Ecological Complexity 31 (2017) 96–103

Original Research Article

Geodiversity increases ecosystem durability to prolonged droughts

Hezi Yizhaqa,b, Ilan Stavia, Moshe Shachakc, Golan Belb,*aDead Sea & Arava Science Center, Yotvata 88820, IsraelbDepartment of Solar Energy and Environmental Physics, Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Sede Boqer Campus8499000, IsraelcMitrani Department of Desert Ecology, Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Sede Boqer Campus 8499000, Israel

A R T I C L E I N F O

Article history:Received 13 April 2017Received in revised form 29 May 2017Accepted 1 June 2017Available online xxx

A B S T R A C T

Interactions between biotic and abiotic factors dictate the response of ecosystems to varying conditionsand disturbances. The importance of the relationship between these factors is demonstrated in theextensively studied interactions between water-limited vegetation and its ecosystem’s physicalcomponents. Landscape geodiversity is often neglected in studies of vegetation dynamics and responseto drought. Here, we combine field studies and mathematical modeling to elucidate the effects ofgeodiversity on shrub mortality following drought. In Israel’s semi-arid northwestern Negev Desert, wefound that homogeneous hillslopes, with little or no stoniness, experienced considerable shrub mortalityfollowing droughts, while neighboring slopes with higher stoniness showed little or no mortality. Amathematical model describing the dynamics of water-limited vegetation and accounting for landscapegeodiversity predicted similar responses. The measurements and the model suggest that geodiversityincreases the amount of water available for the shrubs, thereby increasing their durability. Future climatepredictions of reduced precipitation and increased drought frequency in many regions make studies ofecosystem responses to water deficiency timely. Our findings suggest that future studies should accountfor landscape geodiversity in order to explain local differences in vegetation mortality and to betterassess the possible impacts of climate fluctuations on ecosystem dynamics. In particular, geodiversity hasa great effect on regime shifts and their nature.

© 2017 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Ecological Complexity

journa l home page : www.e l sev ier .com/ loca te /ecocom

1. Introduction

An ecosystem’s responses to extreme climate events aredependent on the interactions between its abiotic and bioticcomponents. The spatial and temporal distribution of the abioticconditions (e.g., geology, topography, soil type and climate)dictates the present biotic organization and the future reorganiza-tion (Ackerly et al., 2010; Anderson and Ferree, 2010; Beier andBrost, 2010). The relationships between abiotic diversity ornature's stage, extreme climatic conditions and biotic reorganiza-tion, on both the population and community levels, are unclear dueto the complex interactions between these three variables (Lawleret al., 2015).

Geology, soil, and topography, as components of geodiversity,all affect the availability of resources for vegetation, modifying

* Corresponding author.E-mail addresses: yiyeh@bgu.ac.il (H. Yizhaq), istavi@adssc.org (I. Stavi),

shachak@bgu.ac.il (M. Shachak), bel@bgu.ac.il (G. Bel).URL: http://mailto:bel@bgu.ac.il (G. Bel)

http://dx.doi.org/10.1016/j.ecocom.2017.06.0021476-945X/© 2017 Elsevier B.V. All rights reserved.

biotic organization and reorganization. Geology is an importantdeterminant of the formation (pedogenesis) and diversity of soils(pedodiversity), through its influence on the chemical and physicalproperties of soil and water flow. These physical properties affectthe organization of plants and microbes that regulate ecosystemprimary production and decomposition (Kruckeberg, 2002).Therefore, pedodiversity is a substantial component of studiesrelated to geodiversity-biodiversity relationships (Ibáñez et al.,2012; Mcbratney, 1992; Phillips 1999; Petersen et al., 2010).

Drought is a primary driver that shifts these relationships. Forinstance, droughts are known to decrease primary productivity(Ciais et al., 2005), drive the reorganization of communities andecosystem function (Engelbrecht et al., 2007), affect water andnitrate dioxide fluxes and reduce plants’ water-use efficiency(Reichstein et al., 2002). Studies have shown that droughts haveboth episodic and continuous effects on woody vegetationmortality rates that induce ecosystem state change (Bowers,2005). In addition to plant mortality, vegetation responses todrought-induced water deficiency, on the landscape level, includemodification of plant spatial patterns (Tongway et al., 2001;Rietkerk et al., 2008; Borgogno et al., 2009).

H. Yizhaq et al. / Ecological Complexity 31 (2017) 96–103 97

Many mathematical models have been developed to explain therelationships between water flow and vegetation pattern forma-tion (Klausmeier 1999; Lejeune et al., 2002; Rietkerk et al., 2002;Gilad et al., 2004). These models are based on short rangefacilitation of vegetation and longer range competition for water,which is the main growth limiting factor. The different modelscapture different facilitation mechanisms. These mechanismsinclude the infiltration contrast—the higher infiltration rate ofsurface water in vegetated spots due to the effect of vegetation onthe soil crust (Rietkerk et al., 2002; Gilad et al., 2004; Getzin et al.,2016); the water uptake feedback—denser vegetation’s higherwater uptake rate due to more developed root systems and higherwater extraction efficiency (Klausmeier 1999; Zelnik et al., 2013;Kinast et al., 2014; Sherratt 2015); and the root augmentationfeedback—the higher growth rate of denser vegetation due to itsspatially extended root system, which allows it to extract waterfrom larger soil volumes (Gilad et al., 2004).

