Drought

Identification

1.	Indicator Description

This indicator measures drought conditions in the United States using several different indices. Drought
can affect agriculture, water supplies, energy production, and many other aspects of society. Drought
relates to climate change because rising average temperatures alter the Earth's water cycle, increasing
the overall rate of evaporation from soil and transpiration from plants. An increase in
evapotranspiration makes more water available in the air for precipitation, but contributes to drying
over some land areas, leaving less moisture in the soil. As the climate continues to change, many
historically wet areas are likely to experience an increased level of moisture and a higher risk of flooding.
Historically dry areas are likely to experience less precipitation and an increased risk of drought (Wehner
etal., 2017).

Components of this indicator include:

•	Average drought conditions in the contiguous 48 states since 1895 based on the Palmer Drought
Severity Index (Figure 1).

•	Average drought conditions in the contiguous 48 states since 1900 based on the five-year-
timescale Standardized Precipitation-Evapotranspiration Index, or SPEI (Vicente-Serrano et al.,
2010) (Figure 2).

•	Trend in the five-year SPEI in the contiguous 48 states, 1900 to 2020 (Figure 3).

•	Percent of U.S. lands classified under drought conditions in recent years, based on an index
called the U.S. Drought Monitor (Figure 4).

The data for Figure 3 are shown for climate divisions, as defined by the National Oceanic and
Atmospheric Administration (NOAA).

2.	Revision History

April 2010:	Indicator published.

December 2012: Added Figure 1, based on the Palmer Drought Severity Index. Updated indicator

with data through 2011.

August 2013:	Updated indicator with data through 2012.

May 2014:	Updated indicator with data through 2013.

June 2015:	Updated indicator with data through 2014.

August 2016:	Updated indicator with data through 2015.

April 2021:	Updated indicator with data through 2020; added Figures 2 and 3 based on the SPEI;

renumbered previous Figure 2 (Drought Monitor) as Figure 4.

Technical Documentation: Drought

1


-------
Data Sources

3.	Data Sources

Data for Figure 1 were obtained from the National Oceanic and Atmospheric Administration's (NOAA's)
National Centers for Environmental Information (NCEI), formerly National Climatic Data Center (NCDC),
which maintains a large collection of climate data online.

Data for Figures 2 and 3 were obtained from the WestWide Drought Tracker, which is a collaboration
between the University of Idaho, the Western Regional Climate Center, and the Desert Research
Institute. The WestWide Drought Tracker based all calculations on measurements and models from
Oregon State University's Parameter-elevation Regressions on Independent Slopes Model (PRISM)
Climate Mapping Program—specifically the AN81m and AN81d data sets.

Data for Figure 4 were provided by the U.S. Drought Monitor, which maintains current and archived
data at: https://droughtmonitor.unl.edu.

4.	Data Availability

Figure 1. Average Drought Conditions in the Contiguous 48 States According to the Palmer Index, 1895-
2020

NCEI's Climate at a Glance interface provides access to monthly values of the PDSI averaged across the
entire contiguous 48 states, which EPA downloaded for this indicator. These data are available at:
www.ncdc.noaa.gov/cag. This website also provides access to monthly PDSI values for nine broad
regions, individual states, and 357 smaller regions called climate divisions (each state has one to 10
climate divisions, except Alaska, which has 13). For accompanying metadata, see:
wwwl.ncdc.noaa.gov/pub/data/cirs/climdiv/divisional-readme.txt.

PDSI values are calculated from precipitation and temperature measurements collected by weather
stations within each climate division. Individual station measurements and metadata are available
through NCEI's website (www.ncdc.noaa.gov/data-access/land-based-station-data).

Figures 2 and 3. Average Drought Conditions and Average Change in Drought in the Contiguous 48 States
According to the SPEI, 1900-2020

EPA obtained the SPEI data in raster format directly from the WestWide Drought Tracker archive, at
both the five-year (i.e., "60 months") and one-year (i.e., "12 months") timescales. The data sets are
publicly available at a variety of spatial extents at: https://wrcc.dri.edu/wwdt/batchdownload.php. The
PRISM data are also made publicly available, via Oregon State University's PRISM Climate Group, by the
Northwest Alliance for Computational Science and Engineering: www.prism.oregonstate.edu. Additional
information on the SPEI can be found at: https://spei.csic.es/home.html.

Figure 4. U.S. Lands Under Drought Conditions, 2000-2020

U.S. Drought Monitor data can be obtained from:

https://droughtmonitor.unl.edu/Data/DataTables.aspx. Select "U.S. States and Puerto Rico" to view the

Technical Documentation: Drought

2


-------
historical data that were used for this indicator. For each week, the data table shows what percentage of
land area was under the following drought conditions:

1.	None

2.	D0-D4

3.	D1-D4

4.	D2-D4

5.	D3-D4

6.	D4 alone

This indicator covers the time period from 2000 to 2020. Although data were available for parts of 2021
at the time EPA last updated this indicator, EPA chose to report only full years.

