Restoring degraded drylands is a worldwide issue. The “land degradation neutrality” target promoted by the United Nations Convention to Combat Desertification (UNCCD) indicates that the progress made with restoration could compensate the impacts of degradation, stressing the importance of a quantitative evaluation process. Studies and attempts to implement restoration strategies in different dry environments are numerous, from rangelands to shrub and forest stands (Camprubi et al., 2015; Cortina et al., 2009; Fuentes et al., 2010; Roa-Fuentes et al., 2015; Zucca et al., 2015a, b), from agricultural ecosystems to mining sites and brownfields (Dickinson et al., 2005; de Moraes Sá et al., 2015; Hasanuzzaman et al., 2014; Oliveira et al., 2011; Stroosnijder, 2009; Toktar et al., 2016; Wong et al., 2015). Though restoring degraded drylands is also a complex issue, it can be pursued by means of several strategies, all of which consider soil characteristics, either directly or indirectly. In fact, soil is a key part of the Earth system, as it controls the hydrological, erosional, biological, and geochemical cycles (Brevik et al., 2015; Keesstra et al., 2012; Smith et al., 2015). If soil-inherent slow-changing soil qualities are of utmost importance in designing ecological restoration strategies, soil dynamic properties can be used to monitor and assess the consequences of restoration activities on ecosystem functioning and services. In any case, finding suitable indicators to monitor ecological restoration activities at different scales, both within ecosystems and in the broader socioeconomic system, requires a full understanding of soil–plant–ecosystem relationships, as well as an interdisciplinary and integrative approach. The integration of different viewpoints from complementary disciplines is, nevertheless, still uncommon in restoration. Drylands' restoration, in particular, due to their idiosyncratic characteristics of high spatial heterogeneity and temporal variability, represents an even greater challenge, requiring restoration indicators able to reflect different spatial and temporal variations.

The objective of this review is to present and discuss soil indicators which show potential to check the effectiveness of restoration activities in drylands at different spatial and temporal scales. The subject is treated from the viewpoints of specialists coming from different disciplines, namely pedology, ecology, hydrology, and land management, all dealing with the practice of ecosystems' restoration. The paper is presented in three parts. The first part introduces linkages between land degradation and ecological restoration, stressing specificities of dryland ecosystems; the second part deals with soil indices and indicators to be used before and after restoration at different scales, and their relationship with soil processes; the third part addresses more integrated assessment of restoration, linking soil and ecological issues with socioeconomic perception. In particular, the paper introduces the purpose of restoration in drylands in Sect. 1.1 and addresses key interactions between plants, soil, and climate in Sects. 2 and 3; having these interactions in mind, a series of soil indicators are discussed from local scale (Sect. 4) to landscape scale (Sects. 5 and 6). More specifically, the nature of soil indicators covers the physical (Sect. 4.1), chemical (Sect. 4.2), biochemical, and biological aspects (Sects. 4.3, 4.4 and 4.5), its integration in a landscape functional approach (Sect. 5), and in a holistic assessment, which also considers socioeconomic indicators (Sect. 6).

