Flash droughts present a new challenge for subseasonal-to-seasonal
prediction
Angeline G. Pendergrass*
1
, Gerald A. Meehl
1
, R oger Pulwarty
2
, Mike Hobbins
3,4
, Andrew
Hoell
4
, Amir AghaKouchak
5
, C éline J.W. B onfils
6
, Ailie J. E . G allant
7,8
, M artin Hoerling
4
,
David Hoffmann
7,8
, Laurna Kaatz
9
, F lavio Lehner
1
, Dagmar Llewellyn
10
, P hilip Mote
11
, R ichard
Neale
1
, Jonathan T. Overpeck
12
, Amanda Sheffield
13
, K erstin Stahl
14
, M ark Svoboda
15
, Matthew
C. Wheeler
16
, Andrew W. Wood
1
, Connie A. Woodhouse
17
1
National Center for Atmospheric Research, Boulder, CO, U S
2
National Oceanographic and Atmospheric Administration, US
3
Cooperative Institute for Research in Environmental Sciences, University of Colorado Boulder,
Boulder, CO, U S
4
NOAA/Earth Systems Research Laboratory/Physical Sciences Division, Boulder, C O, US
5
Department of Civil and Environmental Engineering, and Department of Earth System Science,
University of California, I rvine, CA, U S
6
Program for Climate Model Diagnosis and Intercomparison, Lawrence Livermore National
Laboratory, Livermore, C A, U S
7
School of Earth, Atmosphere and Environment, Monash University, Clayton, Australia
8
ARC Centre of Excellence for Climate Extremes, M onash University, Clayton, A ustralia
9
Denver Water, Denver, CO, U S
10
Albuquerque Area Office, B ureau of Reclamation, Albuquerque, NM, US
11
Oregon Climate Change Research Institute, and Graduate School, Oregon State University,
Corvallis, OR, US
12
School for Environment and Sustainability, University of Michigan, A nn Arbor, M I, U S
13
NOAA/NIDIS, S cripps Institution of Oceanography, La Jolla, CA, U S
14
University of Freiburg, F reiburg, Germany
15
National Drought Mitigation Center, U niversity of Nebraska–Lincoln, Lincoln, NE, US
16
Australian Bureau of Meteorology, Melbourne, Victoria, Australia
17
School of Geography & Development, U niversity of Arizona, T ucson, AZ, U S
Submitted to Nature Climate Change
20 December 2019
*Corresponding author
address:
Angeline
G. P
endergrass, P
.O. B
ox
3000,
Boulder, C
O
80307. E
mail:
apgrass@ucar.edu
1
Flash droughts are a recently recognized type of extreme event distinguished by sudden
onset and rapid intensification of drought conditions with severe impacts. They unfold on
subseasonal to seasonal (S2S) timescales (weeks to months), p resenting a new challenge for
the surge of interest in improving S2S prediction. Here, w e discuss existing prediction
capability for flash droughts and what is needed to establish their predictability. We place
them in the context of synoptic to centennial phenomena, consider how they could be
incorporated into early warning systems and risk management, an d propose two
definitions. T he growing awareness that flash droughts involve particular processes and
severe impacts, and likely a climate change dimension, make them a compelling frontier for
research, monitoring, an d prediction.
Drought is perhaps the most complex and least understood of all “weather and climate
extremes”
1
. Drought can span timescales from a few weeks to decades, and spatial scales from a
few kilometers to entire regions. T heir impacts usually develop slowly, are often indirect and can
linger for long after the end of the drought itself. T he drought risk, therefore, i s often
underestimated and continues to remain a “hidden” hazard
2
. A comprehensive overview of
traditional drought characteristics, processes, mechanisms, and impacts is provided in Ref. 3.
In a future warmer climate droughts are likely to increase in duration and intensity in many
regions of the world
4,5
. A better understanding of drought phenomena, e specially of the physical
processes leading to drought, t heir propagation through the hydrological cycle, the societal and
environmental vulnerability to drought and its wide-ranging impacts, is more important than
2
3
ever. The key challenge is to move from a re-active society responding to impacts to a pro-activ
society that is resilient and adapted to drought risk, i.e. adopts proactive risk management
strategies
3,6
.
Droughts whose impacts arise in part from their long duration, such as the Dust Bowl and the
2011-2015 California drought, have formed strong imagery in the US, and megadroughts lasting
more than 20 years have also been documented in tree-ring records. Much research has been
conducted on aspects of drought that play out over multiple years, but more recently attention
has been drawn to the rapid development of some drought events, in the space of a few weeks -
flash droughts – a specific definition for which we will provide below. These events,
distinguished by their sudden onset and rapid intensification, can have severe impacts
7,8
. Flash
droughts develop on the subseasonal-to-seasonal (S2S) timescale (weeks to months), and presen
a new challenge for prediction efforts on that timescale, which are currently surging in interest
9
.
One flash drought that brought attention to the phenomenon occurred in the US Midwest in
2012
8,10
(Figure 1). The areal extent of abnormally dry conditions expanded from 30% of the
Continental United States (CONUS) in May 2012 to over 60% by August. This event had
significant impacts for agriculture and water-borne transportation in the region. While other
rapidly developing droughts had been identified before
11
, the widespread impacts of the 2012
event caught the attention of the US public and leadership. Flash drought is not confined to the
US
12
. Processes that can produce flash droughts are foci of research in China
13,14
. In southern
Queensland, Australia, a flash drought in early 2018 de-vegetated the landscape and drove
livestock numbers to their lowest level in a century, a significant impact for agriculture
15
.
e
t
4
A drought monitoring and early-warning system is the foundation of effective proactive drought
policy because it enables notice of potential and impending drought conditions. It identifies climate
and water-resource trends and detects the emergence or probability of occurrence and the likely
severity of droughts and their impacts. Reliable information must be communicated in a timely
manner to water and land managers, policy makers and the public through appropriate
communication channels to trigger actions documented in a drought plan, which is particularly
critical for flash droughts. That information, if used effectively, can form the basis for reducing
vulnerability and improving mitigation and response capacities of people and systems at risk.
In this perspective, we build on a recent review of flash droughts
8
and discuss the observational
and predictive skill of key processes with an eye towards impact assessment and early warning
of flash drought. We highlight the current understanding of the physical processes that can drive
flash droughts, the existing capabilities to predict them, and what is needed to make progress to
establish the predictability and effective early warning of flash droughts on S2S timescales.
Following earlier suggestions of possible definitions for flash droughts
8,16
, we propose, for
consideration by the community, two quantitative definitions for flash drought that can be used
for applications related to operations, analysis of observations, model simulations of present and
future climate, and assessing S2S initialized-model predictions.
5
Figure 1. Evolution of a flash drought across the US Midwest in 2012. (a-d) Evaporative
Drought (ED) categories based on two-week Evaporative Demand Drought Index (EDDI) at
five-week intervals during the drought onset. (e-h) US Drought Monitor (USDM). Adapted from
Ref. 17. (i) Percent of High Plains region in USDM categories from 1 June- 3 July 2012.
