Evidential Reasoning Applied to Mapping Regeneration of Forest Stand using Multisource Geospatial Data

The provincial inventory of Quebec suffers from a lack of accurate information in the regenerating stands, at a subparcel scale. Moreover, the information contained in the satellite images bands is not sufficient to be used alone to mitigate this issue. We used the Dempster-Shafer Theory (DST), a data fusion method, to incorporate ancillary information coming from biophysical parameters involved in the regeneration potential. We showed that fusing the spectral bands of the SPOT-5 image with the drainage and surface deposit maps best improved the quality of the mapping of deciduous forest stands. Compared to the results obtained with a maximum likelihood classifier, the increase of the overall classification rate was 8.5 % (from 82.75 % to 90.15 %).