Global scale biome classification and estimation of biome proportions using VEGETATION data

You are here

Context and objectives

Notwithstanding recent advances in optical remote sensing techniques at the level of the multi temporal estimation of biophysical characteristics - such as leaf area index and fraction of absorbed photosynthetically active radiation - robust monitoring as well as early warning systems based on global observation data, are not developed at their full scale capacity at this moment. Especially with respect to shifts in biome distributions, very few global and operational monitoring systems or approaches exist.

The consistent and repeatable derivation of maps with a global coverage, of the current spatial distribution of biome classes, and how these biomes are changing, has become a feasible objective due to recently launched platforms such as SPOT-VEGETATION. The fundamental objective of this proposal is the definition, as well as implementation of a robust classification strategy for the characterisation of major biome distributions at a global scale, based on SPOT-VEGETATION data for a full yearly vegetation cycle. The VGT-PS decadal synthesis NDVI product will be used for this purpose, as the ground segment of the VEGETATION mission produces it quite recently.

To account for the loss of proportional accuracy due to image classification at a coarse resolution, models for improving biome area estimates at a global scale will be developed. The models will be tuned to the major biome classes that will be defined, and integrated into the global classification strategy that will be proposed. Establishing an efficient procedure for improved area estimation from very coarse resolution biome classifications, as derived from the VGT-PS product, will add value to the classified data, and may substantially increase their potential use.

Project outcome

Expected scientific results

On the global scale, and quasi-operational:
- an innovative method for the compression of the immense data sets delivered by SPOT-VGT (or similar sensors)
- an update of the DISCover IGBP-classification of 1992/3, derived from the VGT-imagery of 1998/9 (product and method). In an experimental stage, and mainly tested for Europe:
- a comprehensive evaluation of the performance of different sub-pixel classification strategies applied at the continental scale (part of Europe covered by CORINE + three test zones in Africa)
- sub-pixel proportion estimates for various land-cover classes at 1km x 1km and 10km x 10km resolution for the entire European continent
- aggregated data on land-cover class proportions and maps of proportional error for 19 European countries, at the regional and national levels