£%	United States
Environmental Protection
^yFLrtl Jrm Agency
EPA/600/S-17/256 | October 2017 | www.epa.gov/research
Description of Changes in
Climatic Indices in USA over
25 Years (1989 = 2013)
non-Significant Decrease
Significant Decrease
Significant Increase
non-Significant Increase
Trend of annual number of months with at least one month that
had a day with a minimum temperature < 0°C over 25 years
RESEARCH AND DEVELOPMENT

-------

-------
Description of Changes in Climatic
Indices in USA over 25 Years
(1989-2013)
Prepared by
Maliha S. Nash1, Jay R. Christensen2
1U.S. Environmental Protection Agency, Office of Research and Development,
Ecological & Human Community Analysis Branch, Las Vegas, Nevada, USA
2U.S. Environmental Protection Agency, Office of Research and Development,
Ecosystem Integrity Branch,
Las Vegas, Nevada, USA
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460

-------
ii

-------
Acknowledgements
We are grateful for the valuable inputs and suggestions provided by Drs. Mike McDonald,
James Wickham, and Taylor Jarnagin which improved the comprehensiveness and clarity of this
report.
Notice
The U.S. Environmental Protection Agency (EPA), through its Office of Research and Development (ORD),
funded and performed the research described here. It has been peer-reviewed by the EPA and approved for
publication, but it may not necessarily reflect official Agency policy. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
Ill

-------
iv

-------
Table of Contents
Acknowledgements	iii
Notice	iii
List of Figures	vii
Acronyms and Abbreviations	xi
Introduction	1
Method and Materials	3
Potential Evapotranspiration	3
Moisture Index	4
Apparent Heat	4
Univariate Autoregression	5
Results	7
Monthly: Minimum Temperature, Maximum Temperature, Dew Point Temperature,
Precipitation, Potential Evapotranspiration (PET), Moisture Index (MI), and
Apparent Heat (AT)	7
Annual: Average and Coefficient of Variability for Annual Precipitation, Potential
Evapotranspiration (PET), and Moisture Index (MI)	16
Seasonal: Average and Coefficient of Variability for Seasonal Precipitation, Potential
Evapotranspiration (PET), and Moisture Index (MI)	19
Summary and Discussions	29
Conclusion	33
Implication	35
References	37
v

-------
vi

-------
List of Figures
Figure 1. Pixels (1 km2) with temporal trends in monthly (A) maximum temperature
and (B) minimum temperature (°C/month). The first and last legend classes
represent the 1st and 99th percentiles. The insets are the probability of
temporal trend, green indicates significant increase; red indicates significant
decrease	8
Figure 2. Trend of annual number of months with at least one month that had a day
with a minimum temperature < 0°C over 25 years. Red denotes the
significant (p< 0.05) decrease, green denotes the significant increase, purple
denotes the non-significant decrease, black denotes the non-significant
increase in minimum temperatures. Areas with no color denote not enough
months with minimum temperature < 0°C for trend determination	10
Figure 3. Pixels (1 km2) with temporal trends in monthly dew point temperature
(°C/month). The first and last legend classes represent the 1st and 99th
percentiles. The inset map is the probability of temporal trend, green
indicates significant increase and red indicates significant decrease	10
Figure 4. Pixels (1 km2) with temporal trends in monthly precipitation (mm/month).
The first and last legend classes represent the 1st and 99th percentiles. The
inset map is the probability of temporal trend, green indicates significant
increase and red indicates significant decrease	12
Figure 5. Pixels (1 km2) with temporal trends in monthly PET (mm/month). The first
and last legend classes represent the 1st and 99th percentiles. The inset map
is the probability of temporal trend, green indicates significant increase and
red indicates significant decrease	12
Figure 6. Pixels (1 km2) with temporal trends in monthly (A) MI, the first and last
legend classes represent the 1st and 99th percentiles. (B) Moisture regimes
(Feddema (2005) categories) based on the average monthly moisture index
(MI) over 25 years (1989-2013). The inset map in (A) is the probability of
temporal trend, green indicates significant increase and red indicates
significant decrease	14
Figure 7. Pixels (1 km2) with temporal trends in monthly AT. The first and last legend
classes represent the 1st and 99th percentiles. The inset map is the probability
of temporal trend, green indicates significant increase and red indicates
significant decrease	15
vii

-------
List of Figures (cont.)
Figure 8. Distribution of total number of months (out of 300) that had at least oneAT
value in group:
A)	fatigue possible with prolonged exposure and/or physical activity (26.7°C

-------
List of Figures (cont.)
Figure 16. The 25-year average of MI for seasonal and annual. Legend is the
Feddema et al. (2005) drought classification: very wet [MI > 0.66], wet
[0.66 > MI > 0.33], moist [0.33 > MI > 0.00], Dry [0.00 > MI > -0.33],
semi-arid [-0.33 > MI > -0.66]; and arid [MI < -0.66], Inset maps are the
probability of trend, Red denotes significant decrease and green denotes
significant increase	27
IX

-------
X

-------
Acronyms and Abbreviations
AT	Apparent Heat
CV	Coefficient of Variability (standard deviation/mean)
DP	Dew Point Temperature
FEM	Freshwater Ecosystem Mosaic
MI	Moisture Index
NDVI	Normalized Difference Vegetation Index
PET	Potential Evapotranspiration
PRISM	Parameter-Elevation Regressions on Independent Slopes Model
RF	Rainfall
XI