The success of vegetation establishment in restoration projects of degraded drylands largely depends on the extensive understanding of the relationships between soil characteristics and plant-rooting features. Globally, the soil depth at which different plant growth forms absorb water varies considerably (Canadell et al., 1996). In water-limited ecosystems, root systems' mean depths increase with above ground size: annuals < perennial forbs and grasses < dwarf shrubs < shrubs < trees (Table 1, Fig. 2). Stem succulents are as shallowly rooted as annuals, but have relatively high lateral root spreads (Schenk and Jackson, 2002a). Hence, soil properties that determine water availability along the soil profile largely determine the type of vegetation with potential for establishment. For instance, savannah-like systems of holm oak (Quercus ilex L.) and cork oak (Quercus suber L.) woodlands, found in western Mediterranean Basin drylands, have a grassy understory dominated by annuals, with most of the roots concentrated in the upper 20–30 cm of the soil. In general, this upper layer includes organic soil horizons, where the overall root density is highest, most likely because it stores nutrients and has higher water-holding capacity. However, grassland areas are often intermingled with shrub patches, which evidently obtain water from deeper soil layers. Some of the most prominent shrubs in these systems are the shallow-rooting (30–40 cm) rockroses (Cistaceae family), which have a high lateral root spread. Such root systems may improve water-use efficiency. When soils are deeper, shallow-rooted shrubs may coexist with deeper rooting plants such as the strawberry tree (Arbutus unedo L.) or the mastic tree (Pistacia lentiscus L.) that may get water lower than 2 m (Silva et al., 2002). Deep roots play a fundamental role during the dry season, because they reach deeper layers where water depletion is not as widespread as at the surface. In fact, the dominant oak trees in Mediterranean woodlands seem to get water from even deeper depths (groundwater), particularly during the dry season (Kurz-Besson et al., 2006). Another example is the Ibero-North African dryland steppe, dominated by the perennial alpha grass (Stipa tenacissima L.). Its root system goes no further than 50 cm depth (Cortina et al., 2009), somewhat similar to the aforementioned shallow-rooting shrubs, enabling the species to access upper soil layers after small rainfall events. In these environments, biological soil crusts are a prominent feature covering bare soil. They play an important role by protecting soil surface from wind and water erosion, participating in nutrient cycling, reducing loss of water due to evaporation, and taking part in biotic interactions (e.g., influencing seed germination of vascular plants) (Bowker et al., 2014). Biological soil crusts have been introduced in deserts in several parts of the world in order to help prevent erosion and desertification (e.g., USA, China, Israel).

Soil heterogeneity is reflected in water distribution and availability for root uptake along the soil profile. The major factors affecting this distribution are soil particle size and seasonality of precipitation. Water-limited ecosystems tend to have deeper root systems in coarse-textured soils than in fine-textured soils, because the former have lower water-holding capacity and water tends to percolate more deeply, where groundwater, or a temporary perched water table, may be present. Conversely, the existence of a restrictive soil layer, for instance, in soils with a compacted or cemented layer, or high clay content in the subsoil, or showing shrink-swell properties (Vertisols) may favor shallow-rooted herbs, while limiting the establishment of deeper-rooted species, like perennial grasses or shrubs. Soil information concerning water availability of the different soil horizons, and not only of topsoil, is thus very important in order to adequately select the actions and species used to restore plant cover of degraded sites.

The residence time of water in soil, i.e., the period during which water remains available at a certain soil layer after a precipitation event, is particularly important for plant communities in water-limited ecosystems, especially during the growing season. The longer the period during which water is available, the greater the opportunity for plants to survive, grow, and reproduce. In general, if water is retained in the uppermost soil layers, that may be beneficial for shallow-rooting herbaceous species germination and establishment. On the other hand, if water percolates rapidly to deeper layers, that may favor woody vegetation.

Precipitation distribution and seasonality, i.e., if precipitation is evenly distributed throughout the year or occurs during the cold or warm seasons of the year, play a key role regarding water availability for plants along the soil profile. In drylands, shrubs are more shallowly rooted in climates with summer than winter precipitation regimes (Schenk and Jackson, 2002b). This is because in climates with summer precipitation, the residence time of water in the soil is shorter, and a wider and shallower root system is better able to uptake water before it evaporates. Succulent species are good examples, since they are in general as shallowly rooted as annuals, but have denser lateral root systems, similar to shrubs. These life forms become very widespread when low precipitation amounts are coupled with high temperatures, and hence water residence time is very short.

The assessment of water residence time in soils, and in particular, the information about when and for how long soil water is available to plants, is thus of major importance to predict the most suitable type of plant community for a given site.

The aridity index (rainfall/evapotranspiration ratio, AI) has been taken by the United Nations Convention to Combat Desertification (UNCCD) as a reference for the definition of the areas subjected to desertification. The usefulness of the AI relies upon the relative ease it can be calculated from standard climatic data. However, the AI has several drawbacks. For example, it does not take into account the capacity of the soil to regulate water availability, deep drainage, and runoff, which can vary noticeably inside the same climatic region. This is particularly true in transitional ecozones, such as in the Mediterranean Basin that is characterized by a notable pedodiversity (Ibáñez et al., 2013) and where lands at high and low risk of desertification are very often finely intermingled (Costantini et al., 2009b).

