Irrigation cooling effect on temperature and heat index extremes

Abstract

1] Previous studies of the long-term climate effects of irrigation have focused on average monthly temperatures. Given the importance of temperature (T) extremes to agriculture and human health, we evaluated irrigation induced changes in various metrics of T extremes using daily observations in California and Nebraska. In addition, simulations from a regional climate model were used to evaluate irrigation effects on T and heat index (HI; also known as the discomfort index) extremes in California, with the latter representing a combined measure of T and humidity. Contrary to our expectation that irrigation would have larger effects on hot days when sensible heat fluxes are higher, both observations and a regional climate model indicate that irrigation cools T on the hottest days of the year by a similar magnitude as on an average summer day. The HI is also reduced by irrigation, but by a much smaller magnitude than T because of the higher humidity above irrigated surfaces. Interestingly, HI is influenced less on the most extreme days than on average days, because of the nonlinear effect of humidity on HI at high T.

Full text

Irrigation cooling effect on temperature and heat index extremes
David B. Lobell,
1
Celine J. Bonfils,
2
Lara M. Kueppers,
3
and Mark A. Snyder
4
Received 28 March 2008; accepted 4 April 2008; published 7 May 2008.
[1] Previous studies of the long-term climate effects of
irrigation have focused on average monthly temperatures.
Given the importance of temperature (T) extremes to
agriculture and human health, we evaluated irrigation
induced changes in various metrics of T extremes using
daily observations in California and Nebraska. In addition,
simulations from a regional climate model were used to
evaluate irrigation effects on T and heat index (HI; also
known as the discomfort index) extremes in California, with
the latter representing a combined measure of T and
humidity. Contrary to our expectation that irrigation
would have larger effects on hot days when sensible heat
fluxes are higher, both observations and a regional climate
model indicate that irrigation cools T on the hottest days of
the year by a similar magnitude as on an average summer
day. The HI is also reduced by irrigation, but by a much
smaller magnitude than T because of the higher humidity
above irrigated surfaces. Interestingly, HI is influenced less
on the most extreme days than on average days, because of
the nonlinear effect of humidity on HI at high T.
Citation: Lobell, D. B., C. J. Bonfils, L. M. Kueppers, and
M. A. Snyder (2008), Irrigation cooling effect on temperature
and heat index extremes, Geophys. Res. Lett., 35, L09705,
doi:10.1029/2008GL034145.
1. Introduction
[2] The impacts of climate change on agriculture will
depend, in part, on changes in the frequency and intensity
of extreme events. While most impact assessments have
focused on changes in monthly or growing season aver-
ages, a small but growing number of studies have begun to
quantify the response of cropping systems to temperature
and precipitation extremes [Rosenzweig et al., 2002;
Schlenker and Roberts, 2006; White et al., 2006]. A
current challenge to these efforts is the reliability of
climate model projections of extreme temperat ures in
agricultural areas. More specifically, climate models have
relatively simple treatments of land use that may ignore
important processes affecting extreme events.
[
3] For example, soil moisture is an important control on
heat and water transfer between the land and atmosphere,
which in turn affects the development of extreme he at
events [Ferranti and Viterbo, 2006]. While roughly 17%
of global croplands are irrigated, none of the climate models
included in the fourth assessment report (AR4) of the
Intergovernmental Panel on Climate Change (IPCC)
includes a representation of soil moisture changes due to
irrigation. Meteorological studies have demonstrated effects
of irrigation on surface temperatures, cloud formation and
precipitation at local to continental scales [Adegoke et al.,
2003; Segal et al., 1998]. Analysis of long-term observa-
tions [Bonfils and Lobell, 2007; Mahmood et al., 2006] and
climate or land surface modeling [Haddeland et al., 2006;
Kueppers et al., 2007] efforts have also shown that irriga-
tion can consistently reduce maximum daily temperatures
by up to 7.5°C upon irrigation. However, most of these
studies have not explicitly considered extreme temperatures.
