Geomorphology 234 (2015) 11–18

Contents lists available at ScienceDirect

Geomorphology

j ourna l homepage: www.e lsev ie r .com/ locate /geomorph

Geodiversity, self-organization, and health of three-phase semi-aridrangeland ecosystems, in the Israeli Negev

I. Stavi a,⁎, R. Shem-Tov a, M. Chocron b, H. Yizhaq a,c

a Dead Sea & Arava Science Center, Ketura 88840, Israelb Department of Land of Israel Studies and Archaeology, Bar-Ilan University, Ramat Gan 52900, Israelc Department of Solar Energy and Environmental Physics, Jacob Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Sede Boqer Campus, Midreshet Ben Gurion, Israel

⁎ Corresponding author. Tel.: +972 8 630 6319; fax: +E-mail addresses: istavi@adssc.org, istavi@yahoo.com

http://dx.doi.org/10.1016/j.geomorph.2015.01.0040169-555X/© 2015 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 21 September 2014Received in revised form 25 December 2014Accepted 11 January 2015Available online 17 January 2015

Keywords:Bi-modal patternEcosystem complexityHerbaceous vegetationMesic vs. xeric conditionsSource–sink relationsVegetative pattern

Source–sink, two-phase mosaic-like ecosystems are widespread throughout the world's drylands. Such ecosys-tems are composed of woody vegetation patches and intershrub spaces and have been characterized as havinghigh flexibility and survivability. Recent studies from the semi-arid Negev drylands of Israel reported that live-stock grazing has resulted in the modification of two-phase mosaic-like shrublands into three-phase mosaicrangelands, with livestock trampling routes encompassing a separate, and the most degraded phase, while theshrubs encompass the most improved phase. The objective of this study was, therefore, to reassess this theorythrough the investigation of patch-scale (spatial scale of one to several decimeters) geodiversity and self-organization of these ecosystems. In terms of the effect of type of surface cover (microhabitat), the soil hygro-scopic moisture content and stable aggregate content of the uppermost layer (0–5 cm depth) were significantlyaffected by this factor, and revealed the highest, intermediate, and smallest values for the shrubby patches (3.06%and 77%), intershrub spaces (2.81% and 68%), and the trampling routes (2.63% and 55%), respectively. Anoppositeeffect was recorded for the sand content, revealing 23.9%, 25.3%, and 26.0%, respectively. The clay dispersionindex was also significantly affected by microhabitat, and revealed a higher value for the trampling routes(0.83) than for the intershrub spaces and shrub patches (0.37 for both). At the same time, other soil characteris-tics were not significantly affected bymicrohabitat. Overall, some differences were recorded between north- andsouth-facing hillslopes, proposing somewhat better soil quality in the northern aspects. A conceptual model isproposed, in which moderate livestock pressure increases ecosystem geodiversity at the patch scale, modifyingthe ecosystem's self-organization to encompass a new (dynamic) equilibrium of a tri-modal pattern, and increas-ing ecosystemhealth. Also, a simple numerical simulation is proposed, modeling the effect of livestock tramplingroutes on the redistribution of water at the patch scale, with the resultant modifications in distribution of vege-tation cover. Yet, it is proposed that functioning of three-phase mosaic rangelands is more complex than previ-ously suggested, encompassing several simultaneous effects, of which some may have offsetting impacts.

© 2015 Elsevier B.V. All rights reserved.

1. Introduction

Extensive lands among the world's semi-arid regions have been re-ported as composed of two-phase mosaic-like ecosystems, containingwoody vegetation patches and bare or herbaceous vegetation-coveredground in the intershrub spaces (Carter and O'Connor, 1991). Source–sink spatial relations have been extensively reported to occur betweenthese two types of surface cover (microhabitats) (Merino-Martin et al.,2012),where the intershrub spaces contribute runoffwater and associat-ed dissolved and suspendedmaterials to the vegetation patches,which iswhere these materials are accumulated and utilized for supporting

972 8 635 6634.(I. Stavi).

vegetation production (Imeson and Prinsen, 2004). High functioning ca-pability, i.e., large capacity in retainingwater and soil resourceswithin anecosystem's boundaries while allowing only small leakage, characterizessuch ecosystems, which maintain their production capacity even duringconsecutive drought years (Tongway and Ludwig, 2003). Awide range ofphysical and biotic conditions have led to the formation of several pat-terns of vegetation patchiness, such as stripes, strands, stipples, andothers, efficiently exploiting the limited water and soil resources(Ludwig et al., 1999). Such patterns of self-organization enable the sur-vival of vegetation in drylands, where precipitation regimes could notsupport full vegetation cover (Rietkerk et al., 2002; Borgogno et al.,2009). Among other definitions of ecosystemhealth, one of themost im-portant is its ability to support productivity, (self-)organization, and re-silience, i.e., to carry on vegetation (and animal) growth, to sustaindiversity and interactions among its components, and to buffer

12 I. Stavi et al. / Geomorphology 234 (2015) 11–18

perturbations, respectively (Rapport et al., 2013). Therefore, ecosystemhealth should be the ultimate goal when discussing either naturally oranthropogenically modified environments.

Geodiversity – defined as the natural range of geological, geo-morphic, and soil features (Gray, 2005) – impacts biodiversity(Jačková and Romportl, 2008) and, in addition, affects a range of eco-system services and functions (Gray, 2004). The combination of bio-diversity and geodiversity determines the overall natural diversity(Cañadas and Flaño, 2007). Therefore, a holistic approach to the con-servation of natural diversity can be achieved only if consideringboth the living (biodiversity) and non-living (geodiversity) aspectsof the relevant ecosystems (Pemberton, 2007). Evaluation ofgeodiversity must include the interpretation of processes and rela-tionships (e.g., spatial redistribution of water and soil resources)among its components (Gray, 2004).

