Change detection in Swedish land cover from a year of ERS-1 SAR data

The information for ten agricultural and seven forest land cover classes are analysed using eight ERS-1 SAR images, covering a full vegetation cycle in southern Sweden. Backscattering coefficients (σ o ) were derived for these classes and compared both visually and statistically. Jeffries-Matusita distances were calculated to determine the statistical separabilities between classes and to determine optimal image combinations for given classification applications. It is shown that ERS-1 data can be used to follow changes in agricultural land cover but not forest land cover, aside from possibly distinguishing fresh clearings over time. Data from the ERS-1 SAR may show potential in effectively classifying agricultural land cover, if applied together with optical data. The data appears to be of limited practical benefit if applied alone. It is not suitable in distinguishing forest land cover classes in Swedish environments.