Work Package 2.2 Regional degradation index to identify desertification risk at sub-national to Mediterranean-wide scales (UNIVLEEDS and EC-JRC-EI)

Feeding into WP3.2, the development of a desertification indicator system for Mediterranean Europe, this work package will enable desertification risk to be identified for areas larger than the sub-national, and national areas at which the ESAIS operates. It also uses data from different sources, medium and high resolution remote sensing data, digital elevation models, and global circulation models. As such its use by a different part of the stake-holder community, UNCCD National Committees, Mediterranean-wide regional managers (rather than local land users) is foreseen.

The Regional Degradation Index (RDI) is a methodology to forecast the risk of soil erosion by water. The underlying physical basis is a one-dimensional hydrological model, which is used to estimate potential vegetation cover, if required and storm runoff based on climatic and vegetation data. The forecast runoff, accumulated across the frequency distribution of storms, is used to give a climatic erosion potential, which is then appropriately combined with measures of topography and soil erodibility to estimate the expected rate of soil erosion at a resolution of 250m - 1 km. The methodology is based on work funded by a previous FP IV project (MEDALUS III), a tender under FP IV (MODEM), a current FP V project (PESERA) and prior unfunded research.

Water erosion is known to be directly controlled by a number of factors, including climate, vegetation, soil properties and topography. Each factor is itself complex, and the various factors interact with one another. In creating an estimate of regional erosion risk, it is important to use a clear scientific rationale to combine relevant measures of these factors into composite indicators. This analysis is targeted on such a synthesis, working from a physical basis. The work package contains five components.

1. Model downscaling

The RDI model contains a clear basis for downscaling over space down to single hillslope elements, and over time down to the chosen time-step of the individual day. Downscaling over space is achievable because the basis of the model is a sediment transport equation, which can be evaluated at any point within a landscape. However, the estimates of local basal slope from local relief in a DEM, and of suitably weighted mean slope length inevitably mean that the parameters (especially erodibility) are far from scale free, and that effective parameters at the scale of the entire slope will need scale-dependent correction for applying at local intra-catena resolutions.

Over time, the mean erosion is explicitly integrated over the distribution, so that downscaling to the time step of a day should be relatively straightforward. However, for the finer resolutions which are certainly needed to comprehend the variability inherent in possible distributions of a given rainfall within a given storm total, there is a major intellectual challenge in temporal downscaling as well.

Theoretical investigations of the best way to implement these two aspects of downscaling, the spatial and the temporal, provide the largest area of scientific innovation within the work package. However, it is clear that decisions on how to proceed must be made at an early stage of the project to ensure that progress is made with other aspects of the work plan, and its role within the project.

The particular aspect of downscaling which is most important for DESERTLINKS is the explicit inclusion of indicators which are being developed for ESAs within target areas (WP 2.1). Our objective is to develop a basis for comparison between these scales which is scientifically sound and can be demonstrated to end users.

Part of this downscaling work and validation of the physical model is being explicitly addressed by UNIVLEEDS and other partners within the PESERA project. These studies will be available to DESERTLINKS and are not therefore charged to this project.

2. Development of salinity sub-model

Salinity is the second most important form of desertification in Europe. Increased irrigation is leading to secondary salinisation in many areas, including the Guadalentín. Moving further east or into north Africa, the problems become still more acute.

Preliminary work on salinity has concentrated on models for the 1-dimensional movement of solutes within the soil profile. For inclusion at a scale compatible with the RDI, the results of these investigations must be parameterised in terms of monthly water balances at each point, combined with a linearised model for solute uptake and, where appropriate, re-deposition within the soil profile. These climatic data must be combined with

• data on parent material composition, which defines susceptibility to salinisation,
  
• data on topography and its local relief, defining susceptibility to salinisation from groundwater and
  
• data on water supply and quality, defining potential for salinisation from irrigation waters.
  

With these data, a salinity model can provide data on both current and potential risks, which can be compared with data on saline soils to indicate areas of potentially increased risk under various land use, irrigation and climate scenarios. This phase of model development should follow the downscaling work, but should also be completed in good time to allow map products to be prepared from the finished model.

3. Development of channel delivery model

Channel delivery is important for the off-site impacts of erosion, which often have the highest short term economic impact, causing damage to settlements and loss of life. There are questions about how to manage flood plains, in relation to check dams, urban land use and groundwater recharge.

