Evaluation of Radarsat multitemporal data for crop discrimination in highly cultivated areas

The aim of this paper is to evaluate the potential of Radarsat fine mode multitemporal data for crop discrimination and monitoring. The test site is an intensively cultivated region in Sao Paulo State, Brazil, which has various classes of land use, such as sugarcane, pastures, corn, potato, tomato, fallow fields, beans, urban area, water, bare soil, and natural vegetation. Radarsat F4 data from Jan/05/98, Feb/22/98 and Mar/18/98 were calibrated to produce a multitemporal composition and to extract the backscattering coefficient. Backscattering values statistical analysis produced confidence interval for each class of land use. The statistical analysis and the graphical profile of temporal variation of backscattering of crops presented a significant separability between crops using the multitemporal information. The best single date to discriminate crops was January because of better differences in the phenological stages of crops in the test field: corn starting the ripping, sugarcane still in the vegetative stages, tomato in the early stages, pasture with high biomass, beans in the early stages, bare soil tilled (rough) and summer fallow. Differences in the soil cover, plant height, plant density and crop growth stage contributed to decrease the separability of crops in the same class of land use