Grazing lands cover more than 60 million km2 or 45% of the terres-trial surface of the globe (Reid et al., 2008). Livestock impact, includingthe browsing of vegetation, excretion of feces and urine, and tramplingof soil, considerably affects the functioning and production capacity ofthe rangeland ecosystems (Coughenour, 1991). A set of studies focusingon the three-phase mosaic-like pattern was recently summarized byStavi et al. (2012), highlighting the effect of the non-even distributionof livestock traffic on hillslopes on some characteristics of soil and veg-etation in semi-arid rangeland ecosystems. Nevertheless, the impact oflivestock on patch-scale geodiversity, with a spatial scale of 1 dm to sev-eral decimeters, and its effects on the functioning of rangeland ecosys-tems has still remained greatly unknown. Therefore, the objective ofthis study was to investigate the patch-scale geodiversity of theserangelands, and to assess how it reflects on ecosystem functioning andself-organization.

2. Materials and methods

2.1. Regional settings

The study was implemented at the Lehavim Demonstration Farm,located in the northern semi-arid Negev (31° 20′ N, 34° 46′ E) ofIsrael (Figs. 1 and 2). The area's lithology is chalk of the Eocene, with atopography comprised of rolling hills. The mean altitude ranges be-tween 350 and 500 m above sea level (Perevolotsky and Landau,1988), and the soil is classified as Brown Rendzina (Dan andKoyumdjisky, 1979). The predominant shrub species is Sarcopoteriumspinosum (L.) Spach, while Coridothymus capitatus (L.) Rchb.f. is alsoprevalent in southern-facing hillslopes. The herbaceous vegetation con-sists of a range of grasses, forbs, and legumes. Mean daily temperaturesrange between 11 °C in January and 25 °C in July; mean daily relativehumidity ranges between 67% and 50%, respectively; and mean annualprecipitation is approximately 300 mm (Bitan and Rubin, 1991). Thefarm itself encompasses about 800 ha, where long-term livestock graz-ing has been implemented with a flock of approximately 800 head ofsheep and goats (Stavi et al., 2012).

2.2. Mapping of ground surface cover (microhabitats)

Fieldwork was conducted at the end of the dry season (September)of 2013. Three pairs of north- and south-facing hillslopes were selectedfor the study. This study schemewas implemented due to the prevailingconditions of mesic and xeric habitats in northern and southern aspects,respectively (see: Rigg, 1993), which are assumed to affect the ecosys-tem self-organization and functioning. A location was randomly select-ed along the backslope of each of these hillslopes and utilized for thedelineation of a 10 × 10m plot. This plot sizewas chosen in order to en-sure the capability of randomly selecting the sampling spots withinthem. Theplotswere thenmapped for their different types of cover – in-cluding shrub patches, livestock trampling routes, and the remainder ofthe intershrub spaces – by using a high-resolution (10 cm precision)

GPS apparatus. Mapping was based on delineation of the shrub patches'perimeter and trampling routes' area. The routes were easily identifiedby their exposed surface consisting of mechanical crusts, and by theirpredominant, elongated lateral shape, transecting the hillslopes alongcontours. Data were then digitizedwith ArcGIS software for the calcula-tion of the cover percentage of each of these two types of microhabitats.The cover percentage of the intershrub spaces excluding the tramplingroutes was then calculated by subtracting the areas of shrubby patchesand of trampling routes from the total area of the plot (100 m2).

2.3. Soil sampling and infiltration testing

After mapping, soil samples from the uppermost soil layer (0–5 cmdepth) was obtained in five randomly selected spots of each of thetypes of cover: shrub patches (of S. spinosum only), trampling routes,and intershrub spaces excluding the trampling routes. Tomaintain con-sistency, the soil samplewas taken from each of the shrub patches later-ally rather than upslope or downslope, and on the western side of theshrub center. The samples were carefully placed in a sealed plastic bag.

In addition, in proximity to each spot of the soil sampling, the infil-tration capacity of water to soil under unsaturated conditions was alsotested. This was implemented by using a mini-disk infiltrometer(Decagon®, USA) for 5 min (300 s) per spot.

Upon arrival to the laboratory, the soil sampleswere left to air-dry ina well-ventilated space. Number of spots (n) for the soil samples as wellas for the infiltration tests was: 5 replicates × 3 microhabitats × 3hillslopes × 2 aspects = 90.

2.4. Laboratory analyses

Sub-samples of the soil were put in a drying oven (set to 105 °C,for 24 h) to determine the hygroscopic moisture content. The mainsoil samples were analyzed for texture (by the hydrometer method:Bouyoucos, 1962), electrical conductivity (Richards, 1954), pH(McLean, 1982), aggregate stability index (Herrick et al., 2001), sta-ble aggregate content (by using an aggregate stability apparatus:Eijkelkamp®, the Netherlands), and clay dispersion index. The latterwas determined through the positioning of an aggregate of 3 to5 mm diameter in a Petri plate filled with distilled water, followedby a visual observation of the extent of cloudiness (milkiness) after10 min, and again, after 2 h. Dispersion index scores ranged from 0for no cloudiness at all; 1 for slight cloudiness; 2 for moderate cloud-iness; 3 for strong cloudiness, and 4 for complete cloudiness of theaggregate's clays (adapted from: USDA-NRCS, EFH NOTICE 210-WI-62). These soil characteristics were chosen due to their capacity inrepresenting the overall soil quality.

2.5. Statistical analysis

For analyzing the overall effect of the hillslope aspect, data process-ingwas required. This included the normalizing of data according to therelative cover percentage of each of the types of microhabitats on eachof the hillslopes.

