Dt

Na

b

a

ARR2A

KETSTSPS

1

lacabanokitb

D

m

0d

Agricultural and Forest Meteorology 157 (2012) 77– 85

Contents lists available at SciVerse ScienceDirect

Agricultural and Forest Meteorology

jou rn al h om epa g e: www.elsev ier .com/ locate /agr formet

ynamics of evapotranspiration partitioning in a semi-arid forest as affected byemporal rainfall patterns

aama Raz-Yaseefa,1, Dan Yakira,∗, Gabriel Schillerb, Shabtai Cohenb

Department of Environmental Science and Energy Research, Weizmann Institute of Science, Rehovot, IsraelDepartment of Environmental Physics and Irrigation, Institute of Soil, Water and Environmental Science, A.R.O. Volcani Center, Israel

r t i c l e i n f o

rticle history:eceived 1 August 2011eceived in revised form6 December 2011ccepted 21 January 2012

eywords:vapotranspiration partitioningree transpirationoil evaporationemporal precipitation patternemi-arid

a b s t r a c t

We extend our recent study of the effects of tree density on evapotranspiration (ET) partitioning in asemi-arid pine forest by examining the influence of the temporal patterns in rainfall (P) on the dynamiccontributions of tree transpiration (Tt), soil evaporation (Es) and rainfall interception (IP) to total ET. Soilevaporation accounted for 39% of average annual ET over the four-year period, and was associated withsoil moisture content in the upper 5 cm and solar radiation, therefore peaking during the wetting anddrying seasons (up to 0.75 mm day−1). In the dry summer, Es diminished and as much as 50% of theresidual flux was due to re-evaporation of moisture condensed at night (adsorption). Tree transpirationaccounted for 49% of average annual ET, and was associated with soil moisture at a depth of 10–20 cm.Transpiration peaked only in late spring (1.5 mm day−1), after the accumulation of large storms allowinginfiltration below the topsoil. Moisture at these depths was maintained for longer periods and was evencarried over between rain seasons following a high precipitation year. Interception was 12% of annual ET

ine forestoil water adsorption

but was larger than 20% during the rainy period. The results indicated that both Tt/ET and Es/ET could varybetween 30% and 60% due to their differential response to seasonal environmental drivers. Annual Tt/ET,a major parameter indicating forest productivity and survival, was more influenced by the occurrence oflarge storms (>30 mm; P30/P ratio) than by P itself. In an assessment of the potential warming and dryingtrends predicted for the Mediterranean region in the next century, changes in both total precipitationand in its temporal patterns must be considered.

. Introduction

Evapotranspiration (ET) is an integrated term, composed of theinked vapor fluxes of plant transpiration (T), soil evaporation (Es)nd canopy-intercepted precipitation (IP). Whereas Es is a physi-ally controlled flux, T is strongly influenced by plant physiologynd can be affected by abiotic environmental conditions, but alsoy plant species characteristics, stomatal sensitivity, mycorrhizalssociations, plant disease, atmospheric CO2 concentrations andutrients. A growing awareness of the importance of ecohydrol-gy has motivated efforts to partition ET into its components, as aey to unraveling processes underlying ecosystem water use and

ts response to change (Huxman et al., 2005). Previous studies showhat the T/ET ratio varies greatly among ecosystems and timescales,ut on an annual basis it is mostly in the range 40–70% (Reynolds

∗ Corresponding author. Tel.: +972 8 934 2549; fax: +972 8 934 4124.E-mail addresses: naama.ry@berkeley.edu (N. Raz-Yaseef),

an.Yakir@Weizmann.ac.il (D. Yakir).1 Current address: Department of Environmental Science, Policy and Manage-ent, UC Berkeley, Berkeley, CA, USA.

168-1923/$ – see front matter © 2012 Elsevier B.V. All rights reserved.oi:10.1016/j.agrformet.2012.01.015

© 2012 Elsevier B.V. All rights reserved.

et al., 2000; Mitchell et al., 2009; Moran et al., 2009; Zhongminet al., 2009; Cavanaugh et al., 2010; Staudt et al., 2011). This range ofratios emphasizes that even in water-limited environments, plantsdo not use all of the precipitation input and major water lossesoccur, mainly to Es and runoff. Vegetation type, through its effecton the proportion of the shaded surface fraction, also influencesT/ET, which generally increases from grasses to shrubs to trees(Kostner, 2001; Moran et al., 2009; Raz-Yaseef et al., 2010a; Wanget al., 2010). The T/ET ratio has been shown to vary among ecosys-tems according to the depth of water uptake, and is larger fordeep-rooted trees than for grasses, as well as for soils with bet-ter infiltration regimes than more impermeable soils (Scholes andArcher, 1997; Laio et al., 2001; Kurc and Small, 2004; Cavanaughet al., 2010).

When precipitation is characterized by short and sporadicshowers, such as observed in some semi-arid sites (e.g. Sharon,1972; Sala and Lauenroth, 1982; Lapitan and Parton, 1996; Loiket al., 2004), infiltration depth can be reduced, limiting moisture to

shallow depths. Soils with high densities and fine textures exhibitlow infiltration rates and high water holding capacities, further con-straining infiltration (e.g. Noy-Meir, 1973; Reynolds et al., 2000;Scott et al., 2000; Kochendorfer and Ramírez, 2008). Even in such

7 nd For

cedsdetaacc

pit2tttapt

2

2

boTsleatmtfplit1slNd2o

2

t2Raaatfwt

p

8 N. Raz-Yaseef et al. / Agricultural a

ases, infiltration below the topsoil can occur following infrequentvents with relatively large rain amounts. In some extreme events,eep infiltration can lead to soil moisture “storage” through the dryeason and into the following wet seasons, as has been reported inry ecosystems (Paruelo et al., 2000; Kurpius et al., 2003; Tietjent al., 2009; Raz-Yaseef et al., 2010b). Pulsed precipitation eventshat lead to infiltration deep enough to increase root uptake have

marked effect on above-ground biomass production (Schwinningnd Sala, 2004; Heisler-White et al., 2008; Knapp et al., 2008) andan contribute to ecosystem resilience and survival under droughtonditions.

Quantifying T/ET is key to predicting ecosystem survival androductivity, especially in water-limited regions. This is extremely

mportant in light of the drying and warming trends predicted forhe entire Mediterranean region and the southwestern US (IPCC,007). The objective of this four-year study was first, to definehe dynamics of ET partitioning on seasonal and, within limits ofhe data-collection period, annual time scales. Second, we aimedo identify the main environmental and meteorological conditionsffecting this flux partitioning. Finally, we attempted to connect ETartitioning to the large observed variations in temporal precipita-ion patterns in the semi-arid pine forest ecosystem.