Drought Monitor data are based on a wide variety of underlying sources. Some are readily available
from public websites; others might require specific database queries or assistance from the agencies
that collect and/or compile the data. For links to many of the data sources, see:
https://droughtmonitor.unl.edu/nadm/Home.aspx.

Methodology

5. Data Collection

Figure 1. Average Drought Conditions in the Contiguous 48 States According to the Palmer Index, 1895-
2020

The PDSI is calculated from daily temperature measurements and precipitation totals collected at
thousands of weather stations throughout the United States. These stations are overseen by NOAA, and
they use standard instruments to measure temperature and precipitation. Some of these stations are
automated stations operated by NOAA's National Weather Service. The remainder are Cooperative
Observer Program (COOP) stations operated by other organizations using trained observers and
equipment and procedures prescribed by NOAA. For an inventory of U.S. weather stations and
information about data collection methods, see the technical reports and peer-reviewed papers cited at:
www.ncdc.noaa.gov/data-access/land-based-station-data. and the National Weather Service technical
manuals at: www.weather.gov/coop. This indicator is derived from a specific quality-controlled subset of
long-term stations that NCEI has designated as its nClimDiv data set (www.ncdc.noaa.gov/monitoring-
references/maps/us-climate-divisions.php).

Figures 2 and 3. Average Drought Conditions and Average Change in Drought in the Contiguous 48 States
According to the SPEI, 1900-2020

The PRISM Climate Mapping Program by Oregon State University's PRISM Climate Group is an ongoing
effort to produce and disseminate accurate, high-resolution climate data sets (Daly et al., 1994). PRISM
uses point data, a digital elevation model, and other spatial data sets to generate grid-based estimates
of monthly precipitation and temperature from 1895 to the present day at a scale of about 4 kilometers
(2.5 arc-minutes). Surface stations used to collect point data number nearly 10,000 for temperature and
13,000 for precipitation (Daly et al., 2008). PRISM is frequently updated to map climate in complex
terrain, including rain shadows, high mountains, coastal regions, temperature inversions, and associated

Technical Documentation: Drought

3


-------
complex mesoscale climate processes. More information about the data set can be found at:
www.prism.oregonstate.edu/documents/PRISM datasets.pdf.

The WestWide Drought Tracker collects and analyzes monthly data from the PRISM Climate Mapping
Program. Monthly data are aggregated to create a five-year-timescale data series. Updated data are
processed at the beginning of each month and are available for download in multiple formats at:
https://wrcc.dri.edu/wwdt/about.php.

Figure 4. U.S. Lands Under Drought Conditions, 2000-2020

Figure 4 is based on the U.S. Drought Monitor, which uses a comprehensive definition of drought that
accounts for a large number of different physical variables. Many of the underlying variables reflect
weather and climate, including daily precipitation totals collected at weather stations throughout the
United States, as described above for Figure 1. Other parameters include measurements of soil
moisture, streamflow, reservoir and groundwater levels, and vegetation health. These measurements
are generally collected by government agencies following standard methods, such as a national network
of stream gauges that measure daily and weekly flows, comprehensive satellite mapping programs, and
other systematic monitoring networks. Each program has its own sampling or monitoring design. The
Drought Monitor and the other drought indices that contribute to it have been formulated to rely on
measurements that offer sufficient temporal and spatial resolution.

The U.S. Drought Monitor has five primary inputs:

•	The PDSI.

•	The Soil Moisture Model, from NOAA's Climate Prediction Center.

•	Weekly streamflow data from the U.S. Geological Survey.

•	The Standardized Precipitation Index (SPI), compiled by NOAA and the Western Regional Climate
Center.

•	A blend of objective short- and long-term drought indicators (short-term drought indicator
blends focus on one- to three-month precipitation totals; long-term blends focus on six to 60
months).

At certain times and in certain locations, the Drought Monitor also incorporates one or more of the
following additional indices, some of which are particularly well-suited to the growing season and others
of which are ideal for snowy areas or ideal for the arid West:

•	A topsoil moisture index from the U.S. Department of Agriculture's National Agricultural
Statistics Service.

•	The Keetch-Byram Drought Index.

•	Vegetation health indices based on satellite imagery from NOAA's National Environmental
Satellite, Data, and Information Service.

•	Snow water content.

•	River basin precipitation.

•	The Surface Water Supply Index.

•	Groundwater levels.

•	Reservoir storage.

•	Pasture or range conditions.

Technical Documentation: Drought

4


-------
For more information on the other drought indices that contribute to the Drought Monitor, including
the data used as inputs to these other indices, see:
https://droughtmonitor.unl.edu/About/WhatistheUSDM.aspx.

To find information on underlying sampling methods and procedures for constructing some of the
component indices that go into determining the U.S. Drought Monitor, one will need to consult a variety
of additional sources. For example, as described above for Figure 1, NCEI has published extensive
documentation about methods for collecting precipitation data.