Drought development in the Midwest, 2012
August 7 July 3 June 5 May 1
2-week EDDI US Drought Monitor
Drought developing across
region
Flash drought (including ED3,
ED4) in MO, AR, KS, IL
Persistent intense drought in
region
Persistent intense drought in
region, ED4 area decreasing
D0 in IL, IN, TN; no drought in
MO, AR, OK, NE
Drought expands in region, but
not in intensity
D3 edges into region
D3, D4 emerge over much of
region two months after EDDI
US Drought Monitor
intensity categories
D0 Abnormally dry
D1 Moderate drought
D2 Severe drought
D3 Extreme drought
D4 Exceptional
drought
EDDI categories and
percentile bounds
ED0 70-80
ED1 80-90
ED2 90-95
ED3 95-98
ED4 98+
June June June July
80%
60%
High plains DM
June
12
June
26
June
1
July
3
100%
50%
0
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
12 261
3
100%
40%
20%
0
(a)
(b)
(c)
(d)
Figure 2. US Northern Great Plains flash drought in May 2017. (a) Soil moisture percentile
from the top 1 m from University of Washington simulation of the Variable Infiltration Capacity
(VIC) model
18
forced by an estimate of the time-varying meteorology
19,20
; climatological and
2017 (b) precipitation and (c) daily maximum temperature from a collection of GHCN-D stations
depicted as departures from the long-term climatology (solid black lines). (d) Rank of
accumulated May-July precipitation relative to the 1895-2017 climatology. The timeseries are
data averaged over eastern Montana (demarcated by the dotted line in the right panel). Adapted
from Ref. 21.
1. Physical processes that produce flash drought
To illustrate the physical processes involved with producing a flash drought, we consider another
recent flash drought, in the US Northern Great Plains in 2017. This event shows some recurring
flash-drought characteristics, including precipitation deficit and above-average temperatures
preceding or coinciding with a rapid soil moisture decline (Fig. 2). Precipitation deficits began
6
7
before April, when precipitation would climatologically increase. Soil moisture was nonetheless
high in April, but continued precipitation deficits throughout the month eroded it slowly at first,
before a rapid decline in May.
Figure 3. The response of evaporative demand and evapotranspiration to feedbacks from drying
land. Schematic evolution of evaporative demand (E
0
), evapotranspiration (ET), and surface
moisture availability, starting from a wet (energy-limited, left side) state and developing into a
dry (water-limited, right side) state. Figure adapted from Ref. 22.
We can examine the physical processes driving land surface moisture balance to understand
mechanisms that can lead to rapid drought intensification
23,24
. Moisture flux into the surface is
driven by precipitation. Like other types of drought, precipitation deficit often plays an important
Energy limited
ET ~ E
0
Moisture limited
E
0
> ET
Drought development in time
Evaporation rates (ET and E
0
)
Surface moisture availability
1
0
Surface moisture
availability
Evaporative
demand (E
0
)
Evapotranspiration (ET)
Feedback from
drying land
surface raises E
0
8
role
25
. Moisture flux away from the surface - evapotranspiration (ET) - can also play an
important role in flash drought, driving feedbacks between the land and atmosphere. An
important concept is the demand for moisture from the atmosphere - evaporative demand, whic
is the amount of evaporation that would occur given an unlimited supply of moisture.
Evaporative demand can be thought of as the “thirst” of the atmosphere. It both drives and
responds to ET. Starting from a state with sufficient soil moisture (energy-limited conditions;
Fig. 3), evaporative demand and evaporation vary together - when evaporative demand increase
evaporation follows. With enough evaporation and no replenishment, surface moisture
eventually becomes insufficient to supply further water for evaporation; water becomes the
limiting factor. Under water-limited conditions, further increases in evaporation can no longer
continue, and evaporation decreases. If the same factors that had been driving increases in
evaporative demand persist, then evaporative demand will diverge from evaporation. Meanwhil
sensible heat flux increases instead of evaporation, which increases near surface air temperature
and vapor pressure deficit, and thus also evaporative demand - an amplifying feedback
26–28
.
While much of the focus on flash droughts has been in humid regions, flash droughts and their
impacts are also a concern in semi-arid and arid regions where evaporative demand usually
exceeds evapotranspiration (locations that start on the right side of Fig. 3; Section 5). Starting
from a dry, moisture-limited state, flash droughts in arid regions can be driven by precipitation
deficits, and amplification of warm air temperatures by sensible heat flux feedbacks is also of
concern.
h
s,
e,
9
The local moisture imbalance during flash drought is conditioned by large-scale atmospheric
circulation. The large-scale circulation can modify the frequency and intensity of precipitation,
and it can increase evaporative demand by reducing cloud cover (which increases incoming solar
radiation at the surface), increasing wind speeds and/or increasing temperatures
16,29,30
. In the
midlatitudes in summer, this can involve a persistent “blocking” pattern, with a strong quasi-
stationary ridge of positive geopotential height anomalies and associated anomalously high
surface pressure
16
.
Large-scale atmospheric circulation associated with flash droughts can vary from one event to
the next and between different regions. While moisture-bearing storms were largely absent
during the 2012 US Midwest flash drought, the atmospheric circulation during the event varied
from one month to the next
29
. For the southern US Great Plains, the atmospheric circulation
associated with rapid declines in soil moisture in conjunction with precipitation deficits can be
different from the atmospheric circulation associated with rapid declines in soil moisture in
conjunction with heat waves
30
.
Flash droughts may be triggered or exacerbated by compound extreme events - extremes of
multiple factors that occur simultaneously
31
. A classic example would be an extreme deficit of
precipitation coinciding with a heat wave, such as occurred in southern Queensland in January
2018
15
. If these are superimposed on more slowly evolving factors, like a building soil moisture
deficit, rapid onset or intensification of drought conditions can result.
10
Vegetation type can also influence flash drought through its mediating role in transpiration.
Trees become moisture-stressed over the course of long-term drought, while in contrast, crops
and pasture can be moisture-stressed much more quickly, and might be more sensitive to
moisture in the upper soil layer.
2. The challenge of drought for S2S prediction
Compared to slowly-evolving droughts, the relatively fast development timescale of flash
droughts requires different approaches to monitoring and prediction. Many drought prediction
and monitoring products are updated at monthly or at most weekly timescales. Given a flash
drought’s onset timescale of only a few weeks, these are not sufficient. Instead, products that
update daily are required. This provides an opportunity to leverage synoptic weather forecasts in
combination with seasonal forecasting efforts that have recently become available at shorter
recently, such as the SubX system
32
.
