Flooding in the desert: Assessing the value of satellite observations of inundation from the North American Monsoon with the Pima County Regional Flood Control District

Floods affect more people than any other hazard, and the frequency and magnitude of exposure is growing with demographic and climatic changes. It is increasingly difficult to model floodplains accurately with these changes. Government based floodplain maps from FEMA or developed by local flood control districts may contain errors or underpredict risk. Especially in rural areas with underserved populations this can result in limited or blocked access for residents and emergency response personnel. Yet the increasing availability, frequency, and spatio-temporal resolution of satellite data provides new opportunities to monitor floods and identify locations where floodplain maps may contain error. One of the most challenging places to map floods with satellites is in the desert, where water moves quickly and may fade before the satellite overpass, and smooth sandy surfaces make distinguishing water in radar based water detection algorithms a challenging task. Geomorphic conditions creating areas of broad, shallow floodplain can further complicate mapping by creating poorly defined distributary flow networks. However, high resolution (3-5m) daily optical imagery from commercial data providers such as Planet could provide value for desert environments where clouds clear quickly. We are assessing the value of satellite imagery to provide additional flood information to the Pima County Regional Flood Control District, by mapping major floods from the 2021 and 2022 Monsoon seasons from satellites, and comparing it to existing floodplain maps from both FEMA and the County, along with ground reference data on flood road closures from the Department of Transportation. We will also show the potential for using satellites for near real time flood monitoring of the monsoon in desert conditions by comparing hydrograph peaks from stream gauges to timing of satellite overpasses. We will compare the value of commercial imagery with freely available public satellites including Landsat, Sentinel-2, and Sentinel-1. Inundation will be mapped using deep learning algorithms developed at the SocialPixel lab at the University of Arizona. This project demonstrates how local governments could assess the value of emerging data sources to improve understanding of risk and emergency response with University partnerships.