Climate and Vegetation in the Middle East: Interannual Variability and Drought Feedbacks

Abstract

The Euphrates Plain (EP) experiences large interannual variability in vegetation cover, especially in areas of marginal rain-fed agriculture. Vegetation in this region is primarily limited by available soil moisture, as determined by winter precipitation, spring precipitation, and air temperature. Satellite analyses indicate that the springtime normalized difference vegetation index (NDVI) is negatively correlated with surface albedo, and that interannual variability in albedo in the EP produces an estimated forcing on the radiation balance that peaks at 16.0 W m−2 in May.
Simulations with a regional climate model indicate that surface energy fluxes during a drought year (1999) differed substantially from those during a year with normal precipitation (2003). These differences were geographically specific, with the EP exhibiting increased albedo and decreased sensible heat flux while the neighboring Zagros Plateau region showed no albedo effect, a large increase in sensible heat flux, and an offsetting reduction in latent heat flux. In both the EP and the Zagros there was a potential for positive feedbacks on temperature and drought in late spring, though the most likely feedback mechanisms differed between the two regions: in the EP surface brightening leads to cooling and reduced turbulent heat flux, while in the Zagros region reduced latent heat flux leads to warming and a deepening of the planetary boundary layer.

Full text

Climate and Vegetation in the Middle East: Interannual Variability and
Drought Feedbacks
BENJAMIN F. ZAITCHIK,* JASON P. EVANS,ROLAND A. GEERKEN, AND RONALD B. SMITH
Department of Geology and Geophysics, Yale University, New Haven, Connecticut
(Manuscript received 23 May 2006, in final form 9 November 2006)
ABSTRACT
The Euphrates Plain (EP) experiences large interannual variability in vegetation cover, especially in areas
of marginal rain-fed agriculture. Vegetation in this region is primarily limited by available soil moisture, as
determined by winter precipitation, spring precipitation, and air temperature. Satellite analyses indicate that
the springtime normalized difference vegetation index (NDVI) is negatively correlated with surface albedo,
and that interannual variability in albedo in the EP produces an estimated forcing on the radiation balance
that peaks at 16.0 W m
⫺2
in May.
Simulations with a regional climate model indicate that surface energy fluxes during a drought year (1999)
differed substantially from those during a year with normal precipitation (2003). These differences were
geographically specific, with the EP exhibiting increased albedo and decreased sensible heat flux while the
neighboring Zagros Plateau region showed no albedo effect, a large increase in sensible heat flux, and an
offsetting reduction in latent heat flux. In both the EP and the Zagros there was a potential for positive
feedbacks on temperature and drought in late spring, though the most likely feedback mechanisms differed
between the two regions: in the EP surface brightening leads to cooling and reduced turbulent heat flux,
while in the Zagros region reduced latent heat flux leads to warming and a deepening of the planetary
boundary layer.
1. Introduction
Semiarid regions are subject to regular seasonal dry-
ness and large interannual variability in precipitation.
This results in variable vegetation cover on annual and
interannual time scales, as both natural ecosystems and
nonirrigated crops rely on soil moisture derived from
seasonal rains or springtime snowmelt (Baldocchi et al.
2004; Dall’Olmo and Karnieli 2002; Evans and Geerken
2004; Weiss et al. 2004). Opportunistic annual species
green up rapidly in response to wetting of the soil sur-
face, and their vigor is primarily related to recent rain-
fall events. Winter crops and perennial vegetation have
access to deeper reserves of soil moisture. Growth of
these vegetation types depends on the precipitation
pattern over weeks and months, on evaporative de-
mand, and, for some regions, on temperature con-
straints.
Climate-induced variability in semiarid vegetation is
a matter of both ecological interest and economic con-
cern, as strong sensitivity to climate can result in rapid
land use change (Vanacker et al. 2005) and vulnerabil-
ity to human-induced degradation (Evans and Geerken
2004). Over longer time scales, relatively small shifts in
background climate may have a substantial influence
on the distribution of ecosystems and, perhaps, the vi-
ability of agricultural and pastoral systems (deMenocal
2001; Hole 1994; Weiss and Bradley 2001). Interest in
the climate sensitivities of semiarid vegetation—par-
ticularly crops and rangelands—is evident in the large
body of research devoted to characterizing the relation-
ships between precipitation, soil type, land manage-
ment, and vegetation growth in water-stressed regions
(e.g., Archer et al. 1995; Kremer et al. 1996; Lane et al.
1998; LeHouerou 1996).
There is also a substantial literature concerned with
vegetation feedbacks on climate in semiarid regions. In
an effort to understand the role of land–atmosphere
interactions during the extended Sahel drought of the
early 1970s, Charney (1975) hypothesized a negative
albedo–precipitation feedback in which dry conditions
* Current affiliation: Hydrologic Sciences Branch, NASA God-
dard Space Flight Center, Greenbelt, Maryland.
Corresponding author address: Benjamin F. Zaitchik, Hydro-
logic Sciences Branch, NASA Goddard Space Flight Center,
Code 614.3, Greenbelt, MD 20771.
E-mail: bzaitchik@hsb.gsfc.nasa.gov
3924 JOURNAL OF CLIMATE VOLUME 20
DOI: 10.1175/JCLI4223.1
© 2007 American Meteorological Society
JCLI4223

lead to surface brightening, a decrease in available en-
ergy, and reduced convection. This, in turn, leads to
vegetation die-back and increased albedo. Charney’s
hypothesis has informed numerous studies on albedo–
precipitation feedbacks, including detailed field cam-
paigns (Eltahir 1998; Small and Kurc 2003), satellite
analyses (Brunsell 2006; Courel et al. 1984), and global-
scale modeling exercises (Laval and Picon 1986; Sud
and Fennessy 1982). This line of investigation has re-
vealed important interactions between vegetation, al-
bedo, and climate. The nature of these interactions de-
pends on scale and regional context, and the forcing
mechanisms are not always obvious.
In addition to albedo effects, vegetation is thought to
influence the atmosphere through a number of struc-
tural and physiological mechanisms. Vegetation status
has a significant impact on surface roughness, and the
decrease in roughness associated with drought, land
clearing, or overgrazing can lead to decreases in aero-
dynamic conductivity to surface heat fluxes, limiting
energy transport to the atmosphere, and reducing con-
vection (Sud and Smith 1985; Zheng et al. 2002). Veg-
etation also exerts direct control on latent heat flux
from the surface. An active vegetation cover tends to
increase latent heat flux, due to increased soil infiltra-
tion and subsequent transpiration of otherwise unavail-
able moisture. Increased latent heat flux both humidi-
fies the planetary boundary layer (PBL) and increases
moist static energy (MSE) of near-surface air, increas-
ing the potential for precipitation (Eltahir 1998; Shukla
and Mintz 1982; Sud and Fennessy 1984). Such coupling
between surface conditions and atmospheric processes
is thought to be particularly strong in semiarid regions
(Koster et al. 2004).
This study takes advantage of recent interannual cli-
matic variability to characterize the climatic sensitivi-
ties of vegetation in the Middle East and to investigate
the potential for land–atmosphere feedbacks in the re-
gion. In the first half of this study (sections 2–4) time
series data from the Advanced Very High Resolution
Radiometer (AVHRR), Systeme pour l’Observation
de la Terre (SPOT)-Vegetation, and Moderate Imaging
Spectrometer (MODIS) satellite sensors are used to
describe seasonal and interannual patterns in vegeta-
tion and albedo. Data sources and processing are de-
tailed in section 2. In section 3 the phenology of major
land cover types is described, interannual variability
over the period 1981–2001 is quantified using AVHRR,
and the primary climatic drivers of vegetation variabil-
ity are identified. In section 4 satellite data are used to
test the hypothesis that springtime vegetation growth
has an impact on surface albedo and the surface energy
balance in both spring and summer.
The second half of the study focuses on the years
1999 and 2003—a drought year and a nondrought year,
respectively—for detailed analysis of land–atmosphere
processes during drought, including a functional com-
parison of the lowlands of the Euphrates Plain (EP)
and the uplands of the neighboring Zagros Plateau.
This is accomplished by applying the PSU/NCAR MM5
regional climate model (Dudhia 1993; Grell et al. 1994)
in full-year simulations for both 1999 and 2003, with
satellite-derived datasets used to provide lower bound-
ary information on vegetation cover and surface al-
bedo. Section 5 provides details on the MM5 simula-
tions. Results of the MM5 experiments are presented in
section 6, and in section 7 the potential for feedbacks on
cloud cover and precipitation are discussed. General
conclusions are offered in section 8.
2. Data processing
a. Satellite data
The AVHRR satellite, with its 20-yr data record
(1981–2001) and reasonably high spatial resolution (8
km), provides an excellent tool for the analysis of re-
gional vegetation. AVHRR 10-day composites of sur-
face reflectance and maximum normalized differential
vegetation index (NDVI) were downloaded from the
NASA Distributed Active Archive Center (DAAC).
Next, following a method described by Los (1993), time
series AVHRR data were calibrated against three fairly
time-invariant desert targets located in the Saudi Ara-
bian desert. The method removes effects of sensor deg-
radation and corrects for calibration differences be-
tween different sensor systems. Where indicated, these
corrected 10-day images were averaged to provide
mean monthly values.
Broadband shortwave albedo (
␣
AVHRR
) was calcu-
lated from AVHRR reflectance data (r
1
, r
2
) using the
empirical relationship of Liang et al. (2002):
␣
AVHRR
⫽⫺0.3376r
1
2
⫺ 0.2707r
2
2
⫹ 0.7074r
1
r
2
⫹ 0.2915r
1
⫹ 0.5256r
2
⫹ 0.0035,
where r
1
is red reflectance and r
2
is reflectance in the
near-infrared. This nonlinear formula performed quite
well in initial validation (Liang et al. 2002). In the
present study, reliability of the estimate was confirmed
by comparison with the sophisticated albedo estimates
provided by MODIS on board the Terra satellite. Dur-
ing the 19 months of sensor overlap, AVHRR yielded
albedo estimates for the study area that had similar
variability to those of the MODIS Bidirectional Reflec-
tance Distribution Function (BRDF) albedo product
(mod43B3) (correlation for average EP albedo r ⫽
1AUGUST 2007 Z A I TCHIK ET AL. 3925

