Research Articles

Long-term passive restoration of severely degraded drylands - divergent impacts on soil and vegetation: An Israeli case study

  • Ilan STAVI , 1, 2 ,
  • Eli ARGAMAN 4
  • 1. Dead Sea and Arava Science Center, Yotvata 88820, Israel
  • 2. Eilat Campus, Ben-Gurion University of the Negev, Eilat 88100, Israel
  • 3. Geo-Environmental Research Group, University of Extremadura, Cáceres 10071, Spain
  • 4. Soil Erosion Research Station, Ministry of Agriculture & Rural Development, POB 30, Bet Dagan 50250, Israel

Ilan Stavi, E-mail:

Received date: 2022-05-07

  Accepted date: 2022-10-30

  Online published: 2023-03-21


Land degradation affects extensive drylands around the world. Due to long-term misuse, the Israeli Sde Zin dryland site has faced severe degradation. The study objective was to assess the feasibility of passive restoration in recovering the site. The study was conducted in four land-units along a preservation-degradation continuum: (1) an area that has not faced anthropogenic disturbances (Ecological land); (2) an area that was proclaimed as a national park in the 1970s (Rehabilitation); (3) an area that was prone, until recently, to moderate anthropogenic pressures (Triangle); and (4) a dirt road that was subjected to long-term off-road traffic (Dirtroad). Soil was sampled and analyzed for its properties. The soil physical quality followed the trend of Ecological land > Rehabilitation > Triangle > Dirtroad. Specifically, high soil salinity in the latter three land-units is attributed to long-term erosional processes that exposed the underlying salic horizons. Herbaceous and shrubby vegetation cover was also monitored. The herbaceous vegetation cover followed the trend of Ecological land (86.4%) > Rehabilitation (40.3%) > Triangle (26.2%) > Dirtroad (2.1%), while the shrubby cover was 2.8% in the Ecological land-unit, and practically zero in the other land-units. It seems that despite the effectiveness of passive restoration in recovering the soil’s physical properties, the recovery of vegetation is limited by the severe soil salinity.

Cite this article

Ilan STAVI , Manuel PULIDO FERNÁNDEZ , Eli ARGAMAN . Long-term passive restoration of severely degraded drylands - divergent impacts on soil and vegetation: An Israeli case study[J]. Journal of Geographical Sciences, 2023 , 33(3) : 529 -546 . DOI: 10.1007/s11442-023-2095-9

1 Introduction

Processes of land degradation are prevalent worldwide. Often, causes of land degradation are anthropogenic, and are attributed to the misuse of lands or the implementation of irrational management practices (Morales and Zuleta, 2020). Among the land degradation processes, soil erosion, organic carbon depletion, deterioration of the soil physiochemical quality, soil salinization, reduction in primary productivity, biodiversity loss, and species invasion are predominant (Olsson et al., 2019). Climatic changes, characterized by reduced precipitation, increasing temperatures, long-term droughts, and surges in extreme weather conditions, accelerate the risk of global land degradation, particularly in the world’s drylands (Arneth et al., 2019). Several models have projected degrading climatic conditions of the world’s drylands throughout the 21st century, with the growing frequency and magnitude of droughts and increase in regional aridity on the one hand (Cook et al., 2014; Fu and Feng, 2014; Lickley and Solomon, 2018), while facing rare but intense rainstorms and devastating floods on the other hand (Care Danmark, 2016). Coupled with the forecasted expansion of dryland areas, global land degradation and desertification are expected to further accelerate in the future (Huang et al., 2016).
Passive restoration of degraded lands halts the anthropogenic utilization of the target land, thereby allowing self-restoration processes to take place. The most common passive means is fencing of plots to prevent access by humans or livestock (Morrison and Lindell, 2011; Zahawi et al., 2014). It is expected that once the degrading factor is halted, the physiochemical and biotic components of the target land will gradually recover, and foster each other through positive feedbacks (Aradottir and Hagen, 2013). Such passive schemes are different from active restoration programs, where practices such as tillage (Stavi et al., 2018c), application of soil amendments such as biochar (Stavi, 2012), terracing for runoff harvesting and soil erosion control (Bolo et al., 2019), direct seeding of annual vegetation (Louhaichi et al., 2020), or planting of shrubs (Haddad et al., 2022) and trees are implemented (Baumgärtner, 2012; Mahmoud et al., 2021). Usually, passive restoration schemes are more likely to be successful in sites where degradation processes are not severe, and where climatic conditions can easily support the rehabilitation of ecosystem. At the same time, in extremely degraded sites that cross certain tipping points, or in drylands and other resource-limited areas, active restoration schemes might be preferred (McIver and Starr, 2001; Rohr et al., 2018). Despite this extensive research, knowledge gaps still exist regarding restoration procedures of severely degraded drylands, and specifically, regarding the use of passive schemes.
Across the Israeli Negev drylands, land degradation processes have been generally attributed to anthropogenic activities. This study was conducted in the Sde Zin site - located in the arid central Negev of Israel - that has faced severe degradation processes since the mid 20th century. These degradation processes are attributed to land-use change, under which ‘natural’ lands have been converted to anthropogenic ones, and underwent irrational grazing pressures, long-term irrigated vegetable cropping, heavy traffic by off-road vehicles, and high-pressure outdoor recreation activities (Stavi et al., 2018c). While one certain land unit of the site is still considered ‘natural’, another land unit was proclaimed as a national park in the early 1970s, and two other land units - that were prone over the long run to high and extremely high pressures - were excluded from any anthropogenic use in 2018. Therefore, the study site displays a sequence of passive restoration land units, representing a continuum of preservation-degradation. The objective of this study was, therefore, to assess the soil physiochemical quality and vegetation cover over the land preservation-degradation continuum, in order to determine the feasibility of the passive restoration efforts in rehabilitating the land’s geo-ecosystem functioning. It was hypothesized that despite the relatively severe degradation of the study site, the long-term passive restoration scheme (through the halt of anthropogenic disturbances) has allowed self-restoration processes of soil and vegetation to take place, and subsequently, geo-ecosystem functioning has improved.