All the models mentioned above describe the dynamics overspatial scales that allow continuous description of the vegetationin terms of biomass density (usually on the scale of tens ofcentimeters to tens of meters) and temporal scales that are largerthan the typical time scales for vegetation growth (usually monthsto years). The models predict five basic vegetation pattern statesalong the rainfall gradient (from wet to dry): uniform vegetationcover, a gap pattern (Zelnik et al., 2015; Getzin et al., 2016), alabyrinthine pattern, a spotted pattern, and bare soil. These modelsassume a homogeneous landscape and the spatial heterogeneity ofvegetation results from its interaction with water. However, mostecosystems are not homogeneous, and heterogeneity exists due tosoil texture, stoniness, micro-topography, nutrient distributionand other factors. Our specific objective was to integrate field andmodeling studies in order to elucidate the effects of abioticdiversity on the biotic responses to drought.

The field study was carried out in a Long Term EcologicalResearch (LTER) site where, presumably, severe droughts(during 2008–2009) caused a state change over the lastdecades (Sher et al., 2012; Paz-Kagan et al., 2014; Hoffmanet al., 2016). Most notably, the cover by mature shrubs droppedsubstantially due to high mortality rates of the dominant shrubspecies, Noaea mucronata. The shrub mortality in the studyregion (Fig. 1S and Fig. 2S of the Supplementary materials) hasbeen clearly observed as white patches caused by theaccumulation of snail shells on the ground. Before the sequenceof drought years, the snails fed on and lived in the shrubs. Themortality of the dominant shrub species was estimated as 90%(Sher et al., 2012) of the total population. Increased shrub

Fig. 1. Average annual precipitation in the Sayeret Shaked Park LTER. The blue bars show

average of the previous three years. The average annual precipitation is 164.8 mm/year asignificantly low precipitation rates. (For interpretation of the references to colour in t

mortality in additional sites across the region supports thehypothesis of a climate-driven event (DeMalach et al., 2014;Shoshany and Karnibad 2015).

The abovementioned studies neglected spatial heterogeneityand referred to the mortality across the entire site. However, acareful inspection of the hillslopes in the site reveals that the massshrub mortality has not been spatially uniform and, in fact,strongly depends on the hillslopes’ geodiversity. Two differenttypes of hillslopes (with regard to the surface cover) wereidentified; one with no stones or rock fragments (homogeneoushillslopes) and the second with rock fragment cover (heteroge-neous hillslopes) ranging from �10 to 30%. In the homogeneoushillslopes, the shrub mortality was massive (>90%), whereas in theheterogeneous hillslopes, it was quite small (�7%).

Recently, theoretical studies have suggested that landscapeheterogeneity can increase vegetation survivability and makeregime shifts more gradual (Yizhaq et al., 2014; Yizhaq and Bel2016). However, these theoretical studies have not suggested themechanism by which the heterogeneity influences vegetationdynamics. Here, we attempt to identify this mechanism affectingregime shifts in vegetation ecosystems. To achieve this goal, wecombined field measurements and theoretical studies of a well-explored mathematical model.

2. Methods

2.1. Description of the study site

The field study was implemented in the LTER station of theSayeret Shaked Park, in Israel’s semi-arid northwestern NegevDesert (31�270 N, 34�650 E, 190 ma.s.l., shown in Figs. 1S and 2Sof the Supplementary materials). The LTER site is located in thetransition zone between the semi-arid Negev region and the drysub-humid Mediterranean region of central Israel. The climateconditions in the region are typical semi-arid Mediterranean,with cool wet winters and hot dry summers (>six month dryperiod). Across the region, a rainfall gradient exists from west toeast and from north to south, but inter-annual variability candrive the aridity isohyet farther north in extremely dry years.The LTER station spans an area of �20 ha, which has beensurrounded by a fence since the late 1990s to prevent livestockaccess. The region’s lithology comprised of chalk of the Eoceneand Plio-Pleistocene eolianite. The landform is dominated byrolling hills, with hillslope inclines ranging between 3� and 6�.The soil in the LTER site is classified as loessial Calcic Xerosol,with a sandy loam to loamy sand texture (Singer, 2007).

the measured annual accumulated precipitation and the black bars show the annualnd is marked by the horizontal black line. The letters A, B and C denote years withhis figure legend, the reader is referred to the web version of this article.)

98 H. Yizhaq et al. / Ecological Complexity 31 (2017) 96–103

Long-term averages of the daily temperature are 26 �C for thehottest month and 12 �C for the coldest month. The meancumulative annual precipitation for the period of 1992–2016 is�165 mm/year and the standard deviation is 58 mm/year (Fig. 1).The meteorological data was obtained from a meteorologicalstation located in the LTER site.

2.2. The study design and hillslope survey

Field work for this study started in July 2015, and lasted for anentire year. The study site survey revealed hillslopes with no rockfragment cover at all (Fig. 2a and c) and hillslopes with substantial(�30%) rock fragment cover (Fig. 2b and d). Three hillslopes of eachtype were selected for the study, with a distance of at least 100 mbetween two adjacent hillslopes. Due to the biotic and abioticfeatures of the study site and in order to well represent the studyregion, it was decided to study the hillslope type effect in 400 m2

(20 m � 20 m) plots in the backslope geomorphic unit of each ofthese hillslopes. For negating aspect- and incline-related effects, allplots were delineated at relatively similar azimuth (310 � 29�) andslope (4.5 � 0.5�) values.