0.88). AVHRR-derived albedo was slightly, but consis-
tently, lower than MODIS values by an average of 0.07
for the domain of interest. This offset would not be
expected to affect the statistical analyses applied in this
study.
For MM5 experiments, both albedo and green veg-
etation fraction were derived from 10-day composite
images from the SPOT-Vegetation sensor. SPOT-
Vegetation provides high quality, corrected surface re-
flectance data (Maisongrande et al. 2004), and the sen-
sor was active in both 1999 and 2003, the two years of
analysis in this study. Green vegetation fraction (f
g
)
was calculated from SPOT NDVI using the linear veg-
etation index to percent cover conversion of Gutman
and Ignatov (1998):
f
g
⫽
共
NDVI ⫺ NDVI
0
兲
共
NDVI
⬁
⫺ NDVI
0
兲
.
Minimum and maximum NDVI (NDVI
0
and NDVI
⬁
)
values were set to the minimum and maximum 10-day
values for the study region over the available time se-
ries of SPOT-Vegetation data (1999–2005). Broadband
albedo (
␣
SPOT
) was calculated using the Liang et al.
(2002) empirical reflectance-to-albedo relationship:
␣
SPOT
⫽ 0.3512r
1
⫹ 0.1629r
2
⫹ 0.3415r
3
⫹ 0.1651r
4
,
where r
1–4
are reflectance in the blue, green, red, and
near-infrared bands. SPOT-derived estimates of albedo
were confirmed to be statistically similar to the MODIS
BRDF albedo product over areas of comparison.
Information on land cover type is used only qualita-
tively in this study. Where possible, land use identifica-
tion was accomplished by field visit. For other areas
land cover type was determined using a satellite-de-
rived Fourier filtered cycle similarity (FFCS) technique
detailed in previous papers (Evans and Geerken 2006;
Geerken et al. 2005b).
b. Climate data
Observational records of precipitation and tempera-
ture were extracted from the global summary of month
observations collected at the Climate Prediction Center
(CPC) of the National Centers for Environmental Pre-
diction (NCEP). The study includes 824 stations, at
least 600 of which reported in each month. Station data
were interpolated using a Cressman analysis approach
with variable radius of influence. A 5°Ckm
⫺1
lapse rate
correction was applied for interpolated temperatures,
but no topographic correction was made to precipita-
tion.
3. Vegetation types and variability in the Middle
East
The Middle East is a predominantly semiarid region
that contains a strong north to south precipitation gra-
dient (Fig. 1). Humid regions of Turkey and Transcau-
casia receive more than 1000 mm precipitation per
year, while the deserts south of the Euphrates River
receive 100 mm yr
⫺1
or less. Interannual variability ex-
ceeds mean annual precipitation throughout the south-
FIG. 1. The Middle East: (a) Major geographic features and topography (m); (b) mean annual precipitation from FAO reporting
weather stations (1940–1973) interpolated using Cressman distance weighting (mm).
3926 JOURNAL OF CLIMATE VOLUME 20

ern portion of the region. This variability is of particu-
lar interest because it coincides with the southern limit
of the historical Fertile Crescent agricultural zone and
important rangelands of the Euphrates Plain. Present-
day variability has a significant impact on crop yields
and range productivity (Schmidt and Karnieli 2000;
Weiss et al. 2001), and variability on longer time scales
appears to be associated with the rise and fall of early
civilizations (Weiss and Bradley 2001).
The north to south precipitation gradient is accom-
panied by an ecological gradient ranging from temper-
ate forests and warm season agriculture in the north to
winter crops, grasslands, and eventually shrublands and
desert in the south. These land cover types differ in
vegetation density and phenology (Fig. 2a). Forests in
the highlands of Turkey are characterized by seasonally
high NDVI, with vegetation intensity peaking in early
summer and fading gradually in late summer and au-
tumn (Fig. 2b). Rain-fed agriculture in the northern
portion of the domain follows a similar pattern, return-
ing negative NDVI values in the winter (indicative of
snow cover), a rapid green-up in late spring, and rela-
tively high NDVI throughout the long summer growing
season. Agricultural phenologies in the Fertile Crescent
are quite different. Here agriculture is limited by sum-
mertime dryness rather than wintertime frost, and the
NDVI of rain-fed crops peaks in late spring (Fig. 2b).
Harvest of field crops in this region takes place in May
and June. Orchard crops such as olive also have their
phenologic peak in springtime, but the active growth
period is both longer and more moderate (Fig. 2b).
Along the river valleys and in regions of intense canal
or groundwater irrigation a third field crop phenology
is found, indicative of double-cropping practices (Fig.
2c). In these situations the winter crop may receive
supplementary irrigation and the summer crop is en-
tirely dependent on irrigation water. It should be noted
that in AVHRR-based analysis many agricultural fields
are contained within a single pixel, so a “double crop”
phenology actually represents a composite of winter
cropping, summer cropping, and some fields with active
double cropping. Outside of irrigated areas, land cover
south of the Fertile Crescent, east of the Caspian Sea,
and in much of Iran is limited to sparse grasslands and
shrubs. These lands are utilized seasonally for grazing,
but they have extremely limited vegetation cover for
much of the year. Phenologically, grasses and shrubs
both green up in response to winter rains, peaking in
March or April (Fig. 2c). Senescence comes rapidly in
late spring, though this pattern varies with grazing pres-
sure and the density of shrub cover, as shrubs have
access to deeper moisture reserves and can stay green
FIG. 2. Vegetation intensity in the Middle East. (a) Map of mean annual NDVI maximum, based on monthly averaged AVHRR data,
1981–2001. Color hue indicates the month of peak NDVI and color saturation indicates the magnitude of the NDVI peak. (b), (c)
Phenology of representative pixels for major land cover types, based on SPOT 10-day NDVI composites for 2003.
1A
UGUST 2007 Z A I TCHIK ET AL. 3927
Fig 2 live 4/C

well into the dry season (Geerken et al. 2005b). Finally,
much of the region is essentially barren, with no distinct
NDVI phenology (Fig. 2c).
For the Tigris–Euphrates watershed, interannual
variability in vegetation is greatest along the southern
margin of the Fertile Crescent, an area which includes
both rain-fed agriculture and semiarid rangelands (Fig.
3). In wet years this area is lush with grain crops and
forage material, but in a drought year both crops
and range vegetation can fail entirely. In addition to
this swath of climate-driven variability, Fig. 3a shows
hot spots of anthropogenic effects. These include
the marshes of Mesopotamia, which experienced sig-
nificant human modification over the study period
(Nielsen and Adriansen 2005).
The nature of climate-driven variability can be ex-
plored by analyzing AVHRR data in conjunction with
meteorological data. For each pixel in the study region,
independent linear correlations were calculated be-
tween the annual maximum NDVI in AVHRR data
(n ⫽ 20 yr) and six variables that describe weather
conditions in each hydrologic year: total precipitation
in early winter (November–December), late winter
(January–February), and spring (March–April), as
well as average air temperature for the same three
periods (data are described in section 2). Figure 4
maps the results of this analysis for all pixels with rea-
sonably substantial vegetation coverage (mean annual
NDVI
max
⬎ 0.12) and significant climate sensitivity
(linear correlation between NDVI
max
and at least one
of the six climate variables significant at
␣
⫽ 0.1). Col-
ors on the map indicate the climate variable for which
the coefficient of linear correlation was largest for a
given pixel. In highland regions of northern Iran, for
example, we find that variability in NDVI correlates
most strongly with wintertime temperatures (Fig. 4,
area A): low wintertime temperatures in this area are
associated with strong vegetation growth in spring. The
negative correlation between wintertime temperature
and NDVI
max
can be attributed to the importance of
subzero temperatures for the development of a winter
snowpack, which is the primary source of springtime
soil moisture.
Moving downslope and southward along the precipi-
tation gradient, vegetation growth in the Fertile Cres-
cent and rangelands of the EP correlates more strongly
with precipitation than with temperature (Fig. 4, area
B). The mediating factor is, again, soil moisture, but
in this area snow is not a significant factor in the
water balance of nonirrigated lands. Instead, soil mois-
ture is replenished by winter and spring rains. In
areas where soil moisture storage is sufficient, deep-
rooted woody plants or winter crops are able to ac-
cess moisture that infiltrated during winter rain events
well into the spring, leading to a strong correlation be-
tween January–February precipitation and NDVI
max
.
In areas with shallow soils, or those dominated by op-
portunistic annual grass ecosystems, NDVI
max
corre-
lates most strongly with March–April precipitation, the
period coinciding with the annual vegetation maximum.
FIG. 3. NDVI variability, as captured by AVHRR monthly composites. (a) Map of the standard deviation in annual
maximum NDVI value, 1982–2001. Box indicates the location of the Aleppo Steppe field experiment. (b) Subset of the
AVHRR NDVI time series for a pixel within the Aleppo Steppe, giving an impression of typical annual and interannual
variability at a rangeland location.
3928 JOURNAL OF CLIMATE VOLUME 20