2 Materials and methods

2.1 Regional settings

The study was conducted in Sde Zin (30º86'N, 34º79'E, 475 m. a.s.l.), which is located in the arid central Negev of southern Israel (Figure 1). The site is comprised of a plateau landform (with a mean slope incline of 1.0%), surrounded by rolling hills. The lithology is comprised of Dead Sea Group’s conglomerate, and the soil series is classified as calcic xerosol (see: FAO, 2015). Long term mean daily temperatures in the coldest and warmest months are 9ºC and 24ºC, respectively, and mean cumulative rainfall depth is 93 mm/year (Israel Meteorological Service website: Across the region, the predominant shrubs encompass the perennial plant species Haloxylon scoparium (Pomel) Iljin (previously named as Hammada scoparia (Pomel) Iljin). Since the 1950s, the Sde Zin site has been prone to high anthropogenic pressures imposed by the establishment of the nearby settlement (Kibbutz Sde Boker) and specifically, to intensive vegetable cropping, heavy livestock grazing, and of off-road vehicles traffic.
Figure 1 A map of Israel, with an extension of the study site

2.2 Study design, soil sampling, and laboratory work

The study encompassed four different land units along a land preservation-degradation continuum. These included: (1) an undisturbed ‘natural’ area, which has not faced any modern anthropogenic disturbances such as intensive agriculture, heavy livestock grazing, or off-road vehicle traffic (although it possibly faced some low-pressure grazing under traditional livestock raising regimes until the early 1950s. Entitled as Ecological (mean slope incline 1.1%): Figure 2a); (2) an area that was proclaimed as a national park in the early 1970s, and since then, is under protection from any human activities (Rehabilitation (mean slope incline 1.2%): Figure 2b); (3) an area that was prone, over decades, to livestock grazing, as well as to moderate pressure by hikers, mountain bikers, and off-road vehicles (Triangle (mean slope incline 0.8%), named after the shape of this land unit: Figure 2c); and a wide (~10 m width) dirtroad that was aimed, over the same temporal frame, to funnel the off-road vehicles’ traffic across the region (Dirtroad (mean slope incline 0.8%): Figure 2d). All of the four land units are situated in the plateau landform. In 2018, both the Triangle and Dirtroad land units were surrounded by boulders in order to prevent any further access by off-road vehicles. Therefore, in terms of land status along the preservation-degradation continuum, it is assumed that the Ecological land unit is the most preserved, the Rehabilitation land unit is the second most preserved, the Triangle land unit is the second most degraded, and the Dirtroad land unit is the most degraded. In summer 2020, we delineated five blocks, each containing one 25 m2 (5 × 5 m) plot, in each of the four land units. Depending on the local settings, the distance between two adjacent plots was set to at least 50 m.
Figure 2 Characterizing views of the Ecological land (a), Rehabilitation (b), Triangle (c), and Dirtroad (d) land units. Note the dense (shrubby and herbaceous) vegetation cover in the Ecological land unit, and the developed mechanic crust covering the Dirtroad land unit
Soil monitoring and sampling was conducted in five randomly selected spots in each plot. Soil penetration resistance was monitored on-site using a dynamic penetrometer (Vanags et al., 2004). The number of spots (n) in which we assessed the soil penetration resistance was 4 (land units) × 5 (blocks) × 5 (spots) = 100. Later, in each spot, the soil was sampled from two depths, 0-5 and 5-10 cm. These depths were selected based on several previous studies that were implemented across the region, demonstrating the predominant pedogenic processes in these layers. The number of soil samples (n) was 4 (land units) × 5 (blocks) × 5 (spots) × 2 (depths) = 200. For a schematic illustration of the study design, see Figure 3.
Figure 3 Schematic illustration of the study design
The soil samples were taken to the laboratory to analyze their physical and chemical properties, including gravimetric moisture content (Gardner, 1965), texture (Bouyoucos, 1962), stable aggregate content (using an aggregate stability apparatus: Eijkelkamp®, the Netherlands), slaking index (Herrick et al., 2001), clay dispersion index (USDA-NRCS, EFH Notice 210-WI-62), calcium carbonate content (Loeppert and Suarez, 1996), pH (McLean, 1982), and electrical conductivity (Richards, 1954).

2.3 Ground surface mapping

Ground surface mapping was conducted using an unmanned aerial vehicle (UAV), acquired in June 2020, using a Mavic Pro quadcopter equipped with a 12.35 MP RGB camera (DJI technology Co., Shenzen, China) operated by DJI Ground Station Pro® ( In total, 325 images were obtained from an altitude of 75 m above ground level, with an overlap of 85% for photogrammetric processing by Agisoft Metashpre Pro® (version 1.6.3, Post-processing of the UAV imageries resulted in a surface resolution of 3.0 cm px‒1 of the orthomosaics and 6.0 cm px‒1 of the digital elevation model (DEM). Geospatial referencing of the produced orthomosaics and DEMs was done using 21 ground control points (GCP), which were selected based on high-resolution orthophoto (0.125 m px‒1). Land cover classes were subdivided into the following five pixel based categories: (1) bare soil; (2) sparse annuals; (3) medium-density annuals; (4) dense annuals; and (5) shrubby vegetation. Land unit classes were classified using multiple standard supervised classification algorithms, including [1] Maximum Likelihood (ML) (Figures 4a-4d); [2] Mahalanobis Distance (MhD); [3] Minimum Distance (MD); [4] Parallelepiped (PP); and [5] Spectral Information Divergence (SID). Due to the high spatial resolution of the produced UAV orthophotos, post-classification results were aggregated for adjacent 16 pixels to reduce misclassified cells, while smoothing the boundaries between classes, and identifying clusters with the most frequent neighboring cells.
Figure 4a UAV orthophoto of the Ecological land unit (upper right panel), and extensions of three representing plots, each with a matching aerial image at the left side and the Maximum Likelihood classification product at the right side (the rest of the three panels)
Figure 4b UAV orthophoto of the Rehabilitation land unit (upper right panel), and extensions of three representing plots, each with a matching aerial image at the left side and the Maximum Likelihood classification product at the right side (the rest of the three panels)
Figure 4c UAV orthophoto of the Triangle land unit (upper right panel), and extensions of three representing plots, each with a matching aerial image at the left side and the Maximum Likelihood classification product at the right side (the rest of the three panels)
Figure 4d UAV orthophotos of the Dirtroad land unit (right column), and the corresponding extensions of four representing plots along it, each with a matching aerial image at the left side and the Maximum Likelihood classification product at the right side (left column)
Calculation of the obtained data for the land units generated an overall classification accuracy for each class and for each classification method. Then, this data was visually validated using the orthophoto, as well as against known field points used for the accuracy analysis. A confusion matrix representing the overall accuracy and the kappa coefficient of agreement (Cohen, 1960) was calculated using ENVI version 5.6 (Exelis Visual Information Solutions, Boulder, Colorado). The kappa coefficient varies between 0-1, where 0 represents no agreement and 1 represents perfect agreement, compared to the generated classification maps.