Barnston and Schickedanz [1984] argued that the cooling
effect of irrigation in Texas would be significantly larger on
the hottest days because of reduced humidity relative to cool
days. However, in an analysis of USHCN station data,
Lobell and Bonfils [2008] found similar effects of irrigation
on average and very hot summer days in California.
[
4] The main goal of this paper is to evaluate the long-
term climatological effect of irrigation on extreme temper-
atures, using two independent approaches. First, we conduct
an empirical analysis based on gridded daily temperature
data sets in both California and Nebraska. Second, we
utilize simulations from a previously published regional
climate model (RCM) experiment in California. A second-
ary goal is to evaluate effects of irrigation on extremes in
the heat index (HI), which is a measure of discomfort that
combines temperature and relative humidity (RH) [Schoen,
2005]. We consider the HI because it is often a more useful
predictor of human health effects and mortality than tem-
perature itself, and provides a basis for heat advisories in the
United States [Davis et al., 2003]. For the HI analysis we
focus only on the RCM simulations, as long-term spatially
complete gridded data sets of daily RH observations are
currently unavailable.
2. Methods and Models
2.1. Temperature
[
5] Our analysis of historical observations followed the
method of Bonfils and Lobell [2007] by comparing histor-
ical temperature trends in areas with high levels of irrigation
(>50% of area equipped for irrigation) with trends from a
nearby ‘‘reference’’ area with 0.1 –10% irrigation. Analyses
were conducted for the Central Valley of California (CA)
and the irrigated plains of Nebraska (NE), which represent
the two regions in the United States with the most amount
of land that is intensively (>50% of area) irrigated. In both
regions, only grid cells below an elevation of 500 m were
used to avoid bias from comparing cells in low-lying
irrigated areas to higher eleva tions. A 1/ 12°  1/1 2°
resolution map of area equipped for irrigation [Siebert et
GEOPHYSICAL RESEARCH LETTERS, VOL. 35, L09705, doi:10.1029/2008GL034145, 2008
Click
Here
for
Full
A
rticl
e
1
Program on Food Security and the Environment, Stanford University,
Stanford, California, USA.
2
Lawrence Livermore National Laboratory, Livermore, California, USA.
3
School of Natural Sciences, University of California, Merced,
California, USA.
4
Climate Change and Impacts Laboratory, Department of Earth and
Planetary Sciences, University of California, Santa Cruz, California, USA.
Copyright 2008 by the American Geophysical Union.
0094-8276/08/2008GL034145$05.00
L09705 1of5

al., 2005] was used to delineate the irrigated and reference
regions, which are illustrated in Figure 1.
[
6] For gridded temperatures, we used 1/8°  1/8° resolu-
resolution grids of historical daily minimum (T
min
) and
maximum (T
max
) temperatures, obtained from the Surface
Water Modeling g roup at the University of Washington
(http://www.hydro.washington.edu/Lettenmaier/Data/gridded/).
Two versions of this data set exist, an earlier one that covers
the entire the United States for January 1949 –July 2000
[Maurer et al., 2002], and a more recent one that corrects
for temporal inhomogeneities in station data and extends
from January 1915–December 2003 but covers only select-
ed western states [Hamlet and Lettenmaier, 2005]. In this
study, we use the newer version (UW2) for CA but rely on
the first version (UW1) for NE, which is beyond the current
spatial extent of UW2.
[
7] The irrigation map was re-sampled to the sl ightly
coarser resolution of the temperature data sets, and spatial
averages of daily T
max
were computed for both the irrigated
and reference regions for the relevant study periods (n
CA_irr
=
227 grid cells; n
CA_ref
= 147; n
NE_irr
=9;n
NE_ref
= 428). In
CA, we used the period 1915–1980, as this represents the
period of most rapid irrigation expansion during which
irrigated land area doubled [Bonfils and Lobell, 2007]. In
NE, the coverage of the UW1 data set limited the analysis to
1950–1999, a period in which irrigated area increased by a
factor of mor e than ei ght (http://www.ers.usda.gov/Data/
MajorLandUses/).