Sensitivity to precipitation is also dominant in the
southern Zagros Plateau (area C), where vegetation
includes hillslope grasses and several large agricultural
areas. Somewhat surprisingly, a portion of arid north-
central Saudi Arabia also exhibits sensitivity to precipi-
tation (area D). The most striking vegetation features
in this area are large cropped fields that are supported
almost exclusively by groundwater irrigation (FAO
1997). The area also includes several large wadis, how-
ever, and it is possible that the detected climate sensi-
tivity results from the response of natural vegetation to
rare wadi floods.
Finally, some areas show a significant positive corre-
lation between air temperature and NDVI
max
. These
areas fall in two regions. First, along the Mediterranean
coast and in south-central Turkey (area E) the correla-
tion with temperature reflects the fact that cold winters
limit the growth of winter crops and orchards, which
normally peak early in the year (Fig. 2a). The area is
humid relative to the interior of the Middle East (mean
annual precipitation of 350–700 mm yr
⫺1
), so the cor-
relation between vegetation and precipitation is less im-
portant. The second region of positive temperature sen-
sitivity is Mesopotamia (area F). Land use in this region
includes vast areas of irrigated agriculture, so vegeta-
tion growth is decoupled from local precipitation. In-
stead, warm winters allow for healthy growth of the
cold season crop and a higher peak NDVI.
4. Albedo and the surface energy balance
Vegetation is known to impact surface albedo in sev-
eral ways. The most direct effect is that photosyntheti-
cally active vegetation is dark in the visible range of the
electromagnetic spectrum, where incoming solar radia-
tion is greatest. This means that live plant material re-
flects less solar radiation than most semiarid soils
(Charney et al. 1977). Senescent or dormant vegetation
also influences surface albedo. Dry herbaceous mate-
rial is relatively bright, so the presence of leaf litter over
a wet or dark soil can cause an increase in albedo.
FIG. 4. Linear correlation between annual maximum NDVI and variables of winter/spring climate.
Climate parameters are average air temperature and accumulated precipitation for Nov–Dec, Jan–Feb,
and Mar–Apr in the hydrologic year of the NDVI maximum. Colors indicate the climate variable of
greatest correlation for all pixels with mean annual NDVI
max
⬎ 0.12 and linear correlation is significant
at
␣
⫽ 0.1. Mean correlation coefficient for each class is indicated on the label bar. Dashed lines indicate
regions of coherent correlation type discussed in the text. Climate data are interpolated from the CPC
weather station network; NDVI is derived from AVHRR, 1981–2001.
1A
UGUST 2007 Z A I TCHIK ET AL. 3929
Fig 4 live 4/C

Woody material, on the other hand, tends to be darker
than dry leaves. Dormant trees and shrubs also cast
shadow, reducing the total solar radiation incident on
the soil surface. If the background soil is brighter than
the woody material itself, then this effect will lead to a
decrease in effective albedo and an increase in ab-
sorbed radiation.
Over 20 years of AVHRR data, there was a consis-
tent negative correlation between spring NDVI in the
EP and albedo in both spring and summer; that is,
strong springtime vegetation made for a darker surface
throughout the summer months. This correlation is
strongest in the rangelands and is observed as a non-
significant tendency for the EP on the whole (EP av-
erage statistics: r ⫽⫺0.44, p ⫽ 0.13: Fig. 5a). There are
three reasons for this correlation in the rangelands.
First, at the resolution of an AVHRR pixel, the NDVI
signal can be dominated by wadis and depressions that
remain green throughout the summer in years with
plentiful runoff. These concentrated areas of dark veg-
etation reduce the reflectance values recorded for the
entire pixel. Second, certain succulent shrubs that are
unpalatable to livestock can retain green vegetation
well into the dry season (Geerken and Ilaiwi 2004), and
these shrubs are most green in years with good precipi-
tation. Third, summer albedo is negatively correlated
with summer NDVI due to the structural effects of
woody vegetation described above. Thus, the albedo
effect can persist even after all green vegetation has
senesced. This third phenomenon was confirmed in a
field experiment in the Aleppo Steppe: regular mea-
surements with a portable spectrometer indicated that
high spring NDVI was associated with low summer al-
bedo across a variety of rangeland environments, even
after differences in NDVI had dropped to near zero
(Figs. 5b and 5c; Geerken et al. 2005a).
To carry the analysis a step further, it is possible to
calculate the influence of interannual variability in sur-
face albedo on the absorption of shortwave radiation,
using AVHRR estimates of albedo and the 40-yr NCEP–
NCAR Reanalysis Project (NNRP: Kalnay et al. 1996)
estimates of surface-incident shortwave radiation. Ac-
cording to this calculation, the interannual standard de-
viation in surface albedo is associated with a forcing on
FIG. 5. NDVI and albedo. (a) Scatterplot of June albedo against March NDVI for 20 years of AVHRR data,
averaged for all pixels in the MEP. (b), (c) Relationship between NDVI and broadband albedo for steppe
vegetation (shrub and grass), measured using a portable spectrometer in the Aleppo Steppe (Syria: Fig. 3) in 2001.
(b) Annual maximum albedo plotted against annual maximum NDVI for three subplots at three different field sites
in the steppe. Lines connect the three subplots measured within each field site. Subplots were selected to capture
a representative range of vegetation cover. (c) Weekly NDVI and albedo derived from spectrometer readings for
the three subplots at one study site [the solid diamond (䉬) site in (b)].
3930 JOURNAL OF CLIMATE VOLUME 20

the surface radiation balance that exceeds 10.0 W m
⫺2
in every month from March through August, averaged
monthly for the EP. This forcing peaks at 16.0 W m
⫺2
for the month of May (Fig. 6). The effect is strongest in
areas with large interannual variability in albedo and, it
must be emphasized, is an average that includes much
higher daytime values.
To put this forcing in context, the radiative forcing
associated with a doubling of atmospheric CO
2
is on the
order of 3.7 W m
⫺2
(Myhre et al. 1998). In a regional
climate modeling study, Schär et al. (2004) found that a
17.1 W m
⫺2
increase in net surface radiation was suffi-
cient to produce feedbacks on precipitation in sensitive
regions. Clearly, interannual variability in albedo in the
EP has the potential to have a considerable impact on
local temperatures and, potentially, hydrometeorology
on the regional scale. These effects are investigated in
greater detail in sections 5 and 6.
5. MM5 simulations
a. 1999 and 2003
Between 1998 and 2001 much of the Middle East
experienced an extended drought (DePauw 2004). The
driest year was 1999, when spatially averaged annual
precipitation in the EP was only 91.5 mm [CPC Merged
Analysis of Precipitation (CMAP): Xie and Arkin
(1996)]. The rainy season of 2003 represented a return
to normal hydrologic conditions. The year was not uni-
formly wet throughout the Middle East, but large por-
tions of the region received precipitation that was at or
above the long-term average. The EP received 163.2
mm of precipitation, and precipitation was greater in
2003 than in 1999 for every month of the rainy season
(Fig. 7a). NDVI was correspondingly higher across the
EP and most of the Zagros Plateau.
b. MM5 experiments
The application of MM5 and the NCEP–Oregon
State University–U.S. Air Force–Hydrologic Research
Laboratory (NOAH) land surface model (LSM) to this
region has been described elsewhere (Evans and Smith
2005; Zaitchik et al. 2005). The standard, limited-area,
nonhydrostatic PSU–NCAR MM5 is used (Dudhia
1993; Grell et al. 1994), implemented with the Reisner
mixed-phase explicit moisture scheme (Reisner et al.
1998), the Rapid Radiative Transfer Model (RRTM)
radiation scheme (Mlawer et al. 1997), Grell’s cumulus
scheme (Grell et al. 1994), and the Medium-Range
Forecast (MRF) model’s planetary boundary layer
scheme (Hong and Pan 1996). The MRF scheme is a
first-order, nonlocal scheme optimized to represent
large-eddy turbulence in a well-mixed PBL. In a com-
parison study of MM5 PBL schemes performed for the
southwestern United States—a region that contains
both intense deserts and significant topography, much
like the Middle East—it was found that first-order clo-
sure schemes, including MRF, produced the highest ac-
curacy predictions of PBL depth, CAPE, and vertical
profiles of temperature and mixing ratio (Bright and
Mullen 2002).
The NOAH LSM is used in its coupled form. Noah is
a direct descendent of the Oregon State University
(OSU) LSM (Mahrt and Pan 1984; Mahrt and Ek 1984;
Pan and Mahrt 1987), a sophisticated land surface
model that has been extensively validated in both
coupled and uncoupled studies (Chen and Mitchell
1999; Chen and Dudhia 2001). The NOAH LSM simu-
lates soil moisture, soil temperature, skin temperature,
snowpack depth and water equivalent, canopy water
content, and the energy flux and water flux terms of the
surface energy balance and surface water balance. In its
MM5-coupled form NOAH has a diurnally dependent
Penman potential evaporation (Mahrt and Ek 1984), a
four-layer soil model (Mahrt and Pan 1984), a primitive
canopy model (Pan and Mahrt 1987), modestly com-
plex canopy resistance (Jacquemin and Noilhan 1990),
and a surface runoff scheme (Schaake et al. 1996). Ad-
ditionally, an irrigation scheme was implemented that
FIG. 6. Interannual standard deviation in absorbed short-
wave radiation (W m
⫺2
), averaged over the month of May. Values
are calculated using NNRP incoming shortwave radiation and
AVHRR-derived surface albedo, 1981–2001. Note that maxima
exist in the semiarid pasture lands of the EP, Ustyurt Plateau
(east of the Caspian Sea), and Arabian Peninsula. Shading in-
dicates areas where the interannual standard deviation exceeds
15 W m
⫺2
.
1A
UGUST 2007 Z A I TCHIK ET AL. 3931