2.4 Statistical analysis of soil data

Data on soil properties was processed, assessed for normal distribution, and then statistically analyzed through the Analysis of Variance (ANOVA, by the GLM procedure: SAS Institute, 1990). The model for analysis of the soil penetration resistance data was: land unit (3 df), and plot within land unit (4 df; error term for plot). The model for analysis of the rest of the properties was: land unit (3 df), plot within land unit (4 df; error term for plot); depth (1 df), and the interaction land unit × depth (4 df). For statistically significant interactions, an additional ANOVA was implemented with the GLM’s Slice command. Separation of means was conducted by Tukey’s HSD at a 0.05 probability level.

3 Results and discussion

3.1 Effects on the soil texture

The soil texture varied among the land units (Table 1), and was determined as sandy clay loam in the Ecological land unit, loam in the Rehabilitation land unit, clay loam in the Triangle land unit, and silt loam in the Dirtroad land unit (see the soil texture triangle for a visual demonstration of the different categories: Despite these classifications, textural differences between the Ecological and Rehabilitation land units were comparatively minor. Considering the highest erodibility of the soil’s clay and silt fractions (Parsons et al., 1991; Asadi et al., 2007; Rose et al., 2007), the finer texture of soil in the Dirtroad and Triangle land units compared to the Ecological and Rehabilitation land units is somewhat unexpected, as the assumed considerable erosional processes that occurred over the last decades in the Dirtroad and Triangle land units are expected to preferentially remove the clay and silt fraction.
Table 1 Land unit effect on soil properties
P value Dirtroad Triangle Rehabilitation Ecological land
Clay (%) 0.0918 20.0 a (6.3) 33.8 a (3.6) 23.9 a (2.3) 22.0 a (1.9)
Silt (%) 0.0009 51.3 a (5.4) 33.8 b (2.7) 29.2 b (2.6) 27.1 b (1.3)
Sand (%) 0.0001 28.6 b (2.1) 32.4 b (1.8) 46.9 a (5.0) 50.8 a (3.1)
Soil moisture (%) 0.0001 3.5 a (0.1) 3.6 a (0.1) 3.0 b (0.1) 1.9 c (0.1)
Penetration resistance (MPa) 0.0001 2.66 a (0.00) 1.81 b (0.18) 0.86 c (0.15) 0.81 c (0.13)
Stable aggregates (%) 0.0001 3.2 b (0.5) 5.0 b (0.8) 13.4 a (2.1) 12.6 a (1.8)
Slaking index 0.0001 2.54 a (0.09) 2.28 ab (0.10) 1.98 b (0.16) 1.08 c (0.15)
clay dispersion index 0.0325 4.00 a (0.00) 3.98 ab (0.02) 3.88 ab (0.08) 3.80 b (0.06)
Calcium carbonate (%) 0.087 26.2 a (0.3) 23.9 a (0.3) 27.7 a (0.5) 30.6 a (0.4)
Electrical conductivity (µS/cm) 0.0001 16,332.8 a (474.5) 13,849.9 b (966.4) 4,917.5 c (588.2) 1,321.9 d (235.3)
pH 0.0001 8.20 c (0.01) 8.10 c (0.04) 8.29 b (0.03) 8.44 a (0.03)

Notes: Bold P value indicate a significant effect. Means within the same column followed by a different letter differ at the 0.05 probability level according to Tukey’s Honestly Significant Difference (HSD). Numbers within parentheses are standard error of the means.

Sorting of the finer soil fractions occurs under a wide range of biophysical conditions, where redistribution of mineral material through erosional processes is predominated by selective removal of the lighter particles either by water (Slater and Carleton, 1942) or wind (Lal, 2002).
Yet, this inconsistency can be attributed to the assumed severe soil erosion of these land units over time, which is presumed to be dominated by the transportation of the soil’s coarser fraction (sand). This consists with another study that was implemented in a nearby site, which showed that severe and/or long-term erosion is predominated by rolling processes of medium to large-sized particles (> 150 μm) through interrill erosion, resulting in the removal of coarse fractions and retention of fine fractions (Stavi et al., 2019). This process conforms with similar findings from other studies (Proffitt and Rose, 1991; Durnford and King, 1993; Asadi et al., 2007; Shi et al., 2012).
The soil texture did not significantly change with depth, and mean contents of clay, silt and sand were similar at both depths (0-5 and 5-10 cm depths: Table 2). The effect of the interaction between land unit and depth was not statistically significant for any of the clay, silt, and sand fractions (P = 0.2936, 0.3005, and 0.8611, respectively).
Table 2 Depth effect on soil properties
P value 0-5 cm 5-10 cm
Clay (%) 0.3847 26.6 a (3.1) 23.3 a (3.0)
Silt (%) 0.5198 32.8 a (2.5) 37.8 a (4.4)
Sand (%) 0.5026 40.5 a (3.3) 38.9 a (3.8)
Soil moisture (%) 0.0001 2.6 b (0.1) 3.4 a (0.1)
Stable aggregates (%) 0.2891 9.4 a (1.1) 7.8 a (1.2)
Slaking index 0.3499 2.03 a (0.11) 1.91 a (0.10)
clay dispersion index 0.0913 3.87 a (0.05) 3.96 a (0.02)
Calcium carbonate (%) 0.7586 27.0 a (0.3) 27.1 a (0.4)
Electrical conductivity (µS/cm) 0.0058 8,279.9 b (870.8) 9,931.2 a (731.3)
pH 0.4879 8.26 a (0.02) 8.24 a (0.02)