-------
xii

-------
Introduction
The spatial distribution of long-term changes in climatic factors and its relation with
vegetation cover, human health, hydrology and many other ecosystem processes help to identify
the consequences of climatic factors changes. In recent studies, the significant changes of selected
climatic factors over 25 years has been associated with changes in greenness (Nash et al. 2017),
with shifts in health outcomes (Perry et al. 2011), and with fog formation (Hiatt et al. 2012) as
examples of the impacts of climate factors on ecosystems and societal benefits. This report
concentrates on the methodology and potential future uses for long-term climate factor trends, and
trends of derived climatic factors and indices. Analyses of all factors over the 25 years (1989-
2013) show the magnitude and direction of significant changes on different time scales (monthly,
annual average, and seasonal). Additionally, annual averages, coefficient of variation (CV), first
and 99th percentile were also presented to show the differences in patterns of variability and
extreme values. The climate factors considered are: Minimum, maximum and dew temperatures,
and precipitation. A derived climatic factor considered was: potential evapotranspiration (PET);
and two climatic indices were a moisture index and an apparent temperatures (AT) index, or Heat
Index (Smoyer-Tomic and Rainham 2001). All of our analyses were performed per 1 km2 pixels
for the contiguous U.S. and we present the spatial distribution of temporal changes of:
1-	the significant monthly changes in climatic information over 25 years via maps and a
statistical summary description of results (n=300),
2-	the significant annual changes in climatic information via maps and a statistical summary
description of results (n=25),
3-	significant seasonal changes for precipitation, PET, and MI via maps and a statistical
summary description of results (n=25),
1

-------
4-	comparison of the spatial distribution patterns of 1-3, and
5-	distribution of heat index {AT) groups as it relates to health consequences.
These long-term datasets and trends in temperature and precipitation can be used to study
(among other things) the influence of precipitation and temperature trends on potential shifts in
water storage, hydrological flows, biogeochemical processes or biological shifts on water-
dependent species. Two studies have already begun to utilize the summarization of the 25-year
climate and derivative datasets. A drought resilience study led by EPA NCEA and USGS has
included the 25-year summaries of climate variables to help explain changes in water storage as
measured by inundation maps (Vanderhoof et al. in prep). In another study, the climate variables
are being combined with waterbody data and landscape features to describe the collection of
aquatic habitats for biological species and their patterns of occurrence within a matrix as a
"freshwater ecosystem mosaic" (FEM). Our intent was to make the methodology, data, and
analyses available for research efforts such as these, to better understand the trends of precipitation,
temperature, and their interactions in hydrological, societal, biological and ecological systems.
2

-------
Method and Materials
Monthly averages of precipitation, maximum temperature, minimum temperature, and dew
point temperature were obtained from Parameter-elevation Regressions on Independent Slopes
Model (PRISM; http://www.prism.oregonstate.edu/products/matrix.phtml accessed October
2017). PRISM climatic factors from 1989-2013 at a resolution of 4-km2 grid cell were gridded into
1-km2 grid cells using the inverse distance weighted method in ARC-GIS 9.3.1 (ESRI, Redlands,
California) to match the NDVI resolution (Nash et al. 2017). Downscaling of PRISM climate data
from 4 km2 to 1 km2 was also applied by Thorne et al. (2012).
Potential Evapotranspiration
We derived PET following Hamon (1961) and Enquist et al. (2008) as:
PET = 13.97 dD2Wt	[1]
Wt = 0.0495e° 062T	[2]
where d is number of days in a month, D is mean monthly hours of daylight (in unit of 12 hours),
Wt is saturated water vapor density and Tis the monthly mean temperature in °C. PET is in units
of mm/month. The mean day length (D) was calculated as:
' . pn , . Ln .
sinT80 Sin 180 sin(P
Ln
C0S 180 C0S(P
0 = 0.2163108 + 2 * Arctan {0.9617396 * Tan [0.0086*(./-186)]}
(p = Arcsine(0.39795 * cos6)
where 6 is revolution angle (radians), cp is sun's declination angle (radians),/* is daylight coefficient
(degrees) (=0.8333; Forsythe et al. 1995), L is latitude, and/) is daylight length.
24
D = 24	Arcsin
n
3

-------
Moisture Index
The moisture index (MI), also known as climatic moisture index (Willmott and Feddema,
1992), combines rainfall (RF) and PET as:
MI value ranged from -1 to +1 representing the relative moisture in the environment from
driest to wettest moisture conditions. For annual and seasonal MI, summation of PET and RF for
the year and season were used. Seasons were winter (December, January, February), spring
(March, April, May), summer (June, July, August) and fall (September, October, November).
Feddema (2005) grouped the MI into six moisture regimes: very wet [MI > 0.66], wet [0.66 > MI
> 0.33], moist [0.33 > MI > 0.00], dry [0.00 > MI > -0.33], semi-arid [-0.33 > MI > -0.66]; and
arid [MI < -0.66]; we also used the same groups, but added a zero group as a divider between dry
and wet areas (see map in Fig. 6B).
Apparent Heat
Apparent heat (A'/), derived by Smoyer-Tomic and Rainham (2001) (also known as the
index of heat (Perry et al. 2011)) combines daily temperature and humidity as:
where AT is apparent temperature, T is the air average temperature (°C) and DP is the dew point
temperature (°C). Smoyer-Tomic and Rainham (2001) used A Tto monitor heat waves and grouped
AT into four health impact related classes related to fatigue, two levels of sunstroke, and heatstroke.
Although AT was developed as a daily heat index, our analyses were done monthly to show the
if RF < PET
if RF > PET
if PET = RF = 0
[3]
AT = -2.719 + 0.944 *T + 0.016* DP2
[4]
4

-------
spatial distribution of the number of months with at least one event in Smoyer-Tomic and Rainham
(2001) groups. These four A T groups are described below:
a)	fatigue possible with prolonged exposure and/or physical activity (26.7°C  54.4°C).
The spatial distribution of the significance and direction of the A T trends were also presented.
Annual and seasonal averages of precipitation (with coefficients of variation) were derived
for 25 years (n=25). The four seasons are: Winter (December - February), spring (March - May),
summer (June - August) and fall (September - November). Averages and coefficients of variation
for annual precipitation and PET were also determined. Changes in minimum temperature were
described by equating number of months per year with at least one Day of < 0°C. The trend in the
number of months per year with minimum temperature < 0°C was determined over the USA. The
direction and significance level of the trend were presented in a map view to show the spatial
locations of changes. We also present the 1st and 99th percentiles to spatially locate the extremes.
Univariate Autoregression
We used univariate autoregression to quantify the temporal trend (slope) for each of the
climate factors and derived variables to identify the general pattern of change for each variable
over the 25-year period. The trend values, direction, and probability of each pixel was then
mapped to identify geographic patterns of the trend direction. Time series regression
5

