The Remote Sensing Component
It is widely accepted that satellite remote sensing offers considerable advantages for land degradation assessments. With a comprehensive spatial coverage it is intrinsically synoptic, and provides objective, repetitive data which contribute to resource assessments and monitoring concepts of environmental conditions in drylands. However, only if these observations can be coupled with GIS-based ecological modelling concepts, they may develop their full capacity to be used for modifying and adapting environmental management principles and mitigation strategies in response to present and expected environmental challenges. Taking further into account the importance of human activities which tend to respond primarily to socio-economic triggers it becomes obvious that integrated models need to be developed which are capable to represent land use changes in the linked socio-economic, ecological and physical systems.
It has long been known that surface properties (i.e., vegetation cover and composition, specific properties of parent material and soils) control water availability or the spontaneous emergence and development of new plants in drylands. Consequently, one of the objectives of remote sensing approaches is to focus on this particular interface. Particularly the application of the ecological modelling concept proposed by LADAMER requires spatially distributed estimates of the evaporative coefficient k, the actual vegetation density f (i.e. proportional cover), and preferably a set of georeferenced sample sites were the deviation between f and f0 (the potential vegetation density) can be assumed to be minimal. So far, the primary remote sensing input into the model has been limited to spatially distributed estimates of actual vegetation density (either as fractional cover or Leaf Area Index derived in relation to NDVI) which can be derived with reasonable accuracy from remotely sensed data. In order to meet the prerequisites of LADAMER, this interface, which so far has been based on limited data series obtained from earth observation satellites (e.g. Landsat TM/ETM) must now be extended to accommodate small scale multi-year observations from global monitoring satellites (SPOT VEGETATION, NOAA-AVHRR, MODIS, ASTER). Changes of the vegetation density over time bear important information on land degradation dynamics which are induced by natural or man-made processes. In this respect, the production of suitable small-scale map representations of existing degradation trends, requires the decoupling of long-term trends and cyclic components of vegetation dynamics. The generating of the phase-1-product describing regional degradation trends (WP 2) will be based on the application of Fourier to remove noise and cyclic components induced by phenological effects, and a subsequent classical trend analysis (parametric and non-parametric) of 20 years of pre-processed 8-km AVHRR Pathfinder data. In addition, a significant amount of in-depth research will be required for validating this intermediate product and for extending the time-series-analysis.
The Land Degradation Modelling Component
Recent ecological approaches have been undertaken, in which land degradation is associated with the decline of an ecosystem's capacity to retain and use local resources. In dryland systems this decline is reflected in the local water balance. As the land degrades, the hydrological capacity of the soil deteriorates and 'non-productive' water fluxes, especially surface runoff (Q), become increasingly important at the expense of local water storage (S) and gradual evapotranspiration (Ea). Combining aspects of ecological optimisation theory with empirical work on plant / water relations in dryland environments, EEZA proposed to use the ratio of long-term actual evapotranspiration (Ea) and precipitation (P) as an index of local resource retention and, hence, dryland condition.
The procedure consists mainly of the spatially distributed estimation of Ea on the basis of remotely-sensed deviations between potential (f0) and actual vegetation density (f). A function for the prediction of the potential canopy density, for a given vegetation type and combination of climatic conditions, is established by regression of the actual density against Specht's evaporative coefficient (k) in reference sites where Ea ≈ P. Reference sites are randomly drawn from a population of grid cells with strong topographic constraints on lateral water inputs and are defined as those that have maximum f values for a given value of k. This can be used as a bioclimatic index, but also as a surface conductance. Multivariate regionalisations of target areas, in terms of soil-lithology, terrain and land cover types, are used to reduce the contribution of non-relevant causes of deviations between f and f0, and hence to reduce the uncertainty of the assessment. Experience with the application and qualitative evaluation of this method was obtained in a medium sized area (1000 km2).
The approach is innovative in the sense that it provides a process-oriented, rather than descriptive, procedure for assessing land degradation on the basis of an established ecological theory while meeting most of the mentioned requirements for small scale applications.
As the foregoing approach was designed to obtain an assessment of land condition status in non agricultural areas of dry climates, its adaptation to LADAMER requires the method to be upgraded in a number of aspects. The conceptual basis will be adapted to a wider range of climates, vegetation types, and land use settings. The temporal resolution will be increased from mean annual to annual and, possibly, seasonal to better capture the cover changes of deciduous and annual vegetation types or crops. In both cases, the approach will be based on directly estimating Ea = E0 ⋅ k ⋅ W, where E0 stands for potential evapotranspiration, and W for available soil water storage. The evaporative coefficient k would be estimated, as a surface conductance, by inverting the relation f = F(k) obtained at the reference sites in hydrological equilibrium. Moreover, the conceptual basis and cartographic modelling procedures will be modified to allow application at a range of spatial resolutions (e.g. 30 m - 1 km).
Low-cost validation procedures will be developed for the underpinning of regional applications.
The Land Use / Land Cover Change Modelling Component
The development of integrated assessment models is currently a rapidly expanding activity, accelerated by the revolution in the computing hardware and software since the early eighties. This trend is propelled by the growing understanding that policy-making should be based on integrated approaches. Systems Theory clearly has shown that systems and problems do not exist in isolation, rather that they have dimensions that extend into other domains, other disciplines, other levels of detail, and other temporal and spatial scales. Complexity and Computation Theory has shown that even seemingly weak linkages may have major repercussions on the behaviour of the system as a whole. Policy makers, responsible for the management of regions, watersheds, or coastal zones are confronted with this reality on a daily basis. Confronted with this complexity on the one hand and with better informed, agile recipients of the policies on the other, policy makers have to be able to rely on adequate instruments enabling them to better understand and anticipate the effects of their interventions in the system as fully as possible.
Most relevant in the field of spatial planning and policy making has been the rapid growth of high resolution remote sensing and Geographical Information Systems in the past two decades. As a result, new dynamic modelling techniques have been added to the toolbox of the spatial scientists. Agent based approaches, and in particular Cellular Automata, are rapidly gaining interest.
Cellular Automata (CA) models can be thought of as simple dynamic systems in which the state of each cell in an n-dimensional array depends on its previous state and on the state of the cells within its neighbourhood, according to a set of stated transition rules. While the early applications of CA models in the spatial sciences remained rather conceptual and theoretical, most recent applications are developed with an aim to realistically represent geographical systems, both in terms of the processes modelled and the geographical detail represented. This trend has come with an increase in the complication of the models developed.
Just as important in the context of integrated modelling are the possibilities for linking CA models to other cellular models representing changes in the cellular space -in which the CA dynamics unfold- or to dynamic models operating at a more macroscopic scale. In the latter case, the macro-models will constrain the overall dynamics of the CA. In the EU-projects MODULUS and MEDACTION this has resulted in an integrated model representing the non-homogeneous character of the cellular space by means of sub-models calculating among other: the soil quality and water balance, the quality and volume of the aquifer, the characteristics of the natural vegetation, etc.. On top of these physical layers, the human dynamics unfold changing the land use and land cover. These dynamics are governed by CA decision rules, representing human (spatial) behaviour, socio-economic preferences and decision-making, crop choices, etc.
This is the basis from which LADAMER will start in its effort to integrate physical, ecological and land use models and apply them to the European Mediterranean countries, in an effort to define the 'hot spot' areas which are prone to land degradation.