Extension of environmental parameters over the land surface by improved fuzzy classification of remotely sensed data

Abstract

Recent studies have shown that useful information can be derived from remotely sensed data to extend local environmental measurements over the land surface. A flexible fuzzy classification approach, in particular, was demonstrated to be superior to conventional multivariate regression methods for this purpose. This new extension methodology is described and discussed in detail. Next, some modifications are proposed in order to improve the consideration of the spectral information and to produce per-pixel estimates of the error committed. The performances of the original and modified methodologies are then evaluated in two case studies representative of different parameters to extend and satellite data. The results show the efficiency of the new strategy and, in particular, some improvements obtained by the modified method which testify to the effectiveness of the modifications proposed.