. Methods

.1. Study area

Yatir Forest is located in Southern Israel, at the transition zoneetween sub-humid Mediterranean and arid climates, on the edgef the Judean Mountain ridge (31◦21′N and 35◦02′E, 630 m AMSL).he site is a ca. 45-year-old Pinus halepensis afforestation, currentlypread over an area of 28 km2, with a density of ∼300 trees ha−1,eaf area index (LAI) of ∼1.50 m2 m−2, and sparse understory veg-tation (Maseyk et al., 2008; Grunzweig et al., 2009; Rotenbergnd Yakir, 2010; Sprintsin et al., 2011). Average air tempera-ures for this region are 10 and 25 ◦C for the coldest and hottest

onths, January and July, respectively. Average precipitation forhe last 30 years is 285 ± 88 mm year−1. Nearly all precipitationalls between December and March, followed by an extended dryeriod during the hot summer. Potential ET (1600 mm year−1)

argely exceeds precipitation inputs. The soil at the research sites shallow (20–40 cm), Aeolian-origin loess with a clay-loam tex-ure (0.31 ± 0.02 sand, 0.41 ± 0.10 silt and 0.28 ± 0.04 clay; density:.65 ± 0.14 g cm−3) overlying chalk and limestone bedrock. Deeperoils (up to 1.5 m) are sporadically located at topographic hol-ows. While the natural rocky hillslopes of the semi-arid northernegev are known to produce flash floods, the forest reduces runofframatically (to less than 5% of precipitation, Shachnovich et al.,008). Groundwater is deep (>300 m), eliminating the possibilityf groundwater recharge or utilization.

.2. Meteorological measurements

An instrumented tower was erected in the geographical cen-er of Yatir forest, following Euroflux methodology (Aubinet et al.,000). The system uses a 3D sonic anemometer (Omnidirectional3, Gill Instruments) and a closed path LI-COR 7000 CO2/H2O gasnalyzer (LI-COR Inc.) to measure the evapotranspiration flux (ET)nd net CO2 flux (NEE). The flux tower’s footprint was definedccording to Gockede et al. (2008), indicating that the largest con-ribution to the flux during the day was from an area 34 m awayrom the tower and that 95% of the recorded flux came from an area

ithin 1300 m of the tower. Wind direction was from the northwest

o southwest sector during 64% of the day.Air temperature (Tempa) was measured with temperature

robes at heights of 1, 5, 9, 15 and 19 m. Soil temperatures (Temps)

est Meteorology 157 (2012) 77– 85

at 2 and 6 cm depth were measured at six different points aroundthe tower with thermocouples. Wind speed (Ws), wind direction(WD), relative humidity of the atmosphere (RH), shortwave radi-ation, longwave radiation and photosynthetically active radiationwere measured above and below the canopy, at heights of 15 and1 m, respectively; both the upward and downward components ofradiation were measured (Rotenberg and Yakir, 2011).

Precipitation (P) was measured by (1) a recording rain gauge(Campbell Scientific, USA) positioned at a height of 15 m on top ofthe flux tower, collecting data every half hour, and (2) a standardrain station positioned in a clearing in the forest, at a distance of1.50 km from the tower site, from which data have been manuallycollected on rainy days since 1971. Data from this rain station (Yatirforest, KKL) were used to determine long-term average annualprecipitation for the site. Data from the rain gauge were used tocalculate intercepted precipitation (IP) for individual rain events,based on the equation provided by Shachnovich et al. (2008):

IP(mm) = P (mm) − 0.94 · P (mm) − 0.76 (1)

2.3. Soil water content

Volumetric soil water content (�) was measured at a half-hourtime resolution with three reflectrometry sensors (CS616, Camp-bell Scientific) positioned vertically in the ground and measuring anaverage value for soil depth of 0–30 cm (�0–30). A specific calibrationequation was prepared in the laboratory for these sensors to fit thedense soil at our site (according to the manufacturer’s instructions).The sensitivity of the CS616 measurements to temperature wascorrected by the factory-defined temperature-correction equation.Soil temperatures were measured at depths of 1, 5, 15 and 30 cmin close proximity to the CS616 sensors (HOBO H8 loggers, OnsetComputers). In 2005, time domain reflectometry (TDR) sensors(TRIME, IMKO Inc.) were installed horizontally in three differentpits dug around the tower. The pits varied in depth according tothe soil/bedrock structure at each site. Sensors were installed hor-izontally, at constant depths of 5, 15, 30 (deepest sensor in pit 1),50, 70 (deepest sensor in pit 2) and 125 cm (deepest sensor in pit3). Variability between sensors of similar depths but different posi-tions (pits) was less than 5%. Values of �x (x = 5, 15, 30, 50, 70 or125 cm) presented herein are averages of the one to three sensorsavailable for a particular depth. These values denote soil moisturemeasured at a specific depth, rather than the integral measurementof the previous method (CS616) for depth 0–30 cm (�0–30).

2.4. Transpiration

Sap flux was used to estimate tree transpiration (Tt; the sub-script ‘t’ is added to emphasize that while there is an additional,very small below-canopy component, here only tree transpirationis considered) with two similar techniques: the ‘Tmax’ heat-pulsevelocity method (HPV; Cohen, 1994) and the Granier method(Granier, 1987). HPV temperature signals were converted to massflow rate based on empirical calibration coefficients suitable forYatir trees (type and size; Schiller and Cohen, 1998; Cohen et al.,2008). The HPV system was operated during 2003/2004 and2004/2005; the Granier system was operated during 2005/2006.Measurements were conducted hourly (including night hours) foreight trees representing average forest tree size, age and slopeaspect in the flux-tower footprint. Water uptake (L h−1 tree−1)was converted to Tt (mm h−1) according to forest tree density(300 ha−1). The Granier sensors were shorter than the HPV sensors

(20 and 60 mm, respectively) and did not reach all depths of theconductive sapwood. According to Cohen et al. (2008), conductivesapwood diameter for semi-arid Pinus halepensis is ∼40 mm, butsap velocity decreases with diameter. According to these findings,

nd Forest Meteorology 157 (2012) 77– 85 79

acd(t(su(2iPe

2

2

msflbRfdpsofl

2

(obcet

2

ssetathtmflpa

E

wnmf

mofl

Fig. 1. Daily values of precipitation (P), soil water content averaged for depth

tion, two dry years and one wet year (Table 1). The two years with

N. Raz-Yaseef et al. / Agricultural a

correction factor of 1.72 was applied to the Granier sensors, toompensate for measuring to a depth of only 20 mm. We comparedata from the HPV and Granier systems for an overlapping periodMarch–May 2005) and found a high correlation between thewo methods, with a ratio resembling the above correction factorTt(HPV) = 1.70·Tt(Granier), R2 = 0.95). Fluxes for 2006/2007 wereimulated based on the relationships observed in Yatir (G. Schiller,npublished), simulating Tt (mm day−1) as a function of �0–30%) and potential evapotranspiration (PET; mm day−1), during the-year measuring period. All terms in the model were highly signif-

cant (P < 0.0001), and the overall adjusted R2 of the model was 0.61.ET for Yatir was computed according to the Penman–Monteithquation procedure recommended by Allen et al. (1998).

.5. Soil evaporation

.5.1. Plot-scale flux measurementsMeasurements of soil evaporation (Es) were conducted with a

odified LI-COR 6400-09 soil CO2 flux chamber. Prior to the mea-urements, this methodology was calibrated and tested againstuxes measured with microlysimeters, and showed high relia-ility (for fluxes in the range of −0.02 to 0.40 mm h−1, n = 59,2 = 0.93; Raz-Yaseef et al., 2010a). In the field, Es was measuredrom 14 permanently placed soil collars. Measurements were con-ucted manually on 42 days, spread evenly throughout the researcheriod (October 2004–May 2007) and representing different sea-ons and environmental conditions. Measurements were carriedut between 1100 and 1200 h, the expected time of peak diurnal Es

ux (Baldocchi et al., 2000; Williams et al., 2004).