Prediction efforts focused on flash drought are currently in their infancy. One key challenge is
skillfully forecasting precipitation deficit on the S2S timescale. However, for a successful flash
drought prediction, more is needed than just a forecast of deficient precipitation. Prediction skill
is also required of other potential ingredients of rapid drought onset and intensification: high
temperatures, low humidity, strong winds, and excess solar insolation. In the 2012 US Midwest
event, high temperatures and precipitation deficits may have been driven by a blocking high,
while the substantial soil moisture deficit may have been due to anomalous seasonal circulation
associated with La Niña
33
. For other flash drought events and locations, other processes and
phenomena likely contribute to or affect development, such as land-atmospheric interaction, the
Madden-Julian Oscillation (MJO), the Southern and Northern Annular Modes (SAM, NAM),
11
and the Indian Ocean Dipole (IOD). Each of these have been argued to provide or influence
predictability of surface-climate variables on timescales relevant for flash drought
34–37
and are
fundamental to the prospects of S2S prediction
38
.
Global coupled prediction systems show some S2S skill for precipitation
39–41
and temperature
41–
44
. Seasonal forecasts of evaporative demand are more skillful than for precipitation over the
continental US
45
, and at least as skillful globally
46
; though skill for extreme conditions on
subseasonal timescales, which may be more relevant for flash drought, has not been established.
Predictions are only as accurate as the models that make them. In the case of global climate
models, which are the primary tool for S2S prediction systems, there are significant biases.
For example, a challenge for US flash drought prediction is a summertime dry and warm bias
over the central US in many models
47
. Another key factor for prediction is the fidelity of
teleconnections; some models have biased MJO teleconnections
48
that could play a role in flash
drought predictions. Furthermore, land surface models underestimate characteristics of
evaporative drought
49
.
Establishing predictability and credibility of predictions present considerable challenges. One
aspect is the number of past flash droughts needed to build up a large enough set of samples to
test hindcast efficacy. One property of a flash drought is that it is an unusual event. If the
expected return period were more than a year, then testing predictability using hindcasts would
require at least 20 years of hindcasts; this is more than is available for some current operational
S2S prediction systems
50
. Achieving this will be a continuing challenge with limited computing
12
resources that face competing demands from increased model resolution, ensemble size, and the
number and complexity of physical processes.
Other challenges for flash drought prediction lie in our ability to monitor the current state of the
land surface and soil, and to use this information to initialize forecast models. The initial state of
the soil moisture profile is expected to have greater impact on S2S predictions than on shorter or
longer timescales
51
. Despite recent improvement, accurate monitoring of soil moisture is still
poor compared to many meteorological variables. Perhaps even further afield from operational
systems, but still of potential importance, are interactions between vegetation and the land
surface. Dynamic vegetation models (such as ecodemographic models
52
) are becoming available,
but initializing them in an operational context will present another challenge.
3. Context within longer-timescale droughts and climate change
The factors driving flash drought can change with climate variability and change on longer
timescales, but only a few studies have examined observed regional trends in flash drought
(using varied definitions)
53–55
so how they could be affected by different climatic background
states remains unclear. In this section we consider the context within which flash droughts occur,
and how climate background states, multi-decadal variability, and climate change can influence
flash droughts.
Human influence has been identified on various aspects of hydroclimate, including droughts
56,57
;
external forcing that drives anthropogenic climate change will significantly change the
background climate state as we move further into the 21
st
century. Future changes to
precipitation, temperature and atmospheric circulation will all induce changes to surface water
13
availability and evaporative demand
58
and would thus affect flash drought. Aridity, defined in
terms of evaporative demand, increases in many drought-prone regions in climate-change
projections
59
, and also influences soil moisture
60
. But how evaporative demand is formulated -
via temperature-dependent measures like the Palmer Drought Severity Index (PDSI), versus
more comprehensive measures - can alter its projected response
61–64
. Actual evaporation and its
changes are mediated by vegetation and growing season length, which can counter or exacerbate
increasing evaporative demand
65–68
. How these changes in aridity, evaporation, and land-
atmosphere feedbacks
69
affect flash drought should be a research priority.
Flash droughts can manifest as discrete drought episodes (e.g., the 2012 US Midwest
drought
29,70
), but they may also manifest as a rapid increase in severity from a longer-term
drought already in progress. If they are not terminated, they may continue into a period of
longer-lasting drought (e.g., the 2018 eastern Australian drought
15
). Flash droughts can be
embedded within climate variability occurring at decadal and longer timescales; the
characteristics of the more slowly varying climate will influence the impact of a flash drought.
Centuries-long records of climate from paleoclimatic data are useful
58
for understanding how
short, severe droughts that might have developed rapidly are distributed over longer timescales
and under a variety of climate conditions.
While temporal resolution of even the highest-quality paleoclimatic data is insufficient to capture
subseasonal timescales, these records can nonetheless provide insights on the frequency and
distribution of extreme single year or multi-year drought events. In particular, annually-resolved
tree-ring reconstructions of streamflow for the Colorado River document these extreme
14
occurrences over the past 1200 years under varying baseline climates
71
. If a “paleo flash
drought” is defined as a year with flow less than two standard deviations below average, then it
is possible to identify and characterize periods during which paleo flash droughts occur. For
example, in the medieval period (900-1300 CE), characterized by persistent droughts and
temperatures warmer than during any period until the last few decades in southwestern North
America
72
, the mid-12th century period of persistent drought had no occurrences of flash drought
years in Colorado River streamflow
71
. The 13th century, which was also dry but less persistently
so, contained two flash drought years. However, the most notable cluster of flash drought years
occurred between 1495-1506, a 12-year period during which four flash droughts occurred; this
was not a period of particularly persistent drought. Similar behavior can be found in streamflow
reconstruction of the Upper Rio Grande
73
, where only one of three flash drought clusters in a
record over 500 years long was associated with persistent drought conditions. Furthermore, paleo
reconstructions contain years where runoff efficiency is lower than expected from annual
streamflow alone
74,75
. Notwithstanding reconstruction uncertainties, such years could have
harbored flash droughts that affected runoff efficiency while leaving little imprint on annual
streamflow.
Slowly varying or changing background states present an additional challenge for S2S prediction
of flash drought since the climatic base state can alter S2S predictive skill. For example, a
decline in the predictive skill of Eastern Pacific El Niños in the early 21
st
century has been
attributed to a change in the background state of the tropical Pacific
76
. Potential changes in S2S
skill as the climate baseline evolves need further investigation.
15
4. Proposed definition
We have seen that the physical processes leading to flash drought are a matter of ongoing
research; we will see that specific impacts of flash drought are too. To facilitate the identification
of flash droughts, we adopt three principles to describe them that are broadly consistent with
previously proposed definitions
8,16
and that lend themselves to analysis yet remain useful for
monitoring and prediction. Then, we apply these principles to propose two specific quantitative
definitions of flash drought: one for US operations with the US Drought Monitor
(https://droughtmonitor.unl.edu/AboutUSDM/WhatIsTheUSDM.aspx), and another that can be
used globally for analysis of observations and climate models.
The first principle is that the event should involve a rapid onset and intensification, as
emphasized previously
8
. To adequately reflect the rapid onset rate, the onset period should be
short enough to distinguish flash droughts from the general population of droughts,
encompassing the upper tail of the distribution of events. The second principle is that the
intensification rate should be high, as advocated previously
8
. The third principle is that the event
should end in a state severe enough to qualify as drought. These principles should apply across
drought types, sectors, regions, and seasons, and not only be adapted to definitions based on
different variables (precipitation or drought indices) but also offer broad guidelines for the
development of specific flash-drought definitions. An additional principle that would be
desirable is that the event should have impacts to qualify as a flash drought, but this requires
more work to quantitatively document past and potential future impacts.
