Desertification is defined as the degradation of land in arid, semi-arid, and dry sub-humid areas. It is caused primarily by human activities and climatic variations. It occurs because dryland ecosystems, which cover over one-third of the world's land area, are extremely vulnerable to over-exploitation and inappropriate land use. Combating desertification is essential to ensuring the long-term productivity of inhabited drylands. Unfortunately, past efforts have too often failed, and around the world the problem of land degradation continues to worsen.  Recognizing the need for an internationally coordinated approach, 193 governments have joined, as of October 2009, the United Nations Convention to Combat Desertification.

The desertification and land degradation risk in the north Mediterranean areas is receiving growing attention by the international community, as testified by the numerous national and European projects on the subject carried out during the past few years. Many of them put a lot of emphasis on the exploitation of Earth Observation data. The DesertWatch project of the European Space Agency, which was recently successfully completed, aimed at developing an integrated information system tailored to the specific user’s needs, built on the technological transfer of the most significant results of related research projects.

The DesertWatch Information System (DW IS), which was developed in the course of the project, is a user-friendly tool for monitoring desertification. Automated processing algorithms were included to enable non-specialised users to operate the system and produce the necessary information in all areas with comparable accuracy. The DesertWatch Information System can monitor up to 11 desertification related parameters, ranging from simple geo-referenced indicators (e.g. urban sprawl, irrigated areas, forest fires, vegetation abundance and health, soil erosion, etc.), to complex models that can simulate future scenarios of desertification and risk maps. A comprehensive demonstration phase has been carried out for testing the processing chain results over vast areas of Italy, Turkey and Portugal using data covering the last 20 years.

Monitoring desertification requires the evaluation of a complex set of indicators, related to climatic (e.g. rainfall, evapotranspiration, aridity indexes), biophysical (e.g. morphology, soil and vegetation properties), socio-economic (e.g. population density and age, employment) and management (e.g. policies, protected areas, master plans) factors. The DW IS uses primarily EO data, in combination with some ancillary data, into a seamless data processing facility. To assess the needed indexes the following principal approaches/techniques were used:

1) In desertification monitoring, land cover maps are invaluable instruments as they allow the user to directly extract a number of useful indicators. DW IS provides users with dedicated tools to update in a cost-effective and accurate manner land cover maps, from which to extract a number of additional parameters, as requested by users and UNCCD National Focal Points. All indicators can be spatially aggregated, providing administrative level statistics. The adopted legend is an aggregation of the CLC legend into 12 classes, specifically identified to provide the necessary information for desertification assessment purposes.  From the land cover map a number of indicators are assessed by direct land cover class extraction: Forested areas, Soil sealing, Irrigated lands, Forest Fragmentation, Re-naturalized Areas and Burnt Areas;

2) The semi-empirical model, known as Spectral Mixture Analysis (SMA), describes spectral reflectance signatures as a mixture of few prototype spectra, also called endmembers [5]-[6]. In natural environments, the principal endmembers are typically represented by vegetation, soil, bedrock and shade, which commonly mix at the sub-pixel scale producing mixed-pixel spectra. SMA attempts to unmix such multispectral reflectance in order to assess, for each pixel, the relative abundance of each component. Using this technique it is possible to retrieve the following useful indicators: Vegetation Abundance and normalized Soil/Rock ratio (NSRR). The vegetation abundance, which simply corresponds to the proportion of the vegetation end-member, besides being an indicator in itself, can be used in conjunction with other parameters for the assessment of the vegetation quality index. In Mediterranean ecosystems, NSRR can be seen as a surrogate of soil erosion, especially if the trend analysis shows a decrease of the ratio. This technique can be used with both Landsat and MERIS EO data, for obtaining maps at local and national scales.

3) The Land Degradation Index (LDI) exploits remote sensing images, in conjunction with climatic and physiographic parameters, for assessing the landscape status with respect to its natural resources potential. It works on the assumption that sites in good condition will show outputs in their local water balance mostly through evapotranspiration rather than through runoff. Input data include: EO data, climatic data, topographic data, daily rainfall series and external physiographic layers. Based on the external physiographic layers, homogeneous land systems are identified and their status is expressed in terms of number of standard deviations from the population mean. The final values are then converted into 3 broad classes, which indicate poor, normal and good land conditions. This index can be computed at various resolutions, according to the required mapping scale. For this project both Landsat and MERIS data have been used.

4) A spatial modelling tool has been used for simulating Scenarios of Desertification (named ScenDes). The tool is able to generate possible scenarios for land use, starting with a land use map at a certain time and simulating the land use evolution during the years. The complex simulation model uses local, zonal and neighbouring map-based rules for computing the land use transition potential. For each map pixel, the system computes the potential to change (transition potential) into each of the possible land use classes. In addition to this, based on socio-economic data, such as population, growth rate, job availability, etc., ScenDes evaluates the total demand for each land use. For each simulation step (in these simulations the time step is set to 1 year) the system will try to meet the overall demand for each land use class, starting to change the pixels with highest transition potential, until the total demand is met. The simulation is typically carried out on periods of 10-30 years and allows the visualisation of possible land-use scenarios.

The DesertWatch project follow-up is currently under development. It will exploit the same paradigm as the original project, with three significant highlights: (i) enhanced data processing methodology aimed at improving classification accuracy; (ii) finer resolution, obtained by integrating additional high resolution data sources, such as SPOT and Kompsat; and (iii) extension of the demonstration cases to areas outside the Mediterranean, such as Mozambique and Brazil.