring reduced carbon cycle assessment of the challenges imposed by Rapid-onset impact of the 2012 US summer drought. Proc. Natl. Acad. droughts in the United States. Bull. Amer. Meteor. Soc., 99, Sci. USA, 113, 5880–5885, https://doi.org/10.1073/pnas. 911–919, https://doi.org/10.1175/BAMS-D-17-0149.1. }}, and Coauthors, 2021: Development of a flash drought inten- Yuan, X., L. Wang, P. Wu, P. Ji, J. Sheffield, and M. Zhang, 2019: sity index. Atmosphere, 12, 741, https://doi.org/10.3390/ Anthropogenic shift towards higher risk of flash drought over atmos12060741. China. Nat. Commun., 10, 4661, https://doi.org/10.1038/s41467- Peters, A., E. Walter-Shea, L. Ji, A. Vina,M.Hayes, and M. Svo- 019-12692-7. boda, 2002: Drought monitoring with NDVI-based standardized Zhang, M., and X. Yuan, 2020: Rapid reduction in ecosystem pro- vegetation index. Photogramm. Eng. Remote Sens., 68,71–75. ductivity caused by flash drought based on decade-long Rouse, J. W., Jr.,R.H. Haas, J. A. Schell, and D.W.Deering, FLUXNET observations. Hydrol. Earth Syst. Sci., 24,5579– 1974: Monitoring vegetation systems in the Great Plains with 5593, https://doi.org/10.5194/hess-24-5579-2020. ERTS. Goddard Space Flight Center 3d ERTS-1 Symp., Vol. Zhang, Y., T. F. Keenan, and S. Zhou, 2021: Exacerbated drought 1, Sect. A, NASA Paper A20, 9 pp. impacts on global ecosystems due to structural overshoot. Svoboda, M., and Coauthors, 2002: The Drought Monitor. Bull. Amer. Meteor. Soc., 83, 1181–1190, https://doi.org/10.1175/ Nat. Ecol. Evol., 5, 1490–1498, https://doi.org/10.1038/s41559- 1520-0477-83.8.1181. 021-01551-8.

Journal

Journal of HydrometeorologyAmerican Meteorological Society

Published: Feb 25, 2022

References