-------
(autoregression) was used in both analyses because errors in temporal data may be dependent (e.g.,
consecutive times, annual cycles). If dependency exists and is not corrected, then the standard
error of the estimate (e.g., slope) would be inflated, and the significance level for the slope and
other estimates would be incorrect. Below we detail the analyses.
Trends in the individual climate factors over the 25-year period were addressed for each 1-
km2 pixel by using an autoregression model (Proc Autoreg; SAS/ETS, 1999) with stepwise
selection. Significant autoregressive error was fit to the observed values to define the direction and
p-value for the slope, where:
Yt=6o+0l* Time+fit	[5]
k
t=1
st ~IN(0,cr)
where Y is an individual time series variable (e.g., monthly precipitation, maximum temperature,
minimum temperature, monthly dew point temperature, and MI (n=300 months)). The fitted
autoregression model for the observed variable (Yt) is the same as that of an ordinary least square
regression model (OLS; 0o + 6i *Time), plus the autoregressive error (ut). Coefficients 0o, and Oi
are the intercept and the slope with time, respectively. The time series error term, ut, may be
k
autocorrelated. The term	is the summation of the significant autoregressive parameter
i
(p) times lagged time series error(s), and k is the order of significant lags in the model. The error
term, st, from the autoregressive error model is normally and independently distributed (IN) with
a mean of zero and variance o2. The slope (Oi ) quantifies the rate and direction of change for each
variable over 25 years in each 1 km2 pixel. A significance level of p < 0.05 was used to test
whether the slope differed from zero.
6

-------
Results
Monthly: Minimum Temperature, Maximum Temperature, Dew Point Temperature, Precipitation,
Potential Evapotranspiration (PET), Moisture Index (MI), and Apparent Heat (AT).
Significant maximum and minimum temperature increases covered 12% and 35% of the
contiguous United States, respectively (inset maps in Fig. 1) (Nash et al. 2017). Significant
increases in maximum temperature (inset map in Fig. 1A) were concentrated in New England,
Texas, and Louisiana, but also scattered throughout the western United States. The spatial
distribution of maximum temperature trend values (Fig. 1A) shows that the 1st and the 99th
percentiles of the trend values were + 0.0067°C/month and -0.0069°C/month, respectively. The
clusters of increasing maximum temperature are situated on a diagonal path from Texas to
Washington, with additional clusters in the state of California situated mostly in Stanislaus, Inyo,
and Sequoia national forests. North Dakota, the eastern side of Montana, and the Pacific coast of
Oregon and Washington had a decreasing trend in maximum temperature. The decrease in North
Dakota and eastern side of Montana was not significant (inset map in Fig. 1 A).
Significant minimum temperature (Fig. IB) increases were common throughout the
contiguous United States, except for the upper Midwest. The 1st and the 99th percentiles of the
trend values were -0.0064°C/month and +0.0144°C/month for the minimum temperature (Fig. IB).
The clusters of trend values showed a mosaic pattern in the western states, in contrast to the mid
and eastern states where the clusters are larger in size. This pattern of clustering indicates that
higher variability in minimum temperatures may occur where extremes in higher and lower
nighttime temperatures are more apparent (e.g., in the western U.S. relative to the eastern U.S.).
Most of Oregon and Washington experienced a decrease in minimum temperature, while
California was more of a mosaic. In California, the minimum temperature mostly increased in the
7

-------
Figure 1. Pixels (1 km2) with temporal trends in monthly (A) maximum temperature and (B) minimum temperature
(°C/month). The first and last legend classes represent the 1st and 99th percentiles. The insets are the
probability of temporal trend, green indicates significant increase; red indicates significant decrease.
8

-------
mountainous national forest areas (e.g. Shasta, Trinity, Six Rivers, Lassen, Plumas, Tahoe,
Humboldt, and Los Padres National Forests.) The increasing minimum temperatures over the 25
years in Wyoming, Colorado, and New Mexico coincide with tree mortality and wildfire (Nash et
al. 2014, 2017). In the east, the greatest increase in minimum temperature was in north Maine,
Vermont, New York (Fig. IB) and to a lesser level in Pennsylvania to eastern North Carolina.
The significant decrease in both minimum and maximum temperatures covered an
approximately similar percentage of area (-4%), but were more "patchy." Areas of significant
minimum and maximum temperature decreases were scattered throughout the contiguous United
States, but predominately in the south and the west. The general trend over the last 25 years has
been for monthly minimum temperature to increase significantly for a substantial proportion of the
continental U.S. (inset map in Fig. IB). The annual number of months with < 0°C is decreasing as
well. The trend of annual number of months with at least one month that had a day with a minimum
temperature of < 0°C decreased in 49.9% of the U.S. (Fig. 2) and significantly increased in 23.9%
of the U.S. The number of months with minimum temperature < 0°C did not change significantly
over the 25 years in 26.2% of the U.S. and were mostly in the southern U.S. The number of cold
months increased primarily in eastern-southern part of the U.S. and in a mosaic pattern in
Washington, Oregon, Nevada. The number of months with freezing conditions decreased in the
Northern U.S, especially within the Northern Great Plains and in the Northeastern states.
Significant increases in months with freezing conditions were found in a band from Texas to North
Carolina (Fig.2).
Significant dew point temperature changes were regionally diverse (Fig. 3), exhibiting
significant increases in the east and significant decreases in the west. The contiguous clusters of
pixels (red pixels in the map inset in Fig. 3) with a decrease in dew point temperature were mainly
9