Then, analysis of variance (ANOVA) was conducted with the GLM(general linear model) procedure of SAS (SAS Institute, 1990). Factorsin themodelwere hillslope aspect (1 degree of freedom; df), blockwith-in hillslope aspect (3 df; error term for aspect), type of cover (2 df), andthe interaction hillslope aspect × type of cover (2 df). Statistically signif-icant interactions were subjected to additional ANOVA with the SLICEcommand of PROC GLM. Separation of means was implemented byTukey's HSD at a probability level of 0.05. Pearson correlation coeffi-cients were computed to assess the relations between each pair ofvariables.

Fig. 1.Map of the study site in Israel.

13I. Stavi et al. / Geomorphology 234 (2015) 11–18

3. Results and discussion

3.1. Ground surface cover (microhabitats)

Concordant with previous studies (summarized in Stavi et al., 2012),the hillslope's surface cover was found to be of a patchy nature. Theintershrub spaces were found to have the greatest mean cover(61.1 ± 6.2%), shrub patches, an intermediate cover (28.1 ± 7.1%),and livestock trampling routes, the smallest cover (10.8± 1.9%). The ef-fect of hillslope aspect on themean cover percentage of the differentmi-crohabitats was considerable, with shrubs having much greater meancover percentage in the northern facing hillslopes, and routes havingconsiderably greater mean cover percentage in the southern facinghillslopes (Fig. 3). Two GIS-based maps of representative northern-and southern-facing hillslopes are shown in Fig. 4.

It is noteworthy to mention that the overall mean cover percentageof trampling routes observed in this study was only about a half of thatreported in a recent study (Stavi et al., 2012: ~21%) which was imple-mented in the same region. This could be attributed to the mapping ofpairs of hillslopes in this study different to those which were utilizedin Stavi et al. (2012), suggesting wide heterogeneity among hillslopesin the study region.

3.2. Hillslope effect

The greater mean shrub cover on the northern-facing hillslopes cor-responds with previous studies which showed that compared withsouth-facing hillslopes, the smaller loss of soil moisture through

evaporation on north-facing hillslopes (Shoshany, 2002) results in theformation of mesic conditions (Rigg, 1993), increasing vegetationgrowth and augmenting net primary productivity (NPP) (Bochet andGarcía-Fayos, 2004). The greater NPP in the northern hillslopes is ex-pected to improve the physical and hydraulic characteristics of the sur-face soil (Archer et al., 2002), further increasing the retention of waterand soil resources within the ecosystem boundaries (Andreu et al.,2001). This concept accords with the study results, revealing signifi-cantly greater means of normalized aggregate stability index, stable ag-gregate content, and hygroscopicmoisture content in the north- than inthe south-facing aspects. In addition, the significantly greatermeannor-malized soil pH in the southern, rather than in the northern hillslopes,demonstrates the lower limitation of vegetation productivity due tosoil alkalinity in the north-facing aspects than that in the south. Yet,themean normalized pH level under both of the northern and southernaspects was only slightly alkaline and, presumably, had no impact onthe ecosystem production capacity. At the same time, no significant ef-fect of hillslope aspect was recorded for any of the mean normalizedvalues of soil electrical conductivity, clay dispersion index, and unsatu-rated infiltration capacity (Table 1). However, themean normalized soiltexture was considerably impacted by the hillslope aspect. This was re-vealed by the significantly greater silt content and significantly smallersand content in the northern hillslopes than those in the southernhillslopes (Table 2). Overall, the smaller silt content and larger sandcontent of the surface soil indicate the sorting – through erosional pro-cesses – of the finer fractions (see: Zhang et al., 2014), and suggest thatthe southern aspects are more susceptible to hillslope-scale erosionalprocesses than the northern aspects (e.g., Istanbulluoglu et al., 2008).

Fig. 2. Characteristic landscape of the study region.

14 I. Stavi et al. / Geomorphology 234 (2015) 11–18

3.3. Microhabitat effect

The highly significant effect of type of cover on the mean of each ofthe stable aggregate content and hygroscopic moisture content clearlydemonstrates the existence of three different microhabitats in thistype of ecosystem. The greatest, intermediate, and smallest values ofthese variables for the shrub patches, intershrub spaces, and tramplingroutes, respectively (Table 3), accord with Stavi et al. (2012), who re-ported the same trend of overall quality of soil. Thesefindings are attrib-uted to the source–sink relations, where the trampling routes act asoptimal source areas and the shrubby patches act as optimal sinks,while the intershrub spaces (excluding routes) lay in between thesetwo extremes. The greatest and smallest mean contents of sand fractionunder the trampling routes and shrubby patches, respectively, aswell asthe (though not significantly) smallest clay content in the routes(Table 4), demonstrate the sorting of the finer fractions off the routes.Therefore, it can be assumed that suspended finemineralmaterials, dis-solved materials, and floating organic materials that are generated inthe trampling routes are accumulated in the shrubby patches and to asmaller extent also in the intershrub areas. The latter twomicrohabitatsexperience the improvement in soil structure formation and aggregate

Fig. 3.Mean cover (%) of shrub patches, intershrub spaces, and trampling routes, by hill-slope aspect.

stability, resulting in an increase in the soil hygroscopic moisture con-tent. At the same time, the absence of vegetation in the tramplingroutes, coupled with their smooth surface, negates the retention ofwater and deposition of fine mineral material and organic material ontheir surface, and prevents the development of well-structured soil inthis microhabitat. Over the long run, the reoccurrence of intense live-stock traffic on the trampling routes enables these processes to beself-sustaining. These results strengthen the recently proposed concept(summarized in Stavi et al., 2012), according towhich the considerationof such ecosystems as two-phase mosaics is an over-simplification.