Hillslope mapping was conducted to quantify the surface coverfraction of the different components. This was implemented usingthe line-intercept method (Tongway and Hindley, 2004), in whichthe type of cover was recorded at 25 cm intervals along a 20 mtransect, with three transects per plot (yielding a total of 240reading points per plot). Following the preliminary observationsacross the study region, the types of surface cover included: liveshrub (and its species), dead shrub (and its species, as determinedby the canopy's morphological features), herbaceous vegetation,partially embedded rock fragment, loosely resting rock fragment,exposed soil/biological crust, geophyte, rock exposure, and other(including all other types of cover, such faunal perturbation, etc.).

2.3. Soil sampling, on-site monitoring, and measurements

Soil sampling of the upper layer (0–5 cm) was conducted at thepeak of 2015’s dry season (August), when the soil moisture contentwas at the hygroscopic level. The 0–5 cm depth layer was chosendue to a failure to dig deeper because of the shallow soil layer in thethree heterogeneous hillslopes. In each plot, 10 randomly selectedpatches of N. mucronata and 10 adjacent inter-shrub spots were

Fig. 2. The study area in the Sayeret Shaked Park LTER station. Panels a and b show hexcavation in the plots (c: homogeneous and d: heterogeneous, the green scale length is 3in this figure legend, the reader is referred to the web version of this article.)

sampled. Then, in order to assess the soil depth, we used an augerfor drilling down to the underlying bedrock layer in five otherrandomly selected spots per plot. In order to limit the disturbance,the drilling depth was limited to 100 cm.

The soil stoniness (rock fragment content) was determined inthree randomly selected spots per plot. This was conducted byexcavating 8-l (20 cm � 20cm � 20 cm cube) of soil in each spot. Inlocations with shallower soil depths, the dimensions of excavatedvolume were modified accordingly in order to obtain the sametotal volume.

From December 2015 to May 2016 (covering the rainy season),the soil volumetric moisture content was measured once a month.The measurements were taken using a time-domain reflectometer(TDR: with 7.6-cm pins: Spectrum Technologies©). In each plot, fivedifferent N. mucronata patches and five inter-shrub spots weresampled.

2.4. Laboratory analyses

Gravimetric moisture content was determined for the summer-obtained set of samples by putting them in a drying oven set to105 �C for 24 h. In addition, for the calibration of the winter-obtained TDR readings, representative soil cores (of 100 ml volumeeach) were taken and placed in the drying oven to determinevolumetric moisture content. The bags of the extracted mineralmaterial were sieved through a 2 mm sieve, where all finermaterial was excluded from the sample. Then, the remainingmaterial was gently ground to break down clods and macro-aggregates into <2 mm grains. The ground material was thensieved again, using the same sieve in order to exclude all finematerial from the sample. Lastly, the volume of the remaining rockfragments was measured.

2.5. Statistical analyses

An analysis of variance was conducted with the GLM procedureof SAS (SAS Institute, 1990) to study the effect of hillslope type(heterogeneous vs. homogeneous) and microhabitat (shrubbypatch vs. inter-shrub space) on the soil moisture. Factors used inthe model were hillslope type (1 df), plot within hillslope type (3df; error term for plot), microhabitat (1 df), and the interaction

omogeneous and heterogeneous plots, respectively. Panels c and d show the soil0cm) in order to assess their stoniness. (For interpretation of the references to colour

H. Yizhaq et al. / Ecological Complexity 31 (2017) 96–103 99

hillslope type � microhabitat (1 df). Separation of means wasimplemented by Tukey’s HSD at a significance level of 0.05.

2.6. Mathematical model of vegetation dynamics

We adopted the well-studied model introduced by Rietkerket al. (2002) to describe the dynamics of water-limited vegetation(Rietkerk et al., 2002, 2004; Kéfi et al., 2010). This model describesthe dynamics of three variables: aboveground biomass density,surface-water density and soil-water density. The vegetationdynamics is affected by its growth rate, which depends on thesoil-water density, the natural mortality and the spatial expansiondue to clonal growth or seed dispersion. The dynamics of the soilwater is affected by the vegetation water uptake, evaporation anddrainage, and the infiltration of surface water, as well as by its owndiffusivity. The surface-water dynamics is affected by precipita-tion, infiltration, evaporation and losses due to surface waterrunoff (in the simulations, we assumed a plain topography) andsurface water diffusion. The model captures the infiltrationfeedback (Eldridge et al., 2000; Rietkerk et al., 2002; Gilad et al.,2004), namely the higher infiltration rate of surface water invegetated patches due to cracking of the soil and the surface crustby the vegetation roots (Gilad et al., 2004). The greater soil-waterdensity increases the vegetation growth rate and drives theinstability of the uniform state to non-uniform perturbations(finite wavenumber instability or Turing instability). This feedbackresults in patterns of vegetation and soil-water densities that are“in-phase.” Specifically, the soil-water density is maximal in spotswhere the vegetation density is maximal (Zelnik et al., 2013; Kinastet al., 2014). A detailed description of the model is provided in theSupplementary material.

The effects of heterogeneity in this model were investigated byintroducing a spatially non-uniform maximal surface-waterinfiltration rate (MSWIR, denoted by a in the mathematicalmodel). This type of heterogeneity corresponds to the effect of thestoniness of the upper soil layer (Dunkerley, 1995; Bunte andPoesen, 1994). Partially embedded rock fragments increase theground surface sealing, thereby reducing the infiltration rate, whileloosely resting rock fragments hinder the water’s overland flow,thereby increasing the infiltration rate (Poesen et al., 1990; Laveeand Poesen, 1991).