16
Next, we propose two quantitative definitions of sub-seasonal flash drought that encompass
some of the principles outlined above, using the 2012 US Midwest event as guidance. These
definitions follow from recommendations made previously
8
and are designed to be quantitative
measures than can be evaluated in the context of past flash droughts, used operationally, and also
applied to model simulations analyses and forecast evaluations. Their purpose is not to make
further prescriptions, but rather to provide concrete definitions for scrutiny and analysis by the
community. The eventual goal is to arrive at quantitative, usable definitions – whether these or a
revision.
The first definition is based on Evaporative Demand Drought Index (EDDI,
https://www.esrl.noaa.gov/psd/eddi/), which is an experimental drought monitoring and early-
warning guidance tool based on how anomalous the evaporative demand is for a given location.
The caveat accompanying EDDI is that for a flash drought to develop, the enhanced atmospheric
demand should not be compensated by increased precipitation
8
. The second definition, useful
only for US operations and based on a previous proposal
8
, relies on the US Drought Monitor
(USDM) and can be applied in near real-time for early warning applications:
1--Flash drought (applications: international operations, prediction, research): 50% increase in
EDDI (toward drying) over two weeks, sustained for at least another two weeks
2--Flash drought (application: US operations): Two-category change in the USDM in two weeks,
sustained for at least another two weeks
17
Regarding the second definition above, the USDM is a weekly operational product based on
multiple inputs from observations (e.g., weather, climate, hydrology) and empirical input from
regional observers and expert judgement evaluations from the team of scientists (authors) who
curate the USDM
7
. A caveat limiting application of this definition to the US is that the USDM
involves expert judgement, beyond raw input of observational data, and hence its flash drought
definition cannot be directly applied outside of the US operational setting, although drought
monitors in other countries could also be used (e.g., https://droughtwatch.eu). Even with that
caveat, since the USDM is familiar and widely referenced by stakeholders and other users,
basing a flash-drought index on the USDM categories would have a readily applicable
operational utility not possible from other indices. Following from conditions experienced in
recent events (Fig. 1 and 2), the rapidity of onset of flash drought conditions is reflected by
requiring a two-category change in the drought monitor in a two-week period. Impacts can
emerge on the timescale of weeks during a flash drought, so this definition requires the two-
category change in the drought monitor index to be sustained for at least another two weeks after
it is established. This definition includes no prescription beyond this four-week period - the event
could persist beyond that time or it could terminate. For example, during the US Midwest 2012
event (Fig. 1), 45% of the High Plains went from D0 (“Abnormally dry”) to D2 (“Severe
drought”) between June 12 and June 26, a two category change in the USDM in two weeks.
A more general flash drought definition is the first one listed above, which can be used for
international operations, prediction, analysis of observations and climate model output, research
into future projections, and applications to periods prior to the USDM. This general definition is
based on EDDI, which is multi-scalar and can be calculated at 1-week through 12-month
18
timescales and can capture drying dynamics that operate at the timescales of flash droughts.
EDDI provides information on the emergence and persistence of anomalous evaporative demand
in a region. The rapid onset characteristic is reflected in the EDDI-based definition by requiring
an increase in EDDI of 50 percentiles (toward drying) over two weeks, which must then be
sustained for at least the next two weeks. Again returning to the guidance for this definition
provided by the US 2012 event (Fig. 1, left), there are large areas of the US that experienced at
least a 50% change in the EDDI from June 5 to July 3.
Related to changes in EDDI are changes in soil moisture. The spatial pattern of the frequency of
occurrence of 40-, 50-, and 60-percentile decreases in soil moisture during a 20-day period over
approximately the last 100 years are shown in Figure 4. Large variations in soil moisture are
common over the wettest areas (east of the Mississippi River). Ideally EDDI would see a
comparable change over these periods. In cases where the anomalously dry conditions persist
beyond the initial week, and result in sufficiently dry conditions, these events would qualify as
flash drought, though the threshold for sufficiently dry remains to be assessed.
19
Figure 4. Frequency of different drought intensification rates. Frequency of soil moisture
decreases exceeding 40
th
-, 50
th
-, and 60
th
-percentile thresholds over four pentads for a 100-year
period (1916-2017). Soil moisture is from the same VIC simulations as Fig. 2.
A phenomenon related to but separate from flash drought is rapid-intensification snow drought,
which occurs when snowpack has a sudden and fast decline. These are of particular concern for
regions that rely on snowpack for water supply and power generation. A rapid-intensification
snow drought can be induced, for example, by dust-on-snow, rain-on-snow, or anomalously
warm temperatures
77
. Other processes that drive a rapid decrease in snowpack could also include
advection or sublimation events, for example due to high winds. Because of the cross-timescale
interactions between snowpack loss and impacts, and the substantial differences in processes
from the flash droughts discussed above, we propose that rapid intensification snow drought
should be considered separately from flash drought. Nonetheless, due to its impacts, rapid-
intensification snow drought also requires attention
78
.
The next steps are to test and apply these definitions retrospectively, to verify that they
appropriately encompass events generally described as flash drought events, that they are
(a) (b) (c)
20
sufficiently rare that they describe unusual events, and that they describe events that are
impactful in one or more dimensions. Extending the definition to require that impacts occur
would require quantifying those impacts; this could be addressed by extending the definitions.
Further refinement of flash drought definitions may also be useful for specific regions, seasons,
sectors, and drought types, using criteria of sufficient intensification rate, impact, and rarity
8
.
That said, these definitions are designed as proposals to elicit discussion in the community over
their appropriateness and applicability. It is expected that they would be fine-tuned in the future.
5. Impacts-based early warning
Impacts particular to flash drought arise from its rapid and intense development. Because
drought response plans developed by communities and governments are often designed around
slower-onset events which unfold over the course of months, rapid onset and intensification have
the potential to inhibit the initial response – there may be less time than what is allocated to
prepare or implement mitigation measures. In the 2012 US Midwest flash drought, during the
May-July growing season, dry weather dominated the agricultural areas in the Central Plains and
Midwest. Several states had record dry seasons: Arkansas (April-June and other seasons), Kansas
(May-July), Nebraska (June-August and other seasons), and South Dakota (July-September).
Impacts included, but were not limited to, the reduction in crop yields and commerce-related
activities on major river systems. The Mississippi River had water levels that went below 2-m
depth, and was closed to navigation three times with less loads carried, barges running aground,
slower speeds, and increased dredging costs. The US summer drought of 2012 also contributed
to unusually high acreage burned by wildfires. The 2017 Northern Great Plains flash drought
also brought wildfire and affected water resources and agriculture
21
.