Pedoclimate, that is soil moisture and temperature regimes, has also been used to characterize areas with a certain desertification risk (Eswaran and Reich, 1998). Indeed, the American Soil Taxonomy considers soil moisture regime based on a yearly assessment of the number of days in which the soil moisture control section

The soil moisture control section is the layer in-between the depth to which a dry soil will be wetted by 2.5 cm of precipitation in a 24 h period and the depth to which the same soil will be wetted by 7.5 cm of precipitation in the same period.

Ecosystem services are determined by soil properties and their assessment requires the use of selected indicators (Calzolari et al., 2016). A wide range of soil indicators may be used, depending on the purpose and scale of evaluation. In restoration planning, soil indicators are needed to support both the design and monitoring phases. However, different information is needed for these two purposes. The design phase mainly requires information about soil (and site) attributes that may affect the probability of success of the intervention. The input properties used to work the indicators can be both inherent characteristics (De la Rosa and Sobral, 2008) such as topographic slope angle and aspect, surface rockiness, soil depth, texture, stoniness, structure, presence of subsoil pans, and subsoil wetness conditions, or more dynamic attributes such as acidity and salinity. Planning can be supported by the identification of “optimal” ranges of values of such variables that increase chances of success of restoration and/or decrease risks and costs, and this can be done by means of land suitability schemes. Several approaches are available to create indicators, ranging from traditional categorical or parametric schemes (Costantini, 2009) to more complex approaches integrating multicriteria analysis and decision support frameworks (Yi and Wang, 2013; Uribe et al., 2014).

The soil information needed to monitor and assess restoration depends on the time and spatial scales. In the short term, it might be important to focus on dynamic properties such as soil organic matter, pH, available phosphorus, nitrogen, and other nutrients, and macroporosity. However, because of the large spatial and temporal variability of ecosystems, particularly in drylands, it is critical that indicators focus on “slow variables” (Carpenter and Turner, 2000) so that the assessment of long-term changes and of the sustainability of land management is not confused by short-term variations in land and socioeconomic conditions (Salvati and Baiocco, 2011; Zucca et al., 2013a). Slow indicators can more directly reflect impacts on inherent soil qualities, e.g., through improved structure and porosity and increased topsoil depth and water-holding capacity. Table 2 shows a list of the most frequently used soil indicators, specifying their functional relevance.

Most commonly, mitigation and restoration actions are evaluated by checking vegetation cover and composition. However, functional approaches that also account for the spatial pattern of vegetation, in interaction with the soil nature, seem to be more suited to assess the ecosystem functioning. As previously highlighted, many drylands around the world present a patchy distribution of vegetation following a sink–source spatial pattern. Source areas have a negative balance of resources that accumulate in the sink areas. Beyond this redistribution of resources at the fine scale, a fully functional ecosystem includes the retention within the system. In dry ecosystems, vegetation patchiness can provide a measure of the landscape capacity to conserve water and nutrients (Cerdà, 1997). The assessment of the functionality of these ecosystems should include the description of the spatial distribution of vegetation (size and connectivity of different plant type patches) in combination with soil properties that determine the conservation of resources, especially regarding soil surface attributes. The optimum spatial distribution is both ecosystem-dependent, as the hydrological functioning of plants varies between species, and site-dependent, as resource redistribution at the patch and catchment scale is highly site- and soil-specific; but some attempts have already been done in this regard. Puigdefábregas et al. (1999) suggested that the ratio between sink and source areas in functional ecosystems remains within an optimum range that maximizes functionality. Urgeghe et al. (2010) reported that the collection of runoff by herbaceous patches in a dryland pinyon-juniper ecosystem in southwest USA was maximum when both interpatch bare soil and herbaceous cover were intermediate, suggesting a trade-off between source and sink areas at the finer scale and the existence of herbaceous cover thresholds at the broader catchment scale. Some properties of the sink/source pattern, such as the upslope length and the size of the source area, have been successfully related to the performance of planted seedlings in drylands' restoration actions (Urgeghe and Bautista, 2015).