The MSWIR at each grid point was drawn from a Gaussiandistribution with the same mean, and the standard deviation of thedistribution represents the degree of heterogeneity. In general, theGaussian probability density function allows negative values,

Fig. 3. Mean soil volumetric moisture content throughout the 2015–2016 wet season, adiffer at the 0.05 significance level according to Tukey’s HSD.

which are not physical. Therefore, negative values were replacedby the mean value. The result of this correction is a slightlynarrower distribution and a mean value that is slightly larger thanthe specified value. In all of the realizations that we used in thiswork, we verified that these changes were negligible. It isimportant to note that we neglected spatial correlations of theMSWIR.

The numerical details of the simulations and the modelparameters are provided in the Supplementary material. Theprecipitation rate and the standard deviation of the MSWIR varied,and they are specified for each result.

3. Results

3.1. Physical and biotic features

Meteorological data for the Park Shaked LTER showed that thereis no apparent trend in the cumulative annual precipitation overthe period of 1992–2016 (blue bars in Fig. 1). Three periods ofbelow average precipitation are apparent (marked by the arrowsand the letters A, B and C) during the years 1994 (A), 1998–2000 (B)and 2008–2013 (C). The years 1994, 1999, and 2009 constitute thelowest 12% of the precipitation distribution. Moreover, theydeviate from the mean by more than 1.5 standard deviations,implying that they are at the lowest 7% of the probability densityfunction (PDF) if one assumes a Gaussian distribution of theprecipitation. It is reasonable to assume that the shrubs areaffected by the cumulative precipitation over several years ratherthan the annual cumulative precipitation. Therefore, we calculatedthe three-year average of the annual cumulative precipitation forthe period of 1992–2016 (black bars in Fig. 1). The periods B and Cshowed apparent reductions in the cumulative precipitation overthree year or longer periods. For the three-year precipitationaverage, the years 2000 and 2010 constitute the lowest 9% of thedistribution. Moreover, they deviate from the mean by more than1.45 standard deviations, implying that they are the lowest 8% ofthe PDF if one assumes a Gaussian distribution.

In the heterogeneous hillslopes, the soil was shallow and highlystony, as opposed to the very deep soil layer and complete absenceof stoniness in the homogeneous hillslopes. The means of totalshrubby vegetation species cover (30.8 � 4.1%) and live shrub cover(29.4 � 4.0%) in the heterogeneous hillslopes were more thanfourfold and more than tenfold greater, respectively, than thesemeans in the homogeneous hillslopes (7.5 �1.1% and 2.2 � 1.8%,respectively). The mean number of shrubby vegetation species

ccording to hillslope type. Means of the same month followed by a different letter

100 H. Yizhaq et al. / Ecological Complexity 31 (2017) 96–103

detected in the plant transects was six in the heterogeneoushillslopes, as opposed to two in the homogeneous hillslopes. Inaddition, geophytes, bare soil, rock fragment cover, and faunalperturbation were detected in the heterogeneous hillslopes only.The mean herbaceous vegetation cover in the homogeneoushillslopes (92.5 �1.1%) was more than threefold greater than thatin the heterogeneous hillslopes (29.7 � 2.0%).

3.2. Hillslope type effect on soil moisture regime and vegetationproductivity

Soil moisture content was found to be considerably affected bythe hillslope type. The mean gravimetric moisture content duringthe dry season (of 2015), when the soil moisture was at thehygroscopic level, was significantly (P < 0.0001) greater (by 22%) inthe heterogeneous hillslopes (1.62%) than in the homogeneoushillslopes (1.33%). A significantly greater mean volumetric soilmoisture content in the heterogeneous hillslopes than in thehomogeneous hillslopes was recorded for December 2015 (39%greater in the heterogeneous than in the homogeneous hillslopes,P < 0.0001), January 2016 (13% greater in the heterogeneoushillslopes, P = 0.0241), and February 2016 (34% greater in theheterogeneous hillslopes, P < 0.0001), and then again, for April2016 (19% greater in the heterogeneous hillslopes, P = 0.0002) andMay 2016 (15% greater in the heterogeneous hillslopes, P = 0.001).No significant difference between the two hillslope types wasrecorded for March 2016 (see Fig. 3).

3.3. Microhabitat effect on soil moisture regime

The soil's gravimetric moisture content during the dry seasonwas significantly (P = 0.0032) affected by microhabitat type, inwhich the mean hygroscopic moisture content was 6% greater inthe shrubby patches than in the inter-shrub spaces. However, thedifference between the microhabitats within each of the hillslopetypes was not significant (P = 0.5506) for this soil feature. Asignificantly greater mean volumetric soil moisture content in theinter-shrub spaces than in the shrubby patches was recorded forDecember 2015 (14% greater in the inter-shrub spaces than in theshrubby patches, P = 0.0288), January 2016 (16% greater in theinter-shrub spaces, P = 0.0076), and February 2016 (21% greater inthe inter-shrub spaces, P < 0.0001). In March and April 2016, this

Fig. 4. The total biomass in the simulated domain versus the precipitation rate (R) for ddistribution of the vegetation for the specified precipitation rates (R is given in units of mshows the critical precipitation rate, Rc, at which the vegetation collapses versus the sdisplayed. (For interpretation of the references to colour in this figure legend, the read

trend was reversed, and the mean volumetric moisture contentwas marginally significantly (P = 0.0951 and 0.0516) greater (8%and 9%) in the shrubby patches than in the inter-shrub spaces. Nodifference between the two microhabitat types was recorded forMay 2016. The difference between the microhabitats within eachof the hillslope types was significant (P = 0.0139) only for March2016. The mean values of volumetric soil moisture content in thismonth followed the trend of: shrubby patches on the heteroge-neous hillslopes (11.0%) > either shrubby patches (10.1%) or inter-shrub domains (10.4%) on the homogeneous hillslopes > inter-shrub spaces on the heterogeneous hillslopes (8.9%).