6. Indicator Derivation

Figure 1. Average Drought Conditions in the Contiguous 48 States According to the Palmer Index, 1895-
2020

PDSI calculations are designed to reflect the amount of moisture available at a particular place and time,
based on the amount of precipitation received as well as the temperature, which influences evaporation
rates. The formula for creating this index was originally proposed in the 1960s (Palmer, 1965). Since
then, the methods have been tested extensively and used to support hundreds of published studies. The
PDSI is the most widespread and scientifically vetted drought index in use today.

The PDSI was designed to characterize long-term drought (i.e., patterns lasting a month or more).
Because drought is cumulative, the formula takes precipitation and temperature data from previous
weeks and months into account. Thus, a single rainy day is unlikely to cause a dramatic shift in the index.

PDSI values are normalized relative to long-term average conditions at each location, which means this
method can be applied to any location regardless of how wet or dry it typically is. NOAA currently uses
1931-1990 as its long-term baseline. The index essentially measures deviation from normal conditions.
The PDSI takes the form of a numerical value, generally ranging from -6 to +6. A value of zero reflects
average conditions. Negative values indicate drier-than-average conditions and positive values indicate
wetter-than-average conditions. NOAA provides the following interpretations for specific ranges of the
index:

•	0 to-0.5 = normal

•	-0.5 to-1.0 = incipient drought

•	-1.0 to -2.0 = mild drought

•	-2.0 to-3.0 = moderate drought

•	-3.0 to -4.0 = severe drought

•	< -4.0 = extreme drought

Similar adjectives can be applied to positive (wet) values.

NOAA calculates monthly values of the PDSI for each of the 344 climate divisions within the contiguous
48 states. These values are calculated from weather stations reporting both temperature and
precipitation. As part of its nClimDiv analysis, NOAA uses station data and interpolation between
stations to create a 5-kilometer grid across the contiguous 48 states for each variable in the data set,
including PDSI. Divisional averages are derived by averaging the grid cells within each climate division.
NOAA also combines PDSI values from all climate divisions, weighted by area, to derive a national

Technical Documentation: Drought

5


-------
average for every month. These methods ensure that PDSI values are not biased towards areas that
happen to have more stations clustered close together.

Although NOAA has divided Alaska into 13 climate divisions, PDSI calculations are not available for
Alaska.

EPA obtained monthly national PDSI values from NOAA's website, then calculated annual averages. To
smooth out some of the year-to-year variability, EPA applied a nine-point binomial filter, which is
plotted at the center of each nine-year window. For example, the smoothed value from 2002 to 2010 is
plotted at year 2006. NOAA NCEI recommends this approach. Figure 1 shows both the annual values and
the smoothed curve.

EPA used endpoint padding to extend the nine-year smoothed lines all the way to the ends of the period
of record. As recommended by NCEI, EPA calculated smoothed values as follows: if 2020 was the most
recent year with data available, EPA calculated smoothed values to be centered at 2017, 2018, 2019,
and 2020 by inserting the 2020 data point into the equation in place of the as-yet-unreported annual
data points for 2021 and beyond. EPA used an equivalent approach at the beginning of the time series.

For more information about NOAA's processing methods, see the metadata file at:
wwwl.ncdc.noaa.gov/pub/data/cirs/climdiv/divisional-readme.txt. NOAA's website provides additional
information regarding the PDSI at: www.ncdc.noaa.gov/sotc/drought.

In March 2013, NOAA corrected minor errors in the computer code used to process soil moisture values,
which feed into the computation of the PDSI. This change caused slight revisions to historical data
compared with what EPA presented in Figure 1 prior to August 2013. Although most data were not
substantially changed, minor but discernible differences appeared in data after 2005. NOAA discusses
these improvements in full at: www.ncdc.noaa.gov/sotc/national/2013/3/supplemental/page-7.

Figures 2 and 3. Average Drought Conditions and Average Change in Drought in the Contiguous 48 States
According to the SPEI, 1900-2020

This indicator shows multi-year (five-year) climatological drought/moisture patterns via the SPEI and
examines potential multi-scalar drought trends, spatially and temporally. The SPEI's multi-scalar
characteristics allow for spatial and temporal comparison of drought severity over a wide array of
climates (Vicente-Serrano et al., 2010). Some of the advantages of the SPEI over other drought metrics
are:

•	The index combines multi-timescale aspects of the SPI with information about
evapotranspiration. This ability to represent moisture conditions arguably makes the SPEI a
more complete representation of drought than indices that rely predominantly on temperature.

•	The statistically based index requires only climatological information, without assumptions
about the characteristics of the underlying system.

The SPEI is statistically based and can easily be derived with a concise and customizable calculation
procedure. Using the SPEI enables the analysis of a multitude of drought types (i.e., ecological,
agricultural, meteorological, hydrological, etc.) as well as the potential impacts of climate change
(Vicente-Serrano et al., 2010). Notably, the SPEI is sensitive to the method used to calculate potential
evapotranspiration, or PET (Stagge et al., 2014). The WestWide Drought Tracker's SPEI calculation uses

Technical Documentation: Drought

6


-------
the Thornthwaite method (Thornthwaite, 1948), due to data limitations of the PRISM data set at time of
download and the requirements of other PET calculations. See "Indicator Development" below for
further discussion of PET methods.