21
Severity of drought impacts are not only aggravated by other climatic factors, such as high
temperatures, high winds, and low relative humidity, but also by the timing (i.e., principal season
of occurrence, delays in the start of the rainy season, and occurrence of rains in relation to
principal crop-growth stages) and effectiveness of the rains (i.e., rainfall intensity and number of
rainfall events)
3
. Other impacts may be associated with hazards that compound drought, such as
heatwaves, wildfires, and soil erosion. These may induce public-health effects of heat stress or
air quality degradation due to forest fires. Water quality may degrade, affecting aquatic habitats.
Depletion of water storage, low river flows and associated consequences for water supply
systems and hydropower production can occur with flash drought, though perhaps with some
delays. The recreation sector could feel impacts from wildfire as well as low river flows. This
impact- and sector-specific vulnerability to flash droughts requires more in-depth investigation,
especially as buffers against drought impacts (such as water storage, or grain/feed stores for
livestock agriculture) are used up more quickly than for slower-onset drought.
Even though much of the focus of flash drought has been on humid regions where agriculture is
a primary activity, impacts are also keenly felt in arid and semi-arid regions. A baseline
environment that is already water stressed leaves arid regions more vulnerable to drought with
less buffer until impacts are felt. For example, a flash drought could deplete reservoirs, affecting
both water availability and hydropower generation capacity in places like the Southwest US,
where water is highly managed. Because some physical mechanisms (Section 1) and impacted
systems will differ from humid regions, understanding and predicting flash droughts to provide
early warning in arid regions presents an additional challenge.
22
Overall, some types of flash drought-related impacts will present different challenges from
slower-onset drought. An accelerated "time to impact" from the onset of a meteorological
drought also means that forecasting gains importance compared to monitoring (which remains
important, but not sufficient) in operational drought early warning and risk management.
Furthermore, translating drought development into mitigation action, and predicting the
likelihood of termination versus continuation into long-term drought, are also important. A
systematic assessment of where and when (in terms of seasonal timing) vulnerability to flash
drought is highest is needed in order to guide efforts on where prediction and early warning
would be most useful.
Early warning can enable communities to prepare for impacts. The United Nations office for
Disaster Risk Reduction (UNDRR) has established four key areas of people-centered early
warning: risk knowledge, monitoring and warning, communication, and response capability.
Early warning systems in such contexts are needed not only for event onset, at which a threshold
above some socially acceptable or safe level is exceeded, but also for intensification and
duration
79
. The phrase “early warning information system” can be used to describe an integrated
process of risk assessment, communication, and decision support, of which an early warning is a
central output. An early warning information system involves much more than development and
dissemination of a forecast; it is the systematic collection and analysis of relevant information
about, and coming from, areas of impending risk that (1) informs the development of strategic
responses to anticipate crises and crisis evolution, (2) provides capabilities for generating
problem-specific risk assessments and scenarios, and (3) effectively communicates options to
critical actors for the purposes of decision-making, preparedness, and mitigation
79
.
23
In summary, with improved monitoring and credible S2S timescale predictions, drought early
warning could include flash drought. For risk management before, during and after flash drought
events, improvements in monitoring and also predicting not just onset of flash drought but
termination of events would be beneficial.
6. Ethics of practice in drought research and applications
The ultimate goal of research on flash drought, like many impactful environmental phenomena,
is to avoid or decrease the negative effects of drought on individuals and communities.
Inequalities influence the ability of communities to cope and adapt to disasters
80
. Across the
early warning and response continuum lie three cross-cutting elements: capacity-building,
governance, and gender and social inclusion. These elements are best served through a focus on
procedural justice and the resulting ethics of participation
81,82
. Effective information-based
services engage affected people and multiple perspectives in the development of knowledge, in
decision-making, and as recipients of policies
83,84
. Identifying and understanding how flash
drought and other climate impacts affect communities and individuals requires integrating local
knowledge about impacts. This is knowledge that is inclusive of many different types of
individuals in each community, including people who can successfully and meaningfully engage
with those affected in the research and monitoring process. People from many different identities
are underrepresented in the environmental science workforce; one well-documented example is
women. Women in many parts of the world are at greater risk of harm due to climate-related
disasters
80
, and yet they remain underrepresented among one influential set of climate scientists -
IPCC authors
85
. Improving diversity of the scientific workforce and taking an inclusive approach
to engaging with stakeholders, while remaining mindful of those that are not included, is
24
essential to ethical research on weather and climate in general and droughts, including flash
drought, in particular.
The following objectives are suggested to support the ethical practice of drought research:
- Enhance engagement between users and researchers
- Develop capacity in the segment of the work dedicated to being an interface with
stakeholders and users
- Support individual actions to improve scientific culture
- Make institutional efforts to change the culture of science and its reward system
- Collaborate on interdisciplinary work
- Share research outcomes with society, users, stakeholders
7. Future directions in flash drought research and monitoring
Key areas where progress on flash droughts could be made include improved understanding of
events in the recent and more distant past and their impacts; establishing predictability and
improving prediction of flash drought events; applying these predictions to improve early
warning systems for impending events as well as responding to events as they unfold; and
understanding how flash drought will respond to climate variability and change. We identify
some key challenges and directions for achieving this below.
In order to identify developing flash drought events, monitoring systems must attend to shorter
timescales and more frequent updates than are needed to capture slower, longer-term drought
events. Products that are only updated monthly (including, for example, the North American
Multi-Model Ensemble, NMME
86
) are not very useful for flash-drought monitoring and
25
prediction. Some countries have drought monitoring and Drought Early Warning Systems. In
countries with less monitoring and prediction infrastructure, there is also potential to leverage
systems that provide global hydrological information, such as the Global Flood Awareness
System (GloFAS)
87
, World-Wide Hydrological Predictions for the Environment (HYPE)
88
,
experimental Global Drought Information Systems (GDIS), Global Drought Observatory (GDO),
and Integrated Drought Management Programme (IDMP).
There remain open questions about how to define flash drought. One challenge for identifying
flash drought events is their wide variation in spatial scale. What areal extent is sufficient to
assert that a flash drought is occurring? Assessment of regions and times of year with high
sensitivity to or preponderance for flash drought should also be factored into its identification;
model representation of land use and its change can play a role as well. A better understanding of
flash droughts requires more in-depth research on relevant compound and cascading physical
processes that can trigger or increase the likelihood of a flash drought. These include
relationships among soil moisture, land-atmosphere interactions, their connections to large-scale
meteorological conditions (and precursor conditions), and how these are forced by remote SST
patterns and influenced by internal atmospheric variability. Furthermore, research is needed into
how these conditions will change as the climate base state changes
58,89
, and to incorporate the
changing climate into the definition of flash drought so that flash drought definitions remain
meaningful in the future.
Prediction systems focus mostly on physical quantities like precipitation, but the motivation and
ultimate goal of flash drought monitoring and prediction is to provide as much anticipatory
26
information as possible of impending impacts of flash drought events, and aid response during
and after those events. This requires engagement with relevant stakeholders, building capacity,
establishing ethical practices of research to document potential impacts of flash drought and
when and where these are a concern, and identifying relationships between flash drought
indicators and impacts. Such efforts should cross disciplines and engage researchers and decision
makers at all stages that bridge the weather-climate continuum.