-------
A
0	250 500	1,000 km
	1	i	i	i	I	i	i	i	I
Figure 2. Trend of annual number of months with at least one month that had a day with a minimum temperature
< 0°C over 25 years. Red denotes the significant (p< 0.05) decrease, green denotes the significant increase,
purple denotes the non-significant decrease, black denotes the non-significant increase in minimum
temperatures. Areas with no color denote not enough months with minimum temperature < 0°C for trend
determination.
¦	<= -0.014
¦	-05139 - -0,005
¦	-0.0049 - -0.003
~	-0.0029 - -0,001
¦	-0.0009 • 0
~	0.0001 -0.001
¦	0.0011 -0.005
¦	0.0051 - 0.0095
¦	> 0.0095
Figure 3. Pixels (1 km2) with temporal trends in monthly dew point temperature (°C/month). The first and last legend
classes represent the 1st and 99lh percentiles. The insets maps are the probability of temporal trend, green
indicates significant increase and red indicates a significant decrease.
10

-------
concentrated in the west. The 1st and the 99th percentiles of the trend values for dew point
temperatures were -0.014°C/month and +0.0095°C/month, respectively.
Monthly precipitation trends (Fig. 4) decreased significantly by 8.7% and were mostly
concentrated along the Texas-Oklahoma-Louisiana border and scattered throughout the southern
Appalachians (inset map in Fig. 4). The spatial distribution of precipitation trend values showed that
the highest increases in precipitation were in the north and northeastern part of the nation (Fig. 4)
with a rate (99th percentile) 0.072 mm/month and the lowest were in the southern part of the U.S.
with a rate of (1st percentile) -0.1236 mm/month. Extreme values in increasing and decreasing
precipitation were located in Washington, Oregon, and California. Decreasing precipitation
occurred in Louisiana, Texas, Oklahoma, and southwestern North Carolina (1st percentile, Fig. 4).
Increasing precipitation occurred in Maine and New York (99th percentile, Fig. 4). The overall
changes in precipitation were only significant for small areas in the northeast (-3% significant
increases, Nash et al. 2017) and in the south within Louisiana, Texas, and Oklahoma (-9%
significant decreases, Nash et al. 2017) (see map inset in Fig. 4).
Due to the nearly continent-wide significant increase in monthly temporal trends of minimum
temperature (Fig. IB), PET monthly temporal trends also show a nearly continent-wide increase (Fig.
5). Most of the contiguous U.S. experienced an increase PET trend, with comparatively isolated
clusters of significantly declining PET trends primarily in the Pacific Northwest and California
(Fig. 5 inset). However, clusters of the highest increases in PET were also in the west. The 1st
percentile in decreasing PET was -0.0156 mm/month and the 99th percentile in increasing PET
was 0.0345 mm/month. Many of the changes in PET were significant increases (p < 0.05, inset
map in Fig. 5) and small clusters of significant decreases were scattered in Washington, Oregon,
California, and Florida.
11

-------
¦	<= -0.1236
¦	-01235 - -0.0600
¦	-0.0599 - -0.0200
~ -0.0199 - 0.0000
¦	0.0001 -0.0100
00.0101-0.0200
¦	0.0201 -0.0400
¦	0.0401-0.0720
¦	» 0.0720
Figure 4. Pixels (1 km2) with temporal trends in monthly precipitation (mm/month). The first and last legend classes
represent the I st and 99th percentiles. The insets maps are the probability of temporal trend, green indicates
significant increase and red indicates a significant decrease.
¦	<= -0.0156
¦	-0:0155 - -Q.0120
¦	-0>QU9 - -0.0100
n -0.0099 • 0.0000
¦	0.0001 -0.0100
~ OjOXOI -OJ0150
a 0.0151 -0.0200
¦	0.0201 - 0.0345
¦	> 0.0345
Figure 5. Pixels (1 km2) with temporal trends in monthly PET (mm/month). The first and last legend classes represent
the 1st and 99th percentiles. The inset map is the probability of temporal trend, green indicates significant
increase and red indicates a significant decrease.
12

-------
MI is a composite of precipitation, day length, average temperature and saturated vapor
pressure [Equation 3] and represents the relative wetness (+ MI values) and dryness (- MI values)
of an area. The spatial distribution of monthly MI trend values in Fig. 6A shows the magnitude
and direction of change in monthly MI over the 25 years. While the positive increase in MI
covered 31% of the U.S., only -1% was a significant increase. Sixty-nine percent of the U.S. had
an MI decrease with 14% having a significant decrease (inset map in Fig. 6A). North Dakota and
Montana have the highest increase in MI, large areas in Louisiana and Texas have the lowest MI.
The spatial distribution of the average MI in Fig. 6B presents the gradients in monthly average
moisture conditions/regimes in the nation over the 25 years. The zero value of MI (purple polygons
in Fig. 6B) where PET equals precipitation, stretches from eastern Texas to the northwestern U.S.
and divides the nation into a relatively homogenous moist area to the east and drier areas to the
west. The largest dry areas are in southern Arizona, California, and Nevada; and a mosaic of all
MI groups in the northwest.
The spatial distribution of heat index (AT) (Fig. 7) shows the widespread of increase in AT
values over 25 years. Seventy-eight percent of the U.S. experienced an increase in AT from 1989
to 2013; one-third of the total area experienced a significant increase in AT (inset map in Fig. 7).
Most states experience higher, except for North Dakota, and the coastal area in Washington and
Oregon. Only -3% of the USA had a significant decrease in AT; concentrated primarily in western
Oregon and northern Montana (inset map in Fig. 7). The 1st and 99th percentile for AT were -
0.0049 and 0.0088, respectively. Following Smoyer-Tomic and Rainham (2001), the spatial
distribution of number of months for AT groups a-c (see Material and Method for group
description) are presented in Figs 8A-C. Most areas had at least one month of AT values between
26.7 and 32.2°C (Fig. 6A; 26.7 °C 
-------
u< = -0.00093
¦	-0.00092 - -0.00060
¦	-0.00059 - -0.00030
~-0.00029 --0,00015
¦	-0.00014 -0,00000
~ 0.00001 -0.00010
m 0.00011 -0.00030
¦	0.00031 -0.00054
¦	> 0.00054
¦	- < -0.66 (And)
¦	-0.66 ¦ -033 (Semi-Arid)
O -0 32 • o (Dry)
¦	0.00
¦	0.01 - 033 (Moist)
¦	034 - 045 (Wet)
¦	> = 0,66 (Ve«y Wet)
Figure 6. Pixels (1 km2) with temporal trends in monthly (A) MI. the first and last legend classes represent the 1st
and 99th percentiles. (B) Moisture regimes (Feddema (2005) categories) based on the average monthly
moisture index (MI) over 25 years (1989-2013) a zero group is a divider between dry and wet areas. The
inset map in (A) is the probability of temporal trend, green indicates significant increase and red indicates
a significant decrease.
14

