With the developments of airborne, and recently spaceborne, fully polarimetric synthetic aperture radar (POL-SAR) sensors, microwave remote sensing offers new opportunity to observe and understand geophysical phenomena. However, extracting information from the multi-component complex POLSAR images requires several processing stages, including the POLSAR feature estimation to reveal different backscattering mechanisms, and the fusion of these features to help the end-user to perform detection or classification tasks. In this paper, a new data fusion method based on interactive learning of implicative fuzzy rules is presented. It allows to perform supervised classification by analyzing the training set clusters in the different 2D feature spaces resulting from the different attribute pairs. Experimental results are presented on a real POLSAR data set acquired by the airborne E-SAR system over temperate glaciers in the Alps. They illustrate the interest of the proposed fusion approach for POLSAR data analysis and the potential of the POLSAR imagery for the monitoring of temperate glacier evolution. 1