Arid and semi-arid lands cover approximately one third of the continental surface of the Earth [38]. About two thirds of these areas are located in warm climates between 15 and 35 °C in both hemispheres, and thus belong to the deserts and semi-deserts of tropical and subtropical regions, respectively.

Primarily because of the harsh climatic conditions, drylands are, and always have been extremely sensitive to inappropriate human actions leading to degradation processes. The term “land degradation” is generally used to refer to the loss of valuable land through various processes such as soil erosion, salinisation, depletion of nutrients, deterioration of soil structure and subsequent erosion processes. Desertification may then be conceived as a severe stage of land degradation, in which disturbances have gone beyond the resilience of the land. Although desertification could in the past be attributed to climatological effects, there is little doubt today that desertification is mainly a man-made or man-induced phenomenon developing under the climatic conditions of aridity, leading to varying degrees of vegetation and soil deterioration (e.g., [37]). Both the intensification and extensification of land use can cause degradation depending on the environmental setting.

The extent and spatial distribution of drylands is roughly known at a global scale. However, the knowledge on the current land degradation status or the magnitude of the potential hazard is for the most part incomplete and fragmented, or may even be entirely lacking [8]. This makes it extremely difficult to design and implement mitigation, reclamation and prevention measures. The productivity of dryland systems largely depends on surface properties, which, as they control water availability, the spontaneous emergence and development of new plants and dust production during wind storms, frequently dominate climatic variables (e.g., [30]). It emerges from this conceptual framework that surface properties are considered more indicative than long-term climatic conditions. The ability to draw concise conclusions with respect to land resources and environmental change thus clearly depends on the capability to assess these specific surface characteristics (i.e., vegetation cover and composition, properties of parent material and soils, including effects of mineralogical and biologic crusting).

The high spectral contrast of vegetation relative to soils has provided many successful studies, which use optical remote sensing systems for observing vegetation communities (e.g., [39]). Since plants have distinct spectral signatures with low reflectance in the visible part of the spectrum and high reflectance in the near-infrared region, attempts have been made to use this spectral feature for the estimation of green vegetation cover and biomass (e.g., [40]). It is obvious that sensor systems involving spectral bands in the SWIR region provide much better information to separate green and dry vegetation components. Given the availability of high spectral resolution images, such methodologies like the spectral mixture analysis (SMA), also known as spectral unmixing, have been successfully used for a wide range of related applications (e.g., [3,4,27,36,41,42,44], (ISMA: Iterative Spectral Mixture Analysis) [10]). Because of its direct response to pre-selected component spectra (i.e. typical desert shrubs and appropriate substrate elements) and increased robustness against soil colour differences in the absence of dense vegetation, the method is particularly suited to overcome the limitations of two-band vegetation indices. Consequently, improved estimates can be expected for vegetation abundance in areas with sparse cover of mainly woody plants.

Besides the use of classical indicator variables (LAI (Leaf Area Index), biomass, proportional cover) that can currently be derived from operational Earth observation satellites through empirical and semi-empirical approaches; methods based on the inversion of analytical reflectance models are also explored (e.g., [34]). To invert the reflectance model (i.e. the retrieval of the actual canopy parameters from the hyperspectral satellite measurement), artificial neural networks could be trained to relate suitable combinations of LAI, leaf water content and chlorophyll concentration non-linearly to the spectral canopy reflectance.

Not all studies of dryland ecosystems have relied solely on vegetation cover as an indicator of land suitability or degradation. Escadafal et al. and Hill et al. [21,27], for example, have been able to use parameters indicative of soil conditions and, hence, to apply concepts taken from pedology and geomorphology to obtain qualitative indications for soil degradation or erosion processes. It should be emphasised that soil texture appears to be the key factor for the water balance of arid soils, as it controls infiltration and the capillary rise of water. In fine-textured soils, rainfall can hardly infiltrate and is therefore subject to evaporative losses. More coarsely textured soils (e.g., sandy or stony substrates) permit water to rapidly infiltrate to greater depths where it is well protected against evaporation after the topsoil layer dries out. In addition, almost all infiltrated water is readily available to plants, and salt concentrations in the soil remain low due to the higher leaching efficiency. Relationships between climate and environmental conditions, which are valid at a global scale, thus become problematic when local factors such as topography, lithology and soils determine the redistribution of water [52]. It is therefore mandatory that remote sensing focuses on the precise characterisation of surface physical properties [15,17,18], a perspective which becomes particularly promising using hyperspectral images. Spectral mixture analysis further allows the estimation of relative amounts of rock fragments and soil particles on the surface. Since soil erosion leads to an increase in rock cover in source areas and the accumulation of soil material as colluvium elsewhere, rock content can be used as an indicator of degradation. This corresponds to defining the erosional state of soils as a function of the mixing ratio between developed soil substrates and parent material components. Of course, these must be spectrally distinct in order to be discriminated from each other. As an example, Hill et al. [27] have successfully used this approach to map areas affected by soil degradation and erosion under sub-humid conditions in Mediterranean France. Hill et al. [29] have also demonstrated that spectral unmixing of reflectance signatures permits the successful identification and spatial differentiation of soil substrates and biological crusts in sandy arid ecosystems. More recently, Ben-Dor and Banin [6,7] used the full reflective spectrum to predict soil organic matter and other mineralogical and chemical soil properties based on multivariate statistics. A successful method has been proposed by [30] to retrieve organic carbon in xenomorphic soils, which highlights areas of accumulation and relative stability (sediment sinks). These approaches can already be applied to high-resolution laboratory spectra and real images, provided the latter have been adequately corrected for atmospheric effects. They strongly support the assumption that hyperspectral images can be successfully used to derive more specific soil properties (i.e. soil depth, chemical properties, infiltration capacity, runoff coefficient, etc.). This can be achieved either directly or through indirect relationships to be exploited through the use of pedo-transfer functions.