For flow within the channel network, sediment delivered to the slope base should be routed through the network. Channel sediment transport is generally considered to be strongly buffered by the effects of valley floor sedimentation or erosion, so that periods of high hillslope erosion lead to valley floor sedimentation, and periods of lowered hillslope input lead to channel incision. It is also widely observed that reservoir construction, by impounding sediment from upstream, leads to erosion below dams. In models for long term landscape development, the concept of the effective bedload fraction (ebf) has been developed as a simplifying approach. The ebf is the ratio of actual to capacity sediment transport, reflecting the transition from transport limited by capacity for coarse debris to transport limited by the supply of fines. This therefore depends explicitly on the grainsize distribution of source material and the availability of each grainsize fraction within the channel environment. Thus, as high erosion rates lead to valley floor sedimentation, the ebf rises towards its upper limit of 100%, giving a less than proportional increase in channel sediment transport. Similarly as sediment input falls, the ebf also falls, leading to scour of previously deposited fines and ultimately valley floor incision.

This approach to hillslope-channel sediment coupling offers an innovative re-working of existing data which is consistent with observed patterns of valley sedimentation, both over time and between reaches where there are downstream variation in hillslope sediment delivery. This aspect of the modelling should also follow the initial work on downscaling, and has considerable potential for application to regional problems of offsite sedimentation and reservoir siltation.

4. Provision of updated AVHRR land cover for Mediterranean Europe, and ofvegetation cover for target areas (EC-JRC-EI)

MODEM showed the potential and limitations of direct use of RS-based indicators on their own. We can use vegetation indicators effectively and they can give observable trends in relation to changes in grazing intensity and conversion to/from arable land or tree crops, but this is only one part of the erosion story.

Spectral unmixing techniques have proven their potential to provide semi-quantitative measures of vegetation cover density, which can be more directly compared at different spatial scales than conventional vegetation indices. Vegetation abundance, even if derived from data with quite different spectral and spatial characteristics, appears to be reasonably coherent through scales and thus has a strong potential to be further used to derive land cover and bio-physical vegetation characteristics from the target areas to the region. In this context alternative methods of regional scale (Mediterranean wide) land cover classification and vegetation cover attribute mapping (density, fragmentation, bio-mass etc) will be investigated and applied, based upon spectral mixture analysis applied to multi-temporal signatures of vegetation.

Change detection techniques such as regression and change vector analyses will be applied to identify areas of major changes. The assessment and interpretation of these observed changes will refer to and incorporate land use and climate change scenarios. Using these methods, it will become possible both to extend the coverage for Mediterranean Europe over time, and to add detail for the target areas. This is a crucial resource for both RDI maps and for linkage with ESAIS for target areas.

5. Preparation of European and regional scale maps for erosion and salinity risk

With the developments in modelling, one of the key deliverables is the production of maps for Europe. This requires land cover data, 1-km DEM and the European Soils Data Base.

It is proposed to provide monthly maps of average erosional loss, summed over the frequency distribution of storm events. The methodology also allows production of maps which express the erosion expected in events with an average recurrence interval of 10 or 100 years, which provide a basis for planning the responses to individual extreme events. In addition, the same maps will be generated for a range of likely land use and climate scenarios, to highlight areas where there are likely to deteriorating conditions. The primary basis for land use scenarios at a European scale and for target areas will be from research within the ongoing MEDACTION project. These will be integrated within an interactive web interface to a Synoptic Prediction System which integrates predictions of the climate, physical and socio-economic environment to create scenario-based forecasts of agricultural land-use and land degradation of the Mediterranean regional scale.

Although maps provide an important deliverable from the work package, they lack the flexibility to forecast the impact of all alternative scenarios or policy options. In Work Package 3.2 it is therefore proposed to provide the software to the Annex IV National Committees and other policy makers directly. This will then provide a dynamic tool for evaluating the impact on desertification of policy alternatives. This software will clearly be delivered at a late stage in the project, when it has been validated and debugged as fully as possible.

Results. The output from this work package is intended to be directly useable by policy makers, as end users at both regional and European scales. Maps are provided of erosion risk, vegetation and land use history (Deliverables 2.2a and 2.2b). The results of this work package will also underpin research in WP 3.2 and 3.4, and contribute to deliverables 3.2a-d.