Classification of the Gabon SAR mosaic using a wavelet based rule classifier

The paper mainly focused on a method for semi-supervised classification of the SAR mosaic. However, results explicitly demonstrate the contribution of each input to distinction between classes. It also shows the limitations of using JERS-1 SAR configuration for monitoring of tropical forests. That is, it is mostly sensitive to biomass extremes (ex: savanna-forest) and vegetation moisture (e.g. inundated areas). The decision tree classifier can be used with different inputs. Thus, studies with different texture measures will be performed. The authors also need to compare this classifier to classical ones such as the ML method. The resulting errors and artefact demonstrate the importance of choosing the training sites. This is readily seen when comparing misclassification. Thus, these result must be considered when choosing the training areas.