| dc.description.abstract | The S1 (Sentinel 1) and S2 (Sentinel 2) missions use high-resolution satellite imagery to map the land surface over large swaths. With a minimum 5-day and 6-day repeat time, imagery over earths land can be quickly obtained in S1 synthetic aperture radar imagery and S2 optical imagery. Used for a wide range of studies, the S1 and S2 missions have recently been applied to flood inundation mapping. With a global presence and high degree damage potential of a flood, studies on satellite flood mapping are important to investigate. Based on this, the thesis proposes a flood mapping method on the powerful cloud processing platform Google Earth Engine (GEE). Methodology for the thesis is presented on two cases in Norway: a rain flood in Stjørdalen and a snowmelt flood in Sunndal. The third case study was a monsoon flood in Layyah, Pakistan. Gathered data products of S1 and S2 were initially filtered to enhance and improve quality, then visualized in different polarizations in S1 and calculated in indices in S2, before a flood masking map finally was created. Additional data were gathered from a DSM (Digital Surface Model) where misclassifications from terrain could be corrected. Each study area experienced problems. Clouds were present in Stjørdalen and Layyah, concealing flooded areas in S2 imagery. Snow cover could lead to flood misclassification in Stjørdalen. Different concentrations of suspended sediments found in the floodwaters of Layyah made detection of flood difficult in some areas. Sunndal proved small scale changes in flood could be hard to detect in S1 imagery due to speckle filtering options. Despite the challenges, this study found that implementing reference stacks, pixel and elevation masking and detection type comparisons significantly improved flood detection using S1 and S2 data. The results suggest that optical and SAR-based flood inundation mapping can provide valuable support in detailing flooded areas. | eng |