The mean clay dispersion index, despite being significantly affectedby microhabitat, was similar between the shrub patches and intershrubspaces, demonstrating the complexity of the functioning of such three-phasemosaic-like geo-ecosystems. Oneway or another, the significant-ly and considerably smaller clay dispersion index under these two mi-crohabitats than that under the trampling routes exemplifies theinferior physical quality of soil under the latter (Table 3). Regardless,these results are in accordance with the concept of ‘fertility islands’(Garner and Steinberger, 1989), where vegetative patches are claimedto operate as sinks of water runoff that is generated in the intershrubspaces (Saco et al., 2007). According to this concept, along the temporalaxis, the soil quality of suchmosaic-like ecosystems is getting improvedin the vegetative patches and degraded in the intershrub spaces(Vásquez-Méndez et al., 2010).

At the same time, the effect of type of cover on mean aggregate sta-bility index, pH, and unsaturatedwater infiltration capacity was not sig-nificant (Table 3). To some extent, the absence of significant effect ofmicrohabitat on the unsaturated infiltration capacity may be attributedto the recorded very high variability for this soil feature, which by itself,could be attributed to a finer-scale heterogeneity of the surface soil.Also, despite the considerable differences in themean soil electrical con-ductivity among the various microhabitats, the high variability of thissoil characteristic negated a significant effect. Yet, the much greaterelectrical conductivity of the soil under the trampling routes than thatunder the shrub patches and intershrub spaces could be attributed tothe absence of a fine root system in this microhabitat, decreasing saltleaching from the uppermost soil layer.

The effect of the interaction type of cover × hillslope aspect was sig-nificant only for the soil's mean silt and sand contents. The mean siltcontent was significantly different between the northern and southernhillslopes only for the trampling routes, being greater in the more

Fig. 4. GIS maps of the plot in a representative (the “FOREST”) northern (a) and southern (b) aspects.

15I. Stavi et al. / Geomorphology 234 (2015) 11–18

mesic aspects than in the more xeric aspects. An opposite effect was re-corded for the sand content under each of the intershrub spaces andtrampling routes (Table 5).

3.4. General data integration and knowledge gaps

Over recent years, poor maintenance of the fences surrounding thelivestock-exclusion plots across the Lehavim Demonstration Farm negat-ed the investigation of the actual effect of livestock grazing on the range-land geodiversity. Regardless, the obvious effect of the grazing animals onthe formation andpersistence of the trampling routes highlighted the im-pact of livestock in increasing patch-scale geodiversity. Also, obtainingundisturbed soil coreswas impossible because of the extremely high con-tent of rock fragments in the soil, imposing technical difficulties in inves-tigating the effect of hillslope aspect and microhabitat (type of surfacecover) on the soil's available moisture capacity. Yet, the results of someof the studied soil properties, such as the stable aggregate content, hygro-scopic moisture content, aggregate stability index, clay dispersion index,pH, and texture, highlighted the considerable effect of hillslope aspectand microhabitat on soil quality. Overall, despite some discrepancies,the obtained results affirm the previously proposed concept, suggestingthat livestock trampling routes constitute a separate microhabitat,which causes the two-phase mosaic-like patterns to function as three-phase ecosystems (Stavi et al., 2012).

The concept of natural diversity encompasses two components:(1) the number of different types of objects (e.g., biological speciesand soil types) in a mixture or a sample, and (2) the relative size or

Table 1Effect of hillside aspect on the soil's unsaturated infiltration rate (cm s−1); aggregate stability in(%); clay dispersion index; hygroscopic moisture content (%); electrical conductivity (μS), and

Infiltration rate Stability index Stable aggregate Clay dispe

P value 0.1442 0.0259 0.0001 0.185North 0.00059a (0.00006) 5.77a (0.10) 79.0a (2.0) 0.32a (0.0South 0.00046a (0.00017) 5.40b (0.14) 58.3b (3.4) 0.51a (0.1

Notes:Meanswithin the same column followed by a different letter differ at the 0.05 probabilitymeans.

number of each type of object, as well as its distribution among theother objects (Ibáñez et al., 2012). At the same time, two important con-cepts for the quantification of diversity are: (a) whether the specificgroups are different enough to be considered separate types of objects,and (b)whether the objects in each specific group are similar enough tobe considered the same type (Huston, 1994). According to these con-cepts, the considerable cover of trampling routes (almost 11%), theirspatial reoccurrence, and the remarkable differences between themand the other types of surface cover, make them an important determi-nant of the geo-ecosystem diversity. Regardless, some of the obtainedresults suggest no clear difference between the shrubby patches andthe intershrub spaces excluding the trampling routes. Also, the absenceof a strong correlation (r N 0.50) between any pair of the studied soilcharacteristics further demonstrated the geo-ecosystem's complex na-ture, with the presumably simultaneous impacts of offsetting mecha-nisms between them.

Along the soil quality continuum, the shrubby patches and tramplingroutes represent the maximum and minimum extremes, respectively,with the intershrub spaces lying somewhere between those extremes.The spatial relations among the different microhabitats are proposedto formpositive feedbacks, which strengthen the existing state and con-ditions in each of them (Fig. 5). For example, high-intensity trampling inthe routes is assumed to grind and shear the uppermost soil layer in thismicrohabitat. The ground and sheared mineral material becomes avail-able for suspension in water and to flow downslope with the runoff,where it is deposited either in the shrubby patches or intershrub spacesexcluding routes. Reoccurrence of these processes depletes the fine

dex (1 through 6: the higher the index, the greater the stability); stable aggregate contentpH.

rsion index Hygroscopic moisture Electrical conductivity pH

0.0001 0.9424 0.01019) 3.07a (0.10) 702.1a (36.0) 7.67b (0.06)1) 2.68b (0.06) 706.0a (47.4) 7.88a (0.07)

level according to Tukey'sHSD. Numberswithin parentheses are standard error (SE) of the

Table 2Effect of hillside aspect on soil contents of clay, silt, and sand (in %).