.5.2. From plot- to stand-scale measurementsSpatial variability between the 14 measured soil collars was high

STD = ±47%), and upscaling was achieved according to the method-logy elaborated in Raz-Yaseef et al. (2010a). This methodology isased on the difference between Es fluxes measured below treeanopies and in the gaps between trees, and the ratio of shaded toxposed areas of the forest floor, which varies seasonally accordingo the solar declination.

.5.3. Estimating daily fluxesFull diurnal cycles of Es (starting before sunrise and ending after

unset) were measured in six campaigns representing differenteasons and conditions. On these days, measurements were takenvery half-hour on two adjacent soil collars (2 m apart), one posi-ioned below a tree canopy (shaded) and the other positioned in

gap between the trees (exposed to direct sun). During three ofhe six days, soil-chamber measurements of Es were conductedourly on five additional chambers, confirming that the diurnalrend measured at the two main soil chambers was similar to that

easured at other locations (results not shown). Although dailyuxes and trends varied seasonally, a strong relationship betweeneak noontime values and total daily values of Es was observed forll measured days (R2 = 0.91; n = 12):

DT (mm day−1) = 0.39 · ln[Es-noon (mm h−1)] + 1.49 (2)

here EDT is the daily total soil evaporation and Es-noon is the peakoontime flux of soil evaporation. We used this relationship to esti-ate total daily Es values from measured noontime values on days

or which a full diurnal cycle was not obtained.Good hydrological closure was achieved on daily timescales for

ost days, with 0.97 > [(Es + Tt)/ET] > 0.90 (where Es, Tt, and ET werebtained from soil chambers, sap flow and eddy covariance at theux tower, respectively).

0–30 cm (�0–30), air temperature at 1 m (Tempa) and evaporation fluxes (total evap-otranspiration – ET, tree transpiration – Tt and soil evaporation – Es) during the4-year study period.

2.5.4. Simulating soil evaporationThe 42 days on which Es was measured were divided into four

seasonal groups (autumn, winter, spring and summer). The divi-sion into seasons was conducted according to thresholds in soilwater content and air temperature, and not according to fixedcalendar definitions (see Section 3.2). Using multiple linear cor-relation regression analysis (SPSS 13.0 Software, SPCC Inc., 2004), astepwise selection defined the dominant variables controlling Es

for each season. The environmental parameters included in theanalysis were: soil water content at different depths (�5 to �125),net radiation (Rn), vapor-pressure deficit (VPD), air temperatureat different heights (Tempa 1 m to Tempa 19 m), soil temperature atdifferent depths (Temps 2 cm and Temps 6 cm) and wind speed (Ws).This procedure was conducted for half-hour and daily time steps.The obtained regressions were used together with the continuousmeasurements of the environmental variables at our permanentflux-measurement site to extend our 42 measurement campaignsto a continuous estimated record of Es over the 4-year study period(Fig. 1).

3. Results

3.1. Precipitation and soil water content

Environmental conditions were typical for the region (Fig. 1),with the rainy season usually starting in October with smallrain events (∼5 mm day−1), while the main rain period, betweenDecember and March, had larger storm events (∼20 mm day−1 butup to 60 mm day−1). Rain typically ceased in March/April. Soil watercontent mirrored the seasonality in precipitation: �0–30 respondedalmost instantaneously to the first autumn rains, and remainedabove 0.30 m3 m−3 throughout most winters. Spring drying lastedapproximately two months, during which �0–30 decreased to0.09 m3 m−3. These low soil moisture conditions persisted through-out the approximately half year long dry summer period.

Average annual precipitation during the four year researchperiod was similar to the long-term (35-year) mean value (285vs. 290 mm, respectively), with one year of near-average precipita-

low precipitation, 2003/2004 and 2005/2006, had similar annualprecipitation (231 and 224 mm), but differed in its temporal dis-tribution: a similar number of rain events, 27 and 28 rainy days,

80 N. Raz-Yaseef et al. / Agricultural and Forest Meteorology 157 (2012) 77– 85

F transp( radiar

r1tt

3

ecn2tdaw(odw2cas

TAde

ig. 2. Diurnal cycles of evapotranspiration (ET; half-hour values, flux tower), tree

E1, E2; half-hour values, two adjacent soil collars, soil-chamber technique) and netepresent downward fluxes from the atmosphere to the forest or soil.

espectively, were spread over 132 days in the first year and over79 days in the second year. The long intervals between precipita-ion events during 2005/2006 allowed soil water content to declineo near summer values (Fig. 1).

.2. Dynamics of the evaporation fluxes

Diurnal cycles: During the wet season, a clear diurnal cycle invaporation fluxes was observed (Fig. 2), generally resembling theycle of net radiation. In summer, fluxes were small and the diur-al cycle was less noticeable. During a heat-wave episode (9 June005), tree transpiration values remained similar to those duringhe mid-day depression throughout the day (Fig. 2f). During thery season, small negative fluxes in soil evaporation (i.e. from thetmosphere to the soil; −0.04 to −0.05 mm day−1) were detectedith the soil chamber technique before sunrise and after sunset

Fig. 3, upper panels). A similar phenomenon was measured withur soil chamber during the laboratory calibration process on oven-ried soils, and the negative fluxes measured with the soil chamberere validated by balanced-based measurements (Raz-Yaseef et al.,

010a). These fluxes were consistent with water adsorption from aool and humid atmosphere into the dry air within the soil pores,nd possibly to the hygroscopic soil water. Negative fluxes mea-ured in the field were further supported by the 1.5–3.0 h delay in

able 1nnual sums of precipitation (P), tree transpiration (Tt), soil evaporation (Es) and intercepuring the study period. The ratio between precipitation occurring in storms larger thanach year are also shown. The average value and standard deviation (STD) for the 4-years

P Tt Es

mm year−1 P30/P mm year−1 Tt/ET mm year−1

2003/2004 231 0.56 134 0.57 99

2004/2005 377 0.64 156 0.45 112

2005/2006 224 0.43 111 0.49 93

2006/2007 308 0.55 115 0.44 106

Average 285 0.55 129 0.49 102

STD 72 0.09 21 0.06 8

iration (Tt; half-hour values, average of 8 trees, sap flux method), soil evaporationtion (Rn). Tt was not measured on the first two measurement days. Negative values

diurnal peak soil moisture measured with the TDR, when comparedto atmospheric RH values (the moisture adsorption period, Fig. 3,lower panels; summer night RH > 75%, �5 < 0.10 m3 m−3).

Seasonal cycles: The seasonal cycle of ET was character-ized by low fluxes during the dry season (0.25–0.75 mm day−1,Figs. 1 and 4), and an increase only from the middle of the rainyseason (November/December) peaking in early spring (with typi-cal March ET fluxes of ∼1.50 mm day−1, but up to 3.25 mm day−1).Fluxes rapidly decreased after rain cessation, and reached the lowsummer values by April/May. Seasonal variations in transpirationwere similar to those of ET (Figs. 1 and 4), with a gradual rise in lateNovember, a peak in March/April (typical fluxes of 0.65 mm day−1,but up to 1.50 mm day−1 on some days) and a rapid decrease to alow summer flux of ∼0.20 mm day−1 from June and until the fol-lowing rainy season. Intercepted precipitation was significant onlyduring the rainy season (October–April), and peaked on average inFebruary, at 0.39 mm day−1.