Acknowledgements. The perspectives in this manuscript emerged from an Aspen Global Change
Institute (AGCI) workshop in September 2018; we would like to thank all participants
(https://www.agci.org/event/18s4). This material is based upon work supported by the National Center for
Atmospheric Research, which is a major facility sponsored by the National Science Foundation (NSF)
under Cooperative Agreement No. 1947282. Portions of this study were supported by the Regional and
Global Model Analysis (RGMA) component of the Earth and Environmental System Modeling Program
of the U.S. Department of Energy's Office of Biological & Environmental Research (BER) via NSF IA
1844590. M.H. was supported by a National Oceanic and Atmospheric Administration (NOAA) Joint
Technology Transfer Initiative (JTTI) award and a U.S. Agency for International Development (USAID)-
Famine Early Warning Systems Network (FEWS NET) award (# NA17OAR4320101). C.J.W.B. was
supported by an Early-Mid Career LLNL Laboratory Directed Research and Development award
(tracking code 17-ERD-115) under the auspices of the U.S. Department of Energy by Lawrence
Livermore National Laboratory under Contract DE-AC52-07NA27344. A.J.E.G is supported by
Australian Research Council Discovery Early Career Researcher Award DE150101297. M.C.W. was
partially supported by the Northern Australia Climate Program (NACP), funded by Meat and Livestock
Australia, the Queensland Government, and the University of Southern Queensland. F.L. is also
supported by NSF AGS-0856145 Amendment 87, and by the Bureau of Reclamation under Cooperative
Agreement R16AC00039.
27
Corresponding author. Correspondence and requests for materials should be addressed to Angeline G.
Pendergrass, apgrass@ucar.edu.
Author contributions. A.G.P., G.M., and R.P., coordinated the text and organized the workshop. M.H.
and A.H. provided figures. All authors except A.A. participated in discussions at the workshop, and all
authors contributed to the writing.
Competing Interests. The authors declare that they have no competing financial interests.
Data availability. EDDI is available for the CONUS at
ftp://ftp.cdc.noaa.gov/Projects/EDDI/CONUS_archive and for the globe at
ftp://ftp.cdc.noaa.gov/Projects/EDDI/global_archive. The bottom panel of Fig. 1 is generated from the
USDM (droughtmonitor.unl.edu). The data analyzed in Figs. 2 and 4 are available from
ftp://ftp.cdc.noaa.gov/pub/Public/jeischeid/andy/. The data to generate Fig. 4 is available at
github.com/apendergrass/flashdroughtperspectivefigure.
Code availability. The bottom panel of Fig. 1 is generated from the USDM (droughtmonitor.unl.edu).
Figs. 2 and 4 were generated following the protocol
ftp://192.12.137.7/pub/dcp/archive/OBS/livneh2014.1_16deg/. The code to generate Fig. 4 is available at
github.com/apendergrass/flashdroughtperspectivefigure.
References
1. S. Pulwarty, R. & Sivakumar, M. V. K. Information systems in a changing climate: Early
warnings and drought risk management. Weather Clim. Extrem. 3, 14–21 (2014).
2. UNDRR. Global Assessment Report on Disaster Risk Reduction. (2019).
3. Wilhite, D. A. & Pulwarty, R. S. Drought and Water Crises: Lessons Drawn, Some
Lessons Learned, and the Road Ahead. in Drought and Water Crises: Integrating Science,
28
Management, and Policy (eds. Wilhite, D. & Pulwarty, R. S.) (CRC Press, 2017).
doi:10.1201/b22009
4. Christensen, J. et al. Regional Climate Projections. in Climate Change, 2007: The
Physical Science Basis. Contribution of Working group I to the Fourth Assessment Report
of the Intergovernmental Panel on Climate Change (University Press, 2007).
5. Seneviratne, S. I. et al. Changes in climate extremes and their impacts on the natural
physical environment. in Managing the Risks of Extreme Events and Disasters to Advance
Climate Change Adaptation (eds. Field, C. B. et al.) 109–230 (A Special Report of
Working Groups I and II of the Intergovernmental Panel on ClimateChange (IPCC).
Cambridge University Press, 2012).
6. Wilhite, D. A., Sivakumar, M. V. K. & Pulwarty, R. Managing drought risk in a changing
climate: The role of national drought policy. Weather Clim. Extrem. 3, 4–13 (2014).
7. Svoboda, M. et al. The Drought Monitor. Bull. Am. Meteorol. Soc. 83, 1181–1190 (2002).
8. Otkin, J. A. et al. Flash Droughts: A Review and Assessment of the Challenges Imposed
by Rapid-Onset Droughts in the United States. Bull. Am. Meteorol. Soc. 99, 911–919
(2018).
9. Robertson, A. W. et al. Improving and Promoting Subseasonal to Seasonal Prediction.
Bull. Am. Meteorol. Soc. 96, ES49–ES53 (2015).
10. Hoerling, M. P. et al. Is a Transition to Semipermanent Drought Conditions Imminent in
the U.S. Great Plains? J. Clim. 25, 8380–8386 (2012).
11. Namias, J. Anatomy of Great Plains Protracted Heat Waves (especially the 1980 U.S.
summer drought). Mon. Weather Rev. 110, 824–838 (1982).
12. Yuan, X., Wang, L. & Wood, E. F. 17. Anthropogenic intensification of southern African
flash droughts as exemplified by the 2015/16 season. Bull. Am. Meteorol. Soc. 99, S86–
S90 (2018).
13. Yuan, X., Ma, Z., Pan, M. & Shi, C. Microwave remote sensing of short-term droughts
during crop growing seasons. Geophys. Res. Lett. 42, 4394–4401 (2015).
14. Li, Y. et al. Mechanisms and Early Warning of Drought Disasters: Experimental Drought
Meteorology Research over China. Bull. Am. Meteorol. Soc. 100, 673–687 (2019).
15. Nguyen, H. et al. Using evaporative stress index to monitor flash drought in Australia.
Environ. Res. Lett. (2019). doi:10.1088/1748-9326/ab2103
16. Ford, T. W. & Labosier, C. F. Meteorological conditions associated with the onset of flash
drought in the Eastern United States. Agric. For. Meteorol. 247, 414–423 (2017).
17. Hobbins, M. T., McEvoy, D. J. & Hain, C. R. Evapotranspiration, evaporative demand,
and drought. in Drought and Water Crises: Integrating Science, Management, and Policy
(eds. Wilhite, D. A. & Pulwarty, R. S.) (CRC Press, 2017).
18. Liang, X., Lettenmaier, D. P., Wood, E. F. & Burges, S. J. A simple hydrologically based
model of land surface water and energy fluxes for general circulation models. J. Geophys.
Res. 99, 14415 (1994).
19. Livneh, B. et al. A long-term hydrologically based dataset of land surface fluxes and
states for the conterminous United States: Update and extensions. J. Clim. 26, 9384–9392
(2013).