Clay Silt Sand

P value 0.9023 0.0006 0.0001North 27.3a (0.9) 49.6a (0.9) 23.1b (0.6)South 27.2a (0.6) 45.8b (0.7) 27.0a (0.5)

Notes:Meanswithin the same column followed by a different letter differ at the 0.05prob-ability level according to Tukey's HSD. Numbers within parentheses are standard error(SE) of the means.

Table 4Effect of microhabitat (type of cover) on the soil contents of clay, silt, and sand (in %).

Clay Silt Sand

P value 0.0814 0.0846 0.0522Shrub patches 27.2a (0.6) 48.8a (0.8) 23.9b (0.8)Intershrub spaces 28.8a (0.6) 46.0a (0.7) 25.3ab (0.8)Trampling routes 25.9a (1.3) 48.2a (1.5) 26.0a (0.7)

Notes:Meanswithin the same column followedbya different letter differ at the 0.05prob-ability level according to Tukey's HSD. Numbers within parentheses are standard error(SE) of the means.

16 I. Stavi et al. / Geomorphology 234 (2015) 11–18

mineralmaterial in the route's surface and increases its accumulation inthe remainder of themicrohabitats. At the same time, the surface rough-ness in the intershrub spaces induced by the herbaceous vegetation androck fragment cover, and shrubby patches increases the sinking capacityof water, mineral material, and coarse organic matter in these micro-habitats. These processes stimulate vegetation growth, acceleratingthe retention of the self- (on-site originated) and imported- (off-siteoriginated) resources in the vegetated microhabitats, and negating therestoration of soil quality and production capacity in the tramplingroutes. Reoccurrence of concentrated livestock traffic in the routes, asopposed to the sporadic trampling in the vegetated microhabitats, fur-ther accelerates these feedbacks.

A study implemented in a protected landscape area in the CzechRepublic reported that geodiversity, including (macro-)topographicvariability and relief heterogeneity, positively affected plant taxon rich-ness (Jačková and Romportl, 2008). Recently, a geodiversity index wasdeveloped, enabling comparison among different sites. This index con-sidered the number of physical (including geologic, geomorphic, hydro-logic, and pedogenic) elements involved in the studied site, the surfacearea of the studied site (to a km2 scale), and the roughness of the unit.Yet, this index could not be utilized for smaller-sized aerial units, andis not applicable for determining smaller-scale geodiversity (Cañadasand Flaño, 2007). Moreover, even though geodiversity studies generallyconsider soils, only rarely do they relate their specific features togeodiversity (Ibáñez et al., 2012), mainly focusing on the backgrounddata, such as geology and topography. In our study region, surface het-erogeneity (or diversity) was previously suggested to be reflectedthrough the sharpening of the hillslopes' micro-topographic (to a scaleof several decimeters) step-like profile. This effect was proposed to beassociated with the livestock trampling routes, increasing the disconti-nuity of geomorphic processes, and affecting redistribution of waterand soil resources at the patch- and hillslope-scales (Stavi et al.,2012). It therefore seems that while geological background is promi-nent in determining geodiversity at the macro, landscape scale, the ef-fect of livestock grazing is particularly considerable at the patch scale.

For summarizing the impact of livestock grazing on the rangelandecosystems, we propose a conceptual model which describes the rela-tionship between the stocking rate and each of the patch-scalegeodiversity, the ecosystem self-organization, and the ecosystemhealth(Fig. 6). A long-term moderate stocking rate increases the patch scalegeodiversity, from a two-phase into a three-phase geo-ecosystem, andmodifies the ecosystem's self-organization — from a bi-modal(e.g., Rietkerk et al., 2002) to a tri-modal pattern. It is suggested thatthis new state of (dynamic) equilibrium increases the redistribution ofwater and soil resources at the patch scale. Regardless, the greater

Table 3Effect ofmicrohabitat (type of cover) on the soil's infiltration rate (cm s−1); aggregate stability i(%); clay dispersion index, hygroscopic moisture content (%); electrical conductivity (μS), and

Infiltration rate Stability index Stable aggregate Clay

P value 0.3327 0.109 0.0001 0.017Shrub patches 0.00059a (0.00017) 5.80a (0.07) 77a (0.02) 0.37bIntershrub spaces 0.00064a (0.00023) 5.53a (0.12) 68b (0.03) 0.37bTrampling routes 0.00052a (0.00015) 5.43a (0.16) 55c (0.02) 0.83a

Notes:Meanswithin the same column followed by a different letter differ at the 0.05 probabilitymeans.

geodiversity is proposed to support a wider range of biological species(biodiversity) and activities, improving ecosystem health (see:Rapport et al., 2013). At the same time, it could be assumed that the im-pact of livestock grazing on geodiversity is directly dependent on thestocking rate. In this regard, it is assumed that an excessively high live-stock rate (over-grazing) diminishes geodiversity andmodifies the eco-system into a one-phase form, being either exposed of vegetation(Gamoun et al., 2010) or fully covered with woody vegetation whichis not edible for grazing animals (Schlesinger et al., 1990). Consequent-ly, the ecosystem's self organization is lost and the landform functioningis modified. In the event of the one-phase form consisting of only ex-posed surface, the ecosystem functioning becomes considerably de-graded, as the leaking of water and soil resources off the ecosystemboundaries becomes the most prominent process. As opposed to that,if the one-phase form consists of full cover of woody vegetation, theretaining capacity of water and soil resources within the ecosystemboundaries becomes considerably large, augmenting the ecosystemfunctioning. One way or another, being either exposed or fully coveredwith woody vegetation, the smaller geo- and bio-diversity results in thedegradation of ecosystem health. Also, as shown by Schlesinger et al.(1990), the economic usability of rangelands that become fully coveredwith woody vegetation is lost. Regardless, in the event of edible woodyvegetation, a new state of a two-phase ecosystem may be formed,consisting of shrubby patches and exposed intershrub spaces. Despitepossibly being efficient in resource conservation, the species diversityand ecosystem health are expected to become degraded under thisnew two-phase form.