3.4. Model results for the effect of heterogeneous surface-waterinfiltration rate on vegetation biomass

The stoniness of the soil surface affects the infiltration rate ofsurface water, resulting in a heterogeneous MSWIR. The differentlines in Fig. 4 show the simulated steady state biomass densityversus the precipitation rate for different standard deviations ofthe MSWIR PDF. The more heterogeneous the landscape, the lowerthe precipitation rate under which the vegetation survives. Theinset shows the critical precipitation rate (the precipitation rate atwhich the vegetation collapses) versus the degree of heterogeneityof the MSWIR (the standard deviation of the MSWIR). The panelsshow the spatial distribution of the biomass (green spots representhigher density) for the indicated precipitation rates (mm/day) for alandscape with a standard deviation of the MSWIR equal to0.04 day�1.

3.5. Model results for the effect of the heterogeneous surface-waterinfiltration rate on soil-water density

The heterogeneous MSWIR affects the vegetation through thesoil water. Therefore, we investigated its effect on the soil-waterdensity. In Fig. 5, we show the maximal, minimal and average soil-water density versus the precipitation rate for homogeneous(s = 0) and heterogeneous (s = 0.04/day) MSWIRs. One can see thatthe maximal and minimal soil-water densities are only weaklydependent on the precipitation rate before the vegetationcollapses. In the heterogeneous landscape, the maximal soil-waterdensity is higher and the minimal soil-water density is lower thanin homogeneous landscape. The average soil-water density shows

ifferent degrees of heterogeneity of the MSWIR. The small panels show the spatialm/day). The standard deviation of the MSWIR, s, is given in units of day�1. The insettandard deviation of the MSWIR. The red line is a linear fit whose parameters areer is referred to the web version of this article.)

Fig. 5. The average, maximal and minimal simulated soil-water density (W) versus the precipitation rate (R). The blue lines and symbols correspond to the heterogeneousMSWIR (s = 0.04 day�1) and the black lines and symbol correspond to the homogeneous MSWIR (s = 0 day�1). (For interpretation of the references to colour in this figurelegend, the reader is referred to the web version of this article.)

H. Yizhaq et al. / Ecological Complexity 31 (2017) 96–103 101

stronger dependence on the precipitation rate, which is almostidentical for the homogeneous and the heterogeneous landscapesbefore the vegetation collapses. The curves terminate at theprecipitation rate for which the vegetation collapses (this collapseoccurs at a lower precipitation rate for the heterogeneouslandscape).

3.6. Model results for the correlations between the surface-waterinfiltration rate and the biomass and soil water densities

To better understand the effects of the heterogeneous MSWIRon the soil-water and the biomass densities, we calculated thecorrelations between these variables. In Fig. 6, we present thecross-correlations between the MSWIR and the biomass density,the MSWIR and the soil-water density and the biomass and thesoil-water densities. The cross-correlations were calculated fordifferent precipitation rates and for two degrees of hetero-geneity s ¼ 0:01; 0:04day�1

� �. The correlation between the soil-

water (and the biomass) density and the MSWIR is higher for the

Fig. 6. The correlations between the MSWIR (denoted by a) and the biomass and soil-wheterogeneity of the MSWIR, s = 0.04 day�1, and the dashed lines correspond to lower

more heterogeneous landscape s ¼ 0:04day�1� �

. In both land-scapes, the correlation between the soil-water and the biomassdensities is very high and only drops when the precipitation rateapproaches the critical value at which the vegetation collapses(under these conditions, the biomass density in many regionsapproaches zero, thereby reducing the correlation). An oppositetrend was found for the correlation between the soil-water densityand the MSWIR. As the precipitation rate approaches the criticalvalue, there is a sharp increase in this correlation.

4. Discussion

Field measurements showed that there are many more liveshrubs in the heterogeneous hillslopes than in the homogeneousones. Moreover, there are fewer dead shrubs in the heterogeneoushillslopes. These findings are surprising given the proximity of thehillslopes, implying almost identical climate conditions (especiallyprecipitation values). The studies reporting the mass shrubmortality in response to the droughts during the first decade of

ater densities versus the precipitation rate (R). The solid lines correspond to higherheterogeneity of the MSWIR, s = 0.01 day�1.

102 H. Yizhaq et al. / Ecological Complexity 31 (2017) 96–103

the twenty-first century did not account for this striking differencein mortality between neighboring hillslopes (Shachak 2011; Zaadyet al., 2012; Sher et al., 2012).

Previous theoretical studies (Yizhaq et al., 2014; Stavi et al.,2016; Yizhaq and Bel, 2016) suggested that geodiversity increasesthe durability of vegetation and allows it to survive under harsherconditions. However, the mechanism by which the geodiversityaffects the vegetation was not fully understood. Here, we combinedfield and theoretical studies to show that the survival of shrubbyvegetation under lower precipitation in heterogeneous landscapesis due to the occurrence of patches with higher soil-water density.