20. Livneh, B. et al. A spatially comprehensive, hydrometeorological data set for Mexico, the
U.S., and Southern Canada 1950-2013. Sci. data 2, 150042 (2015).
21. Hoell, A., Perlwitz, J. & Eischeid, J. Drought assessment report: The causes,
predictability, and historical context of the 2017 US Northern Great Plains Drought.
29
(2019).
22. Lukas, J., Hobbins, M. T., Rangwala, I. & EDDI Team, T. The EDDI User Guide. (2017).
23. Hobbins, M. T., Ramírez, J. A. & Brown, T. C. Trends in pan evaporation and actual
evapotranspiration across the conterminous U.S.: Paradoxical or complementary?
Geophys. Res. Lett. 31, n/a-n/a (2004).
24. Ramírez, J. A., Hobbins, M. T. & Brown, T. C. Observational evidence of the
complementary relationship in regional evaporation lends strong support for Bouchet’s
hypothesis. Geophys. Res. Lett. 32, L15401 (2005).
25. Koster, R. D. et al. Flash Drought as Captured by Reanalysis Data: Disentangling the
Contributions of Precipitation Deficit and Excess Evapotranspiration. J. Hydrometeorol.
JHM-D-18-0242.1 (2019). doi:10.1175/JHM-D-18-0242.1
26. Seneviratne, S. I., Lüthi, D., Litschi, M. & Schär, C. Land–atmosphere coupling and
climate change in Europe. Nature 443, 205–209 (2006).
27. Fischer, E. M., Seneviratne, S. I., Lüthi, D. & Schär, C. Contribution of land-atmosphere
coupling to recent European summer heat waves. Geophys. Res. Lett. 34, L06707 (2007).
28. Su, H., Yang, Z.-L., Dickinson, R. E. & Wei, J. Spring soil moisture-precipitation
feedback in the Southern Great Plains: How is it related to large-scale atmospheric
conditions? Geophys. Res. Lett. 41, 1283–1289 (2014).
29. Hoerling, M. et al. Causes and Predictability of the 2012 Great Plains Drought. Bull. Am.
Meteorol. Soc. 95, 269–282 (2014).
30. Mo, K. C. & Lettenmaier, D. P. Precipitation Deficit Flash Droughts over the United
States. J. Hydrometeorol. 17, 1169–1184 (2016).
31. Chiang, F., Mazdiyasni, O. & AghaKouchak, A. Amplified warming of droughts in
southern United States in observations and model simulations. Sci. Adv. 4, eaat2380
(2018).
32. Pegion, K. et al. The Subseasonal Experiment (SubX): A multi-model subseasonal
prediction experiment. Bull. Am. Meteorol. Soc. BAMS-D-18-0270.1 (2019).
doi:10.1175/BAMS-D-18-0270.1
33. Chen, L. G. et al. Flash Drought Characteristics Based on U.S. Drought Monitor.
Atmosphere (Basel). 10, 498 (2019).
34. Dirmeyer, P. A., Gentine, P., Ek, M. B. & Balsamo, G. Land Surface Processes Relevant
to Sub-seasonal to Seasonal (S2S) Prediction. Sub-Seasonal to Seas. Predict. 165–181
(2019). doi:10.1016/B978-0-12-811714-9.00008-5
35. Waliser, D. E. et al. Potential Predictability of the Madden–Julian Oscillation. Bull. Am.
Meteorol. Soc. 84, 33–50 (2003).
36. Hendon, H. H. et al. Australian Rainfall and Surface Temperature Variations Associated
with the Southern Hemisphere Annular Mode. J. Clim. 20, 2452–2467 (2007).
37. Zhao, M. & Hendon, H. H. Representation and prediction of the Indian Ocean dipole in
the POAMA seasonal forecast model. Q. J. R. Meteorol. Soc. 135, 337–352 (2009).
38. National Academies of Sciences, E. Next generation earth system prediction : strategies
for subseasonal to seasonal forecasts.
39. Zhu, H. et al. Seamless Precipitation Prediction Skill in the Tropics and Extratropics from
a Global Model. Mon. Weather Rev. 142, 1556–1569 (2014).
40. Wheeler, M. C., Zhu, H., Sobel, A. H., Hudson, D. & Vitart, F. Seamless precipitation
prediction skill comparison between two global models. Q. J. R. Meteorol. Soc. 143, 374–
383 (2017).
30
41. Wang, L. & Robertson, A. W. Week 3–4 predictability over the United States assessed
from two operational ensemble prediction systems. Clim. Dyn. 52, 5861–5875 (2019).
42. Hudson, D. et al. Forewarned is Forearmed: Extended-Range Forecast Guidance of
Recent Extreme Heat Events in Australia. Weather Forecast. 31, 697–711 (2016).
43. Vitart, F. & Robertson, A. W. The sub-seasonal to seasonal prediction project (S2S) and
the prediction of extreme events. npj Clim. Atmos. Sci. 1, 3 (2018).
44. Lehner, F. et al. Mitigating the Impacts of Climate Nonstationarity on Seasonal
Streamflow Predictability in the U.S. Southwest. Geophys. Res. Lett. 44, 12,208-12,217
(2017).
45. McEvoy, D. J. et al. The Evaporative Demand Drought Index. Part II: CONUS-Wide
Assessment against Common Drought Indicators. J. Hydrometeorol. 17, 1763–1779
(2016).
46. Shukla, S. et al. Examining the Value of Global Seasonal Reference Evapotranspiration
Forecasts to Support FEWS NET’s Food Insecurity Outlooks. J. Appl. Meteorol. Climatol.
56, 2941–2949 (2017).
47. Zhang, C. et al. CAUSES: Diagnosis of the Summertime Warm Bias in CMIP5 Climate
Models at the ARM Southern Great Plains Site. J. Geophys. Res. Atmos. 123, 2968–2992
(2018).
48. Vitart, F. Madden-Julian Oscillation prediction and teleconnections in the S2S database.
Q. J. R. Meteorol. Soc. 143, 2210–2220 (2017).
49. Ukkola, A. M. et al. Land surface models systematically overestimate the intensity,
duration and magnitude of seasonal-scale evaporative droughts. Environ. Res. Lett. 11,
(2016).
50. Vitart, F. et al. The Subseasonal to Seasonal (S2S) Prediction Project Database. Bull. Am.
Meteorol. Soc. 98, 163–173 (2017).
51. Koster, R. D. et al. Contribution of land surface initialization to subseasonal forecast skill:
First results from a multi-model experiment. Geophys. Res. Lett. 37, n/a-n/a (2010).
52. Fisher, R. A. et al. Vegetation demographics in Earth System Models: A review of
progress and priorities. Glob. Chang. Biol. 24, 35–54 (2018).
53. Mo, K. C. & Lettenmaier, D. P. Heat wave flash droughts in decline. Geophys. Res. Lett.
42, 2823–2829 (2015).
54. Wang, L., Yuan, X., Xie, Z., Wu, P. & Li, Y. Increasing flash droughts over China during
the recent global warming hiatus. Sci. Rep. 6, (2016).