realistically approximates flood irrigation practices of
the region (Zaitchik et al. 2005). The use of a coupled
land–atmosphere modeling system was deemed neces-
sary in this study because of the tightly coupled nature
of interactions between surface fluxes, PBL dynamics,
and clouds (Betts 2004). For this reason, the results of
sensitivity studies performed with uncoupled LSMs can
be misleading in the analysis of physical process (Zhang
et al. 2001).
For all simulations the model was implemented at
27-km horizontal resolution and with 23 levels in the
vertical. Initial conditions and lateral boundary condi-
tions were drawn from the NNRP (Kalnay et al. 1996).
To capture the effects of interannual vegetation vari-
ability it was necessary to integrate satellite-derived
datasets for surface albedo and vegetation fraction. As
described in an earlier study (Zaitchik et al. 2005), op-
erational datasets currently available for MM5 provide
only climatological estimates for these properties.
Four 14-month MM5 simulations were performed.
First, control simulations were run for 1 November
1998–31 December 1999 and 1 November 2002–31 De-
cember 2003. In each case the first two months were
treated as spinup and results were recorded for the cal-
endar year. For these simulations surface albedo and
vegetation fraction were drawn from SPOT-Vegetation
data for the corresponding time period (see Section 2),
and they are hereafter referred to as 1999
V1999
and
2003
V2003
, where the subscript indicates the year of sat-
ellite data. It was necessary to prescribe albedo as well
as vegetation fraction because vegetation fields do not
inform the predicted surface albedo field in MM5–
NOAH. Next, “vegetation reversal” simulations were
performed for the same time periods in which SPOT
albedo and vegetation fraction from the nondrought
year (2003) were input to MM5 simulations for the
drought year (1999) and vice versa. These simulations
are hereafter referred to as 1999
V2003
and 2003
V1999
.
It is important to note that the vegetation reversal
experiments include the reversal of vegetation fraction,
which is explicitly a property of vegetation, and surface
albedo, which may be influenced by a number of fac-
tors, notably soil moisture. It is the authors’ contention
that variability in the albedo field is associated prima-
rily with vegetation rather than soil moisture. The
SPOT images used to derive the surface albedo were
acquired in late morning under clear-sky conditions.
For the highly evaporative conditions that predominate
in the Middle East in late spring (the season of primary
interest), the soil surface is generally dry by this time of
day, minimizing the impact that residual soil moisture
can have on remotely sensed surface albedo. For this
reason it can be stated that vegetation reversal experi-
ments have the potential to capture vegetation–climate
feedbacks, but not soil moisture–climate feedbacks,
both of which may operate during a drought. If our
assumption is incorrect and albedo is under strong con-
trol of soil moisture, then soil moisture–climate feed-
backs related to albedo are included in the analysis,
while those related to the evaporation of soil water are
not. In either case, the inclusion of albedo in the veg-
etation reversal experiment leads to results that are dif-
ferent from vegetation reversal studies that have not
included albedo (e.g., Matsui et al. 2005).
c. Statistical analysis
For results presented as spatial averages, the signifi-
cance of differences between simulations was tested us-
ing a Student’s t test. To minimize the influence of spa-
tial autocorrelation, only a subset of grid points was
included in our statistical analysis, such that there was a
minimum distance of 108 km (four grid cells) separating
any two points included in the analysis. This resulted in
n ⫽ 18 points for both the EP and northern Zagros
Plateau (NZ) subregions. The analysis cannot assess
FIG. 7. (A) CMAP (black lines) and MM5-simulated precipitation (gray lines), averaged monthly for the EP in 1999 (dashed lines)
and 2003 (solid lines). MM5-predicted total precipitation for (b) 1999 and (c) 2003, based on control simulations. The heavy contour
in (b) and (c) corresponds to 250 mm yr
⫺1
, the theoretical limit for rain-fed agriculture in the region.
3932 JOURNAL OF CLIMATE VOLUME 20

the significance of results relative to the uncertainty in
atmospheric drivers, but it does establish the signifi-
cance of the differences relative to regional variability
in the MM5 simulations. All reported t tests are for
monthly averaged MM5 output.
6. Results of MM5 simulations
As expected, the MM5 returned drier conditions in
1999
V1999
than in 2003
V2003
. Precipitation results were
reasonably similar to the CMAP integrated observa-
tions (Fig. 7a), and precipitation was, in general, greater
in 2003 across the Euphrates Plain and the Zagros Pla-
teau (Figs. 7b and 7c). For the EP the difference in
precipitation was present only in the rainy and transi-
tional seasons, as June–September was almost com-
pletely dry in both years. Strong wintertime precipita-
tion in 2003 led to an extensive snowpack that persisted
through April in highland areas of Turkey and Iran.
For both years, snow retreated from most of the Tau-
rus and Zagros regions by May. May is an interesting
month for analysis because it lies in the transition be-
tween the rainy season of winter and spring and the
persistent aridity of summer. The intrusion of large
storm systems into the region is reduced relative to
November–April, allowing for greater local influence
on the atmosphere, but significant precipitation is still
observed in most years, indicating a potential for moist
convection. For these reasons, May is a period during
which precipitation processes are expected to be sensi-
tive to a land–atmosphere forcing involving persistent
spring vegetation or soil moisture.
From the perspective of land–atmosphere interac-
tions, it is important to note that the 1999 drought led
to an increase in albedo only in the semiarid lowland
areas of the EP and, in the northeast of the study re-
gion, the Caspian steppe. In the Zagros foothills and
plateau—where soil moisture is greater, soils are
darker, and vegetation is less sensitive to precipitation
drought—there was no detectable difference in May
surface albedo between 1999 and 2003 (Fig. 8a; Table
1). In August a small difference between 1999 and 2003
albedo in the northern Zagros did develop (⫹0.03, not
shown), but this difference was at all times smaller than
that in the EP, and it occurred at a time of year when
precipitation is strongly inhibited by large-scale circu-
lations (Rodwell and Hoskins 1996; Ziv et al. 2004),
limiting the potential for local feedbacks. This spatial
variability in albedo is important because of its impli-
cations for the surface radiation balance. If we consider
the two neighboring subregions of the EP and the NZ,
the MM5 simulations indicate that net radiation avail-
able at the surface (R
net
) was reduced in the EP ( p ⬍
0.0001) but not in the NZ (Fig. 8b). In fact, R
net
tended
to be slightly greater in 1999 for the Zagros, due to
clear-sky conditions and an increase in solar radiation
incident on the ground surface, though this tendency
was not significant ( p ⫽ 0.359). This contrast in R
net
was associated with a contrast in ground temperature.
In the EP the warm synoptic conditions of 1999 were
offset by a decrease in local R
net
, such that the average
ground temperature in May was no higher in 1999 than
in 2003 (t test for difference: p ⫽ 0.960). In the NZ, in
contrast, the increase in R
net
in 1999 combined with a
FIG. 8. Average difference for the month of May, 1999–2003, in (a) surface albedo and (b) net surface
radiation, according to MM5 simulations. Boxes EP and NZ indicate the Euphrates Plain and Northern
Zagros subregions described in the text and in Tables 1 and 2. In (b), regions of positive difference are
outlined by solid lines and regions of negative difference are outlined by dashed lines.
1A
UGUST 2007 Z A I TCHIK ET AL. 3933