Soil evaporation measurements indicated high fluxes during thetransition wetting and drying periods (autumn and spring, respec-tively; maximum daily fluxes of 0.75 mm day−1) and lower fluxes

in the range of 0.05–0.25 mm day−1 during both the wet winter anddry summer (Fig. 5). We therefore divided soil evaporation into sea-sons of characteristic fluxes, and used multivariable correlation onthe predefined seasonal basis to estimate daily Es. These seasons

ted precipitation (IP) and their partial contribution to total evapotranspiration (ET) 30 mm and total P (P30/P) and the water-use efficiency index (WUE; NEE/ET) fortudy period are noted below.

IP ET WUE

Es/ET mm year−1 IP/ET mm year−1 gCO2 kg H2O−1 year−1

0.42 27 0.11 235 0.930.33 39 0.11 343 1.060.41 26 0.11 227 0.740.40 33 0.13 263 0.89

39 31 0.12 267 0.900.04 6 0.01 53 0.13

N. Raz-Yaseef et al. / Agricultural and Forest Meteorology 157 (2012) 77– 85 81

Fig. 3. Upper panels: diurnal cycles of soil evaporation (Es) and net radiation (Rn) measured over two summer days. Soil adsorption (negative fluxes of Es) was measuredduring the hours before sunrise and after sunset. Lower panels: relative humidity (RH) and soil moisture in the topsoil (�5) on the same days as in upper panels. A 1.5–3 hd ere sc he top

wmttsatc(tssDD8

FtrEf

elay was observed between peak �5 and peak RH, but otherwise diurnal trends wooler, wetter night-time atmosphere, which produced an increase in moisture in t

ere as follows: summer (the water-limited season) where soiloisture content at the topsoil, �5 was above 0.06 m3 m−3; win-

er (the energy-limited season) with daily average Tempa < 15 ◦C;he autumn and spring transition seasons (both energy and surfaceoil water content are abundant), in which 0.06 < �5 < 0.23 m3 m−3

nd daily average Tempa > 15 ◦C. Applying these environmentalhresholds resulted in variations in the timing of the seasonalhanges among years. The summer season started on day of yearDOY) 142 ± 10 days. The autumn transition season was not iden-ifiable in two of the study years, implying a direct shift fromummer to winter conditions. During the other two years, the tran-ition to “autumn” conditions was on DOY 356 (2005/2006) and

OY 301 (2006/2007). Transition to “winter” conditions was onOY 351 ± 21. Transition to “spring” conditions occurred on DOY9 ± 10.

! " #$ %&

0.00

0.25

0.50

0.75

1.00

Oct Dec Feb Apr Jun Aug

Re

lative

co

ntib

utio

n

to E

T

Tt Es Ip

0.00

0.50

1.00

1.50

Eva

po

ratio

n F

lux

(m

m d

-1)

Tt Es Ip ET

ig. 4. Upper panel: monthly averages (mm day−1) of the separate fluxes of treeranspiration (Tt), soil evaporation (Es) and intercepted precipitation (IP) for theesearch period. Lower panel: relative contribution of each of these fluxes to totalT. The seasonal cycle, as well as the relative contribution of each flux, was differentor each of the fluxes.

imilar. This suggests that the negative values measured for Es originated from thesoil.

The resulting season-specific multivariable linear correlationsbetween Es and environmental parameters were high and signif-icant (0.68 < R2 < 1.00, P < 0.05, Table 2). The analysis defined therelative contribution of each parameter to the regressions (ˇ%). Soilwater content in the topsoil was found to control soil evaporationduring the wetting period ( ̌ = 81%). During the winter, there wasno one important variable, but air temperature was the most influ-ential one ( ̌ = 28%). Soil water content ( ̌ = 40%) and wind speed( ̌ = 33%) were the largest contributing variables during the dryingseason. During the summer, the main factor affecting Es was netradiation ( ̌ = 35%), and soil water content was insignificant. Notethat similar correlations were not statistically significant when theabove seasonality was not applied.

Annual cycles: On an annual basis, ET varied between 227and 343 mm year−1 with a normalized standard deviation (NSTD;STD/average) of 0.20, comparable to that for precipitation(NSTD = 0.25). The ET/P ratio varied between 1.02 and 0.85, andwas negatively correlated with precipitation: evaporation losseswere slightly larger than precipitation inputs during the dry years.Despite the large interannual variability in precipitation and ET,

the hydrological budget was nearly closed in all years, and on aver-age for the four year research period, [(Es + Tt + IP)/ET] = 1.00 ± 0.09,validating our experimental approach.

Fig. 5. Seasonal trend of air temperature (Tempa) and soil moisture in the topsoil(�5), averaged over the three years. Es fluxes were high during the wetting and dryingtransition periods and low during the winter period (energy-limited, Tempa < 15 ◦C)and summer period (water-limited, �5 < 0.06 m3 m−3).

82 N. Raz-Yaseef et al. / Agricultural and Forest Meteorology 157 (2012) 77– 85

Table 2Analytical results of seasonal multiple linear regressions between noontime soil evaporation (Es) and environmental and meteorological variables. A list of all variables usedfor this analysis appears in the methods section. The dominant variables for each regression were chosen automatically by stepwise procedure: water content in the topsoil(�5), net radiation (Rn), vapor-pressure deficit (VPD), air temperature (Tempa) and wind speed (Ws). The number of days used for the analysis (n), correlation coefficient withmeasured parameters (R2) and significance (P) are presented for each regression; the coefficients (B) and their relative contributions (ˇ%) are presented for each variable.

Wetting season Winter Drying season Summer

n R2 P n R2 P n R2 P n R2 P

7 0.682 0.045 16 0.813 0.034 5 1.000 – 14 0.977 0.046

Wetting season Winter Drying season Summer

B ˇ% B ˇ% B ˇ% B ˇ%

Cons. −6.57E−02 −4.90E−02 −4.22E−02 1.03E−01�5 1.22E+00 81 3.20E−01 22 2.07E−01 40Rn 4.82E−05 16 −5.05E−03 35VPD 1.03E−02 19 −1.24E−02 22 −2.41E−03 10 −2.53E−03 24

28

13

1oCv

3p

t1(aep

tltwtwfspab

drctewpswii

yPpsde

study by focusing on the hypothesis that at a given forest tree den-sity, the temporal patterns of precipitation, and consequently thedistribution of soil moisture, also have an important influence on ETpartitioning and dynamics. This, in turn, must have consequences

Fig. 6. Relationships between the evaporation fluxes and soil water content at dif-

Tempa 2.31E−03

Ws −8.09E−03

Annual values of tree transpiration varied in the range of11–156 mm year−1 (NSTD = 0.16). Annual values of soil evap-ration varied between 93 and 112 mm year−1 (NSTD = 0.08).alculated values of intercepted precipitation (IP; using Eq. (1))aried between 26 and 39 mm year−1 (NSTD = 0.17).

.3. ET partitioning as affected by the temporal precipitationattern

On average for the whole research period, tree transpira-ion was the largest water flux, making up 49% of ET with29 ± 21 mm year−1. Soil evaporation accounted for 39% of ET102 ± 8 mm year−1), and intercepted precipitation was 12% of ETt 31 ± 6 mm year−1. Annual values of soil water adsorption werestimated to be in the range of 10–15 mm year−1, adding ∼5% torecipitation inputs.