Notes: Bold P value indicates a significant effect. Means within the same column followed by a different letter differ at the 0.05 probability level according to Tukey’s Honestly Significant Difference (HSD). Numbers within parentheses are standard error of the means.

3.2 Effects on the soil physical quality

Although soil moisture at the hygroscopic level is not accessible for plant uptake, it encompasses a positive indication of the physical quality of soil (Stavi et al., 2018c). Apparently, this seems contradictory as the highest mean soil moisture content was measured in the Dirtroad and Triangle land units, whilst the lowest mean soil moisture content was found in the Ecological land unit (Table 1). A possible explanation to this effect could be that soils covered with developed mechanical crusts face lower evaporation rates from the underlying layers. This effect is attributed to the disruption of capillarity when the crust shrinks, breaks apart, and separates from the underlying soil (Haliburton et al., 1977). This accords with observations across the study site, revealing the predominance of a thick (~3-4 mm thickness) mechanical crust covering the Dirtroad and Triangle land units, as opposed to the comparatively thin (~1 mm thickness) crust cover in the Rehabilitation land unit, and the marginal to non-mechanical crust cover in the Ecological land unit. This mechanism also explains the significantly greater mean soil moisture content in the deeper depth than that in the shallower depth (Table 2). Further, this mechanism is demonstrated by the significant interaction between land unit and depth (P = 0.0263), which revealed the greater mean soil moisture content in the deeper depth than that in the shallower depth within each of the land units (Figure 5).
Figure 5 Effect of the interaction between land unit and depth on the gravimetric moisture content of soil. Notes: error bars are standard error of the means. Bars with different letters differ at 0.05 probability level.
The effect of land unit on the physical quality of soil is further demonstrated by the soil’s mean penetration resistance, which was maximal in the Dirtroad, intermediate in the Triangle, and lowest in the Rehabilitation and Ecological land units (Table 1). The extreme soil compaction of the Dirtroad unit is attributed to the intensive off-road vehicle traffic. The severe compaction is expected to hinder plant root penetrability and growth capacity (Bengough et al., 2011).
Specifically, the physical quality of soil is predominantly determined by its soil aggregates, which are formed by the attractive and disruptive forces that act on the soil particles to cause greater cohesion among some particles, as well as among some groups of particles. Particularly, aggregate stability determines soil erodibility (Nimmo, 2004). The approximately threefold to fourfold greater mean content of stable aggregates in soil of the Rehabilitation and Ecological land units than that of the Triangle and Dirtroad land units (Table 1), exemplifies the higher risk of soil erosion in the latter two land units. Aggregate stability was slightly, though not significantly, greater in the shallower depth than that in the deeper depth (Table 2). The effect of the interaction between land unit and depth was not significant for aggregate stability.
An opposite trend to that of aggregate stability was recorded for the mean slaking index (Table 1). Slaking, a process of aggregate breakdown upon rapid wetting, happens when the pressure of air trapped inside the aggregate exceeds its cohesion capacity, or when differential swelling stresses create planes of weakness within the aggregate (Zaher and Caron, 2008). The trend of clay dispersion (index) - in which soil particles are detached from each other when the soil is wetted (Chibowski, 2011) - was similar to that of the slaking index (Table 1). Together, these soil properties determine the formation of mechanical crust on the ground surface, regulating the water infiltration capacity, and controlling water overland flow and soil erosion (Nimmo, 2004; Zaher and Caron, 2008; Basga et al., 2018).
Despite the differences in soil’s aggregation-related properties among the land units, the mean calcium carbonate content was rather similar for all of them (Table 1). One way or another, the relatively moderate content of calcium carbonate at all land units does not seem to limit microbial activity (Bashan and Vazquez, 2000) or macronutrient availability (Bhargavarami Reddy et al., 2013).
The effect of depth (Table 2), as well as the effect of the interaction between land unit and depth, were not significant for any of the properties of aggregate stability, slaking index, clay dispersion index, and calcium carbonate content.