substantial reduction in soil moisture to enhance large-
scale advective warming, producing radiative ground
temperatures that were 3.1°C warmer in May 1999 rela-
tive to May 2003 (Table 1; p ⫽ 0.003). This MM5 result
is consistent with NNRP estimates of the 2-m air tem-
perature for the same period: according to the NNRP,
May 1999 was only 0.1°C warmer than May 2003 for the
EP, while it was 1.3°C warmer for the NZ.
The difference in drought impact on R
net
in the EP
versus the NZ also has an impact on turbulent heat flux
via the surface energy balance: R
net
⫽ H ⫹
␭
E ⫹ G,
where H is net surface sensible heat flux,
␭
E is surface
latent heat flux, and G is conductive heat flux from the
surface to subsurface. The drought-related reduction in
R
net
in the EP for 1999 was associated with a decrease
in H of 6.1 W m
⫺2
relative to 2003 ( p ⬍ 0.0001), aver-
aged over the month of May, while the clear-sky-
associated increase in R
net
for the NZ was associated
with an increase in H of 32.4 W m
⫺2
(p ⬍ 0.0001).
Latent heat flux (
␭
E), meanwhile, was reduced in 1999
for both the EP (by 9.6 W m
⫺2
; p ⫽ 0.0007) and the NZ
(by 29.7 W m
⫺2
; p ⬍ 0.0001) due to reduced soil mois-
ture. The effect was considerably larger in the NZ area
because soil moisture in the EP was quite low in May
even in the relatively wet year of 2003.
Vegetation reversal
Vegetation reversal experiments have the potential
to clarify the relative importance of synoptic forcing
and local vegetation properties during drought. The
“reversals” of the surface properties performed in this
experiment capture the effects of drought on albedo
and latent heat flux. Surface roughness effects were
not captured by these experiments, as roughness is not
a predicted field in the MM5–NOAH simulations, and
the parameter was left at land cover defaults for all
simulations. Latent heat flux effects could arise as a
result of greater access to soil moisture, which is a func-
tion of vegetation fraction in MM5–NOAH (Chen
and Dudhia 2001), or as a secondary product of albedo
since an increase in net radiation at the surface would
be expected to increase the total turbulent energy
transfer. In application, however, vegetation reversal
had no significant effect on the May latent heat flux
in the EP (Table 2; p ⫽ 0.960). By this time of the
year model soil moisture was depleted in the EP in
both 1999 and 2003, leaving little possibility for vegeta-
tion to influence latent heat flux. What difference there
was in
␭
E between 1999 and 2003 (Table 1) primarily
arose from direct soil evaporation after rain events, and
TABLE 1. Surface fluxes and PBL properties for MM5 control simulations. Values are averages for the month of May for albedo (
␣
);
near-surface soil moisture (SM
0–10cm
); incoming and reflected solar radiation at the surface (S
↓
, S
↑
); surface-incoming and surface-
emitted longwave radiation (L
↓
, L
↑
); net surface radiation (R
net
); ground temperature (T
ground
); sensible heat flux (H ); latent heat flux
(
␭
E ); conductive heat flux to the subsurface (G); depth and temperature of the planetary boundary layer (Depth
PBL
, T
PBL
); turbulent,
radiative, and total surface heat fluxes (Q
t
, Q
R
, Q
Total
); local contribution to the moist static energy density of the PBL (具MSE典
local
);
and precipitation.
Variable Unit
Euphrates Plain Northern Zagros
1999 2003 1999–2003 1999 2003 1999–2003
␣
— 0.31 0.25 0.06 0.19 0.19 0.00
SM
0–10cm
m
⫺3
m
⫺3
0.132 0.160 ⫺0.028 0.148 0.215 ⫺0.067
S
↓
Wm
⫺2
344.7 330.5 14.2 331.8 301.9 29.9
S
↑
Wm
⫺2
106.9 82.6 24.2 63.0 57.4 5.7
L
↓
Wm
⫺2
337.4 344.2 ⫺6.8 306.4 311.3 ⫺4.9
L
↑
Wm
⫺2
459.3 459.3 0.0 424.2 406.6 17.6
R
net
Wm
⫺2
116.0 132.8 ⫺16.8 151.0 149.3 1.7
T
ground
°C 300.0 300.0 0.0 294.1 291.0 3.1
H Wm
⫺2
99.5 105.6 ⫺6.1 114.3 81.9 32.4
␭
E Wm
⫺2
6.6 16.2 ⫺9.6 25.4 55.1 ⫺29.7
G Wm
⫺2
6.4 7.4 ⫺1.0 8.0 9.9 ⫺1.9
␤
— 15.08 6.52 8.56 4.50 1.49 3.01
Depth
PBL
m 1270 1397 ⫺127 1477 1207 270
Q
t
Wm
⫺2
106.1 121.8 ⫺15.7 139.7 137.0 2.7
T
PBL
* °C 293.1 292.1 1.0 281.8 278.9 2.9
Q
R
Wm
⫺2
⫺142.9 ⫺137.2 ⫺5.7 ⫺103.0 ⫺99.1 ⫺3.9
Q
total
Wm
⫺2
⫺36.8 ⫺15.4 ⫺21.4 36.7 37.9 ⫺1.2
具MSE典
local
Wm
⫺3
⫺0.029 ⫺0.011 ⫺0.018 0.025 0.031 ⫺0.007
Precip mm 0.6 10.1 ⫺9.5 4.6 20.3 ⫺15.7
*T
PBL
is taken at a height near the top of the average PBL: 850 hPa for EP and 700 hPa for NZ.
3934 JOURNAL OF CLIMATE VOLUME 20

this flux was not significantly affected by vegetation
status.
In the NZ, modeled
␭
E for May was substantial in
both 1999 and 2003. In May 2003, in fact, soil moisture
was so great that actual evapotranspiration (AET) fre-
quently approached the atmospherically limited rate of
potential evapotranspiration (PET). Under such moist
conditions, the access to deeper reserves of soil mois-
ture afforded by healthy vegetation does not have a
substantial impact on
␭
E. In 1999, when there was little
precipitation in May, the soil surface dried out and
␭
E
was reduced by more than 50% (Table 1). Under these
drier conditions AET was well below PET due to mois-
ture limitation. Vegetation access to subsurface soil
moisture thus has the potential to influence the parti-
tioning of surface energy. In 1999, the average MEP
␭
E
was reduced by 5.6 W m
⫺2
in the simulation with
drought vegetation (1999
V1999
) relative to that with
nondrought vegetation (1999
V2003
), though the differ-
ence was not statistically significant (p ⫽ 0.140).
Vegetation reversal had a substantial impact on ra-
diation fluxes in the EP for both 1999 and 2003 simu-
lations. In both years, drought vegetation was associ-
ated with increased albedo and reduced R
net
(p ⬍
0.0001). Reduced R
net
led to lower ground temperature
(significant for 2003, p
2003
⫽ 0.050; nonsignificant for
1999, p
1999
⫽ 0.109), reduced sensible heat flux ( p ⬍
0.0001), and a shallower PBL (p ⬍ 0.0001) (Figs. 9a–c).
Drought vegetation also led to a reduction in total tur-
bulent heat flux from the surface (Q
t
: Table 2) and an
elevated lifting condensation level (LCL: Fig. 9d), the
implications of which will be discussed in the following
section. In the NZ, where the 1999 drought did not
affect the May surface albedo, the vegetation reversal
experiment had no significant impact on radiation
fluxes in either year (R
net
p
1999
⫽ 0.140, p
2003
⫽ 0.597).
Ground temperature, sensible heat flux, and the depth
of the PBL were also unaffected (p ⬎ 0.2 for all analy-
ses). As in the interannual comparison, the vegetation
reversal experiments did yield some effects on albedo
and R
net
for the NZ in July and August. These effects
were small relative to those in the EP in May and, as
mentioned above, they arise during a season when a
land–atmosphere forcing on precipitation is unlikely to
occur.
7. Feedbacks on cloudiness and precipitation
Two hypotheses for a drought feedback on precipi-
tation are considered: a mechanism associated with re-
duced R
net
(Eltahir 1998) and a mechanism involving
deepening and drying of the planetary boundary layer
(Betts and Ball 1998). These hypotheses have been re-
viewed and compared in detail by Small and Kurc
(2003) and Zaitchik et al. (2006). Briefly, in the mecha-
nism proposed by Eltahir (1998) a drought-related in-
crease in albedo and decrease in soil moisture leads to
a reduction in R
net
and an associated reduction in total
TABLE 2. Impact of vegetation reversal on MM5 simulations. Values reported are differences that indicate the impact of drought
vegetation (1999
V1999
, 2003
V1999
) as opposed to healthy vegetation (1999
V2003
, 2003
V2003
) in each simulation year. Symbols are defined
in Table 1.
Variable Unit
Euphrates Plain Northern Zagros
1999
V1999
–1999
V2003
2003
V1999
–2003
V2003
1999
V1999
–1999
V2003
2003
V1999
–2003
V2003
␣
— 0.06 0.06 0.00 0.00
SM
0–10cm
m
⫺3
m
⫺3
⫺0.002 ⫺0.002 ⫺0.004 ⫺0.004
S
↓
Wm
⫺2
1.1 3.5 4.3 4.1
S
↑
Wm
⫺2
21.0 20.9 0.8 0.8
L
↓
Wm
⫺2
⫺-2.9 ⫺4.1 ⫺2.8 ⫺1.4
L
↑
Wm
⫺2
⫺6.2 ⫺6.7 0.0 0.0
R
net
Wm
⫺2
⫺16.6 ⫺14.8 0.7 1.9
T
ground
°C ⫺1.0 ⫺1.1 0.0 0.0
H Wm
⫺2
⫺17.0 ⫺15.7 5.3 2.5
␭
E Wm
⫺2
⫺0.3 ⫺0.3 ⫺5.6 ⫺0.6
G Wm
⫺2
⫺0.2 1.1 0.0 ⫺0.3
Depth
PBL
m ⫺165 ⫺156 ⫺54
Q
t
Wm
⫺2
⫺17.3 ⫺16.0 ⫺0.3 1.9
T
PBL
* °C ⫺0.5 ⫺0.5 ⫺0.1 ⫺0.1
Q
R
Wm
⫺2
⫺0.8 ⫺1.2 0.5 0.5
Q
total
Wm
⫺2
⫺18.1 ⫺17.2 0.2 2.4
具MSE典
local
Wm
⫺3
⫺0.016 ⫺0.015 0.000 0.002
Precip mm 0.0 ⫺0.9 ⫺0.4 ⫺0.8
* T
PBL
is taken at a height near the top of the average PBL: 850 hPa for EP and 700 hPa for NZ.
1A
UGUST 2007 Z A I TCHIK ET AL. 3935