On a monthly and seasonal basis, the relative contributions ofhe individual evaporation fluxes to total ET varied greatly (Fig. 4,ower panel). Soil evaporation values were similarly high duringhe wetting and drying seasons, but the relative contribution of Es

as largest during the wetting season (Es/ET up to 60%). Duringhe drying season, when photosynthesis and transpiration fluxesere highest, Es accounted for less than 40% of ET. The contribution

rom tree transpiration, Tt/ET varied, between 30% and 60%, and itseasonal cycle differed from that of Tt (Tt peaked in March and Tt/ETeaked in May). As previously noted, intercepted precipitation was

relatively small component on an annual scale (12% on average),ut values accounted for up to 20% of ET during the rainy months.

The differential seasonal variations in Tt and Es also reflected theifferential changes in soil moisture at different depths. Soil evapo-ation correlated best with �5, soil moisture at the topsoil, while Tt

orrelated best with �15 (Fig. 6). The latter correlation was consis-ent with maximal root density at a depth of 10–20 cm (Grunzweigt al., 2009: fine root density at depths 0–10, 10–20 and 20–30 cmas 3.40, 5.30 and 1.60 mg cm−3, respectively). Interannual com-arison showed that �5 was consistently high throughout the weteason and could support high soil evaporation fluxes when energyas available. In contrast, soil moisture in the main root zone, �15,

ncreased only toward the middle of the winter season, as reflectedn the timing of peak Tt flux.

This effect can be demonstrated by two consecutive years, a wetear (2004/2005: P = 377 mm), followed by a dry one (2005/2006:

= 224 mm). In the topsoil layer, the duration of excess water (the

eriod during which soil moisture increased above the low, con-tant, summer values) was similar for both years, despite the largeifference in P (Fig. 7). In contrast, at a depth of 30 cm, the period ofxcess water was much shorter during the low precipitation year.

7.56E−04 17 −8.08E−04 263.66E−02 33 −4.27E−04 14

Furthermore, the period of excess soil moisture at depth in the wetyear extended into the following hydrological year, constituting‘water storage’ between years.

4. Discussion

Recently, we showed (Raz-Yaseef et al., 2010a) that ET parti-tioning, mainly between Tt and Es, is related to LAI, with largerproportions of Es for open forests, consistent with other studies(Yepez et al., 2005; Lawrence et al., 2007; Mitchell et al., 2009;Newman et al., 2010). Thus an increase in LAI (representing tree sizeand density) will increase Tt and decrease Es, both because of thelarger tree canopy and because of the larger shaded under-canopyfraction. The 65% canopy cover at our study site was previouslyshown to produce a near optimal Tt/ET ratio, as greater tree den-sity and canopy cover would lead to a negative hydrological budgetand tree mortality (Raz-Yaseef et al., 2010a). Here we extended that

ferent depths. Es was synchronized with soil water content in the topsoil (�5; upperpanel), while Tt correlated best with soil water content in the main root zone (�15;lower panel). Below the graphs, a table with the linear regression coefficients (R2)between � at different depths and Es or Tt are shown, expressing the best fits arebetween Es and �5 and Tt and �15.

N. Raz-Yaseef et al. / Agricultural and For

Fig. 7. The period of excess soil water in the soil profile over two consecutive years: awet year (2004/2005, P=377 mm) and a dry year (2005/2006, P=224 mm). Thresholdsfor excess soil water content were defined for each layer based on summer values:� > 0.10, � > 0.16, � > 0.19, � > 0.25, � > 0.27 and � > 0.30 m3 m−3. Wet soilcb

fct

oddcdiwndppaaRdd(iab

ctomWwmso

5 15 30 50 70 125

onditions in the deepest soil layer following the wet year were still observed at theeginning of the following dry year.

or forest productivity and tree survival in semi-arid zones, whichould greatly expand under predicted climate change scenarios forhe Mediterranean and other regions.

A key observation in our results was the differential responsef the individual evaporation flux components to environmentalrivers. This resulted in large variations in the contribution of theifferent components, Tt, Es and IP, to total ET during the seasonalycle (Fig. 4). What, then, are the main processes affecting theynamics of ET partitioning on a seasonal timescale? Not surpris-

ngly, results showed that soil water content in the topsoil layeras the dominant factor in Es variations (Table 2). Although a largeumber of parameters were used for these regressions, our analysisemonstrated that measurements of topsoil water content, tem-erature, net radiation and VPD are sufficient to estimate Es. Thearameterization of these variables, however, changed seasonally,s generally consistent with a two-stage process-based modelingpproach (e.g. the two-stage evaporative model of Ritchie, 1972;itchie et al., 2009). Wind speed influenced Es mainly during therying season (spring), but its effect was variable throughout theay and between days, similar to results reported by Gentine et al.2007). Meaningful estimates of Es are dependent on the divisionnto seasons, highlighting the importance of the large seasonal vari-bility, which leads to a shift in the dominant environmental driversetween seasons.

The results presented in Fig. 6 clearly demonstrate that Es isontrolled by soil water content in the topsoil, �5, whereas Tt is con-rolled by soil water content in the root zone (10–20 cm depth atur site, Maseyk et al., 2008; Grunzweig et al., 2009). While our siteostly lacks the grass component, the above process resembles thealkers’ two layers hypothesis for grass-tree savanna coexistence,

here grasses use the topsoil moisture and trees use the subsoiloisture (Archer, 1971; Scholes and Archer, 1997). Our research

hows that this depth-separation further affects the sensitivityf the different ecosystem components to inter-annual climatic

est Meteorology 157 (2012) 77– 85 83

variability in precipitation. The results presented in Fig. 7 demon-strate that while soil moisture in the topsoil remains similar, even inyears with different precipitation regimes, soil moisture in the sub-soil root zone varies markedly between years, and this differenceincreases with depth. Furthermore, when large storms and deepinfiltration did occur, we found evidence for soil moisture carry-over (‘legacy effect’) between hydrological years, a process that isoften neglected in modeling efforts, either for the lack of such evi-dence or because it allows modeling each season independentlywith greater computational efficiency.

A four-year period of study is insufficient to produce robustconclusions on what influences interannual variations in ET par-titioning. But it does provide some preliminary indications ofeffects that warrant further research. For example, 2003/2004 and2005/2006 were relatively dry hydrological years with compara-ble amounts of precipitation. But the two years differed in theirtemporal precipitation patterns, such that P30/P (the fraction ofprecipitation in storms >30 mm) was 0.56 for 2003/2004, but only0.43 for 2005/2006 (Table 1). As expected, the higher storm inten-sity of the first year generated high soil moisture content in theroot zone and, in turn, higher Tt/ET (0.57 vs. 0.49). Another exam-ple is in comparing 2003/2004 to 2006/2007. The first was a dryyear (P = 231 mm) while the second was a wet year (P = 308 mm),but both had comparable P30/P values: 0.56 and 0.55, respectively,and the dry year had even higher Tt/ET (0.57 vs. 0.44), likely dueto deeper infiltration depth during that year. The importance ofprecipitation temporal dynamics and not merely total precipita-tion amounts, has been previously shown (Pitt and Heady, 1978;Nordbotten et al., 2007; Miranda et al., 2001), but was limited toits effect on grasses. Our results show, that P30/P was clearly moreimportant than P itself in influencing tree transpiration. Below, weexplain how in this ecosystem precipitation patterns effect deepinfiltration and tree water availability.