[
8] We then computed several metrics of temperature
extremes using daily T
max
. Following Hegerl et al.
[2004], we computed average T
max
on the hottest 1, 5, 10,
and 30 days of each year, and also the average T
max
for
June–August (JJA) for comparison. An index of heat wave
duration was computed using the warm spell duration index
(WSDI), which is one of the extreme indices defined by the
Expert Team on Climate Change Detection and Indices
(ETCCDI, http://cccma.seos.uvic.ca/ETCCDI/list_27_indices.
shtml.) The WSDI is defined as the annual count of days
with at least six consecutive days when T
max
exceeds the
90th percentile of the climatology. The climatology was
defined for each day using values from 1961–1990 for a
five day window centered on that day (giving a total of 150
values for each day). Linear trends in all extremes were
computed using ordinary least squares regression for the
entire study period, 1915–1980 in CA and 1950–1999 in
NE, and are reported in °C decade
1
. To estimate the effects
of irrigation, we calculated trends in the differences of
extremes between irrigated and reference regions. In this
manner, climate variations or trends (ascribable to natural
modes of climate variability or resulting from other forcing
agents) that are common to both irrigated and reference
regions do not affect the analysis [Bonfils and Lobell, 2007].
The 95% confidence interval was computed non-paramet-
rically via bootstrap resampling (n = 50) of the original time
series. The different data sets and time periods used in the
two regions caution against direct comparison of the values
derived in each region, and instead we focus on a qualitative
comparison of results in each region.
[
9] To complement the observational study, we ana-
lyzed RCM simulations from the experiment described by
Kueppers et al. [2007]. Briefly, we performed two simu-
lations with RegCM3 [Pal et al., 2007] at a horizontal
resolution of 30km, using NCEP/DOE Reanalysis II as lateral
boundary conditions for the period 1979 –2000. One simu-
lation used California land cover defined as potential natural
vegetation (NAT) and one used modern day land cover that
included irrigated cropland (MOD), where soil moisture for
irrigated land is maintained at field capacity all year round.
For the current study, we spatially averaged daily T
max
for all
grid cells converted to irrigated agriculture in the Central
Valley (between 24– 30°N and 118–124°W; n = 27 grid
cells) in the MOD run for each simulation, and computed
temperature extremes for the final 20 years of the simulation.
Heat wave extremes i n MOD were computed using the
climatology from NAT. We calculated differences between
MOD and NAT 20-year averages for each T and HI metric to
estimate the effect of irrigation on extremes in the RCM.
2.2. Heat Index
[
10] As described by Kueppers et al. [2007], the simu-
lated cooling of average monthly T
max
is largely driven by
higher latent heat fluxes, which also leads to substantial
increases in RH. A relevant issue is therefore the net effect
on HI, which decreases with lower T but increases with
higher RH. We computed the daily maximum HI (HI
max
)for
the MOD and NAT runs using the definition of Schoen
[2005], which is an empirical fit to a data table generated
from a model of human physiology:
HI ¼ T  1:0799 e
0:03755T
1  e
0:0801 D14ðÞ
hi
ð1Þ
where T is temperature and D is dewpoint, which varies
with RH and T. Figure 2 illustrates the relationship between
HI and RH for different levels of T, demonstrating that RH
has a stronger effect on HI on warmer days.
3. Results and Discussion
3.1. Temperature
[
11] Observed trends in T
max
were negative for irrigated
grid cells in both CA and NE over the study periods, and
slightly positive for the reference region in CA but negative
for the reference region in NE (Figure 3). The time series of
Figure 1. The (left) California (CA) and (right) Nebraska
(NE) study regions. Dark blue cells indicate the irrigated
region where greater than 50% of the grid cell area is
equipped for irrigation. Light brown indicates the associated
reference region with 0.1 – 10% area equipped for irrigation,
according to the maps of Siebert et al. [2005]. Both regions
were restricted to grid cells with mean elevation below
500m.