Unlike biodiversity evaluation, standardizedmethods for evaluatinggeodiversity have yet to be established (Jačková and Romportl, 2008).This study revealed that suchmethods are specifically absent for the as-sessment of small-scale geodiversity. Particularly, for better under-standing the impact of livestock on patch-scale geodiversity andecosystem self-organization of mosaic-like patterned rangelands, addi-tional studies are needed to examine the actual effects of differentstocking rates. This could be implemented by using livestock enclosuresand applying several grazing regimes, comparing them to long-termgrazing exclusion plots as a reference treatment. In addition, so far,geodiversity studies are almost absent in mathematical models of vege-tation patterns (e.g., Borgogno et al., 2009).

One possible way to quantify the three-phase mosaics is by modify-ing the model proposed by Kéfi et al. (2010), which described the for-mation of vegetation patterns in water-limited environments. In thismodel, the pattern-forming feedback is based on the infiltration con-trast between vegetated and bare-soil domains, which is dictated by

ndex (1 through 6: the higher the index, the greater the stability); stable aggregate contentpH.

dispersion index Hygroscopic moisture Electrical conductivity pH

7 0.0001 0.2583 0.7856(0.11) 3.06a (0.08) 701.3a (23.9) 7.75a (0.06)(0.10) 2.81b (0.07) 682.5a (36.8) 7.79a (0.05)(0.17) 2.63c (0.06) 783.8a (72.8) 7.76a (0.06)

level according to Tukey'sHSD. Numberswithin parentheses are standard error (SE) of the

Table 5Effect of the interaction between hillside aspect and microhabitat (type of cover) on thesoil's silt and sand contents (in %).

Silt Sand

P value 0.0017 0.0059North aspect × shrub patches 48.1abc (1.1) 23.5b (1.4)North aspect × intershrub spaces 48.4abc (0.9) 22.6b (1.1)North aspect × trampling routes 52.4a (2.3) 23.1b (0.7)South aspect × shrub patches 49.5ab (1.2) 24.4b (0.8)South aspect × intershrub spaces 43.6c (0.5) 28.0a (0.5)South aspect × trampling routes 44.1bc (1.3) 28.7a (0.6)

Notes:Meanswithin the same column followed by a different letter differ at the 0.05prob-ability level according to Tukey's HSD. Numbers within parentheses are standard error(SE) of the means.

Fig. 6. Conceptual model of the effects of livestock rate on the rangeland ecosystems'geodiversity, self-organization, and health: low livestock rate has no effect on the existingtwo-phase system, which supports the bi-modal self-organization, characterized by a fairstate of health;moderate livestock ratemodifies the geodiversity to a three-phase pattern,resulting in the formation of a tri-modal self-organization,which is characterized by a highstate of health; high livestock rate eliminates the vegetation cover, resulting in the loss ofself-organization and the state of poor ecosystemhealth. *Note: in specific occasions, a fullcover of inedible woody vegetation could be formed, but yet, the overall health of suchecosystems would be rather low.

17I. Stavi et al. / Geomorphology 234 (2015) 11–18

the parameter α that stands for maximum soil water infiltration (see:Kéfi et al., 2010; Yizhaq et al., 2014). The concept is to define the tram-pling routes with lower α values than the background. Figs. 7 and 8show the vegetation biomass and the soil water distribution, respective-ly, for a domain with five trampling routes and under four different αvalues. The greater the α value, the larger the effect of trampling routes.The three-phase mosaic can be easily observed for the soil water distri-bution, where the lowest values exist in the trampling routes, interme-diate values in the bare soil, and the highest values in the vegetationpatches. The trampling routes act as a strong source for the water, in-creasing its redistribution, and augmenting the provision of water forthe nearby vegetation patches. Yet, it should be emphasized that thismodel is of a simple nature, and has to be thoroughly elaborated inorder to more precisely describe the role of trampling routes in vegeta-tion pattern formation. Regardless, future efforts shouldmodel the rela-tions among patch-scale geodiversity, self-organization, and ecosystemhealth in water-limited environments. Moreover, for wider verificationof the concept of three-phasemosaics, similar studies have to be imple-mented in additional semi-arid rangelands around the world.

4. Conclusions

This study highlighted the role of livestock trampling routes in deter-mining geodiversity at the patch scale of semi-arid rangelands. The inten-sive trampling along certain trails modifies their physical characteristics,making them optimal source areas of resources. These are accumulated

Fig. 5. Soil quality continuum (in grey) and feedback relations (in black) at the patch scale,by type of surface cover.

in the remainder of the intershrub spaces and in the shrub patches,which act as sink of these resources. Concordant with the modificationsin the physical characteristics of the routes, their chemical and bio-chemical characteristics are also modified. The resultant increasedgeodiversity of the hillslopes considerably regulates the spatial distribu-tion of vegetation and modifies the functioning of the rangeland geo-ecosystem. Unlike the common perception of bi-modal self-organizationpatterns, such rangelands encompass tri-modal patterns, resulting ingreater ecosystem health. Yet, compared to previous studies, the presentstudy suggests that even the consideration of such shrublands as tri-

