Chini et al. in [1], developed an algorithm to automatically map urban floods from Synthetic Aperture Radar (SAR) data, which can function in near-real time worldwide. It has been tested on the extreme Pakistani floods, which started in the summer of 2022 and lasted for several months. This test case also served to study the characteristics of interferometric phase (in-phase) Standard Deviation (SD), to ensure the preservation of the original information. Then, using four classification methods: thresholding, region growing (RG), Bayesian network fusion (BNF) and BNF with RG, all the possible combinations between intensity, coherence and in-phase are tested to map floods in urban areas. For smaller areas (tens of square kilometers), it appears that in-phase SD does not add significantly more information. However, for larger areas (more than hundreds of square kilometers), the combination of the three SAR features in a BNF provides the best results.