3.3 Effects on the soil chemical quality

The soil’s mean electrical conductivity clearly indicates the effect land unit has on the soil’s chemical quality. Electrical conductivity - representing the overall soil salinity - followed the trend of Dirtroad > Triangle > Rehabilitation > Ecological land unit. Mean electrical conductivity in the Ecological land unit is similar to the normal salinity level of intact lands across the region (Stavi et al., 2019). However, mean electrical conductivity in the Rehabilitation land unit was almost fourfold greater than that in the Ecological land unit, indicating severe soil salinity. Further, mean electrical conductivity of soil in the Dirtroad and Triangle land units were ~ threefold greater than that in the Rehabilitation land unit, and ~ tenfold greater than that in the Ecological land unit (Table 1). The severe salinity in the Rehabilitation land unit and the extremely severe salinity in the Triangle and Dirtroad land units is assumed to be attributed to the moderate or intense erosion of the non-saline surface horizons, with the resulted in exposure of the underlying salic horizons (personal communications with Yoav Avni). To some extent, this process is similar to the scalding mechanism, where erosion of the soil’s A-horizon - by either wind or water action - exposes the underlying saline layer (Clark, 1985; Queensland Government, 2020). A complementary mechanism could be the respective deteriorated or severely deteriorated hydraulic conductivity of the soil in these land units - as indicated by the concordant degraded or severely degraded physical quality of soil - which is expected to reduce leaching of salts to deeper depths (Stavi et al., 2019) and inhibit plant growth (Stavi et al., 2018b). At the same time, the high physical quality of soil in the Ecological land unit (and to some extent, also the moderate soil quality in the Rehabilitation land unit) enables the effective leaching of salts to deeper layers (Stavi et al., 2019), beyond the reach of plant roots (Stavi et al., 2018b).
One way or another, the soil salinity levels in the Rehabilitation, Dirtroad, and Triangle land units seem to impose a serious risk of physiological drought for plants, negatively impacting vegetation establishment and growth (Stavi et al., 2021a). This conforms with previous studies, which stated that high salinity levels disrupt the uptake of water by plants because of the increased soil-water osmotic potential (McFarlane et al., 2016; Sedaghathoor and Zare, 2019). Also, the dissolved salts might cause the excessive accumulation of ions in the plant tissues, toxifying the plants and resulting in their mortality (Minhas, 1996; Deb et al., 2013).
Unexpectedly, an opposite trend was recorded for soil pH, which followed the trend of Ecological > Rehabilitation > Triangle ≈ Dirtroad. Yet, all measured pH values are considered relatively moderate compared with calcareous dryland soils, and do not limit vegetation growth (Medeiros and Drezner, 2012).
Among the soil chemical properties, the depth effect was significant for the electrical conductivity only, where its value in the deeper depth was ~ 20% greater than that in the shallower depth (Table 2). This effect could be attributed to rainfalls that cause the leaching of salts from the ground surface, which are then precipitated in the deeper depth (Osman, 2012). The effect of the interaction between land-use and depth was not significant for any of the chemical soil properties.

3.4 Effects on vegetation

Large differences in vegetation cover were recorded among the four land units. When assessing the vegetation cover models, the highest classification’s overall accuracy and reliability (kappa coefficient) was found for the ML algorithm (Table 3). According to this algorithm, the shrubby vegetation cover followed the trend of Ecological (2.8%) > Rehabilitation (0.2%) ≈ Triangle (0.1%) > Dirtroad (~0.0%) land unit. The overall herbaceous vegetation cover - including dense, medium, and sparse cover - was 86.4% in the Ecological, 40.3% in the Rehabilitation, 26.2% in the Triangle, and only 2.1% in the Dirtroad land unit. Among these, the cover of dense and medium-density herbaceous vegetation in the Ecological land unit was ~ 5-20 times greater than that in the Rehabilitation and Triangle land units. The sparse herbaceous vegetation cover was similar in the Ecological and Rehabilitation land units, and ~ twice the cover in the Triangle land unit. The cover of bare soil showed an opposite trend, and followed the trend of Dirtroad (97.9%) > Triangle (73.7%) > Rehabilitation (59.5%) > Ecological (10.9%) land unit (Figure 6 and Table 4).
Table 3 Kappa coefficient of agreement and overall accuracy results of types of surface cover, according to the different land units and classification methods
Classification method
Land unit Kappa Acc. [%] Kappa Acc. [%] Kappa Acc. [%] Kappa Acc. [%] Kappa Acc. [%]
Ecological land 0.87 89.93 0.83 87.29 0.63 71.93 0.25 37.34 0.72 79.01
Triangle 0.67 79.87 0.69 78.84 0.66 76.43 0.05 31.88 0.75 83.16
Rehabilitation 0.79 93.46 0.65 81.68 0.55 77.21 0.02 32.27 0.63 81.39
Dirtroad 0.82 93.83 0.47 95.03 0.49 95.55 0.83 99.14 0.49 95.45

Notes: ML - Maximum Likelihood; MhD - Mahalanobis Distance; MD - Minimum Distance; PP - Parallelepiped; SID - Spectral Information Divergence. Kappa values were divided into five categories: [1] 0.01-0.2: slight agreement; [2] 0.21-0.40: Fair agreement; [3] 0.41-0.60: Moderate agreement; [4] 0.61-0.80: Substantial agreement (underlined); and [5] 0.81-1.0: perfect agreement (bold).

Figure 6 Types of ground surface cover according to land unit
Table 4 Surface cover classification’s area and percentage for the Maximum Likelihood (ML) supervised classification
Land unit
Ecological land Triangle Rehabilitation Dirtroad
Surface cover class Area (m2) Cover (%) Area (m2) Cover (%) Area (m2) Cover (%) Area (m2) Cover (%)
Dense annuals 18,650.1 26.9 2,097.7 5.2 1,023.0 1.2 0.0 0.0
Moderate annuals 19,452.6 28.1 1,101.1 2.7 5,496.8 6.1 30.8 2.1
Sparse annuals 21,727.9 31.4 7,451.8 18.3 31,872.6 33.0 0.0 0.0
Shrubby vegetation 1,905.7 2.8 58.3 0.1 107.1 0.2 0.3 0.0
Bare soil 7,542.2 10.9 29,932.9 73.7 59,279.4 59.5 2,212.6 97.9
Total area 69,278.5 100.0 40,641.8 100.0 97,778.8 100.0 2,243.7 100.0
In severely degraded lands, the natural recovery of shrubby vegetation through recruitment processes is expected to be slow, particularly in drylands, where self-restoration processes are primarily limited by the low availability of water. This consists with Qinfeng (2004), who reported for the Sonoran Desert in Arizona, the United States, that the recovery of perennial vegetation in heavily-grazed lands lasted over 50 years. This may explain the marginal shrub cover in the Rehabilitation land unit, which was practically identical or similar that in the Triangle and Dirtroad land units, respectively. Specifically, the marginal shrub cover in these land units could be attributed to the very challenging environmental conditions, which are predominantly determined by the respective severe or extremely severe soil salinity.
At the same time, herbaceous vegetation can recover much faster after eliminating the degrading factor, i.e., negating agricultural activities / intensive livestock grazing / access of off-road vehicles, etc. This explains the considerable herbaceous vegetation cover in the Triangle land unit (after ~ two years of exclosure), as well as the much greater herbaceous vegetation cover in the Rehabilitation land unit (after > 40 years of exclosure). This conforms with Atsbha et al. (2020), who reported a considerable increase in productivity of herbaceous vegetation after five to seven years of grazing exclusion in degraded rangelands of the Ethiopian southern Tigray. A complementary explanation could be the assumed comparatively lower sensitivity of annual vegetation to soil’s high salinity levels. One way or another, the coupling of short-term exclosure (~ two years) with the long-term severe degradation processes in the Dirtroad, explain the extremely low herbaceous vegetation cover in this land unit.