-------
Figure 7. Pixels (1 km2) with temporal trends in monthly AT. The first and last legend classes represent the 1st and
99th percentiles. The inset map is the probability of temporal trend, green indicates significant increase and
red indicates a significant decrease.
concentrated in eastern states, Florida, and western states (Texas, New Mexico, Arizona, Nevada
and California). AT values between 32.2 C to 40.6 C are clustered in a smaller geographic area
(Fig. 8B). Areas in the top 10th percentile were concentrated in the southern coastal area extending
from Florida to Texas. For/4r bet ween 40.6°C and 54.4°C, an even smaller area was covered, and
the top 10th percentile areas were found mainly in southern Texas, Arizona, and California (Fig.
8C).
Average and variability are often used to describe climatic factors changes over time. We
present the overall average and coefficient of variability (CV) of the monthly minimum, maximum,
and average temperatures in Fig. 9. Highest average temperatures are clustered in the southern
part of the U.S. and decreased toward the northern part of the U.S. mainly in North Dakota and
Minnesota. While the average temperature decreases, highest variability occurs in northern
15

-------
Midwestern states and in several clusters within the states of Utah, Idaho, and Montana. (Changes
in temperatures using average and CV values based on different temporal scales are discussed
below).
Annual: Average and Coefficient of Variability for Annual Precipitation, Potential
Evapotranspiration (PET), and Moisture Index (MI).
The spatial distribution of average and CV for annual precipitation and annual PET are
shown in Fig 10. Average precipitation was higher in the eastern and the north Pacific coastal areas
than most of the western states (Fig. 10A). The general pattern of the precipitation was in
longitudinal parallel bands at mid U.S. with clusters in the southern states (Alabama, Mississippi,
and Louisiana) and western Washington and Oregon. Highest precipitation values are within these
clusters. The variability (CV) in precipitation is the highest in areas with low precipitation:
California, southern Nevada, and eastern Arizona on the southern border of New Mexico and
Arizona. Variability radially decreases as one moves from the southwest toward the northeast.
PET ranged from 435 mm (-17 inches) in light blue areas in Fig. 10B (Northeast, areas
around the Great Lakes, and in the high Rocky Mountains) to highest PET of at least 1000 mm (39
inches) in dark blue area (Florida, south Texas, south Arizona, southwest California and southern
Nevada). This potential loss may or may not be compensated by the amount of precipitation. In
order to consider the balance of PET with available precipitation, we combined the yearly
precipitation and PET (Equation 3) to calculate the yearly MI. The annual average and CV of MI
are shown in Fig. 11. The pattern of distribution for average annual MI (Fig. 11A) in the eastern
part of U.S. was more dominated by precipitation (Fig. 10A & Fig. 10B) than PET (Fig. 10C &
Fig. 10D). Highest variability in annual MI (Fig. 1 IB, in blue) extended in parallel bands from the
north along the eastern border of the Dakotas to the eastern side of Texas.
16

-------
Figure 8. Distribution of total number of months (out of 300) that had at least one .17' value in group: A) fatigue
possible with prolonged exposure and/or physical activity (26.7°C < TA < 32.2°C) B) sunstroke, heat
cramp, and heat exhaustion possible with prolonged exposure and/or physical activity (32.2°C 
-------
-9-0
1-3
4-6
7-10
11 -23
2-13
14-16
17-20
21 -24
25-33
33 - 45
46 - 60
61 -78
79 - 275
1-2-7
8 • 10
11 -14
15-17
18-25
650 1,300
-100
1100 - 300
300 - 500
500
2,600 km
_i	i
Figure 9. Average and coefficient of variability (CV) of monthly minimum temperature (A & B). monthly maximum
temperature (C & D) and monthly average temperature (E & F) (n=300). The CV legend represents the
absolute value (e.g. 100 is abs (±100)). Higher CV values represent higher variability in climatic factor.
18

-------
3«
2,000 km
¦ ¦ '
435
518
565
606
654
719
801
896
1,002
1-1,681
C
1-10
11-13
14 - 16
7 - 19
!0 - 23
'4-27
8 - 32
Figure 10. Precipitation annual average (A) and coefficient of variability (B), PET annual average (C) and
coefficient of variability (D) (n=25).
Seasonal: Average and Coefficient of Variability for Seasonal Precipitation, Potential
Evapotranspiration (PET), and Moisture Index (MI).
The spatial distribution pattern for precipitation varies with the temporal scales, which can
be seen by comparing the average (Fig. 12) and CV (Fig. 13) for seasonal precipitation with that
of the average of annual precipitation (Fig. 10A). The average amount of precipitation was higher
in the winter season, followed by fall, spring, and summer (Table 1). The general spatial
distribution of seasonal average precipitation (Fig. 12A-D) shows that higher precipitation was on
the eastern side of the nation rather than that on the western side, except for the northern Pacific
19

-------
coastal area which had low precipitation in summer only. The top 10th percentile of precipitation
in the south-eastern states moves to southern coastal areas in summer (Fig. 12).
A significant increase in winter precipitation was mainly clustered in Michigan, Wisconsin,
and Minnesota (Fig. 12A). For the spring season, a significant increase in precipitation -occurs in
northern Minnesota, in most of North Dakota and northeastern Montana (inset maps in Fig. 12B).
For summer and fall precipitation, smaller clusters with significant increases were within Maine
and eastern New York (summer), mid-Montana and southern Nevada (fall). About 39% of the
areas had significantly decreasing all season precipitation while increasing precipitation occurred
in only 23% of the areas. Spring precipitation decreased more than in other season (13%) and it
was mainly within Louisiana, Arizona and southern Nevada and California. The significant
decrease in summer precipitation covered fewer areas than in the spring seasons (12%) (Fig. 12C).
The significant decrease in fall precipitation (5%) was clustered mainly in South Carolina and
Oklahoma (inset map in Fig. 12C). The overall monthly trend for precipitation showed that much
of the significant decrease in precipitation was clustered in Oklahoma and Texas (Fig. 4).
Coefficient of variability (Fig. 13) describes the distribution of variability in precipitation
at different temporal scales. In general, the variability of seasonal precipitation (Fig. 13) is greater
in the western part of the nation than in the eastern part of the nation. Higher variability occurs in
areas where the precipitation is low, at the center of the nation (Fig. 13) extending from north to
south, reflecting the pattern in the distribution of average seasonal precipitation. Average seasonal
precipitation (Fig. 12), closer to median value, is located at the center of the nation extending from
north to south.
20