turbulent heat flux from the surface (Q
t
⫽ H ⫹
␭
E).
This means that the local land surface contributes less
to the moist static energy (MSE) of the PBL during
drought than it would under normal conditions. The
density of MSE in the PBL is directly associated with
conditional instability (Zawadzki et al. 1981), so a re-
duction in the local contribution to MSE leads to re-
duced potential for convective precipitation. The Betts
and Ball (1998) hypothesis focuses on total local heat-
ing of the PBL. Because latent heat flux is not released
locally—condensation might occur some distance
downwind—it is excluded from consideration. Instead,
the hypothesis states that an increase in H, possibly
accompanied by an increase in local radiative heating
(Q
R
), causes a deepening of the PBL due to more vig-
orous mixing and entrainment at the top of the bound-
ary layer. The combination of deepening and entrain-
ment of dry, low MSE air from the free troposphere
leads to reduced MSE density (具MSE典) and a higher
lifting condensation level (LCL), thus reducing the po-
tential for precipitation.
Comparing 1999 to 2003, the average R
net
for May in
the EP was 16.8 W m
⫺2
less during the drought year of
1999, resulting in a 15.7 W m
⫺2
decrease in Q
t
(Table
1). Additionally, a slightly cooler ground temperature
in 1999 was associated with a 5.7 W m
⫺2
reduction in
the net radiative exchange (Q
R
) between the surface
and the lower atmosphere. Here Q
R
was calculated as a
simple radiative balance, Q
R
⫽␧
a
L
↑
⫺␧
s
L
↓
, where ␧
a
and ␧
s
are the emissivity of the lower atmosphere and
surface, and longwave upwelling (L
↑
) and downwelling
(L
↓
) terms are solved based on ground temperature
and air temperature near the top of the PBL, respec-
tively (Liou 2002; Zaitchik et al. 2006). The change in
Q
t
and Q
R
sum to a reduction of 22.5 W m
⫺2
in the total
local energy transfer between the surface and the lower
atmosphere (Q
Total
⫽ Q
t
⫹ Q
R
). A forcing of this mag-
nitude is on the order of forcings that have triggered
precipitation feedbacks in previous modeling studies
(e.g., Schär et al. 1999), though the prevailing aridity of
the Middle East may limit the potential for feedback in
this region.
In the NZ, in contrast, the 1999 drought was associ-
ated with very slight increases in average R
net
(1.7 W
m
⫺2
) and Q
t
(2.7 W m
⫺2
) for the month of May. The
small change in Q
t
is the result of large offsetting
changes in H (⫹32.4 W m
⫺2
in the drought year) and
␭
E (⫺29.7 W m
⫺2
). The minor increase in Q
t
was offset
by a decrease in Q
R
, and the difference in Q
Total
be-
tween 1999 and 2003 (⫺1.2Wm
⫺2
) was far too small to
trigger any forcing on precipitation via the Eltahir
(1998) mechanism. The MM5 results for the NZ are
consistent with the Betts and Ball (1998) feedback hy-
pothesis, however, as otal local heating (H ⫹ Q
R
) was
enhanced by 28.5 W m
⫺2
in 1999 relative to 2003 and
the average depth of the PBL was increased by 87 m.
This leads to a deeper, drier PBL, with reduced poten-
tial for moist convection.
Vegetation reversal
Cloud thickness in MM5 can be estimated by sum-
ming the column-integrated fields of all condensed wa-
ter species (ice, snow, rainwater, and cloud water). Av-
eraged over a month, this “cloudiness” calculation is a
proxy for total cloudiness, sensitive to changes in either
cloud frequency or cloud thickness. The spatial pattern
of the MM5 cloudiness compares relatively well with
the MODIS-derived monthly cloud fraction (e.g., Fig.
10).
Comparing simulations 1999
V2003
and 1999
V1999
(Fig. 11a), one finds that cloudiness was greater over
the northern Zagros and east of the Caspian in the
1999
V2003
simulation. There was essentially no differ-
ence in cloudiness over the EP. In 2003, when the back-
ground atmospheric humidity and local evaporation
were both higher than in 1999 (Table 2), the presence of
nondrought vegetation (2003
V2003
– 2003
V1999
) results
in increased cloudiness over broad areas of the EP,
Saudi Arabia, Turkey, the Zagros Plateau, and east of
the Caspian Sea (Fig. 11b). For the EP this discrepancy
between the 1999 and 2003 vegetation reversal ex-
FIG. 9. Results of MM5 vegetation reversal experiment for 1999,
expressed as the difference 1999
V1999
⫺ 1999
V2003
: (a) ground
temperature (°C), (b) sensible heat flux (W m
⫺2
), (c) PBL depth
(m), and (d) lifting condensation level (m). Contour intervals are
0.4 (a), 10.0 (b), (d), and 50.0 (c), centered on zero, with negative
contours dashed.
3936 JOURNAL OF CLIMATE VOLUME 20

periments is most pronounced in May (Fig. 11c). This
supports the suggestion that May is a key month for
land–atmosphere forcings on hydrometeorology (sec-
tion 5). Over the NZ differences in cloudiness are
present from April through August in both 1999 and
2003 vegetation reversal experiments, with increased
cloudiness observed in simulations with healthy vegeta-
tion (1999
V2003
and 2003
V2003
).
The cloudiness result for the EP emphasizes the im-
portance of abiotic factors—primarily soil moisture and
mesoscale atmospheric conditions—in determining
whether a vegetation forcing will trigger a feedback on
the atmosphere. In the wetter, convectively active con-
ditions of 2003 the presence of a healthy vegetation
cover had the potential to enhance the production of
clouds relative to a simulation with drought vegetation.
The dry, convectively stable background conditions of
1999, meanwhile, were relatively resistant to the intro-
duction of an artificially healthy vegetation cover
(1999
V2003
), and cloudiness effects were not realized in
the EP.
It is difficult to assess the significance of the cloudi-
ness and precipitation results relative to background
variability without performing an ensemble of simula-
tions, which was not possible in the presented research.
Nonetheless, differences in cloudiness in the vegetation
reversal experiments suggest that a vegetation-induced
feedback is operating. This evidence of feedback in the
MM5 can be attributed to the reversal of vegetation
fraction and albedo fields because they were the only
variables manipulated in the experiment and because
no perturbations were applied to the forcing data.
For the EP, healthy vegetation cover is associated
with reduced albedo, enhanced sensible heat flux, and
FIG. 11. Column-integrated condensed water content (g m
⫺2
): (a) 1999
V1999
–1999
V2003
, (b) 2003
V1999
–2003
V2003
, average
values for the month of May. Contour interval is8gm
⫺2
, centered on zero, with negative contours dashed. (c) Time series
of monthly averages for 1999
V1999
–1999
V2003
(solid line) and 2003
V1999
–2003
V2003
(dashed line), averaged over
the EP.
FIG. 10. Cloud cover for May 2003: (a) MODIS monthly average cloud fraction, composited
at 1° from MODIS Terra and MODIS Aqua data; (b) MM5 2003
V2003
column-integrated
condensed water content (g m
⫺2
), averaged for the month of May and smoothed for com-
parison with MODIS.
1A
UGUST 2007 Z A I TCHIK ET AL. 3937