3.5 Insights and implications

In ecological restoration projects, the ultimate goal is to reestablish ecological systems that existed prior to the disturbance (Finch et al., 2016). Further, there is a general consensus that true restoration not only improves the structure or composition of the ecosystem, but also facilitates the recovery of processes needed to sustain the structure or composition for the long-run (McIver and Starr, 2001).
Passive restoration - defined as the removal of stresses that cause the degradation processes - may be appropriate for ecosystems that are relatively slightly impaired (McIver and Starr, 2001). Regardless, the ecological outcomes of passive restoration are often uncertain (Brancalion et al., 2016), as the typical recovery time of such schemes is usually much longer than that needed for active schemes (Zahawi et al., 2014). Therefore, active approaches may be preferable in highly degraded ecosystems (McIver and Starr, 2001). Despite the comparatively higher costs involved in active restoration schemes (Rohr et al., 2018), long-term experience shows that many times, the generated benefits are greater than the related costs (De Groot et al., 2013). This seems to be particularly relevant for resource-limited environments, such as drylands, where primary productivity and other ecosystem functions are primarily determined by the low availability of water (D’Odorico et al., 2007; Schaffer et al., 2018; Lian et al., 2021).
It seems that the effectiveness of the long-term (over 40 years) exclosure of the Rehabilitation land unit is controversial. On the one hand, the soil’s physical properties of this land unit proved to be similar to those of the Ecological land unit, suggesting the effective restoration of soil quality. At the same time, the soil salinity in the Rehabilitation land unit was far beyond normal levels. Nevertheless, the somewhat improved conditions of the Rehabilitation land unit may have provided better habitat settings, which would be reflected by the coinciding increase in plant cover. Yet, despite the several decades of exclosure, the cover of annual herbaceous vegetation in the Rehabilitation land unit was less than half that in the Ecological land unit, and only about one-third greater than that in the Triangle land-unit. More importantly, the shrubby vegetation cover in the Rehabilitation land unit revealed very meager (or practically no) restoration, and was identical or similar to that in the Triangle or Dirtroad land units, respectively. The importance of perennial shrubby vegetation for overall ecosystem quality has been widely acknowledged (e.g., Durán Zuazo et al., 2008; Campanella et al., 2018; Xiao et al., 2019). Specifically, in drylands, shrubs were reported to encompass a crucial component in determining the geo-ecosystem functioning (Stavi et al., 2010; 2018a; 2021b). Therefore, the obtained results only partially accord with the study’s hypothesis.
Insights of this study suggest that in severely degraded lands, or where climatic conditions are relatively harsh, self-restoration processes occur faster in the ecosystem’s soil component than in its vegetal component. This accords with Stavi et al. (2018c), who reported for another (active) restoration project in a nearby site that the soil factor was restored at a more pronounced rate than the vegetation factor. Regardless, it could be expected that annual vegetation cover is predominantly regulated by yearly precipitation. This concurs with Stavi et al. (2018c), who showed that the growth of annual herbaceous vegetation was predominantly determined by yearly rainfall, regardless of the site-specific implementation of (active) restoration means. One way or another, in order to accelerate the recovery rate of the vegetation community in severely degraded drylands, the implementation of active restoration practices should be considered.

4 Conclusions

A passive restoration scheme of a severely degraded dryland, through the long-term exclosure of a target land area, was proved effective in restoring the physical quality of soil. Yet, the soil salinity of the exclosed land unit was still far beyond normal levels, imposing considerable limitations to plant establishment and growth. Specifically, although the long-term exclosure allowed moderate recovery of annual herbaceous vegetation, the perennial shrubby vegetation of this land unit was negligible. Overall, the results question the effectiveness of passive means in enabling self-restoration processes to take place, and in rehabilitating the overall geo-ecosystem functioning of such lands. Therefore, where drylands have faced severe degradation processes, active restoration schemes should be considered.


Fieldwork was funded by the Nature and Parks Authority, and laboratory works was funded by the Israel Science Foundation (ISF) Grant No.602/21. The Dead Sea and Arava Science Center is supported by the Ministry of Science and Technology. The authors gratefully acknowledge Michelle Finzi for proofreading of the manuscript.
Aradottir A L, Hagen D, 2013. Ecological restoration: Approaches and impacts on vegetation, soils and society. Advances in Agronomy, 120: 173-222.

Arneth A et al., 2019. IPCC special report on climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems: Summary for policymakers, approved draft. Intergovernmental Panel on Climate Change, Geneva.

Asadi H, Ghadiri H, Rose C W et al., 2007. An investigation of flow driven soil erosion processes at low stream powers. Journal of Hydrology, 342: 134-142.


Atsbha T, Wayu S, Gebretsadkan N et al., 2020. Exclosure land management for restoration of herbaceous species in degraded communal grazing lands in Southern Tigray. Ecosystem Health and Sustainability, 6: 1-11.

Basga S D, Tsozue D, Temga J P et al., 2018. Land use impact on clay dispersion/flocculation in irrigated and flooded vertisols from Northern Cameroon. International Soil and Water Conservation Research, 6: 237-244.


