Authors: Sadra Seyvani, Sagy Cohen, Parvaneh Nikrou – University of Alabama
Title: Assessing Accuracy and Limitations: Current InSAR-Based Flood Inundation Mapping Algorithms Across Diverse Terrains Characteristics and Conditions
Abstracts: Floods are one of the natural disasters their occurrence rate and intensity have increased in recent years due to climate change, land use changes, etc. As a result, producing accurate flood inundation mapping is vital to estimate the damage, provide relief after a flood, study to provide suitable civil infrastructure to prevent future damages, etc. In recent years, thanks to the increase in the availability of InSAR satellite data, as well as significant improvements in the spatial and temporal resolution of these data, the use of these data for flood inundation mapping offers many advantages, including high accuracy and almost global coverage. However, these algorithms still face challenges in accurately flood inundation mapping in some areas such as urban areas and areas with dense vegetation due to complex scattering patterns. This study aims to compare the flood inundation maps produced by the current InSAR-based flood inundation mapping algorithms with the reference maps provided by NASA for the ETCI 2021 Competition on Flood Detection and address the limitations of using these algorithms in different terrain characteristics and conditions. For this study, we will create flood inundation maps using a current algorithm and then identify any areas that are incorrectly labeled as flooded or non-flooded. Then, we will use image processing techniques and apply various image filters to the generated flood inundation maps, as well as maps of precipitation, temperature, land use, land cover, topography, and other geomorphological maps of the study area to generate new data. Finally, we will analyze the data to identify any logical trends using statistical and machine learning methods in areas that are incorrectly labeled as flooded or non-flooded. so that by using its results, new and accurate algorithms can be presented in future studies.