Classifications of the SPEI values are defined by McKee et al. (1993). More information on the
calculation and standardization of the SPEI is available at: https://spei.csic.es/home.html and
https://climatedataguide.ucar.edu/climate-data/standardized-precipitation-evapotranspiration-index-
spei. The WestWide Drought Tracker's method for calculating SPEI is further documented in Abatzoglou
et al. (2017).

Figure 2 of this indicator shows average drought conditions from 1900 to 2020 for the contiguous 48
states, based on the five-year-timescale SPEI values on a July to June time span. For example, the 2019
SPEI values contain the mean SPEI values per 4-kilometer pixel within the contiguous 48 states from July
2014 to June 2019, inclusive. EPA chose to analyze the SPEI values and trends with a five-year timescale
for added perspective in assessing long-term drought conditions in the United States. A multi-year
drought index is less affected by interannual variability. Consequently, it may better reflect changes in
long-term dry/wet cycles, which is especially relevant when considering climate trends and potential
connections of drought to climate change (Abatzoglou et al., 2017; Koch et al., 2012). Negative SPEI
values indicate more prevalent drought conditions over the preceding five-year time period. Conversely,
positive SPEI values indicate that higher moisture levels were more prevalent over the preceding five-
year time period.

Figure 3 presents a map of the change in the five-year SPEI values by NOAA climate division over the
time period of 1900 to 2020 (Vose et al., 2014). EPA calculated these trends by conducting ordinary least
squares regressions of the five-year SPEI values for each climate division over the full time period. As
with Figure 1, the five-year SPEI values are based on a July-to-June time span (e.g., the 2019 value
considers SPEI from July 2014 to June 2019). Climate divisions in brown display a trend toward lower
SPEI values; blue represents a trend toward higher SPEI. Each pixel's average annual rate of change
(regression slope) has been multiplied by the length of the period of record (120 years) to derive and
depict an estimate of total change.

Over the complete period of record, EPA aggregated the 4-kilometer SPEI pixel values within each
climate division as well as the contiguous 48 states, for both one- and five-year-timescale calculations.
For each one- and five-year SPEI raster, EPA used the "Calculate Zonal Statistics as Table" tool in ESRI
ArcGIS with the NOAA climate division layer as zone data to determine SPEI values per climate division.
Because SPEI value output is multiplied by 100, as per WestWide Drought Tracker settings, EPA divided
the "mean" values by 100 using the "Calculate Field" tool in ESRI ArcGIS. Analysis at the climate division
level can provide more insight into regional drought trends. Climate division data were obtained from
NOAA's NCEI.

Figure 4. U.S. Lands Under Drought Conditions, 2000-2020

The National Drought Mitigation Center at the University of Nebraska-Lincoln produces the U.S.

Drought Monitor with assistance from many other climate and water experts at the federal, regional,
state, and local levels. For each week, the Drought Monitor labels areas of the country according to the
intensity of any drought conditions present. An area experiencing drought is assigned a score ranging
from DO, the least severe drought, to D4, the most severe. For definitions of these classifications, see:
https://droughtmonitor.unl.edu/About/WhatistheUSDM.aspx.

Technical Documentation: Drought

7


-------
Drought Monitor values are determined from the five major components and other supplementary
factors listed in Section 5. A table on the Drought Monitor website

(https://droughtmonitor.unl.edu/About/WhatistheUSDM.aspx) explains the range of observed values
for each major component that would result in a particular Drought Monitor score. The final index score
is based to some degree on expert judgment, however. For example, expert analysts resolve
discrepancies in cases where the five major components might not coincide with one another. They
might assign a final Drought Monitor score based on what the majority of the components suggest, or
they might weight the components differently according to how well they perform in various parts of
the country and at different times of the year. Experts also determine what additional factors to
consider for a given time and place and how heavily to weight these supplemental factors. For example,
snowpack is particularly important in the West, where it has a strong bearing on water supplies.

From the Drought Monitor's public website, EPA obtained data covering the contiguous 48 states plus
Alaska, Hawaii, and Puerto Rico, then performed a few additional calculations. The original data set
reports cumulative categories (for example, "D2-D4" and "D3-D4"), so EPA had to subtract one
category from another in order to find the percentage of land area belonging to each individual drought
category (e.g., D2 alone). EPA also calculated annual averages to support some of the statements
presented in the "Key Points" for this indicator.

No attempt has been made to portray data outside the time and space in which measurements were
made. Measurements are collected at least weekly (in the case of some variables like precipitation and
streamflow, at least daily) and used to derive weekly maps for the U.S. Drought Monitor. Values are
generalized over space by weighting the different factors that go into calculating the overall index and
applying expert judgment to derive the final weekly map and the corresponding totals for affected area.