Bashan Y, Vazquez P, 2000. Effect of calcium carbonate, sand, and organic matter levels on mortality of five species of Azospirillum in natural and artificial bulk soils. Biology and Fertility of Soils, 30: 450-459.


Baumgärtner J, 2012. Reforestation techniques in the Mediterranean:Woody plant propagation and establishment. Bachelor thesis submitted to the University of Applied Sciences, Berlin.

Bengough A E, McKenzie B M, Hallet P D et al., 2011. Root elongation, water stress, and mechanical impedance: A review of limiting stress and beneficial root tip trails. Journal of Experimental Botany, 62: 59-68.


Bhargavarami Reddy C H, Guldekar V D, Balakrishnan N, 2013. Influence of soil calcium carbonate on yield and quality of Nagpur mandarin. African Journal of Agricultural Research, 8: 5193-5196.

Bolo P O, Sommer R, Kihara J et al., 2019. Rangeland degradation: Causes, consequences, monitoring techniques and remedies. Publication No.478. International Center for Tropical Agriculture (CIAT). Nairobi.

Bouyoucos G J, 1962. Hydrometer method improved for making particle size analyses of soils. Agronomy Journal, 54: 464-465.


Brancalion P H S, Schweizer D, Gaudare U et al., 2016. Balancing economic costs and ecological outcomes of passive and active restoration in agricultural landscapes: The case of Brazil. Biotropica, 48: 856-867.


Campanella M V, Rostagno C M, Videla L S et al., 2018. Land degradation affects shrub growth responses to precipitation in a semiarid rangeland of north-eastern Patagonia (Argentina). Austral Ecology, 43: 280-287.


Care Danmark, 2016. Fleeing climate change: Impacts on migration and displacement. Zøbenhavn.

Chibowski E, 2011. Flocculation and dispersion phenomena in soils. In: Gliński J, Horabik J, Lipiec J (eds.) Encyclopedia of Agrophysics. Dordrecht: Springer.

Clark A N, 1985. Longman Dictionary of Geography: Human and Physical. Essex: Longman.

Cohen J, 1960. A coefficient of agreement for nominal scales. Educational and Psychological Measurements, 20: 37-46.


Cook B I, Smerdon J E, Seager R et al., 2014. Global warming and 21st century drying. Climate Dynamics, 43: 2607-2627.


De Groot R S, Blingnaut J, van der Ploeg S et al., 2013. Benefits of investing in ecosystem restoration. Conservation Biology, 27: 1286-1293.


Deb S K, Sharma P, Shukla M K et al., 2013. Drip-irrigated seedlings response to irrigation water salinity. Hortscience, 48: 1548-1555.


D’Odorico P, Caylor K, Okin G S et al., 2007. On soil moisture-vegetation feedbacks and their possible effects on the dynamics of dryland ecosystems. Journal of Geophysical Research, 112: G04010.

Durán Zuazo V H, Rodríguez Pleguezuelo C R et al., 2008. Harvest intensity of aromatic shrubs vs. soil erosion: An equilibrium for sustainable agriculture (SE Spain). Catena, 73: 107-116.


Durnford D, King J P, 1993. Experimental study of process and particle-size distributions of eroded soil. Journal of Irrigation and Drainage Engineering, 119: 383-398.


FAO, 2015. World reference base for soil resources 2014: International soil classification system for naming soils and creating legends for soil maps, update 2015. Food and Agriculture Organization of the United Nations, Rome.

Finch D M et al., 2016. Rangeland drought:Effects, restoration, and adaptation. In: Vose J M et al. (eds.). Effects of Drought on Forests and Rangelands in the United States: A Comprehensive Science Synthesis. Report WO-93b. USDA Forest Service, Washington, DC: 156-194.

Fu Q, Feng S, 2014. Responses of terrestrial aridity to global warming. Journal of Geophysical Research Atmosphere, 119: 2014JD021608.

Gardner W G, 1965. Water content. In: Black C A (ed.). Methods of Soil Analysis, No.9. American Society of Agronomy, Madison, WI.82-127.

Haddad M, Strohmeier S M, Nouwakpo K et al., 2022. Rangeland restoration in Jordan: Restoring vegetation cover by water harvesting measures. International Soil and Water Conservation Research. (in press)

Haliburton T A et al., 1977. Effects of mechanical agitation on drying rate of fine-grained dredged material. Technical Report D-77-10. United States Army Waterways Experiment Station. Vicksburg, MO.

Herrick J E, Whitford W G, de Soyza A G et al., 2001. Field soil aggregate stability kit for soil quality and rangeland health evaluations. Catena, 44: 27-35.


Huang J, Yu H, Guan X et al., 2016. Accelerated dryland expansion under climate change. Nature Climate Change, 6: 166-171.


Lal R, 2002. Biophysical controls that make erosion-transported soil carbon a source of greenhouse gases. Applied Sciences, 12: 8372.


Lian X, Piao S, Chen A et al., 2021. Multifaceted characteristics of dryland aridity changes in a warming world. Nature Reviews Earth & Environment, 2: 232-250.

Lickley M, Solomon S, 2018. Drivers, timing and some impacts of global aridity change. Environmental Research Letters, 13: 104010.


Loeppert R H, Suarez D L, 1996. Carbonate and gypsum. In: Sparks D L et al. (eds.). Methods of Soil Analysis, Part 3, Volume 5, Chemical Methods, SSSA Special Pub., Madison, WI:437-474.

Louhaichi M, Gamoun M, Belgacem A Q, 2020. Direct seeding: A fast and cost-effective method for large-scale restoration of degraded rangelands. The International Center for Agricultural Research in the Dry Areas (ICARDA), Beirut.

Mahmoud A M A, Mohd Hasmadi I, Alias M S et al., 2021. Rangeland restoration analysis on the south slope of the Al-Jabal Al-Akhdar, northeast Libya. Journal of Rangeland Science, 11: 26-43.

McFarlane D J et al., 2016. Salinity in dryland agricultural systems:Challenges and opportunities. In: Farooq M, Siddique K H M (eds.). Innovations in Dryland Agriculture. Berlin: Springer, 521-547.