L09705 LOBELL ET AL.: IRRIGATION COOLING EFFECT L09705
2of5

differences between the irrigated and reference regions
exhibited significant negative trends for both states.
[
12] The cooling effect on the hottest days, although
slightly larger, was indistinguishable from that on the
average summer day. We therefore find little evidence that
irrigation has a larger or smaller effect on particularly hot
days.
[
13] Trends in WSDI wer e not signifi cantly different
from zero for the irrigated and reference regions, or for
the difference between the two (Figure 3). A significant
effect of irrigation was observed for the number of heat
waves in CA (the number of times per year that six or more
consecutive days had T
max
above the 90th percentile), but
not in NE (not shown). Overall, there was therefore no clear
signal of changes in heat waves resulting from irrigation,
despite significant effects of irrigation on T
max
extremes. A
possible explanation is that WSDI has many years with a
value of zero (e.g., 25 out of 66 years in CA reference
region), which leads to an odd statistical distribution with
several values of zero and all other values greater than five.
Trend detection in time series of WSDI is therefore difficult
as the values can appear quite noisy, which has led some to
exclude heat wave indices altogether from their analysis
[Kiktev et al., 2003]. Others, however, have found signifi-
cant trends in WSDI for many regions, although not CA or
NE [Alexander et al., 2006].
[
14] The results of the RCM simulations were generally
consistent with those from the observational analysis.
Namely, T
max
extremes were reduced by irrigation, but by
an amount that differed little from the effect on an average
summer day (Figure 4a). The primary mechanism respon-
sible for this cooling was an increase in latent heat flux and
corresponding reduction in sensible heat flux, as described
in detail by Kueppers et al. [2007]. That the increase in
latent heat flux was the same on average and very hot days
indicates that evapotranspiration in RegCM3 for conditions
typical of California summer is limited by a factor other
than temperature, such as solar radiation, stomatal conduc-
tance, or wind speed. Indeed, there is little day-to-day
correlation between evapotranspiration and daily tempera-
ture in the model, both for MOD and NAT experiments (not
shown). The WSDI was reduced in the RCM from an
average of 8.9 days per year in NAT to 1.7 days per year
in MOD, a decrease equivalent to roughly one heat wave
per year.
[
15] The magnitude of T
max
decreases in the RCM from
irrigation, for example 7.0°C for JJA, was substantially
larger than the 0.13°C per decade (or roughly 1.0°C over
the 1915–1980 period) in the UW analysis for CA. These
observed and simulated changes are however not directly
comparable. One obvious reason for this is that of the
75% of land area that is currently irrigated in the irrigated
region, half was already irrigated by 1915. Thus, irrigation
wasintroducedinonly38% of the land area in the
irrigatedregionover1915–1980[Bonfils and Lobell,
2007]. If one assumes that the regional effect of irrigation
scales linearly with irrigated area, this implies a roughly
2.5°C cooling for 100% irrigation, still well below the RCM
value. Other possible reasons for the discrepancy include
Figure 2. The relationship between heat index (HI) and
relative humidity (RH) for different values of temperature
(T), according to the equations of Schoen [2005].
Figure 3. Decadal trends in the average of the 1, 5, 10, and
30 highest daily T
max
values per year, along with the
average June–August (JJA) T
max
and warm spell duration
index (WSDI) for (a) California and (b) Nebraska. Error
bars indicate 95% confidence interval based on bootstrap
resampling. Units are °C decade
1
for all indices except
WSDI, which is expressed as day decade
1
. Irr-Ref
represents the trends in the difference between values in
the irrigated and reference regions, used as a measure of the
effect of irrigation.
L09705 LOBELL ET AL.: IRRIGATION COOLING EFFECT L09705
3of5

artificial smoothness in the gridded data sets that arises from
the interpolation of station data and obscures the full effect
of irrigation, downwind effects of irrigation on the nearby
reference regio n grid cells, and unrealistically high and
stable soil moisture in the RCM simulation [Bonfils and
Lobell, 2007].