increased MSE density within the PBL (Table 2). This
is consistent with the R
net
drought feedback proposed
by Eltahir (1998) and with results of the interannual
model comparison described above. In the NZ the veg-
etation reversal experiments had little impact on sur-
face radiation and total turbulent heat flux. This sug-
gests that vegetation cover on its own does not signifi-
cantly influence land–atmosphere interactions during
drought. Instead, soil moisture may play a more domi-
nant role, through its effects on soil surface properties
and its relevance to evapotranspiration. This being the
case, differences in NZ cloud cover obtained in vegeta-
tion reversal simulations (Figs. 11a and 11b) can be
explained in two ways. First, the NZ lies downwind of
the EP, so any hydrometeorological forcing in the EP
has the potential to impact cloud formation in the NZ.
Second, for mountainous areas such as the NZ, the veg-
etation reversal experiments include some soil moisture
effects. This is because portions of the NZ were snow
covered through March 2003, while the area was nearly
snow-free in March 1999. This snow cover discrepancy
affected the SPOT-derived surface albedo, with the re-
sult that simulations using the 2003 albedo (i.e.,
1999
V2003
and 2003
V2003
) had less early spring evapora-
tion and thus greater residual soil moisture in late
spring than the simulations that used the 1999 albedo.
Finally, it should be noted that secondary effects of
vegetation-induced cloud cover—reduced incident
shortwave radiation and enhanced downwelling long-
wave radiation—were small and offsetting in both re-
gions for both 1999 and 2003 (Table 2).
The distinction between the EP and the NZ is not
particularly surprising. In the EP surface moisture is
extremely limited in late spring, and evapotranspiration
is limited as well. The greatest difference between a
drought year and a wet year, then, is that in a wet year
the surface albedo is reduced due to greater vegetation
cover over bright, dry soils. In the NZ, as indicated in
satellite analysis (Fig. 6), interannual variability in al-
bedo is relatively small. Instead, a year of below-
average precipitation results in substantial reductions in
soil moisture and evapotranspiration. These soil-
moisture-induced changes in the surface energy balance
form the basis for variability in land–atmosphere inter-
actions.
8. Conclusions
In this study it was found that interannual fluctua-
tions in climate lead to considerable variability in veg-
etation for much of the Middle East. The greatest vari-
ability was found in a sensitive transitional climate zone
that includes much of the Fertile Crescent and its neigh-
boring rangelands. In this area vegetation intensity is
strongly correlated with surface albedo both during and
after the prime growing season. An externally imposed
drought, then, leads to reduced vegetation, increased
albedo, and the potential for feedbacks associated with
changed land surface conditions.
During the drought of 1999, the impact of drought on
vegetation, albedo, and soil moisture led to conditions
consistent with surface feedbacks on precipitation. In
the Euphrates Plain (EP), increased albedo led to re-
duced local heat flux, producing a shallow, stable plan-
etary boundary layer (PBL) with low conditional insta-
bility (Fig. 12). In the neighboring northern Zagros re-
gion (NZ), drought led to reduced soil moisture and
increased sensible heat flux, causing a deep, dry PBL to
develop. This condition is associated with enhanced en-
trainment of dry air at the top of the PBL and a reduc-
tion in conditional instability (Fig. 12). In “vegetation
reversal” sensitivity experiments it was found that the
drought-related feedback tendency in the EP relied
heavily on vegetation status and albedo, while that in
the NZ appeared to be more strongly associated with
near-surface soil moisture.
The EP is substantially more arid than the NZ. Dif-
ferences in drought impacts and feedbacks between the
two subregions are analogous to differences observed
during drought in predominantly arid zones versus
more humid regions. In dry areas, vegetation cover is
sparse and soils are typically bright. Vegetation drought
therefore causes an increase in albedo and a reduced
energy environment (Charney et al. 1977; Eltahir 1998).
In more humid regions, drought is not necessarily as-
sociated with an increase in albedo; senescent vegeta-
tion tends to be brighter than live vegetation, and dry
FIG. 12. Surface processes relevant to drought–precipitation
forcings in the EP and NZ (schematic): symbols as defined in
Table 1. Solid gray line represents the height of the PBL under
nondrought conditions and the dashed line represents the height
of the PBL during drought. The black line is the land surface.
3938 JOURNAL OF CLIMATE VOLUME 20

soils are brighter than wet soils (Eltahir 1998), but these
effects can be mitigated by resilient vegetation and off-
set by increased exposure of soils that are relatively
dark even when dry (Zaitchik et al. 2006). Under these
circumstances feedbacks related to evapotranspiration
are expected to dominate, as drought transforms a nor-
mally moisture-rich landscape into a moisture-limited
environment (e.g., Heck et al. 1999; Sud et al. 2003).
In the Middle East, climate variability is largely a
product of external factors, and drought is a common
occurrence for both the Euphrates Plain and the more
humid—but still drought-prone—Zagros Plateau. In a
region that contains steep precipitation gradients and
large areas of marginal rain-fed agriculture, any local
processes that enhance or mitigate drought are of in-
terest. These processes can include local forcings on air
temperature, feedbacks on vapor pressure and cloudi-
ness, and modification of precipitation events. The
present study revealed strong evidence for a local in-
fluence on temperature, water vapor, and cloudiness
during moderate drought, with inconclusive results on
precipitation. Further studies of Middle East drought,
including analysis of historically extreme or persistent
droughts, will further our understanding of land–atmo-
sphere interactions in this environmentally sensitive re-
gion.
Acknowledgments. This study was carried out with
the financial support of NASA Grant NNG05GB36G
and with NCAR computer resources made available
under NSF Grant ATM-0112354. Preprocessed
AVHRR data were provided by Prasad Thenkabail and
Praveen Noojipady of the International Water Manage-
ment Institute (Colombo, Sri Lanka).
REFERENCES
Archer, S., D. S. Schimel, and E. A. Holland, 1995: Mechanisms of
shrubland expansion: Land-use, climate or CO
2
. Climatic
Change, 29, 91–99.
Baldocchi, D. D., L. K. Xu, and N. Kiang, 2004: How plant func-
tional-type, weather, seasonal drought, and soil physical
properties alter water and energy fluxes of an oak-grass sa-
vanna and an annual grassland. Agric. For. Meteor., 123, 13–
39.
Betts, A. K., 2004: Understanding hydrometeorology using global
models. Bull. Amer. Meteor. Soc., 85, 1673–1688.
——, and J. H. Ball, 1998: FIFE surface climate and site-average
dataset 1987–89. J. Atmos. Sci., 55, 1091–1108.
Bright, D. R., and S. L. Mullen, 2002: The sensitivity of the nu-
merical simulation of the Southwest monsoon boundary layer
to the choice of PBL turbulence parameterization in MM5.
Wea. Forecasting, 17, 99–114.
Brunsell, N. A., 2006: Characterization of land-surface precipita-
tion feedback regimes with remote sensing. Remote Sens. En-
viron., 100, 200–211.
Charney, J. G., 1975: Dynamics of deserts and drought in Sahel.
Quart. J. Roy. Meteor. Soc., 101, 193–202.
——, W. J. Quirk, S. H. Chow, and J. Kornfield, 1977: Compara-
tive study of effects of albedo change on drought in semi-arid
regions. J. Atmos. Sci., 34, 1366–1385.
Chen, F., and K. Mitchell, 1999: Using GEWEX/ISLSCP forcing
data to simulate global soil moisture fields and hydrological
cycle for 1987–1988. J. Meteor. Soc. Japan, 77, 1–16.
——, and J. Dudhia, 2001: Coupling an advanced land surface-
hydrology model with the Penn State–NCAR MM5 modeling
system. Part II: Preliminary model validation. Mon. Wea.
Rev., 129, 587–604.
Courel, M. F., R. S. Kandel, and S. I. Rasool, 1984: Surface albedo
and the Sahel drought. Nature, 307, 528–531.
Dall’Olmo, G., and A. Karnieli, 2002: Monitoring phenological
cycles of desert ecosystems using NDVI and LST data de-
rived from NOAA-AVHRR imagery. Int. J. Remote Sens.,
23, 4055–4071.
deMenocal, P. B., 2001: Cultural responses to climate change dur-
ing the Late Holocene. Science, 292, 667–673.
DePauw, E., 2004: Drought early warning systems for the Near
East. Challenges and Strategies for Dryland Agriculture, S. C.
Rao and J. Ryan, Eds., American Society of Agronomy and
Crop Science Society of America, 93–111.
Dudhia, J., 1993: A nonhydrostatic version of the Penn State–
NCAR mesoscale model: Validation tests and simulation of
an Atlantic cyclone and cold front. Mon. Wea. Rev., 121,
1493–1513.
Eltahir, E. A. B., 1998: A soil moisture rainfall feedback mecha-
nism 1. Theory and observations. Water Resour. Res., 34, 765–
776.
Evans, J. P., and R. Geerken, 2004: Discrimination between cli-
mate and human-induced dryland degradation. J. Arid Envi-
ron., 57, 535–554.
——, and R. B. Smith, 2005: Classifying precipitation events in the
Fertile Crescent. Preprints, 16th Conf. on Climate Variability
and Change, San Diego, CA, Amer. Meteor. Soc., CD-ROM,
9.4.
——, and R. Geerken, 2006: Classifying rangeland vegetation type
and coverage using a Fourier component based similarity
measure. Int. J. Remote Sens., 105, 1–8.
FAO, 1997: Irrigation in the Near East region in figures. Water
Reports 9, U.N. Food and Agriculture Organization, 292 pp.
[Available online at http://www.fao.org/docrep/w4356e/
w4356e00.htm.]
Geerken, R., and M. Ilaiwi, 2004: Assessment of rangeland deg-
radation and development of a strategy for rehabilitation.
Remote Sens. Environ., 90, 490–504.
——, N. Batikha, D. Celis, and E. Depauw, 2005a: Differentiation
of rangeland vegetation and assessment of its status: Field
investigations and MODIS and SPOT VEGETATION data
analyses. Int. J. Remote Sens., 26, 4499–4526.
——, B. Zaitchik, and J. P. Evans, 2005b: Classifying rangeland
vegetation type and coverage from NDVI time series using
Fourier filtered cycle similarity. Int. J. Remote Sens., 26,
5535–5554.
Grell, G. A., J. Dudhia, and D. R. Stauffer, 1994: A description of
the fifth-generation Penn State/NCAR Mesoscale Model
(MM5). NCAR Tech. Note NCAR/TN-398⫹STR, Boulder,
CO, 117 pp.
Gutman, G., and A. Ignatov, 1998: The derivation of the green
vegetation fraction from NOAA/AVHRR data for use in
1A
UGUST 2007 Z A I TCHIK ET AL. 3939