The mathematical model captures two important factors invegetation dynamics. On one hand, vegetation increases thesurface-water infiltration rate and thereby has a positive feedbackon its growth. On the other hand, there is competition for the soilwater, which is the main growth-limiting resource. These twomechanisms result in spatially patterned vegetation and soil-waterdensities. Several characteristics of the landscape that are oftenheterogeneous are captured in the theoretical model. In particular,the soil-water diffusivity, the evaporation and drainage rate andthe MSWIR affect the biomass and soil-water densities. In thisstudy, we focused on the effects of the MSWIR’s heterogeneity onthe vegetation dynamics. However, the stoniness that is found inthe field may be represented by additional parameters in themodel. For example, the stoniness of the soil is expected to affectthe evaporation rate of surface water, the soil-water diffusivity andthe evaporation and drainage rate of soil water. The effects ofheterogeneity in the latter parameters were previously studied(Yizhaq et al., 2014; Yizhaq and Bel, 2016) and showed similareffects on the vegetation (the correlations between the soil waterand the heterogeneous parameters were not previously investi-gated). Other factors, such as the soil texture, the micro-topography, and grazing, may introduce heterogeneity in the fieldand in the model parameters. A more realistic modeling of anactual ecosystem should involve heterogeneity in all theseparameters and the appropriate correlations between them. Ourresults suggest that the main effects of the heterogeneity (in thetheoretical model) are well captured by our simplified approachthat focused on one parameter.

In the field, we found greater mean volumetric moisturecontent of soil in the heterogeneous hillslopes throughout almostthe entire wet season. In addition, despite being too low for uptakeby the vegetation, the higher mean hygroscopic moisture contentin the heterogeneous hillslopes (relative to the homogeneousones) indicates better physical quality of the soil with resultingimproved living conditions for the shrubs (Stavi et al., 2015). Thehigher soil-water density in the heterogeneous hillslopes is alsopredicted by the mathematical model simulations. The theoreticalresults suggest that the stoniness of the soil does not significantlychange the average soil-water density but, rather, changes thespatial distribution of the water. According to the modelsimulations, the maximal soil-water density only weakly dependson the precipitation rate and remains much higher than theaverage soil-water density throughout the range of precipitationrates that supports vegetation. Domains with larger MSWIRscontain more water, thereby allowing shrubs to survive droughtsthat are severe enough to cause significant mortality in thehomogeneous hillslopes. The correlations between the MSWIR andthe soil-water and biomass densities indicate that geodiversityaffects vegetation survivability by providing domains with enoughsoil-water even under precipitation rates for which vegetation in ahomogeneous landscape would collapse. This increased durabilityappears despite the fact that the homogeneous and heterogeneouslandscapes have the same average characteristics (in the theoreti-cal study). In order to fully understand the mechanism and theimportance of the different aspects of geodiversity, controlled

experiments in which the various factors are isolated should becarried out, for example, by examining the relative importance ofsoil-surface stoniness versus soil-profile stoniness.

This study focused on the dynamics of shrubs in semi-aridlandscapes. However, the ecosystem also includes herbaceousvegetation. In particular, the homogeneous landscape hosts muchdenser herbaceous vegetation that covers most of the hillslopes.Obviously, the existence of different types of vegetation affects thedynamics of each of them and imposes both facilitation andcompetition effects. Previous records of the study site show that inthe past, prior to the sequence of droughts presented in Fig. 1, boththe homogeneous and the heterogeneous landscapes were coveredby shrubby vegetation (Shachak, 2011). However, massivemortality of shrubs has been only observed in the homogeneoushillslopes. This change in the vegetation population of theecosystem can be attributed to the stronger effect of drought onthe shrubs, while the herbaceous vegetation is mostly sensitive tothe seasonal precipitation. To fully understand the interactions ofthe two plant lifeforms, further field studies, which involvecontrolled removal of herbaceous vegetation in designated plots,and an extension of the model to account for both types ofvegetation are needed.

Climate models predict a significant decrease in the precipita-tion over large regions in the Middle East and Eastern Mediterra-nean (Evans, 2009). There are also predictions of increased risk ofdroughts in these areas (IPCC, 2013). The decrease in precipitationand increase in drought frequency are expected to affect not onlyrainfed agriculture but also many ecosystems and, in particular,shrubby vegetation. Our results emphasize the need to account forthe geodiversity of ecosystem landscapes when assessing theirvulnerability to water deficiency and drought.

In this work, we combined field measurements and atheoretical study of a water-limited vegetation ecosystem in orderto identify the mechanism by which geodiversity–expressed bystoniness–affects vegetation survivability under prolongeddrought. The results suggest that the missing link is the soil-water density. We found that geodiversity increases the soil-watercontent. The theoretical model suggests that the maximal soil-water density only weakly depends on the precipitation ratepreceding the collapse of the vegetation. Therefore, in heteroge-neous landscapes, domains with higher soil-water density, andthereby better living conditions for the vegetation, exist over awider precipitation range. These domains improve the survival ofvegetation and its resilience under drought conditions. Thesefindings suggest that the geodiversity of ecosystems constitutes animportant factor in their vulnerability to climate changes andanthropogenic disturbances.

Authorship

HY and GB performed the modeling, IS, HY and GB designed andperformed the field measurements, GB, HY and IS developed theconceptual model, MS designed the ecological context, all authorscontributed to the writing of the manuscript.