3.2. Heat Index
[
16] Values of the extreme indices and JJA averages
computed from HI
max
rather than T
max
reveal some inter-
esting differences between the two variables (Figure 4b).
First, the irrigation-induced decrease in average summer
HI
max
was just 2.7°C, less than half the value for T
max
.
This result is driven by the increase in summer RH from
an average 0.45 in NAT to 0.69 in MOD, which counters
but does not completely balance the effect of reduced T on
HI. More interestingly, the most extreme HI
max
values
were affected much less by irrigation than the average
summer day, with the highest value of HI
max
not statisti-
cally different between the NAT and MOD simulations.
[
17] The diminished response of extreme heat days to
irrigation is explained by the temperature dependence of
the relationship between HI and RH (Figure 2). For
example, an increase in RH from 0.45 to 0.69 leads to
an increase in HI of 3.1°C, 5.6°C, and 10.1°C at T equal to
30°C, 35°C, and 40°C, respectively. In the NAT simula-
tion, T
max
averages roughly 36°C in JJA and reaches
above 40°C on the hottest days. A cooling of 7°Cin
T
max
from irrigation is therefore larger than the effect of
RH increases on an average summer day, but not on the
hottest days of the year. The level of discomfort to humans
is accordingly diminished by irrigation for average summer
days, but the increase of humidity on the hottest days
makes these days just as uncomfortable as they would be
without irrigation.
4. Summary and Conclusions
[18] The results of both the long-term observational
and modeling studies presented here indicate that maxi-
mum temperatures for two major irrigated regions in the
United States are similarly reduced by irrigation on
average and hot summer days. While this conclusion
implies that the frequency and leng th of heat waves
should also be diminished by irrigation, we observed a
significant effect only in the modeling study. The lack of
a detectable trend in WSDI in the observations likely
results from the high percentage of years in the current
data sets with WSDI equal to zero.
[
19] Values of the maximum daily heat index, HI
max
,in
the RCM simulations were reduced much less by irrigation
than T
max
, as a result of substantial increases in RH. Values
on extremely uncomfortable days (highest HI
max
of the
year) were particularly unaffected by irrigation because HI
is more sensitive to RH at higher temperatures. Future work
to synthesize observations of relative humidity in relation to
land cover and land use change –perhaps by pursuing a
similar effort to that for temperature by digitizing, quality
checking, and interpolating data to fine mesh grids– will be
useful for further understanding extreme values of HI
beyond the modeling results presented here. Overall, we
conclude that irrigation has a similar cooling effect on
relatively average and hot summer days in terms of T
max
,
but that cooling effects on HI
max
are more pronounced on
average days.
[
20] Acknowledgments. We thank four anonymous reviewers for
helpful comments on the manuscript. This work was supported in part by
the California Energy Commission. C.B. was supported by a Distinguished
Scientist Fellowship awarded to B. Santer by the U.S. DOE, Office of
Biological and Environment Research.
References
Adegoke, J. O., et al. (2003), Impact of irrigation on midsummer surface
fluxes and temperature under dry synoptic conditions: A regional atmo-
spheric model study of the U.S. high plains, Mon. Weather Rev., 131,
556 – 564.
Alexander, L. V., et al. (2006), Global observed changes in daily climate
extremes of temperature and precipitation, J. Geophy s. Res., 111,
D05109, doi:10.1029/2005JD006290.
Barnston, A. G., and P. T. Schickedanz (1984), The effect of irrigation on
warm season precipitation in the Southern Great Plains, J. Appl. Meteorol.,
23, 865 – 888.
Bonfils, C., and D. Lobell (2007), Empirical evidence for a recent slow-
down in irrigation-induced cooling, Proc. Natl. Acad. Sci. U. S. A., 104,
13,582 – 13,587.
Davis, R. E., et al. (2003), Changing heat-related mortality in the United
States, Environ. Health Perspect., 111, 1712 – 1718.
Ferranti, L., and P. Viterbo (2006), The European summer of 2003: Sensi-
tivity to soil water initial conditions, J. Clim., 19, 3659 – 3680.