55. Zhang, Y., You, Q., Chen, C. & Li, X. Flash droughts in a typical humid and subtropical
basin: A case study in the Gan River Basin, China. J. Hydrol. 551, 162–176 (2017).
56. Barnett, T. P. et al. Human-induced changes in the hydrology of the Western United
States. Science (80-. ). 319, 1080–1083 (2008).
57. Marvel, K. et al. Twentieth-century hydroclimate changes consistent with human
influence. Nature 569, 59–65 (2019).
58. Cook, B. I., Mankin, J. S. & Anchukaitis, K. J. Climate Change and Drought: From Past to
Future. Current Climate Change Reports 4, 164–179 (2018).
59. Seager, R. et al. Model projections of an imminent transition to a more arid climate in
southwestern North America. Science 316, 1181–4 (2007).
60. Zhou, S. et al. Land–atmosphere feedbacks exacerbate concurrent soil drought and
atmospheric aridity. Proc. Natl. Acad. Sci. U. S. A. 116, 18848–18853 (2019).
61. Roderick, M. L., Greve, P. & Farquhar, G. D. On the assessment of aridity with changes
31
in atmospheric CO2. Water Resour. Res. 51, 5450–5463 (2015).
62. Milly, P. C. D. & Dunne, K. A. Potential evapotranspiration and continental drying. Nat.
Clim. Chang. 6, 946–949 (2016).
63. Feng, S. et al. Why Do Different Drought Indices Show Distinct Future Drought Risk
Outcomes in the U.S. Great Plains? J. Clim. 30, 265–278 (2017).
64. Lehner, F. et al. Projected drought risk in 1.5°C and 2°C warmer climates. Geophys. Res.
Lett. 44, 7419–7428 (2017).
65. Swann, A. L. S., Hoffman, F. M., Koven, C. D. & Randerson, J. T. Plant responses to
increasing CO2 reduce estimates of climate impacts on drought severity. Proc. Natl. Acad.
Sci. 113, 10019–10024 (2016).
66. Bonfils, C. et al. Competing Influences of Anthropogenic Warming, ENSO, and Plant
Physiology on Future Terrestrial Aridity. J. Clim. 30, 6883–6904 (2017).
67. Yang, Y., Roderick, M. L., Zhang, S., McVicar, T. R. & Donohue, R. J. Hydrologic
implications of vegetation response to elevated CO2 in climate projections. Nat. Clim.
Chang. 9, 44–48 (2019).
68. Mankin, J. S., Seager, R., Smerdon, J. E., Cook, B. I. & Williams, A. P. Mid-latitude
freshwater availability reduced by projected vegetation responses to climate change. Nat.
Geosci. (2019). doi:10.1038/s41561-019-0480-x
69. Dirmeyer, P. A. et al. Evidence for Enhanced Land–Atmosphere Feedback in a Warming
Climate. J. Hydrometeorol. 13, 981–995 (2012).
70. Otkin, J. A. et al. Assessing the evolution of soil moisture and vegetation conditions
during the 2012 United States flash drought. Agric. For. Meteorol. 218–219, 230–242
(2016).
71. Meko, D. M. et al. Medieval drought in the upper Colorado River Basin. Geophys. Res.
Lett. 34, L10705 (2007).
72. Woodhouse, C. A., Meko, D. M., MacDonald, G. M., Stahle, D. W. & Cook, E. R. A
1,200-year perspective of 21st century drought in southwestern North America. Proc.
Natl. Acad. Sci. U. S. A. 107, 21283–8 (2010).
73. Woodhouse, C., Stahle, D. & Villanueva Díaz, J. Rio Grande and Rio Conchos water
supply variability over the past 500 years. Clim. Res. 51, 147–158 (2012).
74. Woodhouse, C. A. & Pederson, G. T. Investigating Runoff Efficiency in Upper Colorado
River Streamflow Over Past Centuries. Water Resour. Res. 54, 286–300 (2018).
75. Lehner, F., Wahl, E. R., Wood, A. W., Blatchford, D. B. & Llewellyn, D. Assessing
recent declines in Upper Rio Grande runoff efficiency from a paleoclimate perspective.
Geophys. Res. Lett. 44, 4124–4133 (2017).
76. Zhao, M. et al. Weakened Eastern Pacific El Niño Predictability in the Early Twenty-First
Century. J. Clim. 29, 6805–6822 (2016).
77. Huning, L. S. & AghaKouchak, A. Mountain snowpack response to different levels of
warming. Proc. Natl. Acad. Sci. U. S. A. 115, 10932–10937 (2018).
78. Harpold, A., Dettinger, M. & Rajagopal, S. Defining Snow Drought and Why It Matters.
Eos (Washington. DC). (2017). doi:10.1029/2017EO068775
79. Pulwarty, R. S. & Verdin, J. P. Crafting integrated early warning information systems:
The case of drought. in Measuring Vulnerability to Natural Hazards: Towards Disaster
Resilient Societies Second Edition (ed. Birkmann, J.) (United Nations University Press,
2013).
80. Cutter, S. et al. Managing the risks from climate extremes at the local level. in IPCC
32
Special Report on Managing the Risks of Extremes and Disaster to Advance Climate
Change Adaptation 291–338 (Cambridge Press, 2012).
81. Shrader-Frechette, K. S. Environmental justice : creating equality, reclaiming democracy.
Environmental Ethics and Science Policy (Oxford University Press, 2002).
82. Jamieson, D. Ethics and the Environment: An Introduction. (Cambridge University Press,
2008).
83. Pulwarty, R. S. et al. El Niño Events, Forecasts and Decision-making. in Mapping
Vulnerability: Disasters, Development and People (eds. Bankoff, G. & Frerks, G.)
(Routledge, 2004). doi:10.4324/9781849771924-14
84. Allis, E. et al. The Future of Climate Services. World Meteorol. Organ. Bull. 68, (2019).
85. Gay-Antaki, M. & Liverman, D. Climate for women in climate science: Women scientists
and the Intergovernmental Panel on Climate Change. Proc. Natl. Acad. Sci. U. S. A.
201710271 (2018). doi:10.1073/pnas.1710271115
86. Kirtman, B. P. et al. The North American Multimodel Ensemble: Phase-1 Seasonal-to-
Interannual Prediction; Phase-2 toward Developing Intraseasonal Prediction. Bull. Am.
Meteorol. Soc. 95, 585–601 (2014).
87. Alfieri, L. et al. GloFAS - global ensemble streamflow forecasting and flood early
warning. Hydrol. Earth Syst. Sci. 17, 1161–1175 (2013).
88. Arheimer, B. et al. Global catchment modelling using World-Wide HYPE (WWH), open
data and stepwise parameter estimation. Hydrol. Earth Syst. Sci. Discuss. 1–34 (2019).
doi:10.5194/hess-2019-111
89. Yuan, X. et al. Anthropogenic shift towards higher risk of flash drought over China. Nat.
Commun. 10, 4661 (2019).
