Land Cover/land Use Classification in a Semiarid Environment in East Africa Using Multi-temporal Alternating Polarisation Envisat Asar Data

In central Kenya, in the ecological transition zone from semi -humid to semi-arid, climatic and human induced land cover changes have an immense impact on the ecosystem. Therefore, investigations were made to analyse the potential to separate 10 detailed land use/land cover classes in this area from multitemporal ENVISAT ASAR data. Besides mean, different texture measures like variance, coefficient of variation and semivariogram were calculated to generate additional input layers for a classification procedure. A quantitative separability criterion was used to determine the contribution of each measure to the overall class separability and for specific classes. The results show that the contribution of the different texture measures is class dependent. However, multitemporality provides most class discrimination. While the different polarizations individually do not show a distinct influence on one specific class, still an overall better separability is obtained when using both polarization images of one date.