Evaluation of SAR and Optical Data for Flood Delineation Using Supervised and Unsupervised Classification

Precise and accurate delineation of flooding areas with synthetic aperture radar (SAR) multi-spectral (MS) data is challenging because flooded are inherently heterogeneous as emergent vegetation (EV) turbid water (TW) common. We addressed these challenges by developing applying a new stepwise sequence unsupervised supervised classification methods using both SAR MS data. The signatures land targets in the study area were evaluated prior to identify classes that could be delineated. based on simple thresholding method provided satisfactory estimate total but did not perform well surface water. To deal heterogeneity fragmentation patches, approach combination segmentation (CThS) was developed. Since sandy classified SAR-based methods, random forest (RF) applied time series co-event data, combined separated. tested for determining flood extent two events Italy. results showed all outperformed ones Although RF may lead better accuracies, CThS (unsupervised) achieved precision accuracy comparable RF, making it more appropriate rapid mapping due its ease implementation.