For more information about how the Drought Monitor is calculated, including percentiles associated
with the occurrence of each of the D0-D4 classifications, see Svoboda et al. (2002), along with the
documentation provided on the Drought Monitor website at: http://droughtmonitor.unl.edu.

Indicator Development

EPA has conducted several analyses to compare methods and metrics at the national scale:

•	Figure TD-1 shows the influence of timeframe on the SPEI, with a comparison of one- and five-
year SPEI values derived from the same source using the same underlying PET method
(Thornthwaite). Broad patterns are highly similar; the one-year version just shows more year-to-
year variability, as one would likely expect.

•	Figure TD-2 compares the PDSI and SPEI, both at one-year timescales over the same spatial and
temporal extent. Visually, this graph shows general agreement between the indices as far as
national-scale patterns are concerned.

•	Figure TD-2 also shows two versions of the one-year SPEI, reflecting different methods of
accounting for PET: Thornthwaite, as used in this indicator, and another commonly used
method, Penman-Monteith (e.g., Allen et al., 1998). The two methods are correlated for the
period of overlap, with R2 = 0.45.

•	Figure TD-3 compares the influence of PET method on SPEI at a four-year timescale—the longest
timeframe readily available from the source that provided SPEI based on Penman-Monteith.
Here, the correlation has R2 = 0.90.

Technical Documentation: Drought

8


-------
These results all reflect national aggregation. It is very possible that differences between methods may
be wider at smaller spatial scales and for shorter or specific time periods.

Figure

1.5

1

0.5

-0.5

-1

TD-1. Comparison of One- and Five-Year SPEI Values for the Contiguous 48 States, 1900-2020

-1.5

1900	1920	1940

Data source: WestWide Drought Tracker

I960
Year

1980	2000	2020

Technical Documentation: Drought

9


-------
Figure TD-2. Comparison of One-Year PDSI and SPEI Values for the Contiguous 48 States, 1900-2020

8 i	i	i	i	1	i	1	1	i	i	i	r	r 1.5

o o

O

cr> en

o

CM

cr>

o
m
en

o

O)

o

LO

Ch

o

cr>

Year

o
r^
cn

o

oo

(T)

o

cn

O)

o
o
o

rsj

O

fN

1

0.5

0 £

IS)

-0.5
-1

-1.5

o

c\i
O
fN

	1-Year PDSI Value

— CONUS 1-Year SPEI Value (Thornthwaite method)

—CONUS 1-Year SPEI Value (Penman-Monteith method)

Data sources:

•	PDSI from NO A A NCEI

•	SPEI (Thornthwaite method) from WestWide Drought Tracker

•	SPEI (Penman-Monteith method) from SPEIbase (https://spei.csic.es/database.html)

Technical Documentation: Drought

10


-------
Figure TD-3. Comparison of Four-Year SPEI Values for the Contiguous 48 States, 1900-2018

2

1.5
1

0.5

S n
o_ 0

i/i

-0.5
-1
-1.5
-2

1900	1920	1940	1960	1980	2000	2020

Year

—CONUS 4-Year SPEI Value (Thornthwaite method)

—CONUS 4-Year SPEI Value (Penman-Monteith method)

Data sources:

•	PDSI from NO A A NCEI

•	SPEI (Thornthwaite method) from WestWide Drought Tracker

•	SPEI (Penman-Monteith method) from SPEIbase (https://spei.csic.es/database.html)

The Thornthwaite method estimates PET based on mean temperature, whereas the Penman-Monteith
equation combines temperature with wind speed, relative humidity, and solar radiation. Because these
other variables all influence evapotranspiration, studies and experts suggest a preference for using the
Penman-Monteith method where it is available. For example, Stagge et al. (2014) conducted
comparisons and sensitivity analyses and recommended using more robust options such as Penman-
Monteith where possible, with the observation that PET calculations are particularly sensitive to the
method of accounting for radiation. However, several key inputs to the Penman-Monteith equation
(wind speed, relative humidity, and solar radiation) were not recorded during the first half of the 20th
century, and even when they were measured actively during the second half of the century, the density
of stations with complete data for analysis was sparse. SPEI data based on Penman-Monteith are
nonetheless available—including the data shown in the figures above, which come from an international
consortium website called SPEIbase (https://spei.csic.es/database.html). Yet these data do not have the
high geographic resolution needed for the PRISM analysis that feeds into EPA's indicator, particularly
when climate division mapping is concerned. Given the desired spatial and temporal coverage for this
indicator, EPA elected to use a Thornthwaite-based SPEI data source.

Technical Documentation: Drought

11


-------
EPA notes that temperature-based PET methods (such as Thornthwaite) have been shown to perform
relatively well in climatological applications because air temperature is correlated with net radiation and
humidity at weekly, monthly, and sub-annual timescales (Sheffield et al., 2012). The apparent
similarities between methods at the scales shown in Figures TD-2 and TD-3 support this observation.

