SENTINEL-1 AND COSMO-SKYMED DATA FUSION FOR HYDROPERIOD ESTIMATION IN A COASTAL WETLAND AREA

The inundation time of temporary wetlands is a key parameter affecting plant and animal communities. Reliable detection methodologies are needed for their monitoring, especially under climate change, to understand the trends and dynamics ongoing. In this study, conducted on a protected coastal wetland (“Zone Umide della Capitanata”), Italy, we propose an approach based on random forest to fuse Sentinel- 1 and COSMO-SkyMed data to obtain a time series of water maps with both high spatial and temporal resolutions, by using topographic and climatic predictors. Validation shows a high Overall Accuracy (0.91), demonstrating a good result in general, however, both User’s Accuracy and F1-score of water class (0.47 and 0.59) reveal an overestimation of the presence of water, confirmed by a low Kappa coefficient (0.54). In conclusion, this approach represents a promising attempt at improving hydroperiod estimation, although there is still room for improvement.