McIver J, Starr L, 2001. Restoration of degraded lands in the interior Columbia River basin: passive vs. active approaches. Forest Ecology and Management, 153: 15-28.


McLean E O, 1982. Soil pH and lime requirements. Methods of Soil Analysis, Part 2: Chemical and Microbiological Properties, American Society of Agronomy, Madison, WI, USA.

Medeiros A S, Drezner T D, 2012. Vegetation, climate, and soil relationships across the Sonoran Desert. Ecoscience, 19: 148-160.


Minhas P S, 1996. Saline water management for irrigation in India. Agricultural Water Management, 30: 1-24.


Morales N S, Zuleta G A, 2020. Comparison of different land degradation indicators: Do the world regions really matter. Land Degradation & Development, 31: 721-733.


Morrison E B, Lindell C, 2011. Active or passive forest restoration? Assessing restoration alternatives with avian foraging behavior. Restoration Ecology, 19: 170-177.


Nimmo J R, 2004. Aggregation:Physical aspects. In: Hillel D (ed.). Encyclopedia of Soils in the Environment. London, Academic Press.

Olsson L et al., 2019. Land degradation. IPCC SRCCL. Intergovernmental Panel on Climate Change, Geneva.

Osman K T, 2012. Factors and processes of soil formation. In: Osman K T (ed.). Soils: Principles, Properties and Management. Berlin: Springer, 17-30.

Parsons A J, Abrahams A D, Luk S H, 1991. Size characteristics of sediment in interrill overland-flow on a semiarid hillslope, southern Arizona. Earth Surface Processes and Landforms, 16: 143-152.


Proffitt A P B, Rose C W, 1991. Soil erosion processes: 2. Settling velocity characteristics of eroded sediment. Australian Journal of Soil Research, 29: 685-695.

Qinfeng G, 2004. Slow recovery in desert perennial vegetation following prolonged human disturbance. Journal of Vegetation Science, 15: 757-762.


Queensland Government, 2020. Types of erosion.

Richards L A, 1954. Diagnosis and improvement of saline and alkali soils. In: USDA Agricultural Handbook. Vol. 60, Washington, D.C.

Rohr J R, Bernhardt E S, Cadotte M W et al., 2018. The ecology and economics of restoration: when, what, where, and how to restore ecosystems. Ecology and Society, 23: 15.

Rose C W, Yu B, Ghadiri H et al., 2007. Dynamic erosion of soil in steady sheet flow. Journal of Hydrology, 333: 449-458.


SAS Institute, 1990. SAS/STAT User’s Guide. Version 6, 4th ed. SAS Institute: Cary, NC.

Schaffer B E, Rodriguez-Iturbe I, 2018. Water-limited vegetated ecosystems driven by stochastic rainfall: feedbacks and bimodality. Proceedings of the Royal Society A Mathematical and Physical Engineering, 474: 20170649.

Sedaghathoor S, Zare S K A, 2019. Interactive effects of salinity and drought stresses on the growth parameters and nitrogen content of three hedge shrubs. Cogent Environmental Science, 5: 1682106.


Shi Z H, Fang N F, Wu F Z et al., 2012. Soil erosion processes and sediment sorting associated with transport mechanisms on steep slopes. Journal of Hydrology, 454/455: 123-130.


Slater C S, Carleton E A, 1942. Variability of eroded material. Journal of Agricultural Research, 65: 209-219.

Stavi I, 2012. The potential use of biochar in reclaiming degraded rangelands. Journal of Environmental Planning and Management, 55: 657-665.


Stavi I, Gusarov Y, Halbac-Cotoara-Zamfir R, 2019. Collapse and failure of ancient agricultural stone terraces: On-site geomorphic processes, pedogenic mechanisms, and soil quality. Geoderma, 344: 144-152.


Stavi I, Perevolotsky A, Avni Y, 2010. Effects of gully formation and headcut retreat on primary production in an arid rangeland: Natural desertification in action. Journal of Arid Environments, 74: 221-228.


Stavi I, Rachmilevitch S, Hjazin A et al., 2018a. Geodiversity decreases shrub mortality and increases ecosystem tolerance to droughts and climate change. Earth Surface Processes and Landforms, 43: 2808-2817.


Stavi I, Rozenberg T, Al-Ashhab A et al., 2018b. Failure and collapse of ancient agricultural stone terraces: On-site effects on soil and vegetation. Water, 10: 1400.


Stavi I, Siegal Z, Drori B et al., 2018c. Single session of chiseling tillage for soil and vegetation restoration in severely degraded shrublands. Water, 10: 755.


Stavi I, Thevs N, Priori S, 2021a. Soil salinity and sodicity in drylands: A review of causes, effects, monitoring, and restoration measures. Frontiers in Environmental Science, 9: 712831.


Stavi I, Yizhaq H, Szitenberg A et al., 2021b. Patch-scale to hillslope-scale geodiversity alleviates susceptibility of dryland ecosystems to climate change: insights from the Israeli Negev. Current Opinion in Environmental Sustainability, 50: 129-137.


USDA-NRCS. Dispersive Clay Soils. EFH Notice 210-WI-6.

Vanags C, Minasny B, McBratney A B, 2004. The dynamic penetrometer for assessment of soil mechanical resistance. In:Proceedings of the SuperSoil 2004, 3rd Australian New Zealand Soils Conference, Sydney.

Xiao J, Eziz A, Zhang H et al., 2019. Responses of four dominant dryland plant species to climate change in the Junggar Basin, northwest China. Ecology and Evolution, 9: 13596-13607.


Zahawi R A, Reid J L, Holl K D, 2014. Hidden costs of passive restoration. Restoration Ecology, 22: 284-287.


Zaher H, Caron J, 2008. Aggregate slaking during rapid wetting: Hydrophobicity and pore occlusion. Canadian Journal of Soil Science, 88: 85-97.