Soil moisture distribution in the time-depth dimension isaffected by the interactions between precipitation pattern and thephysical nature of the soil. Soil characteristics, which are constantover the time scale of this study, can help translate temporal pre-cipitation patterns to predicted patterns of soil moisture content,and ultimately to timing of peak Tt. Differences in soil character-istics, for example, explain the differences in the results obtainedby Cavanaugh et al. (2010), who reported that transpiration corre-lated with soil moisture at depths of 37.5 and 75 cm in two creosotebush sites with sandy loam, high gravel content and better infiltra-tion compared to the Yatir site (which has a loess with clay-loamtexture). In a previous study (Raz-Yaseef et al., 2010b), we calcu-lated soil water holding capacities for each layer, and found thatdue to the clayish nature of the soils at the research site, stormssmaller than 30 mm cannot penetrate below the upper 5 cm layer.Accordingly, we used a 30 mm storm size threshold to characterizeannual precipitation patterns at our research site. Other sites arelikely to differ according to interactions between the depth defin-ing the topsoil, initiating mainly soil evaporation and the deeperroot zone, providing plant water uptake, and between the stormsize needed to overcome this boundary.

Most precipitation events (81% of storms) at our research sitewere smaller than this 30 mm threshold. Therefore, while �5 iscontinuously high during the wet seasons in all years, �15 isdictated by the temporal pattern of the precipitation and the occur-rence of larger precipitation events. These more rare event oftenaccount for most of the interannual variability in total precipita-tion (Schwinning and Sala, 2004), and can explain the observedlarger interannual variability in Tt compared to Es (NSTD of 16 vs.

8%). In our case, we found that the partial contribution of largestorms was not always correlated to total annual precipitation, andtherefore years with similar precipitation amounts experienced dif-ferent water flux dynamic. This further emphasizes the control of

8 nd For

tw

isdRrowa2pR

cte22tetidltlptflflioaVbttopm

5

oNptcs

bssittpw

A

i

4 N. Raz-Yaseef et al. / Agricultural a

he temporal precipitation pattern on ecosystem functioning in thisater-limited system.

In this semi-arid ecosystem, P30/P is probably the most signif-cant component influencing forest productivity and ultimately,urvival. It is not surprising, therefore, that when incorporatingata from the CO2 flux exchange at our site (Maseyk et al., 2008;otenberg and Yakir, 2010) to estimate water-use efficiency (theatio of net CO2 uptake to ET), a tight linear correlation wasbserved for the research period with P30/P (R2 = 0.98), but notith P (R2 = 0.63). Similar conclusions have been recently made for

rid grassland using rain manipulation experiment (Thomey et al.,011), showing these environments will benefit from extreme rainrecipitation events (as previously suggested by Knapp et al., 2008;az-Yaseef et al., 2010a).

Summers are dry and hot in semi-arid regions, presenting ahallenge to forest survival. Low tree transpiration fluxes duringhe summer are likely influenced by residual soil moisture lev-ls in the deep soil horizons or bedrock crevices (Schwinning,010), soil water storage following wet years (Raz-Yaseef et al.,010b), hydraulic resistance (Klein et al., 2011), and possibly mois-ure redistribution, as suggested for ponderosa pine (Fernandezt al., 2008; Warren et al., 2008). While it is possible to explainhe low levels of Tt that persist under dry summer conditions, its more difficult to explain the maintenance of Es, however low,uring the period in which soil moisture approaches hygroscopic

evels. Our results clearly indicated soil water adsorption duringhe dry season, and although small, these fluxes were not neg-igible when related to the residual Es flux during the summereriod. We estimated that re-evaporation from night-time adsorp-ion can account for up to 50% of the daily residual summer Es

uxes (adsorption of −0.05 mm day−1, while typical summer Es

uxes were 0.10 mm day−1). The handful of research efforts focus-ng on this process have measured soil adsorption within the rangef 0.10–1.44 mm day−1, varying mainly as a result of soil texturend canopy cover (Jacobs et al., 2000; Ninari and Berliner, 2002;erhoef et al., 2006). Whereas these studies were conducted atare-soil sites, the lower values measured at Yatir can be attributedo the reported negative influence of trees and shrubs on adsorp-ion amounts (Kosmas et al., 2001; Ramírez et al., 2007). The effectf adsorption on tree water uptake is likely to be negligible, but thisrocess may provide some protection by decelerating the upwardovement of deeper soil water.

. Conclusions

Evapotranspiration in vegetated ecosystems is often treated as ane-source component, mainly due to methodological difficulties.evertheless, partitioning ET is critical to better understanding therocesses underlying variations in this flux. Indeed, we showed thathe contribution of tree transpiration, soil evaporation and inter-epted precipitation change on diurnal, seasonal and interannualcales.

Temporal precipitation patterns and the proportional contri-ution of large rain events to total precipitation (P30/P) appear totrongly influence the partitioning of ET, with increasing tree tran-piration and ecosystem water-use efficiency being associated withncreasing P30/P. This ratio may therefore be at least as important asotal precipitation for the survival and productivity of forest ecosys-ems in dry environments. This is significant in light of persistentredictions of drying in the entire Mediterranean region and else-here, which is also often linked to increasing storm intensities.

cknowledgements

The long-term operation of the Yatir Forest Research Field Sites supported by the Cathy Wills and Robert Lewis Program in

est Meteorology 157 (2012) 77– 85

Environmental Science. Financial support from the JNF, KKL, IsraelMinistry of Agriculture and GLOWA-JR (Israel–Germany ministriesof Science) is gratefully acknowledged. We thank Dr. M. Sprintsinand D. Elmowitz for help with sap flow measurements, the entireYatir team for technical support, and the local KKL personnel fortheir cooperation.

References

Allen, R.G., Pereira, L.S., Raes, D., Smith, M.L., 1998. Crop Evapotranspiration – Guide-lines for Computing Crop Water Requirements. FAO, Rome.

Aubinet, M., Grelle, A., Ibrom, A., Rannik, Ü., Moncrieff, J., Foken, T., Kowalski, A.S.,Martin, P.H., Berbigier, P., Bernhofer, C., Clement, R., Elbers, J., Granier, A., Grün-wald, T., Morgenstern, K., Pilegaard, K., Rebmann, C., Snijders, W., Valentini,R., Vesala, T., 2000. Estimates of the annual net carbon and water exchangeof forests: the EUROFLUX methodology. Advances in Ecological Research 30,113–175.

Baldocchi, D.D., Law, B.E., Anthoni, P.M., 2000. On measuring and modeling energyfluxes above the floor of a homogeneous and heterogeneous conifer forest. Agri-cultural and Forest Meteorology 102, 187–206.

Cavanaugh, M.L., Kurc, S.A., Scott, R.L., 2010. Evapotranspiration partitioning in semi-arid shrublandecosystems: a two-site evaluation of soil moisture control ontranspiration. Ecohydrology, doi:10.1002/eco.157.