-------
Figure 11. Yearly average for MI (A) and coefficient of variability (CV) (B) (n=25 years). The legend in (B)
represents the absolute value of CV (e.g. <= 10 is <= abs (±10)).
21

-------
~1
n2
2
2
3
3
129
166
167
216
217
2&4
255
282
283
320
321
364
365
1.410
197 - 138
139-186
|lB7 - 228
229 - 263
264 - 291
292 - 326
327- 1,78ft
0	6S0 1,300	2,600 km
	1	i	i	i	i	i	i	i	i
63-93
94 - 146
147¦ 198
199 - 240
241 -292
293 - 386
387-2.673
Figure 12. The ten quantiles for the average seasonal precipitation for (A) winter, (B) spring, (C) summer, and
(D) fall.
22

-------
¦a
-K
»
-»
•*
-»
-44
SO
•«
m
-	M
-»
¦	it
-	17
-	39
•m
¦	Sl
»
179
Figure 13. Coefficient of variability for (A) winter precipitation, (B) spring precipitation, (C) summer precipitation
and (D) fall precipitation. Winter precipitation (Dec - Feb), spring precipitation (Mar - May), summer
precipitation (Jun - Aug), fall precipitation (Sep - Nov).
Average seasonal value, seasonal trend and significant change in PET are presented in Figs
14 & 15. In all seasons, the highest PET annual average was located in the southern part of the
nation, southern California, Nevada, Arizona, Texas, Alabama, Georgia, and Florida. While PET
decreased in most areas in winter with less areal coverage in spring, PET increased significantly
in approximately all areas in summer followed by fall (insets in Fig. 15). The spatial distribution
of the average annual PET (Fig. IOC) was close in spatial distribution to that of the spring season
(Fig. 14B).
23

-------
000

0 01 -
3.70
3 71 -
11 09
11 1 ¦
23.11
23.12
-39.75
39 76
- 58.23
5824
¦ 7395
73 96
- 87 81
87 82
- 11092
110 93-235 7
10 50 - 85 93
8594- 107 69
107 7- 120 74
120 75- 130 89
130 9-143 95
143 96 - 158 46
158 47- 180 21
180 22-207 77
207 78-241 14
241 15-380 39
Figure 14. Average seasonal PET for (A) winter, (B) spring. (D) summer, and (E) fall. Winter precipitation (Dec
- Feb), spring precipitation (Mar - May), summer precipitation (Jun - Aug), fall precipitation (Sep -
Nov).
650
	i	
108.78-
120 82-
137 37
158 44-
18102-
12081
137 36
158 43
181 01
205 09
1,300
_j	I	i
2051 -242 72
42.73 - 414 3
2,600 km
_l	1
24

-------
-2 33 - -0 57
•0 58--0 39
-0 38-0 27
-0.26- -0.19
-0 18- -0 13
-0 12- -O.M
-0 07 - -0,02
-0 01 - 0 06
0 07 • 0 22
0.23-2.71
Figure 15. PET seasonal trend value for (A) winter, (B) spring, (C) summer and (D) fall. The inset figures are the
probability of trend.
The distribution of the MI (Fig. 16), seasonally characterized by low MI values (drier) in
the summer season, were spatially extended from mid-western to western U.S. Higher MI values
(wetter) in winter extended from north to south (Fig. 16). Regardless of being seasonal or annual,
25

-------
the eastern part of the nation was wetter than the west. The spatial distribution of the annual MI
is more likely to be between fall/spring pattern, but it was much different than winter and summer
seasons. MI increases significantly mostly in the northern part of the nation for winter and spring
but decreases significantly in the southern/western parts (inset maps in Fig. 16).
26

-------
Winter
Spring
Summer
Annual
¦	< -0.66 (Arid)
¦	-0.66 - -0.33 (Semi-Arid)
O -0.32 - 0 (Dry)
~ 0.01 -0.33 (Moist)
¦	0.34 -0,65 (Wet)
¦	> = 0.66 (Very Wet)
Figure 16. The 25-year average of MI for seasonal and annual. Legend is the Feddema et al. (2005) drought
classification: very wet [MI > 0.66], wet [0.66 > MI > 0.33], moist [0.33 > MI > 0.00], Dry [0.00 > MI
> -0.33 [, semi-arid [-0.33 > MI > -0.66]; and arid [MI < -0.66], Inset maps are the probability of trend.
Red denotes significant decrease and green denotes significant increase.
27