Haddeland, I., et al. (2006), Effects of i rrigation on the water and
energy balances of the Colorado and Mekong river basins, J. Hydrol.,
324, 210 – 223.
Figure 4. (a) Average differences between the final 20
years of the MOD and NAT simulations (MOD–NAT) for
the 1, 5, 10, and 30 highest daily T
max
values per year, along
with the average June–August (JJA) T
max
and warm spell
duration index (WSDI). Units are °C for all indices except
WSDI, which is expressed as days. (b) Same as in Figure 4a
except using HI
max
instead of T
max
. Averages are taken over
all 27 grid cells converted to irrigated agriculture in MOD.
Error bars indicate ± 2 standard errors.
L09705 LOBELL ET AL.: IRRIGATION COOLING EFFECT L09705
4of5

Hamlet, A. F., and D. P. Lettenmaier (2005), Production of temporally
consistent gridded precipitation and temperature fields for the continental
United States, J. Hydrometeorol., 6, 330 – 336.
Hegerl, G. C., et al. (2004), Detecta bility of anthropogenic changes in
annual temperature and precipitation extremes, J. Clim., 17, 3683 – 3700.
Kiktev, D., et al. (2003), Comparison of modeled and observed trends in
indices of daily climate extremes, J. Clim., 16, 3560 – 3571.
Kueppers, L. M., M. A. Snyder, and L. C. Sloan (2007), Irrigation cooling
effect: Regional climate forcing by land-use change, Geophys. Res. Lett.,
34, L03703, doi:10.1029/2006GL028679.
Lobell, D. B., and C. Bonfils (2008), The effect of irrigation on regional
temperatures: An analysis of spatial and temporal trends in California,
1934 – 2002, J. Clim., in press.
Mahmood, R., et al. (2006), Impacts of irrigation on 20th century tempera-
ture in the northern Great Plains, Global Planet. Change, 54, 1 – 18.
Maurer, E. P., et al. (2002), A long-term hydrologically based dataset of
land surface fluxes and states for the conterminous United States, J. Clim.,
15, 3237 – 3251.
Pal, J. S., et al . (2007), Regional climate modeling for the developing
world: The ICTP RegCM3 and RegCNET, Bull. Am. Meteorol. Soc.,
88, 1395 – 1409.
Rosenzweig, C., et al. (2002), Increased crop damage in the US from
excess precipitation under climate change, Global Environ. Change,
12, 197 – 202.
Schlenker, W., and M. J. Roberts (2006), Nonlinear effects of weather on
corn yields, Rev. Agric. Econ., 28, 391 – 398.
Schoen, C. (2005), A new empirical model of the temperature– humidity
index, J. Appl. Meteorol., 44, 1413 – 1420.
Segal, M., et al. (1998), On the potential impact of irrigated areas in North
America on summer rainfall caused by large-scale systems, J. Appl.
Meteorol., 37, 325 – 331.
Siebert, S., et al. (2005), Development and validation of the global map of
irrigation areas, Hydrol. Earth Syst. Sci., 9, 535 – 547.
White, M. A., et al. (2006), Extreme heat reduces and shifts United States
premium wine production in the 21st century, Proc. Natl. Acad. Sci.
U. S. A., 103, 11,217– 11,222.

C. J. Bonfils, Lawrence Livermore National Laboratory, Livermore, CA
94551, USA.
L. M. Kueppers, School of Natural Sciences, University of California,
Merced, CA 95343, USA.
D. B. Lobell, Program on Food Se curity and the Environment,
Environment and Energy Building, MC4205, 473 Via Ortega, Stanford,
CA 94305, USA. (dlobell@stanford.edu)
M. A. Snyder, Climate Change and Impacts Laboratory, Department of
Earth and Planetary Sciences, University of California, Santa Cruz, CA
95064, USA.
L09705 LOBELL ET AL.: IRRIGATION COOLING EFFECT L09705
5of5