numerical weather prediction models. Int. J. Remote Sens.,
19, 1533–1543.
Heck, P., D. Luthi, and C. Schär, 1999: The influence of vegeta-
tion on the summertime evolution of European soil moisture.
Phys. Chem. Earth, 24B, 609–614.
Hole, F., 1994: Environmental instabilities and urban origins.
Chiefdoms and Early States in the Near East: the Organiza-
tional Dynamics of Complexity, Prehistory Press, 121–151.
Hong, S. Y., and H. L. Pan, 1996: Nonlocal boundary layer verti-
cal diffusion in a medium-range forecast model. Mon. Wea.
Rev., 124, 2322–2339.
Jacquemin, B., and J. Noilhan, 1990: Sensitivity study and valida-
tion of a land surface parameterization using the HAPEX-
MOBILHY data set. Bound.-Layer Meteor., 52, 93–134.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-
analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Koster, R. D., and Coauthors, 2004: Regions of strong coupling
between soil moisture and precipitation. Science, 305, 1138–
1140.
Kremer, R. G., E. R. Hunt, S. W. Running, and J. C. Coughlan,
1996: Simulating vegetational and hydrologic responses to
natural climatic variation and GCM-predicted climate change
in a semi-arid ecosystem in Washington, USA. J. Arid Envi-
ron., 33, 23–38.
Lane, D. R., D. P. Coffin, and W. K. Lauenroth, 1998: Effects of
soil texture and precipitation on above-ground net primary
productivity and vegetation structure across the central grass-
land region of the United States. J. Veg. Sci., 9, 239–250.
Laval, K., and L. Picon, 1986: Effect of a change of the surface
albedo of the Sahel on climate. J. Atmos. Sci., 43, 2418–2429.
LeHouerou, H. N., 1996: Climate change, drought and desertifi-
cation. J. Arid Environ., 34, 133–185.
Liang, S. L., C. J. Shuey, A. L. Russ, H. L. Fang, M. Z. Chen, C. L.
Walthall, C. S. T. Daughtry, and R. Hunt, 2002: Narrowband
to broadband conversions of land surface albedo: II. Valida-
tion. Remote Sens. Environ., 84, 25–41.
Liou, K. N., 2002: An Introduction to Atmospheric Radiation.2d
ed. International Geophysics Series, Vol. 84, Academic Press,
583 pp.
Los, S. O., 1993: Calibration adjustment of the NOAA AVHRR
normalized difference vegetation index without recourse to
component channel-1 and channel-2 data. Int. J. Remote
Sens., 14, 1907–1917.
Mahrt, L., and M. Ek, 1984: The influence of atmospheric stability
on potential evaporation. J. Climate Appl. Meteor., 23, 222–
234.
——, and H. L. Pan, 1984: A two-layer model of soil hydrology.
Bound.-Layer Meteor., 29, 1–20.
Maisongrande, P., B. Duchemin, and G. Dedieu, 2004: VEGETA-
TION/SPOT: An operational mission for the Earth monitor-
ing; presentation of new standard products. Int. J. Remote
Sens., 25, 9–14.
Matsui, T., V. Lakshmi, and E. E. Small, 2005: The effects of
satellite-derived vegetation cover variability on simulated
land–atmosphere interactions in the NAMS. J. Climate, 18,
21–40.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A.
Clough, 1997: Radiative transfer for inhomogeneous atmo-
sphere: RRTM, a validated correlated-k model for the long-
wave. J. Geophys. Res., 102, 16 663–16 682.
Myhre, G., E. J. Highwood, K. P. Shine, and F. Stordal, 1998: New
estimates of radiative forcing due to well mixed greenhouse
gases. Geophys. Res. Lett., 25, 2715–2718.
Nielsen, T. T., and H. K. Adriansen, 2005: Government policies
and land degradation in the Middle East. Land Degradation
Development, 16, 151–161.
Pan, H. L., and L. Mahrt, 1987: Interaction between soil hydrol-
ogy and boundary-layer development. Bound.-Layer Meteor.,
38, 185–202.
Reisner, J., R. J. Rasmussen, and R. T. Bruintjes, 1998: Explicit
forecasting of supercooled liquid water in winter storms using
the MM5 mesoscale model. Quart. J. Roy. Meteor. Soc., 124B,
1071–1107.
Rodwell, M. J., and B. J. Hoskins, 1996: Monsoons and the dy-
namics of deserts. Quart. J. Roy. Meteor. Soc., 122, 1385–
1404.
Schaake, J. C., V. I. Koren, Q. Y. Duan, K. Mitchell, and F. Chen,
1996: A simple water balance model (SWB) for estimating
runoff at different spatial and temporal scales. J. Geophys.
Res., 101, 7461–7475.
Schär, C., D. Luthi, U. Beyerle, and E. Heise, 1999: The soil–
precipitation feedback: A process study with a regional cli-
mate model. J. Climate, 12, 722–741.
——, P. L. Vidale, D. Luthi, C. Frei, C. Haberli, M. A. Liniger,
and C. Appenzeller, 2004: The role of increasing temperature
variability in European summer heatwaves. Nature, 427, 332–
336.
Schmidt, H., and A. Karnieli, 2000: Remote sensing of the sea-
sonal variability of vegetation in a semi-arid environment. J.
Arid Environ., 45, 43–59.
Shukla, J., and Y. Mintz, 1982: Influence of land-surface evapo-
transpiration on the earth’s climate. Science, 215, 1498–1501.
Small, E. E., and S. A. Kurc, 2003: Tight coupling between soil
moisture and the surface radiation budget in semiarid envi-
ronments: Implications for land-atmosphere interactions.
Water Resour. Res., 39, 1278–1291.
Sud, Y. C., and M. Fennessy, 1982: A study of the influence of
surface albedo on July circulation in semi-arid regions using
the GLAS GCM. J. Climatol., 2, 105–125.
——, and M. J. Fennessy, 1984: Influence of evaporation in semi-
arid regions on the July circulation—A numerical study. J.
Climatol., 4, 383–398.
——, and W. E. Smith, 1985: The influence of surface-roughness
of deserts on the July circulation (a numerical study).
Bound.-Layer Meteor., 33, 15–49.
——, D. M. Mocko, K. M. Lau, and R. Atlas, 2003: Simulating the
midwestern U.S. drought of 1988 with a GCM. J. Climate, 16,
3946–3965.
Vanacker, V., M. Linderman, F. Lupo, S. Flasse, and E. Lambin,
2005: Impact of short-term rainfall fluctuation on interannual
land cover change in sub-Saharan Africa. Global Ecol. Bio-
geogr., 14, 123–135.
Weiss, E., S. E. Marsh, and E. S. Pfirman, 2001: Application of
NOAA-AVHRR NDVI time-series data to assess changes in
Saudi Arabia’s rangelands. Int. J. Remote Sens., 22, 1005–
1027.
Weiss, H., and R. S. Bradley, 2001: What drives societal collapse?
Science, 291, 609–610.
Weiss, J. L., D. S. Gutzler, J. E. A. Coonrod, and C. N. Dahm,
2004: Long-term vegetation monitoring with NDVI in a di-
verse semi-arid setting, central New Mexico, USA. J. Arid
Environ., 58, 249–272.
Xie, P. P., and P. A. Arkin, 1996: Analyses of global monthly
precipitation using gauge observations, satellite estimates,
and numerical model predictions. J. Climate, 9, 840–858.
Zaitchik, B. F., J. Evans, and R. B. Smith, 2005: MODIS-derived
3940 JOURNAL OF CLIMATE VOLUME 20

boundary conditions for a mesoscale climate model: Appli-
cation to irrigated agriculture in the Euphrates basin. Mon.
Wea. Rev., 133, 1727–1743.
——, A. K. Macalady, L. R. Bonneau, and R. B. Smith, 2006: Eu-
rope’s 2003 heat wave: A satellite view of impacts and land–
atmosphere feedbacks. Int. J. Climatol., 26, 743–769.
Zawadzki, I., E. Torlaschi, and R. Sauvageau, 1981: The relation-
ship between mesoscale thermodynamic variables and con-
vective precipitation. J. Atmos. Sci., 38, 1535–1540.
Zhang, H., A. Henderson-Sellers, A. J. Pitman, J. L. McGregor,
C. E. Desborough, and J. J. Katzfey, 2001: Limited-area
model sensitivity to the complexity of representation of the
land surface energy balance. J. Climate, 14, 3965–3986.
Zheng, Y. Q., G. Yu, Y. F. Qian, M. Q. Miao, X. M. Zeng, and
H. Q. Liu, 2002: Simulations of regional climatic effects of
vegetation change in China. Quart. J. Roy. Meteor. Soc., 128,
2089–2114.
Ziv, B., H. Saaroni, and P. Alpert, 2004: The factors governing the
summer regime of the eastern Mediterranean. Int. J. Clima-
tol., 24, 1859–1871.
1A
UGUST 2007 Z A I TCHIK ET AL. 3941