Landscape Function Analysis (LFA) (Tonway and Hindley, 2004) incorporates both vegetation and soil survey in the evaluation of dryland patchy ecosystems, using functional indicators instead of direct measures of key features. The LFA uses semi-quantitative field-based indicators (Table 6) to evaluate soil surface conditions at the hillslope scale in every identified type of patches and interpatches, targeting surface properties that control stability, nutrient cycling, and infiltration processes. The stability index provides an idea of the vulnerability to erosion and the ability to recover after stresses, the infiltration/runoff index indicates the ratio of rainfall water available to plants and export by runoff, and the nutrient cycling index informs about the in situ recycling of organic matter. For every single type of patch or interpatch, the scores of the quantitative indicators that have an impact on a particular index are summed and referred to the maximum possible score. The final value of the index is calculated by weighing the attained values in all patch and interpatch types by its representativeness in the working area.

Maestre and Puche (2009) observed significant relationships of the indices calculated through LFA with measured soil variables in alpha grass steppes in southeast Spain. These authors found that the infiltration index was positively related to soil-water-holding capacity and negatively to soil compaction, and the nutrient cycling and stability indices were positively related to soil-nutrient variables and microbial activity. However, the sensitivity of the indices might vary depending on the scale and the contrast between different situations.

LFA assessment represents a cheap, rapid, accurate, and repeatable methodology for the evaluation of soil functioning properties, especially in patchy drylands, and it is especially useful as a relative indicator when areas of contrasted histories and disturbance regimes of a similar ecosystem are compared. It has been used, for instance, to monitor the impacts on ecosystem functioning of restoration actions using exotic plant species (Derbel et al., 2009) or fodder shrubs (Zucca et al., 2013b), and also to monitor the effects of grazing and reforestation. LFA infiltration and nutrient cycling indexes have been observed to relate significantly to perennial species richness in Mediterranean drylands (Maestre and Cortina, 2004). In addition, some of the LFA indices, especially infiltration and nutrient cycling, show good correlations with remote sensing indices such as the NDVI (Gaitán et al., 2013). The combination of these two approaches at such different scales may provide useful information on ecosystem functioning and might be a good tool for dryland management by selecting and prioritizing areas to restore.

Besides LFA, there are other possible metrics dealing with the connectivity of water fluxes, bare soil, or interpatch areas, which try to link plant spatial distribution and hydrology. For instance, the Flowlength is a spatially based index that effectively relates the connectivity of source areas and vegetation distribution and topography (Mayor et al., 2008). Borselli et al. (2008) developed a GIS-based connectivity index and a field validation procedure that assesses the links between source and sink areas at the hillslope and landscape scales. The leakiness index by Ludwig et al. (2007) aims at quantifying the ability of the system to retain key resources, such as water and soil, within the system.

Integrated assessment protocols combine field observations of key ecosystem attributes, socioeconomic surveys, and remote sensing (RS)-based geospatial information. Particularly, to conduct the evaluation over wider areas, RS should be employed for land cover change and ecosystems' natural temporal pattern detection, land degradation assessment, and analysis of the impacts of land restoration (Zucca et al., 2015b; Ramos et al., 2015). The quantification of the photosynthetically active herbaceous and shrub biomass production in rangelands and savannahs is one of the most widely used metrics.