-------
28

-------
Summary and Discussions
The general pattern of increase in temperature in the 48 contiguous states is consistent with
other studies of climate change, where the trend for increasing average temperatures was higher
for 1992-2008 than that of 1961-1979 (observational data, Karl and Melillo, 2009). An increasing
trend in daily minimum temperatures (or nighttime temperature) was higher than the trend for daily
maximum temperatures (daytime temperature) in western and central North America (Robeson,
2004). Millett et al. (2009) analyzed temperature and precipitation in the Central Plains of the
U.S. in the 20th century and found increases in precipitation and minimum temperatures. The
number of months per year that contain at least one month with a < 0°C temperature decreased
significantly; Vincent and Mekis (2006) found a similar trend in Canada between 1950-2003.
Although it is not within our study boundary, Alaska experienced the largest warming trend in
temperature from 1951 to 2001 where the temperature increased from 0.8°C to 1.9°C, respectively.
This warming occurred mostly in the winter season and coincided with the 1977 Arctic
atmospheric and ocean regime shift (Hartmann and Wendler, 2005). Robles and Enquist (2011)
found that precipitation in the United States increased over the past five decades by 5%; our
findings suggest that < 3% of the U.S. had a significant increase in monthly precipitation during
1989-2013.
Dew-point is an essential element in fog formation. Hiatt et al. (2012) indicated that despite
low rainfall, vegetation productivity did not decrease in the central and southern coastal regions in
California due to fog deposition. Wet deposition of atmospheric nitrogen and sulfur via fog
formations have been well documented (Fenn et al. 2000, 2003; Klemm and Wrzesinsky, 2007;
Lovett and Tear, 2008; Polkowska et al. 2011). Spatial distribution of the dew point temperature
changes can help inform fog formation locations.
29

-------
Potential Evapotranspiration (PET) represents the amount of water loss via plant and soil
transpiration. PET is estimated from temperature, water saturated vapor pressure (Equations 1 &
2) and daylight length which varies with season and spatial locations. Thus, it is important to look
at precipitation and PET in concert to determine the magnitude and direction of water loss or gain
across seasons and spatial location. Hinzman et al. (2005) reported that the difference between
precipitation and PET decreased during the warm season from 1960 to 2000 in Alaska and was
mainly related to an increasing in air temperature (precipitation did not change significantly). On
the Tibetan Plateau where daytime temperature trend is unchanged, Shenbin et al. (2006) found
that the negative trend in PET is more related to the regional monsoon circulation. Millett et al.
(2009) in the Prairie Pothole Region of the U.S. found that an east to west moisture gradient
steepened as temperature and precipitation shifted in the 20th century, resulting in a wetter eastern
section and a drier western section. The Moisture index (MI) formally combines PET and
precipitation, thus MI can demonstrate resulting effects of interactions of temperature and
precipitation both across space and changes/shifts through seasons. Grundstein (2009) showed
significant increases in annual MI (wetter conditions) for climate divisions in the Southeast and in
a band across the north, stretching from South Dakota to New York. Areas that may not show
annual shifts in MI may still have seasonal MI shifts which may be important due to hydrological,
biological and biogeochemical dependence on the timing of available water. For example,
decreases in snowfall or earlier snow melt in the Western U.S. impacts water storage, the timing
of stream flows and the biota that rely on those streams (Hamlet et al. 2005; Mote et al. 2005).
Likewise, an increase in precipitation in Great Yellowstone Ecosystem since 1977 has not been
enough to compensate for water loss by PET. This deficit has been exacerbated by the increase in
30

-------
minimum temperature which impacted snowpack and has resulted in a reduced flow through
streams, especially in summer season (Chang and Hansen, 2014).
31

-------
32

-------
Conclusion
Our analysis per pixel for the entire U.S. offers a means of monitoring changes in trend
direction and magnitude for a specific area of interest. Choosing the appropriate time scale
(monthly, annual or seasonal) in climatic factors in building a relation with a response (e.g.,
vegetation, fills and spills from water bodies) is an important step to explore. The spatial and
temporal analyses of these climatic factors combined with factors of interest (fires, health factors,
environmental factors, etc.) can be used to examine different responses to direct impacts and/or
climate changes in an area.
33

-------
34

-------
Implication
We envision the use of these data sets in diverse multidisciplinary studies and research. In
previous studies, we used pixel-based univariate versus multivariate autoregression models with
these climatic factors to demonstrate that long-term monitoring can potentially detect broad-scale,
slow changes, such as those caused by climate change over decades, as well as more local and
rapid changes such as those caused by fire, agriculture, land clearing, and habitat restoration over
time. (Nash et al. 2014, 2017). Such monitoring can provide environmental decision-makers with
early warning signals for widespread general trends, as well as a means to identify specific areas
where land conditions are degrading or improving. Using a similar multivariate approach, we
assessed the dynamic interaction between climate, argan trees, local communities, rural household
welfare, and forest conservation and sustainability in Morocco (Lybbert et al. 2011). We have
applied Thornthwaite and Mather's (1957) approach using the monthly PET, precipitation, and
depth of soil to calculate water balance in selected forested areas in U.S. (Nash unpublished data).
The distribution of significant changes, extremes, variability and averages at different temporal
scales provide a road map to the availability and important trends of temperature, rainfall and
available moisture that have been occurring across the U.S.
35

-------
36

-------
References
Chang, T. and A.J. Hansen, 2014. Climate Change Brief: Greater Yellowstone Ecosystem.
Landscape. Climate Change Vulnerability Project, p. 1-8.
http://www.montana.edu/lccvp/documents/LCCVP GYE ClimateBrief.pdf. Accessed
Dec 7, 2017.
Enquist, C.A.F., E.H. Girvetz, and D.F. Gori, 2008. A climate change vulnerability assessment for
biodiversity in New Mexico, Part II: Conservation implications of emerging moisture stress
due to recent climate changes in New Mexico. The Nature Conservancy.
Feddema, J.J., 2005. A revised Thornthwaite-type global climate classification. Physical
Geography, 26(6): 442-466.
Fenn, M.E., M.A. Poth, S.I	Schilling, and D.B. Grainger, 2000. Throughfall and fog deposition
of nitrogen and sulfur at an N-limited and N-saturated site in the San Bernardino
Mountains, southern California. Canadian Journal of Forest Research, 30:1476-1488.
Fenn, M.E., R. Haeuber, G.S. Tonnesen, J.S. Baron, S. Grossman-Clarke, D. Hope, D. A. Jaffe,
S. Copeland, L. Geiser, H.M. Rueth, and J.O. Sickman, 2003. Nitrogen Emissions,
Deposition, and Monitoring in the Western United States. Bioscience, 53(4): 391-403.
Forsythe, W.C., E.J. Rykiel Jr., R.S. Stahl, H.Wu, and R.M. Schoolfield, 1995. A model
comparison for daylength as a function of latitude and day of the year. Ecological Modeling
80:87-95.
Grundstein, A., 2009. Evaluation of climate change over the continental United State using a
moisture index. Climate Change, 93:103-115. DOI 10.1007/s 10584-008-9480-3.
37

