Fuzzy classification of JERS-1 SAR data: an evaluation of its performance for soil salinity mapping

Remote sensing of surface features has been used intensively to identify and map salt-affected areas. Salt-tolerant vegetation is among the indicators used to separate saline-alkaline areas from non-affected ones. However, this type of vegetation causes spectral confusion and erroneous labelling between salinity and alkalinity classes when working with optical sensors such as Landstat TM or Spot. Accordingly, this paper evaluates the capabilities of the microwave range to map saline and alkaline areas. Fuzzy sets are used to model the information classes, and a fuzzy overlay model is implemented to classify the JERS-1 radar satellite image. The study shows that fuzzy classification of JERS-1 SAR data provides reliable detection (overall accuracy equal to 81%) of areas degraded by salinity-alkalinity processes. The main problems appear to be due to the interaction between soil roughness and radar backscattering, which determined erroneous allocation of alkaline and saline-alkaline areas to non-affected areas.