7. Quality Assurance and Quality Control

Figure 1. Average Drought Conditions in the Contiguous 48 States According to the Palmer Index, 1895-
2020

Data from weather stations go through a variety of quality assurance and quality control (QA/QC)
procedures before they can be added to historical databases in their final form. NOAA's nClimDiv data
set follows strict QA/QC procedures to identify errors and biases in the data and then either remove
these stations from the time series or apply correction factors. Procedures for nClimDiv are summarized
at: www.ncdc.noaa.gov/monitoring-references/maps/us-climate-divisions.php. Specific to this indicator,
Karl et al. (1986) describe steps that have been taken to reduce biases associated with differences in the
time of day when temperature observations are reported. Other procedures include:

•	Removal of duplicate records.

•	Procedures to deal with missing data.

•	Testing and correcting for artificial discontinuities in a local station record, which might reflect
station relocation or instrumentation changes.

Figures 2 and 3. Average Drought Conditions and Average Change in Drought in the Contiguous 48 States
According to the SPEI, 1900-2020

QA/QC measures for temperature and precipitation data include the removal of duplicate records,
missing data corrections, and testing and correcting for artificial discontinuities in a local station record,
which might reflect station relocation or instrumentation changes. Monthly PRISM data follow a variety
of QA/QC procedures before they are distributed in their final form. PRISM adjusts for complex terrain,
including rain shadows, high mountains, coastal regions, temperature inversions, and associated
complex mesoscale climate processes in the QA/QC procedure. All station data are spatially QCed, and
all procedures are peer-reviewed. Methodology for the PRISM data set is described in further detail in
Daly etal. (2008).

QA/QC procedures behind the WestWide Drought Tracker calculations are not readily available. Each
underlying data source has an associated methodology, which typically includes some degree of QA/QC
and is available online.

Figure 4. U.S. Lands Under Drought Conditions, 2000-2020

QA/QC procedures for the overall U.S. Drought Monitor data set are not readily available. Each
underlying data source has its own methodology, which typically includes some degree of QA/QC. For
example, precipitation and temperature data are verified and corrected as described above for Figure 1.
Some of the other underlying data sources have QA/QC procedures available online, but others do not.

Technical Documentation: Drought

12


-------
Analysis

8. Comparability Over Time and Space

Figure 1. Average Drought Conditions in the Contiguous 48 States According to the Palmer Index, 1895-
2020

PDSI calculation methods have been applied consistently over time and space. In all cases, the index
relies on the same underlying measurements (precipitation and temperature). Although fewer stations
were collecting weather data during the first few decades of the analysis, NOAA has determined that
enough stations were available starting in 1895 to calculate valid index values for the contiguous 48
states as a whole.

Figures 2 and 3. Average Drought Conditions and Average Change in Drought in the Contiguous 48 States
According to the SPEI, 1900-2020

The calculation methods of the SPEI have been applied consistently over both dimensions. In all cases,
the index relies on the same underlying measurements and calculations. The SPEI values used for this
indicator are based on weighted PRISM data by location (Daly et al., 2008) including the 4-kilometer
pixel grid values. PRISM calculates a climate-elevation regression for each digital elevation model grid
cell, and stations entering the regression are assigned weights based primarily on the physiographic
similarity of the station to the grid cell. Factors considered are location, elevation, coastal proximity,
topographic facet orientation, vertical atmospheric layer, topographic position, and orographic
effectiveness of the terrain (Daly et al., 2008).

A potential source of variability relates to aggregating the original resolution of the data (i.e., raster
data). The PRISM temperature data model was revamped in late October of 2019. This change is made
less important by natural data smoothing associated with temporal aggregation. Data in this indicator
rely on the current model iteration, whereas future SPEI data will incorporate temperature data
collected with the newer versions of the model and any associated changes to the parameters.

Figure 4. U.S. Lands Under Drought Conditions, 2000-2020

The resolution of the U.S. Drought Monitor has improved over time. When the Drought Monitor began
to be calculated in 1999, many of the component indicators used to determine drought conditions were
reported at the climate division level. Many of these component indicators now include data from the
county and sub-county level. This change in resolution over time can be seen in the methods used to
draw contour lines on Drought Monitor maps.

The drought classification scheme used for this indicator is produced by combining data from several
different sources. Different locations may use different primary sources—or the same sources, weighted
differently. These data are combined to reflect the collective judgment of experts and in some cases are
adjusted to reconcile conflicting trends shown by different data sources over different time periods.

Though data resolution and mapping procedures have varied somewhat over time and space, the
fundamental construction of the indicator has remained consistent.

Technical Documentation: Drought

13


-------
9. Data Limitations

Factors that may impact the confidence, application, or conclusions drawn from this indicator are as
follows:

1.	The indicator focuses on a broad overview of drought conditions in the contiguous United
States. The spatial averaging of drought metrics over such a large area can obscure drought
extremes occurring at regional scales. It is not intended to replace local or state information that
might describe conditions more precisely for a particular region. Local or state entities might
monitor different variables to meet specific needs or to address local problems. As a
consequence, there could be water shortages or crop failures within an area not designated as a
drought area, just as there could be locations with adequate water supplies in an area
designated as D3 or D4 (extreme or exceptional) drought.