Integrated assessment protocols are oriented to more holistically assess the impacts of land management and restoration measures, i.e., to identify their ecological, economic, and sociocultural effects, both over the short term and long term as well as on- and off-site. The WOCAT network (www.wocat.net) has developed such methods in order to document and evaluate SLM technologies and approaches applied in the field. WOCAT is an international network founded in 1992 by land management specialists in order to document and share local sustainable land management practices at the global scale. The methods are internationally standardized and since 2014, have been accredited by the UNCCD as their documentation- and knowledge-sharing platform. The role of science in monitoring and assessing desertification, as well as mitigation/restoration actions, is to produce evidence of their impacts on natural resources and to assess the implications of these impacts on local societies. However, sophisticated and detailed assessment is often expensive and time-consuming and depends on the availability of skilled experts. On the other hand, stakeholder engagement in assessment of indicators is still rare or limited in scope. These are the reasons why a comprehensive but practical assessment tool, like WOCAT is providing, is needed, enabling scientific data to be combined with local experience. In order to evaluate mitigation/restoration practices, performance indicators – e.g., the impact of a given practice on degradation and its economic, ecological, and sociocultural benefits or disadvantages – should be assessed. These are mostly not available quantitatively, but can only be assessed qualitatively by experts, ideally according to predefined response categories (such as “no/negligible” for 0–5, “little” for 5–20, “medium” for 20–50, and “high” for > 50 % of change) in order to ensure comparability over practices, sites, and time. However, where available, quantitative data should be included as well (Schwilch et al., 2011, 2014). Soil- and vegetation-related indicators used in the WOCAT SLM technology questionnaire and assessed in the above described way include soil moisture, evaporation, surface runoff, soil cover, biomass/above-ground C, nutrient cycling, soil organic matter, soil loss, plant diversity, invasive species, beneficial species, etc. Another important aspect is the evaluation of the technical function, such as whether the practice works though an improvement of ground cover, surface roughness, soil structure, water availability, vegetation varieties. Socioeconomic impacts are equally recorded with quantitative and qualitative assessments. These include advantages and disadvantages of the SLM technology regarding, for example, production, income, workload, food security, recreational opportunities, aesthetic and cultural values, community strengthening or conflicts, health. Each SLM technology documented is assessed with the indicators listed and in the way described above. To date, over 500 such SLM technologies have been documented worldwide and are accessible in the WOCAT database (https://qt.wocat.net/qt_report.php).

Based on such assessments, conclusions can be drawn as to whether and how the documented practices address key threats in drylands, i.e., by means of improved water management, reduced soil degradation, diversified and enhanced production, resilience towards climate change and variability, and by providing sociocultural benefits including conflict mitigation and prevention of out-migration (Giger et al., 2015; Schwilch et al., 2014).

The development of methods for assessing the success of the actions to combat desertification is considered as a priority by the scientific community. The failure of restoration plans is often caused by the choice of plants or practices that are not suited to the site. The success of restoration plans instead relies on a proper and detailed knowledge of the relationships between soil and plant properties and ecology in drylands. One of the main challenges is to select the different species to be used for restoration which have a pattern of the root system matching the horizon characteristics of the soil profile, as well as the specific climate and hydrology of the site. Dryland restoration is a site-specific activity, which implies considering soil spatial and temporal heterogeneity before plant placement. However, ecological restoration of degraded lands is more than the mere recovery of soil ability to support vegetation. In addition to biomass production, restoration strategies should target restoration of ecosystem processes (e.g., nutrient cycling, decomposition), increasing additional ecosystem services such as biodiversity, carbon stock increase, greenhouse gases reduction, flood and sediment regulation, etc.

The understanding of dryland ecosystem processes stems from the very detailed scale of soil observation and analysis. A number of soil indicators support the design of measures and the assessment/monitoring phases. Such soil indicators need to refer to soil properties, which can actually be modified through management or restoration activities. Soil organic matter, in particular, is a key attribute for many ecosystem services and one of the main factors affecting water availability in drylands. Soil dynamic properties related to the forms of organic matter, as well as biochemistry, microbiology and mesobiology, are very sensitive to restoration activities. Although the functional forms of soil organic matter and related biological activities and organisms are still not completely understood and characterized, they are promising candidate indicators that may be utilized to assess the effectiveness of restoration strategies in dryland ecosystems.

A recent approach in assessing the effectiveness of restoration strategies in dryland ecosystems is combining the analysis of spatial pattern of vegetation with qualitative soil surface indicators. This simplified but effective methodology, specifically tailored for the surface patterns of drylands, allows the monitoring of landscape functioning variations in space and time, and it is particularly suitable for the assessments carried out at the intermediate territorial scales. On broader scales, effective strategies to combat desertification should be based on integrated biophysical and socioeconomic evaluation methods. Evaluation and monitoring of progress and success are expected to demonstrate the benefits of sustainable management, establish cost-effective thresholds for intervention alternatives, and identify priority areas for action. Recent approaches propose the assessment and evaluation of the effectiveness of management and restoration programs based on indicators that relate to ecosystem integrity and services, but also to socioeconomic and cultural variables associated to human well-being, both over the short term and long term, as well as on- and off-site. To this aim, there is a need for interaction and dialog among the diverse set of scientists and stakeholders involved, which can result in a co-production of new knowledge and, at the same time, in the formulation of new knowledge needs.