Fig. 7.Numerical simulation of biomass density (g m−2) in a unit area with five tram-pling routes, applied to the model by Kéfi et al. (2010) and by Yizhaq et al. (2014).Panels a, b, c, and d correspond to different values of α in the trampling routeswhich is the maximum soil water infiltration, α = {0.2, 0.18, 0.12, 0.06 d−1} re-spectively. The precipitation rate (R) is 1.56 mm d−1 and the spatial domain is50 × 50 m. All other parameters (see Kéfi et al., 2010 for the model details) are identical

in all panels and are given by: c ¼ 10; gmax ¼ 0:05mm−1m−2; k1 ¼ 5 mm; d ¼ 0:25d−1

; k2 ¼ 5grm−2; W0 ¼ 0:2; rw ¼ 0:2d−1; i0 ¼ 0:06d−1

; Dp ¼ 0:005m2d−1; Dw ¼ 0:1

m2d−1; Ds ¼ 25m2d−1.

Fig. 8. Soil water distribution (in mm) for the simulation described in Fig. 7. Panels a, b, c,and d correspond to different values of α in the trampling routes which is the maximumsoil water infiltration, α = {0.2, 0.18, 0.12, 0.06 d−1} respectively. The three-phase pat-tern is clearly visible for the lower values of alfa: the soil–water content is minimal atthe trampling routes, maximal under the vegetation patches, and intermediate at thebare soil.

18 I. Stavi et al. / Geomorphology 234 (2015) 11–18

modal ecosystems may be an over-simplification, demonstrating the in-herent complexity of the functioning of semi-arid rangelands.

Acknowledgments

The authors are grateful for ICA in Israel (the JCA Charitable Founda-tion), for participating in funding this study. The authors kindly ac-knowledge Professor Xulong Wang and two additional anonymousreviewers for their very helpful comments on a previous version ofthis manuscript.

References

Andreu, V., Imeson, A.C., Rubio, J.L., 2001. Temporal changes in soil aggregates and watererosion after a wildfire in a Mediterranean pine forest. Catena 44, 69–84.

Archer, N.A.L., Quinton, J.N., Hess, T.M., 2002. Below-ground relationships of soil texture,roots and hydraulic conductivity in two-phase mosaic vegetation in South-eastSpain. J. Arid Environ. 52, 535–553.

Bitan, A., Rubin, S., 1991. Climatic Atlas of Israel for Physical and Environmental Planningand Design. Department of Geography, Tel Aviv University, Tel Aviv, and Israel Mete-orological Service, Ministry of Transport, Bet Dagan.

Bochet, E., García-Fayos, P., 2004. Factors controlling vegetation establishment and watererosion on motorway slopes in Valencia, Spain. Restor. Ecol. 12, 166–174.

Borgogno, F., D'Odorico, P., Laio, F., Ridolfi, L., 2009. Mathematical models of vegetationpattern formation in ecohydrology. Rev. Geophys. 47, RG1005. http://dx.doi.org/10.1029/2007RG000256.

Bouyoucos, G.J., 1962. Hydrometer method improved for making particle size analyses ofsoils. Agron. J. 54, 464–465.

Cañadas, S.E., Flaño, R.P., 2007. Geodiversity: concept, assessment and territorial applica-tion. The case of Tiermes–Caracena (Soira). Bol. Asoc. Geógrafos Esp. 45, 389–393.

Carter, A.J., O'Connor, T.G., 1991. A two-phase mosaic in a savanna grassland. J. Veg. Sci. 2,231–236.

Coughenour, M.B., 1991. Spatial components of plant–herbivore interactions in pastoral,ranching, and native ungulate ecosystems. J. Range Manag. 44, 530–542.

Dan, J., Koyumdjisky, H., 1979. Israel soil classification. Special Publication No. 137.Volcani Center, Bet Dagan, Israel (in Hebrew).

Gamoun, M., Tarhouni, M., Belegacem, A.O., Hanchi, B., Neffati, M., 2010. Effect of grazingand trampling on primary production and soil surface in North African rangelands.Ekol. Bratislava 29, 219–226.

Garner, W., Steinberger, Y., 1989. A proposed mechanism for the formation of “fertileislands” in the desert ecosystem. J. Arid Environ. 16, 257–262.

Gray, M., 2004. Geodiversity: Valuing and Conserving Abiotic Nature. John Wiley & Sons,Chichester, p. 448.

Gray, M., 2005. Geodiversity and geoconservation: what, why, and how? In: Santucci, V.L.(Ed.), Geodiversity & Geoconservation. The George Wright Forum 22, pp. 4–12.

Herrick, J.E., Whitford, W.G., de Soyza, A.G., Van Zee, J.W., Havstad, K.M., Seybold, C.A.,Walton, M., 2001. Field soil aggregate stability kit for soil quality and rangelandhealth evaluations. Catena 44, 27–35.

Huston, M.A., 1994. Biological Diversity — The Coexistence of Species. Cambridge Univer-sity Press, Cambridge, MA, p. 704.

Ibáñez, J.J., Krasilnikov, P.V., Saldañ, A., 2012. Archive and refugia of soil organisms: apply-ing a pedodiversity framework for the conservation of biological and non-biologicalheritages. J. Appl. Ecol. 49, 1267–1277.

Imeson, A.C., Prinsen, H.A.M., 2004. Vegetation patterns as biological indicators for iden-tifying runoff and sediment source and sink areas for semi-arid landscapes in Spain.Agric. Ecosyst. Environ. 104, 333–342.

Istanbulluoglu, E., Yetemen, O., Vivoni, E.R., Gutiérrez-Jurado, H.A., Bras, R.L., 2008. Eco-geomorphic implications of hillslope aspect: Inferences from analysis of landscapemorphology in central New Mexico. Geophys. Res. Lett. 35, L14403. http://dx.doi.org/10.1029/2008GL034477.

