Datensatz

Zusammenfassung

Different approaches have been developed and used for global land cover mapping. Differences, heterogeneities and problems are largely understood and the logical next step is to combine a suite of datasets into a "best" map based on accuracy data. The prototype activity here uses global probabilities derived in a comparative accuracy assessment in combination with the spatial homogeneity to generate a synthetic new dataset of higher quality. A pixel-probability was derived from a comparative validation of the four input datasets and used to derive a global best map prototype with coarse thematic detail. We present the approach using 5 generic classes.