Research project T4/DD/02 (Research action T4)
Formulation of the problem
The demand for information about the global distribution of land-cover types and the changes in global land-cover is very large. To be able to accurately measure these changes, sensors on satellite platforms may be used, which are able to deliver data on vegetation at regular points in time for the whole of the globe. One of these sensors is the VEGETATION instrument, which is scheduled for launch on the SPOT 4 platform, for which the image processing centre and the global archive is located at the Centre for Teledetection of VITO. A major problem in transforming the medium resolution data delivered by the VEGETATION instrument towards data which are useful to the description of world biomes at a global scale, is the degradation of these data to data at a much coarser level of resolution. In this degradation, information is lost, especially with regard to the proportions of biome-classes. The objective of this project is to limit these information losses through the combined use of advanced classification methods and mathematical-statistical models.
Proposed methods for solution
The core idea behind this project is to develop a global method for biome classification which will be based on phenological differences between the different cover types. To account for the loss of proportional accuracy due to spatial degradation of image data two different approaches for the estimation of land-cover proportions will be explored:
(a) modelling of the relationship between the spatial characteristics of land-cover distribution and the scale-dependent changes in the proportion of different cover types;
(b) modelling of the proportion of component land-cover classes for each pixel in the coarse resolution image, using fuzzy classification.
Several experimental designs will be realised in order to calibrate, verify and validate the models developed in the course of this project in an optimal way.
Rationale for the research
Many scientists (a.o. geographers, researchers on the biosphere, ecologists and people involved in global change research) are in need of accurate and up-to-date information about the global distribution of major cover types, to be produced through the use of the models elaborated in this project. The project serves as a base for the development of an operational method for the production of land-cover data that can be applied to the data delivered by the VEGETATION instrument and may be of use to a large number of customers.