Cohen, Y., 1994. Thermoelectric methods for measurement of sap flow in plants.In: Stanhill, G. (Ed.), Advances in Bioclimatology, vol. 3. Springer, Heidelberg,Germany, pp. 63–88.

Cohen, Y., Cohen, S., Cantuarias-Aviles, T., Schiller, G., 2008. Variations in the radialgradient of sap velocity in trunks of forest and fruit trees. Plant and Soil 305,49–59.

Fernandez, M.E., Gyenge, J., Licata, J., Schlichter, T., Bond, B.J., 2008. Belowgroundinteractions for water between trees and grasses in a temperate semiarid agro-forestry system. Agroforestry Systems 74, 185–197.

Gentine, P., Entekhabi, D., Chehbouni, A., Boulet, G., Duchemin, G., 2007. Analysis ofevaporative fraction diurnal behavior. Agricultural and Forest Meteorology 143,13–29.

Gockede, M., Foken, T., Aubinet, M., Aurela, M., Banza, J., Bernhofer, C., Bonnefond,J.M., Brunet, Y., Carrara, A., Clement, R., Dellwik, E., Elbers, J., Eugster, W., Fuhrer,J., Granier, A., Grunwald, T., Heinesch, B., Janssens, I.A., Knohl, A., Koeble, R., Lau-rila, T., Longdoz, B., Manca, G., Marek, M., Markkanen, T., Mateus, J., Matteucci,G., Mauder, M., Migliavacca, M., Minerbi, S., Moncrieff, J., Montagnani, L., Moors,E., Ourcival, J.-M., Papale, D., Pereira, J., Pilegaard, K., Pita, G., Rambal, S., Reb-mann, C., Rodrigues, A., Rotenberg, E., Sanz, M.J., Sedlak, P., Seufert, G., Siebicke,L., Soussana, J.F., Valentini, R., Vesala, T., Verbeeck, H., Yakir, D., 2008. Qualitycontrol of CarboEurope flux data – Part I: Footprint analyses to evaluate sites inforest ecosystems. Biogeosciences 5, 433–450.

Granier, A., 1987. Evaluation of transpiration in a Douglas fir stand by means of sapflow measurements. Tree Physiology 3, 309–320.

Grünzweig, J.M., Hemming, D., Maseyk, K., Lin, T., Rotenberg, E., Raz-Yaseef, N.,Faloon, P.D., Yakir, D., 2009. Water limitation to soil CO2 efflux in a pine forest atthe semiarid timberline. Journal of Geophysical Research – Biogeosciences 114,G03008.

Heisler-White, J.L., Knapp, A.K., Kelly, E.F., 2008. Increasing precipitation eventsize increases aboveground net primary productivity in a semi-arid grassland.Oecologia 158, 129–140.

Huxman, T.E., Wilcox, B.P., Breshears, D.D., Scott, R.L., Snyder, K.A., Small, E.E., Hul-tine, K., Pockman, W.T., Jackson, R.B., 2005. Ecohydrological implications ofwoody plant encroachment. Ecology 86, 308–318.

IPCC, 2007. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt,K.B., Tignor, M., Miller, H.L. (Eds.), Climate Change 2007: The Physical ScienceBasis. Contribution of Working Group 1 to the Fourth Assessment Report of theIntergovernmental Panel on Climate Change. Cambridge University Press, UK.

Jacobs, A.F.G., Heusinkveld, B.G., Berkowicz, S.M., 2000. Dew measurements alonga longitudinal sand dune transect, Negev Desert, Israel. International Journal ofBiometeorology 43 (4), 184–190.

Klein, T., Cohen, S., Yakir, D., 2011. Hydraulic adjustments underlying drought resis-tance of Pinushalepensis. Tree Physiology 31 (6), 637–648.

Knapp, A., Beier, C., Briske, D., Classen, A., Luo, Y., Reichstein, M., Smith, M., Smith,S., Bell, J., Fay, P., Heisler, J., Leavitt, S., Sherry, R., Smith, B., Weng, E., 2008.Consequences of more extreme precipitation regimes for terrestrial ecosystems.BioScience 58, 811–821.

Kochendorfer, J.P., Ramírez, J.A., 2008. Ecohydrologic controls on vegetation densityand evapotranspiration partitioning across the climatic gradients of the centralUnited States. Hydrology and Earth System Sciences Discussions 5, 648–700.

Kosmas, C., Marathianou, M., Gerontidis, S., Detsis, V., Tsara, M., Poesen, J., 2001.Parameters affecting water vapor adsorption by the soil under semi-arid climaticconditions. Agricultural Water Management 48, 61–78.

Kostner, B., 2001. Evaporation and transpiration from forests in Central Europe – rel-

evance of patch-level studies for spatial scaling. Meteorology and AtmosphericPhysics 76, 69–82.

Kurc, S.A., Small, E.E., 2004. Dynamics of evapotranspiration in semiarid grasslandand shrubland ecosystems during the summer monsoon season, central NewMexico. Water Resources Research 40 (9), WO9305.

nd For

K

L

L

L

L

M

M

M

M

N

N

N

N

P

P

R

R

R

R

R

R

N. Raz-Yaseef et al. / Agricultural a

urpius, M.R., Panek, J.A., Nikolov, N.T., McKay, M., Goldstein, A.H., 2003. Partitioningof water flux in a Sierra Nevada ponderosa pine plantation. Agricultural andForest Meteorology 117, 173–192.

aio, F., Porporato, A., Ridolfi, L., Rodriguez-Iturbe, I., 2001. Plants in water-controlledecosystems: active role in hydrologic processes and response to water stress: II.Probabilistic soil moisture dynamics. Advances in Water Resources 24, 707–723.

apitan, R.L., Parton, W.J., 1996. Seasonal variabilities in the distribution of themi-croclimatic factors and evapotranspiration in a shortgrass steppe. Agriculturaland Forest Meteorology 79, 113–130.

awrence, D.M., Thronton, P.E., Oleson, K.W., Bonan, G., 2007. The partitioning ofevapotranspiration into transpiration, soil evaporation, and canopy evaporationin a GCM: impacts on land-atmosphere interaction. Journal of Hydrometeorol-ogy 8, 862–880.

oik, M.E., Breshears, D.D., Lauenroth, W.K., Belnap, J., 2004. A multi-scale perspec-tive of water pulses in drylandecosystems:climatology and ecohydrology of thewestern USA. Oecologia 141, 269–281.

aseyk, K., Grünzweig, J., Rotenberg, E., Yakir, D., 2008. Respiration acclimationcontributes to high carbon-use efficiency in a seasonally dry pine forest. GlobalChange Biology 14, 1553–1567.

iranda, J.D., Armas, C., Padilla, F.M., Pugnaire, F.I., 2001. Climatic change and rain-fall patterns: effects on semi-arid plant communities of the Iberian Southeast.Journal of Arid Environments 75, 1302–1309.

itchell, P.J., Veneklaas, E., Lambers, H., Burgess, S.S.O., 2009. Partitioning of evap-otranspiration in a semi-arid eucalypt woodland in south-western Australia.Agriculture and Forest Meteorology 149, 25–37.

oran, M.S., Scott, R.L., Keefer, T.O., Emmerich, W.E., Hernandez, M., Nearing, G.S.,Paige, G.B., Cosh, M.H., O’Neill, P.E., 2009. Partitioning evapotranspiration insemiarid grassland and shrubland ecosystems using time series of soil surfacetemperature. Agricultural and Forest Meteorology 149, 59–72.