Acknowledgments

This study was supported by the Israel Science Foundation,Grant No.1260/15. We thank Evyatar Nurik, Shimon Rachmilevitch,Reut Berger-Tal, Amjad Hijazen, and Ofir Altstein.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.ecocom.2017.06.002.

H. Yizhaq et al. / Ecological Complexity 31 (2017) 96–103 103

References

Ackerly, D.D., Loarie, S.R., Cornwell, W.K., Weiss, S.B., Hamilton, H., Branciforte, R.,Kraft, N.J.B., 2010. The geography of climate change: implications forconservation biogeography. Divers. Distrib. 16, 476–487.

Anderson, M.G., Ferree, C.E., 2010. Conserving the stage: climate change and thegeophysical underpinnings of species diversity. PLoS One 5 (7), e11554.

Beier, P., Brost, B., 2010. Use of land facets to plan for climate change: conserving thearenas, not the actors. Conserv. Biol. 24, 701–710.

Borgogno, F., D'Odorico, P., Laio, F., Ridolfi, L., 2009. Mathematical models ofvegetation pattern formation in ecohydrology. Rev. Geophys. 47, 1–36.

Bowers, J.E., 2005. Effects of drought on shrub survival and longevity in the northernSonoran Desert. J. Torrey Bot. Soc. 132, 421–431.

Bunte, K., Poesen, J., 1994. Effects of rock fragment size and cover on overland flowhydraulics: local turbulence and sediment yield on an erodible soil surface.Earth Surf. Proc. Landforms 19, 115–135.

Ciais, P.H., Reichstein, M., Viovy, N., Granier, A., Ogee, J., Allard, V., Aubinet, M.,Buchmann, N., Bernhofer Chr, Carrara, A., Chevallier, F., De Noblet, N., Friend, A.D., Friedlingstein, P., Grunwald, T., Heinesch, B., Keronen, P., Knohl, A., Krinner,G., Loustau, D., Manca, G., Matteucci, G., Miglietta, F., Ourcival, J.M., Papale, D.,Pilegaard, K., Rambal, S., Seufert, G., Soussana, J.F., Sanz, M.J., Schulze, E.D.,Vesala, T., Valentini, R., 2005. Europe-wide reduction in primary productivitycaused by the heat and drought in 2003. Nature 437, 529–533.

DeMalach, N., Kigel, J., Voet, H., Ungar, E.D., 2014. Are semiarid shrubs resilient todrought and grazing? Differences and similarities among species and habitats ina long-term study. J. Arid Environ. 102, 1–8.

Dunkerley, D.L., 1995. Surface stone cover on desert hillslopes – parametrizingcharacteristics relevant to infiltration and surface runoff. Earth Surf. Proc.Landforms 20, 207–218.

Eldridge, D.J., Zaady, E., Shachak, M., 2000. Infiltration through three contrastingbiological soil crusts in patterned landscapes in the Negev, Israel. Catena 40,323–336.

Engelbrecht, B.M.J., Comita, L.S., Condit, R., Kursar, T.A., Tyree, M.T., Turner, B.L.,Hubbell, S.P., 2007. Drought sensitivity shapes species distribution patterns intropical forests. Nature 447, 80–82.

Evans, J.P., 2009. 21st century climate change in the Middle East. Clim. Change 92(3), 417–432.

Getzin, S., Yizhaq, H., Bell, B., Erickson, T.E., Postle, A.C., Katra, I., Tzuk, O., Zelnik, Y.R.,Wiegand, K., Wiegand, T., Meron, E., 2016. Discovery of fairy circles in Australiasupports self-organization theory. PNAS 113, 3551–3556.

Gilad, E., von Hardenberg, J., Provenzale, A., Shachak, M., Meron, E., 2004. Ecosystemengineers: from pattern formation to habitat creation. Phys. Rev. Lett. 93 098105(1-4).

Hoffman, O., deFalco, N., Yizhaq, H., Boeken, B., 2016. Annual plant diversitydecreases across scales following widespread ecosystem engineer shrubmortality. J. Veg. Sci. 27, 578–586.

IPCC, 2013. In: Stocker, T.F., Qin, D., Plattner, G.-K., Tignor, M., Allen, S.K., Boschung, J.,Nauels, A., Xia, Y., Bex, V., Midgley, P.M. (Eds.), Climate Change 2013: ThePhysical Science Basis. Contribution of Working Group I to the Fifth AssessmentReport of the Intergovernmental Panel on Climate Change. CambridgeUniversity Press, Cambridge, United Kingdom and New York, NY, USA (1535 pp).

Ibáñez, J.J., Krasilnikov, P.V., Saldaña, A., 2012. REVIEW: archive and refugia of soilorganisms: applying a pedodiversity framework for the conservation ofbiological and non-biological heritages. J. Appl. Ecol. 49, 1267–1277.

Kéfi, S., Eppinga, M.B., de Ruiter, P.C., Rietkerk, M., 2010. Bistability and regularspatial patterns in arid ecosystems. Theor. Ecol. 3 (4), 257–269.

Kinast, S., Zelnik, Y.R., Bel, G., Meron, E., 2014. Interplay between Turing mechanismscan increase pattern diversity. Phys. Rev. Lett. 112 (7), 078701.

Klausmeier, C.A., 1999. Regular and irregular patterns in semiarid vegetation.Science 284, 1826–1828.

Kruckeberg, A.R., 2002. Geology and Plant Life, Seattle. University of WashingtonPress, WA.