Jačková, K., Romportl, D., 2008. The relationship between geodiversity and habitat rich-ness in Šumava National Park and Křivoklátskopla (Czech Republic): a quantitativeanalysis approach. J. Landsc. Ecol. 1, 23–38.

Kéfi, S., Eppinga, M.B., de Ruiter, P.C., Rietkerk, M., 2010. Bistability and regular spatial pat-terns in arid ecosystems. Theor. Ecol. 3 (422), 257–269.

Ludwig, J.A., Tongway, D.J., Marsden, G., S.G., 1999. Stripes, strands or stipples: modellingthe influence of three landscape banding patterns on resource capture and productiv-ity in semi-arid woodlands, Australia. Catena 37, 257–273.

McLean, E.O., 1982. Soil pH and lime requirements. In: Page, L.A. (Ed.), Methods of SoilAnalysis, Part 2: Chemical and Microbiological Properties. Am. Soc. Agron., Madison,pp. 199–224.

Merino-Martin, L., Breshears, D.D., Moreno-de las Heras, M., Villegas, J.C., Perez-Domingo,S., Espigares, T., Nicolau, J.M., 2012. Ecohydrological source–sink interrelationshipsbetween vegetation patches and soil hydrological properties along a disturbance gra-dient reveal a restoration threshold. Restor. Ecol. 20, 360–368.

Pemberton, M., 2007. A Brief Consideration of Geodiversity and Geoconservation. Depart-ment of Primary Industries and Water, Tasmania (http://www.proceedings.com.au/quarrying2007/papers/paper_pemberton.pdf).

Perevolotsky, A., Landau, Y., 1988. Improvement and development of the Bedouin stockindustry in the Northern Negev — Lehavim demonstration farm. Professional Report1982–1988. Department of Natural Resources, Institute of Field and Garden Crops.Agricultural Research Organization, Bet Dagan (in Hebrew).

Rapport, D.J., Fayfe, W.S., Costanza, R., Spiegel, J., Yassi, A., B hm, G.M., Patil, G.P., Lannigan,R., Anjema, C.M., Whitford, W.G., Horwitz, P., 2013. Ecosystem health. Definitions, as-sessment, and case studies. Encyclopedia of Life Support Systems, Ecology vol. II(http://www.eolss.net/sample-chapters/c09/e6-70-08-06.pdf).

Reid, R.S., Galvin, K.A., Kruska, R.S., 2008. Global significance of extensive grazing landsand pastoral societies: an introduction. In: Galvi, K.A., Reid, R.S., Behnke, R.H.,Hobbs, N.T. (Eds.), Fragmentation in Semi-arid and Arid Landscape: Consequencesfor Human and Natural Landscapes. Springer, Dordrecht, pp. 1–24.

Richards, L.A., 1954. Diagnosis and improvement of saline and alkali soils. USDA Agricul-tural Handbook 60. United States Department of Agriculture, Washington D.C.

Rietkerk, M., Boerlijst, M.C., van Langevelde, F., HilleRisLambers, R., van de Koppel, J.,Kumar, L., Prins, H.H.T., de Roos, A.M., 2002. Self-organization of vegetation in aridecosystems. Am. Nat. 160, 521–530.

Rigg, L., 1993. Study of Soil Properties Along a Hillslope in Parson's Parcel, Boulder, CO.University of Colorado, Boulder, CO (https://www-static.bouldercolorado.gov/docs/347_Rigg_Lesley_Study-1-201307151559.pdf).

Saco, P.M., Willgoose, G.R., Hancock, G.R., 2007. Eco-geomorphology of banded vegetationpatterns in arid and semi-arid regions. Hydrol. Earth Syst. Sci. 11, 1717–1730.

SAS Institute, 1990. SAS/STAT User's Guide. Version 6. 4th ed. SAS Inst., Cary, NC.Schlesinger, W.H., Reynolds, J.F., Cunningham, G.L., Huenneke, L.F., Jarrell, W.M., Virginia,

R.A., Whitford, W.G., 1990. Biological feedbacks in global desertification. Science 247,1043–1048.

Shoshany, M., 2002. Landscape fragmentation and soil cover changes on south- andnorth-facing slopes during ecosystems recovery: an analysis frommulti-date air pho-tographs. Geomorphology 45, 3–20.

Stavi, I., Lavee, H., Ungar, E.D., Sarah, P., 2012. Grazing-inducedmodification of a semi-aridrangeland from a two-phase to a three-phase mosaic geo-ecosystem. Arid Land Res.Manag. 26, 79–83.

Tongway, D.J., Ludwig, J.A., 2003. The nature of landscape dysfunction in rangelands. In:Ludwig, J.A., Tongway, D.J., Freudenberger, D., Noble, J., Hodgkinson, K. (Eds.), Land-scape Ecology Function and Management. CSIRO Publishing, Canbbera, pp. 49–61.

USDA-NRCS, Dispersive clay soils. EFH NOTICE 210-WI-62. https://prod.nrcs.usda.gov/Internet/FSE_DOCUMENTS/nrcs142p2_024606.pdf.

Vásquez-Méndez, R., Ventura-Ramos, E., Oleschko, K., Hernandez-Sandoval, L. Parrot, JF,Nearing, M.A., 2010. Soil erosion and runoff in different vegetation patches fromsemiarid Central Mexico. Catena 80, 162–169.

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

Zhang, J.H., Wang, Y., Zhang, Z.H., 2014. Effect of terrace forms on water and tillage ero-sion on a hilly landscape in the Yangtze River Basin, China. Geomorphology 216,114–124.