2.	Because this indicator focuses on national trends, it does not show how drought conditions vary
by region. For example, even if half of the country suffered from severe drought, Figure 1 could
show an average index value close to zero if the rest of the country was wetter than average.
Thus, Figure 1 might understate the degree to which droughts are becoming more severe in
some areas, while other places receive more rain as a result of climate change.

3.	Although the PDSI is arguably the most widely used drought index, it has some limitations that
have been documented extensively in the literature. While the use of just two variables
(precipitation and temperature) makes this index relatively easy to calculate over time and
space, drought can have many other dimensions that these two variables do not fully capture.
For example, the PDSI loses accuracy in areas where a substantial portion of the water supply
comes from snowpack.

4.	This SPEI calculation uses the Thornthwaite PET equation, given the limitation on variables
available from the monthly PRISM data set. With the Thornthwaite approach, variables other
than temperature that can affect PET, such as wind speed, surface humidity, and solar radiation,
are not considered. However, at present, these additional variables are not as widely available
as temperature and precipitation for the same temporal and spatial scope of this multi-decadal
period of record. As discussed in Section 6, the Thornthwaite method performs reasonably well
at the scale of this indicator.

5.	Some studies have found that the PRISM data set may report anomalous high-elevation
warming trends. Some extreme warming observed at higher elevations is potentially the result
of systematic artifacts and not climatic conditions (Oyler et al., 2015).

6.	The extreme conditions associated with the Dust Bowl in the 1930s may unduly influence the
long-term trends in some parts of the country. Widespread and exceptional drought in the early
part of the time series may accentuate the appearance of a wetter trend more recently. This
may be especially true for climate divisions in the center part of the country.

7.	Indices such as the U.S. Drought Monitor seek to address the limitations of the PDSI by
incorporating many more variables. The Drought Monitor is relatively new, however, and cannot
yet be used to assess long-term climate trends.

Technical Documentation: Drought

14


-------
8. The drought classification scheme used for Figure 4 is produced by combining data from several
different sources. These data are combined to reflect the collective judgment of experts and in
some cases are adjusted to reconcile conflicting trends shown by different data sources over
different time periods.

10.	Sources of Uncertainty

Error estimates are not readily available for national average PDSI, the U.S. Drought Monitor, or the
underlying measurements that contribute to this indicator.

As far as the SPEI calculations are concerned, it is not clear how much uncertainty might be associated
with the weighting of different parts of the PRISM model, as well as the accompanying calculation.
However, an in-depth analysis of PRISM is conducted in Daly et al. (2008). Estimating the true levels of
uncertainty associated with spatial climate data sets is challenging and influenced by its own set of
errors. With the exception of a relatively small number of observations, the true climate field is largely
unknown. Even the known observations are subject to uncertainties. For more information on
uncertainty with respect to temperature and precipitation data, see Daly et al. (2008).

An additional source of uncertainty exists within the spatial aggregation at both the climate division and
nationwide scales. The NOAA climate division layer resolution used by this indicator is finer than that
used by the WestWide Drought Tracker (the source of SPEI data). This resulted in small differences in
output, as certain pixels were attributed to differing climate divisions, thereby influencing aggregative
SPEI values. In coastal areas, EPA excluded SPEI values from 4-kilometer grid cells if a majority of the
area lay outside the climate division layer. On the border between two or more climate divisions, EPA
attributed the grid cell values to the climate division that most contained the pixel.

It is not clear how much uncertainty might be associated with the component indices that go into
formulating the Drought Monitor or the process of compiling these indices into a single set of weekly
values through averaging, weighting, and expert judgment.

11.	Sources of Variability

Conditions associated with drought naturally vary from place to place and from one day to the next,
depending on weather patterns and other factors. This indicator addresses spatial variability by
presenting trends by climate division as well as aggregate national trends. Figures 1, 2, and 3 address
temporal variability by using indices that are designed to measure long-term drought and are not easily
swayed by short-term conditions. Figure 1 provides an annual average, along with a nine-year smoothed
average. Figures 2 and 3 use a multi-year drought index, which minimizes annual and monthly variability
and helps to better reflect changes in long-term drought/wet cycles. Examining trends over the data
set's full period of record (1900-2020) accounts for any sizable interannual variability. Figure 4 smooths
out some of the inherent variability in drought measurement by relying on many indices, including
several with a long-term focus. While Figure 4 shows noticeable week-to-week variability, it also reveals
larger year-to-year patterns.

12.	Statistical/Trend Analysis

This indicator does not report on the slope of the trend in PDSI values over time, nor does it calculate
the statistical significance of this trend. This information is currently not available from NOAA's NCEI.