Many indices have been proposed to describe the susceptibility of drylands to desertification and land-degradation processes. Most of these indices are only minor modifications of climatic indices describing the degree of aridity, which, although implicitly related to other key variables such as the fractional vegetation, are not useful indicators for degradation processes. The crucial issue for global environmental assessment and monitoring strategies is the ability to identify, understand and map degradation processes under a wide range of environmental conditions. Ground-based methods are generally time-consuming and lack the degree of standardisation that would render them sufficiently reproducible, transferable, up-to-date and cost-effective to cover large areas (e.g., [8,28]). While standard methods for undertaking such measurements are imperfect and require heavy costs, satellite remote sensing, based on spaceborne hyperspectral imaging systems, offers considerable advantages. With a comprehensive spatial coverage, it is intrinsically synoptic, and provides objective, repetitive data with a continuous spectral coverage that is ideally suited for resource assessments and monitoring of environmental conditions in drylands. Repeated resource assessments and degradation monitoring in environmentally sensitive areas, coupled with GIS-based ecological model simulations, may provide the key to modify and adapt environmental management principles and mitigation strategies in response to present and expected environmental changes in drylands [24].

Indeed, hyperspectral imaging observation through time at a spatial scale on the order of classical Earth observation systems such as Landsat and SPOT would permit the retrieval of the quantitative biogeochemical/biogeophysical parameters that are to be incorporated into erosion and dryland degradation models to (a) monitor land-degradation processes, (b) assess land-degradation status, and (c) gain indicators for characterising specific surface properties related to water cycles, erosion processes, and plant productivity in drylands. In order to make efficient use of the hyperspectral information for measuring vegetation density, biodiversity, geochemistry, soil properties and other subsequent parameters relevant to erosion and land degradation status, it is necessary to assimilate the extracted quantitative parameters into the models. Remote sensing-based methods are especially suited for the detection of change, but need to be further optimised for detecting specific object properties within changing natural contexts. While it has been shown that hyperspectral imagery is able to provide better and more detailed information on proportional vegetation cover (including its dry/senescent components) than broad-band images [19], a further advantage lies in the fact that high spectral resolution imaging sensors also deliver spectral information on surface properties. This can be translated into specific parameters describing substrate characteristics required for pre-stratification strategies (e.g., lithology of parent material, runoff coefficient), or collateral model improvements (infiltration capacity, mineral and biogenic crusting, organic matter concentration, carbonate and iron content), or can be combined with models describing the erosional processes frequently involved in land degradation dynamics.

The spectral molecular absorption and scattering properties of materials form the basis of the identification and determination of the abundances of surface and atmospheric constituents in spectrometry. Considerable efforts have been taken to build hyperspectral imaging systems, which are able to observe the Earth from space. However, the existing systems (Fig. 2), such as Hyperion (part of NASA's EO-1 platform), and MERIS on board the ENVISAT platform (ESA), still collect data in a somewhat limited fashion in terms of radiometric and/or spectral performance, swath width, and revisit time (Hyperion) as well as spectral range with rather moderate spatial detail (MERIS). A spaceborne hyperspectral imaging system such as the envisioned EnMAP one (Environmental Mapping and Environmental Program) (Fig. 2), with its improved analytical capacity vis-à-vis lithological, soil and dry plant properties, could contribute substantially, with the support of the science community in promoting its Earth observation capabilities, to solve a number of environmental issues. The implementation of such a dedicated system in space is further justified through the success of long-term retrospective monitoring studies, the need to derive qualitative and quantitative indicators of surface properties [8,29], and the necessity to foster integrated approaches, which link surface process models with quantitative approaches in the analysis of remotely sensed data.