Lavee, H., Poesen, J.W.A., 1991. Overland flow generation and continuity on stone-covered soil surfaces. Hydrolog. Process 5, 35–360.

Lawler, J.J., Ackerly, D.D., Albano, C.M., Anderson, M.G., Dobrowski, S.Z., Gill, J.L.,Heller, N.E., Pressey, R.L., Sanderson, E.W., Weiss, S.B., 2015. The theory behind,and the challenges of, conserving nature’s stage in a time of rapid change.Conserv. Biol. 29, 618–629.

Lejeune, O., Tlidi, M., Couteron, P., 2002. Localized vegetation patches: a self-organized response to resource scarcity. Phys. Rev. E 66, 010901.

Mcbratney, A.B., 1992. On variation: uncertainty and informatics in environmentalsoil management. Soil Res. 30, 913–935.

Paz-Kagan, T., Panov, N., Shachak, M., Zaady, E., Karnieli, A., 2014. Structural changesof desertified and managed shrubland landscapes in response to drought:spectral, spatial and temporal analyses. Remote Sens. 6, 8134–8164.

Petersen, A., Gröngröft, A., Miehlich, G., 2010. Methods to quantify the pedodiversityof 1 km2 areas — results from southern African drylands. Geoderma 155 (3–4),140–146.

Phillips, J.D., 1999. Earth Surface Systems. Blackwell (180 pp).Poesen, J., Ingelmo-Sanchez, F., Mucher, H., 1990. The hydrological response of soil

surfaces to rainfall as affected by cover and position of rock fragments in the toplayer. Earth Surf. Proc. Landforms 15, 653–671.

Reichstein, M., Tenhunen, J.D., Roupsard, O., Ourcival, J., Rambal, S., Miglietta, F.,Peressotti, A., Pecchiari, M., Tirone, G., Valentini, R., 2002. Severe drought effectson ecosystem CO2 and H2O fluxes at three Mediterranean evergreen sites:revision of current hypotheses? Glob. Change Biol. 8 (10), 999–1017.

Rietkerk, M.G., Boerlijst, M., Langevelde, van, F., HillerisLambers, R., Koppel, van de,J., Kumar, L., Prins, H.H.T., de Roos, A.M., 2002. Self-organization of vegetation inarid ecosystems. Am. Nat. 160, 524–530.

Rietkerk, M., Dekker, S.C., de Ruiter, P.C., van de Koppel, J., 2004. Self-organizedpatchiness and catastrophic shifts in ecosystems. Science 305, 1926–1929.

Rietkerk, M., van de Koppel, J., 2008. Regular pattern formation in real ecosystems.Trends Ecol. Evol. 23, 169–175.

SAS Institute, 1990. SAS/STAT User's Guide. Version 6, 4th ed. SAS Inst., Cary, NC.Shachak, M., 2011. Ecological textures: ecological systems in the northern Negev as a

model. Ecol. Environ. 1, 18–29 (in Hebrew).Sher, Y., Zaady, E., Ronen, Z., Nejidat, A., 2012. Nitrification activity and levels of

inorganic nitrogen in soils of a semi-arid ecosystem following a drought-induced shrub death. Eur. J. Soil Biol. 53, 86–93.

Sherratt, J.A., 2015. Using wavelength and slope to infer the historical origin of semi-arid vegetation bands. PNAS 112, 4202–4207.

Shoshany, M., Karnibad, L., 2015. Remote sensing of shrubland drying in the south-east Mediterranean: 1995–2010: water-use-efficiency-cased mapping ofbiomass change. Remote Sens. 7, 2283–2301.

Singer, A., 2007. The Soils of Israel. Springer Verlag, Berlin (pp. 306.).Stavi, I., Shem-Tov, R., Chocron, M., Yizhaq, H., 2015. Geodiversity, self-organization,

and health of three-phase semi-arid rangeland ecosystems in the Israeli Negev.Geomorphology 234, 11–18.

Stavi, I., Rachmilevitch, S., Yizhaq, H., 2016. Small-scale geodiversity regulatesfunctioning, connectivity, and productivity of shrubby, semi-arid rangelands.Land Degrad. Dev. doi:http://dx.doi.org/10.1002/ldr.2469.

Tongway, D.J., Hindley, N.L., 2004. Landscape Function Analysis: Procedures forMonitoring and Assessing Landscapes. CSIRO Publishing, Brisbane.

Banded Vegetation Patterning in Arid and Semiarid Environments. In: Tongway, D.,Valentin, C., Seghieri, J. (Eds.), Springer-Verlag, New York.

Yizhaq, H., Bel, G., 2016. Effects of quenched disorder on critical transitions inpattern-forming systems. New J. Phys. 18, 023004.

Yizhaq, H., Sela, S., Svoray, T., Assouline, S., Bel, G., 2014. Effects of heterogeneoussoil-water diffusivity on vegetation pattern formation. Water Resour. Res. 50,5743–5758.

Zaady, E., Shir, Y., Ronen, Z., Nujedat, A., 2012. The effect of consecutive drought inthe northern Negev on soil characteristics and plant mortality. Ecol. Environ. 3,44–52 (in Hebrew).

Zelnik, Y., Kinsat, S., Yizhaq, Y., Golan, B., Meron, E., 2013. Regime shifts in models ofdryland vegetation. Philos. T. Roy. Soc. A 371, 20120358.

Zelnik, Y.R., Meron, E., Bel, G., 2015. Gradual regime shifts in fairy circles. PNAS 112(40), 12327–12331.