Technical Documentation: Drought

15


-------
As a first-order screening tool, EPA applied an ordinary least-squares linear regression to the national
five-year SPEI time-series in Figure 2 and identified an overall trend toward wetter conditions at a rate
of +0.0036 SPEI units per year. This trend is statistically significant (p < 0.0001). EPA also applied an
ordinary least-squares regression to the five-year SPEI index data from 1900 to 2020 for each NOAA
climate division. The slopes of the trendlines (change in SPEI value per year) are transformed into total
change and colorized in Figure 3. Of a total of 344 climate divisions, 75 display negative trends in SPEI
values and 269 display positive trends. A total of 252 of these trends (73 percent of climate divisions)
are significant to a 95 percent level (p < 0.05).

Because data from the U.S. Drought Monitor are only available since 2000, this metric is too short-lived
to be used for assessing long-term climate trends. With continued data collection, future versions of this
indicator should be able to paint a more statistically robust picture of long-term trends in Drought
Monitor values.

References

Abatzoglou, J.T., D.J. McEvoy, and K.T. Redmond. 2017. The West Wide Drought Tracker: Drought
monitoring at fine spatial scales. B. Am. Meteorol. Soc. 98:1815-1820.

Allen, R., L. Pereira, D. Raes, and M. Smith. 1998. Crop evapotranspiration. FAO irrigation and drainage
paper 56. Food and Agriculture Organization of the United Nations.

Daly, C., R.P. Neilson, and D.L. Phillips. 1994. A statistical-topographic model for mapping climatological
precipitation over mountainous terrain. J. Appl. Meteor. 33:140-158.

Daly, C., M. Halbleib, J.I. Smith, W.P. Gibson, M.K. Doggett, G.H. Taylor, J. Curtis, and P.A. Pasteris. 2008.
Physiographically-sensitive mapping of temperature and precipitation across the conterminous United
States. Int. J. Climatol. 28:2031-2064.

Karl, T.R., C.N. Williams, Jr., P.J. Young, and W.M. Wendland. 1986. A model to estimate the time of
observation bias associated with monthly mean maximum, minimum, and mean temperatures for the
Unites States. J. Clim. Appl. Meteorol. 25(1):145-160.

Koch, F.H., J.W. Coulston, and W.D. Smith. 2012. Mapping drought conditions using multi-year windows.
In: Potter, K.M., and B.L. Conkling (eds.). 2012. Forest health monitoring: 2009 national technical report.
Gen. Tech. Rep. SRS-167. Asheville, NC: U.S. Department of Agriculture.

McKee, T.B., N.J. Doesken, and J. Kleist. 1993. The relationship of drought frequency and duration to
time scales. Proceedings of the Eighth Conference on Applied Climatology. Boston, MA: American
Meteorological Society.

Oyler, J.W., S.Z. Dobrowski, A.P. Ballantyne, A.E. Klene, and S.W. Running. 2015. Artificial amplification
of warming trends across the mountains of the western United States. Geophys. Res. Lett. 42:153-161.

Palmer, W.C. 1965. Meteorological drought. Res. Paper No. 45. Washington, DC: U.S. Department of
Commerce.

Technical Documentation: Drought

16


-------
Sheffield, J., E. Wood, and M. Roderick. 2012. Little change in global drought over the past 60 years.
Nature 491:435-438. doi:10.1038/naturell575

Stagge, J.H., L.M. Tallaksen, C.-Y. Xu, and H.A.J. Van Lanen. 2014. Standardized Precipitation-
Evapotranspiration Index (SPEI): Sensitivity to potential evapotranspiration model and parameters. In:
Daniell, T.M. (ed.). Hydrology in a changing world: Environmental and human dimensions. Proceedings
of FRIEND-Water 2014. IAHS Publ. No. 363. Wallingford, UK: IAHS Press.

Svoboda, M., D. Lecomte, M. Hayes, R. Heim, K. Gleason, J. Angel, B. Rippey, R. Tinker, M. Palecki, D.
Stooksbury, D. Miskus, and S. Stephens. 2002. The drought monitor. B. Am. Meteorol. Soc. 83(8):1181-
1190.

Thornthwaite, C.W. 1948. An approach toward a rational classification of climate. Geogr. Rev. 38(1):55-
94.

Vicente-Serrano, S.M., S. Begueria, and J.I. Lopez-Moreno. 2010. A multi-scalar drought index sensitive
to global warming: The Standardized Precipitation Evapotranspiration Index. J. Clim. 23:1696-1718.

Vose, R.S., S. Applequist, M. Squires, I. Durre, M.J. Menne, C.N. Williams Jr., C. Fenimore, K. Gleason, and
D. Arndt. 2014. Improved historical temperature and precipitation time series for U.S. climate divisions.
J. Appl. Meteorol. Climatol. 53:1232-1251.

Wehner, M.F., J.R. Arnold, T. Knutson, K.E. Kunkel, and A.N. LeGrande. 2017. Droughts, floods, and
wildfires. In: USGCRP (U.S. Global Change Research Program). 2017. Climate science special report:
Fourth National Climate Assessment, volume I. Wuebbles, D.J., D.W. Fahey, K.A. Hibbard, D.J. Dokken,
B.C. Stewart, and T.K. Maycock (eds.). https://science2017.globalchange.gov.

Technical Documentation: Drought

17


-------