-------
Hamlet, A.F., P.W. Mote, M P. Clark, and D P. Lettenmaier, 2005. Effects of temperature and
precipitation variability on snowpack trends in the western United States. Journal of
Climate, 18(21), 4545-4561.
Hartmann, B., and G., Wendler, 2005. The Significance of the 1976 Pacific Climate Shift in the
Climatology of Alaska. Journal of Climate, 18: 4824-4839.
Hamon, W.R., 1961. Estimating potential evapotranspiration. Journal of the Hydraulics Division,
American Society of Civil Engineers 87:107-120.
Hiatt, C., D. Fernandez, and C. Potter, 2012. Measurements of fog water deposition on the
California central coast. Atmospheric and Climate Science 2: 525-531.
Hinzman, L.D., N.D. Bettez, W.R. Bolton, et al. 2005. Evidence and implications of recent climate
change in northern Alaska and other Arctic Regions, Climatic Change, 72: 251-298. DOI:
10.1007/sl0584-005-5352-2.
Karl, T. R., J. T. Melillo, and T. C. Peterson, 2009: Global Climate Change Impacts in the United
States. T.R. Karl, J.T. Melillo, and T.C. Peterson, Eds. Cambridge University Press, 189
pp. Online at:
http://downloads.globalchange.gov/usimpacts/pdfs/climate-impacts-report.pdf
Klemm, O., and T. Wrzesinsky, 2007. Fog deposition fluxes of water and ions to a mountainous
site in Central Europe. Tellus, 59: 705-714.
Lovett, G.M., and T.H. Tear, 2008. Threats from Above: Air Pollution Impacts on Ecosystems and
Biological Diversity in the Eastern United States. The Nature Conservancy and the Cary
Institute of Ecosystem Studies.
38

-------
Lybbert, T.J., A. Aboudrare, D.J. Chaloud, N. Magnan, and M.S. Nash, 2011. Booming Markets
for Moroccan argan oil appear to benefit some rural households while threatening the
endemic argan forest. Proceeding of the National Academy of Science 108:13963-13968.
Millett, B.V., W.C. Johnson, and G.R. Guntenspergen, 2009. Climate trends of the North
American prairie pothole region 1906-2000. Climatic Change, 93: 243-267.
Mote, P.W., A.F. Hamlet, M.P. Clark, and D.P. Lettenmaier, 2005: Declining mountain
snowpack in western North America. Bull. Amer. Meteor. Soc., 86, 39-49.
Nash, M.S., D.F. Bradford, J.D. Wickham, and T.G. Wade, 2014. Detecting change in landscape
greenness over large areas: An example for New Mexico, USA. Remote Sensing of
Environment. 150: 152-162.
Nash, M.S., J. Wickham, J. Christensen, and T. Wade, 2017. Changes in landscape greenness and
climatic factors over 25 years (1989-2013) in the USA. MDBIRemote Sensing, 9 (3), 295;
doi:10.3390/re9030295.
Perry, A.G., M.J. Korenberg, G.G. Hall, and K.M. Moore, 2011. Modeling and syndromic
surveillance for estimating weather-induced heat-related illness. Journal of Environment
and Public Health, I -10, doi: 10.1 155/201 1/750236.
Polkowska, Z., T. Gorecki, and J. Namiesnik, 2011. Determination of atmospheric pollutants in
wet deposition. Environmental Reviews, 19: 185-213. https://doi.org/10.1139/al 1-006
Robeson, S.M., 2004. Trends in time-varying percentiles of daily minimum and maximum
temperature over North America. Geophysical Research Letters, 31, DOI:
1029/2003GL019019.
39

-------
Robles, M.D., and C. Enquist, 2011. Managing changing landscapes in the Southwestern United
States. The Nature Conservancy. Tucson, Arizona. 26 pp. Monitoring Systems Laboratory,
Office of Research and Development, U.S. Environmental Protection Agency, Cincinnati,
Ohio.
Shenbin, C., L. Yunfeng L., and A. Thomas, 2006. Climatic change on the Tibetan plateau:
potential evapotranspiration trends from 1961-2000. Climatic Change, 76: 291-319. DOI:
10.1007/s 105 84-006-9080-z
Smoyer-Tomic, K.E., and D.G.C. Rainham, 2001. Beating the heat: development and evaluation
of a Canadian hot weather health-response plan. Environmental Health Perspectives.
109(12) 1241-1248.
Thorne, J., R. Boynton, L. Flint, A. Flint, and T. Le, 2012. Development and Application of
Downscaled Hydroclimatic Predictor Variables for Use in Climate Vulnerability and
Assessment Studies. California Energy Commission. Publication number: CEC-500-2012-
010.
Thornthwaite, C.W., J.R. Mather, and D.B. Carter, 1957. Instruction and tables for computing
potential evapotranspiration and the water balance. Drexel Institute of Technology,
Philadelphia. Publications in climatology. 10(3): 185-311.
Vincent, L. and E. Mekis, 2006. Changes in daily and extreme temperature and precipitation
indices for Canada over the twentieth century. Atmosphere-Ocean, 44, 177-193.
Willmott, C.J. and J.J. Feddema, 1992. A More Rational Climatic Moisture Index. The
Professional Geographer, 44(1), 84-88.
40

-------

-------
&EPA
United States
Environmental Protection
Agency
Office of Research
and Development (8101R)
Washington, DC 20460
Please make all necessary changes on the below label,
detach or copy, and return to the address in the upper
left-hand corner.
If you do not wish to receive these reports CHECK HERE ~;
detach, or copy this cover, and return to the address in the
upper left-hand corner.
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA PERMIT No. G-35
Official Business
Penalty for Private Use
$300
EPA/600/S-17/434
October 2017
www.epa.gov
Recycled/Recyclable
Printed with vegetable-based ink on
paper that contains a minimum of
50% post-consumer fiber content
processed chlorine free

-------