ewman, B.D., Breshears, D.D., Gard, M.O., 2010. Evapotranspiration Partitioning ina Semiarid Woodland: ecohydrologic Heterogeneity and Connectivity of Vege-tation on Patches. Vadose Zone Journal 9, 561–572.

inari, N., Berliner, P.R., 2002. The role of dew in the water and heat balance of bareloess soil in the Negev Desert: quantifying the actual dew deposition on the soilsurface. Atmospheric Research 64, 323–334.

ordbotten, J.M., Rodriguez-Iturbe, I., Celia, M.A., 2007. Stochastic coupling of rain-fall and biomass dynamics. Water Resources Research 43, W01408.

oy-Meir, I., 1973. Desert ecosystems: environment and producers. Annual Reviewof Ecology and Systematics 4, 25–51.

aruelo, J.M., Sala, O.E., Beltran, A.B., 2000. Long-term dynamics of water and carbonin semi-arid ecosystems: a gradient analysis in the Patagonian steppe. PlantEcology 150, 133–143.

itt, M.D., Heady, H.F., 1978. Responses of annual vegetation to temperature andrainfall patterns in Northern California. Ecology 59 (2), 336–350.

amírez, D.A., Bellot, J., Domingo, F., Blasco, A., 2007. Can water responses in Stipate-nacissima L. during the summer season be promoted by non-rainfall water gainsin soil? Plant Soil 291, 67–79.

az-Yaseef, N., Rotenberg, E., Yakir, D., 2010a. Effects of spatial variations in soilevaporation caused by tree shading on water flux partitioning in a semi-aridpine forest. Agricultural and Forest Meteorology 150, 454–462.

az-Yaseef, N., Yakir, D., Rotenberg, E., Schiller, G., Cohen, S., 2010b. Ecohydrologyof a semi-arid forest: partitioning among water balance components and itsimplications for predicted precipitation changes. Ecohydrology 3 (2), 143–154.

eynolds, J.F., Kemp, P.R., Tenhunen, J.D., 2000. Effects of long-term rainfall variabil-ity on evapotranspiration and soil water distribution in the Chihuahuan Desert:a modeling analysis. Plant Ecology 150, 145–159.

itchie, J.T., 1972. Model for predicting evaporation from a row crop with incompletecover. Water Resources Research 8, 1204–1213.

itchie, J.T., Porter, C.H., Judge, J., Jones, J.W., Suleiman, A.A., 2009. Extension of anexisting model for soil water evaporation and redistribution under high watercontent conditions. SSSAJ 73 (3), 792–801.

est Meteorology 157 (2012) 77– 85 85

Rotenberg, E., Yakir, D., 2011. Distinct patterns of changes in surface energy budgetassociated with forestation in the semiarid region. Global Change Biology 17 (4),1536–1548.

Rotenberg, R., Yakir, D., 2010. Contribution of semi-arid forests to the climate system.Science 327, 451–454.

Sala, O.E., Lauenroth, W.K., 1982. Small rainfall events: an ecological role in semiaridregions. Oecologia 53, 301–304.

Schiller, G., Cohen, Y., 1998. Water balance of Pinushalepensis Mill. afforestation inan arid region. Forest Ecology and Management 105, 121–128.

Scholes, R.J., Archer, S.R., 1997. Tree-grass interactions in savannas. Annual Reviewof Ecological Systems 28, 517–544.

Schwinning, S., 2010. The ecohydrology of roots in rocks. Ecohydrology 3 (2),238–245.

Schwinning, S., Sala, O.E., 2004. Hierarchy of responses to resource pulses in aridand semi-arid ecosystems. Oecologia 141, 211–220.

Scott, R.L., Shuttleworth, W.J., Keefer, T.O., Warrick, A.W., 2000. Modeling multi-year observations of soil moisture recharge in the semiarid American southwest.Water Resources Research 36, 2233–2247.

Shachnovich, Y., Berliner, P.R., Bar, P., 2008. Rainfall interception and spatial distribu-tion of throughfall in a pine forest planted in an arid zone. Journal of Hydrology349, 168–177.

Sharon, D., 1972. The spottiness of rainfall in a desert area. Journal of Hydrology 17,161–175.

Sprintsin, M., Cohen, S., Maseyk, K., Rotenberg, E., Grünsweig, J., Karnieli, A.,Berliner, P., Yakir, D., 2011. Long term and seasonal courses of leaf area indexin a semi-arid forest plantation. Agricultural and Forest Meteorology 151,565–574.

Staudt, K., Serafimovich, A., Siebicke, L., Pyles, R.D., Falge, E., 2011. Vertical struc-ture of evapotranspiration at a forest site (a case study). Agricultural and ForestMeteorology 151, 709–729.

Thomey, L.L., Collins, S.L., Vargas, R., Johnson, J.E., Brown, R.F., Natvig, D.O.,Friggens, M.T., 2011. Effect of precipitation variability on net primary produc-tion and soil respiration in a Chihuahuan Desert. Global Change Biology 17,1505–1515.

Tietjen, B., Zehe, E., Jeltsch, F., 2009. Simulating plant water availability in dry landsunder climate change: a generic model of two soil layers. Water ResourcesResearch 45, WO1418.

Verhoef, A., Diaz-Espejo, A., Knight, J.R., Villagarcia, L., Fernandez, J.E., 2006. Adsorp-tion of water vapor by bare soil in an olive grove in southern Spain. Journal ofHydrometeorology 7 (5), 1011–1027.

Wang, L., Caylor, K.K., Villegas, J.C., Barron-Gafford, G.A., Breshears, D.D., Huxman,T.E., 2010. Partitioning evapotranspiration across gradients of woody plantcover: assessment of a stable isotope technique. Geophysical Research Letters37, L09401.

Warren, J.M., Brooks, J.R., Meinzer, F.C., Eberhart, J.L., 2008. Hydraulic redistributionof water from Pinus ponderosa trees to seedlings: evidence for an ectomycor-rhizal pathway. New Phytologist 178, 382–394.

Williams, D.G., Cable, W., Hultine, K., Hoedjes, J.C.B., Yepez, E.A., Simonneaux, V.,Er-Raki, S., Boulet, G., de Bruin, H.A.R., Chehbouni, A., Hartogensis, O.K., Tim-ouk, F., 2004. Evapotranspiration components determined by stable isotope, sapflow and eddy covariance techniques. Agricultural and Forest Meteorology 125,241–258.

Yepez, E.A., Huxman, T.E., Ignaceb, D.D., English, N.B., Weltzin, J.F., Castellanos, A.E.,Williams, D.G., 2005. Dynamics of transpiration and evaporation following amoisture pulse in semiarid grassland: a chamber-based isotope method forpartitioning flux components. Agricultural and Forest Meteorology 132 (3–4),

359–376.

Zhongmin, H., Guirui, Y., Yanlian, Z., Xiaomin, S., Yingnian, L., Peili, S., Yanfen, W., Xia,S., Zemei, Z., Li, Z., Shenggong, L., 2009. Partitioning of evapotranspiration andits controls in four grassland ecosystems: application of a two-source model.Agricultural and Forest Meteorology 149, 1410–1420.