A High-Resolution Flood Inundation Archive (2016–the Present) from Sentinel-1 SAR Imagery over CONUS

Abstract

Most existing inundation inventories are based on surveys, news, or passive remote sensing imagery. Affected by spatiotemporal resolution or weather conditions, these inventories are limited in spatial details or coverage. Satellite Synthetic Aperture Radar (SAR) data has recently enabled flood mapping at unprecedented spatiotemporal resolution. However, the bottleneck in producing SAR-based flood maps is the requirement of expert manual processing to maintain acceptable accuracy by most SAR-driven mapping techniques. To fill the vacancy, we generate a high-resolution (10 m) flood inundation dataset over the contiguous United States (CONUS) from nearly the entire Sentinel-1 SAR archive (from January 2016 to the present), using a recently developed automated Radar Produced Inundation Diary (RAPID) system. RAPID uses U.S. Geological Survey (USGS) water watch system and accumulated precipitation to identify SAR images that potentially overlap a flood event. The dataset includes inundation events with the temporal scale from several days to months. Concluded from all 559 overlapping images in the period from 2016 to the first half of 2019, the comparison of the proposed dataset against the USGS Dynamic Surface Water Extent (DSWE) product yields an overall, user, producer agreements, and critical success index of 99.06%, 87.63%, 91.76%, and 81.23%, respectively, demonstrating the high accuracy of the proposed dataset and the robustness of the automated system. We anticipate this archive to facilitate many applications, including large-scale flood loss and risk assessment, and inundation model calibration and validation.

Automatic Open Water Flood Detection from Sentinel-1 Multi-Temporal Imagery

“…On the other hand, if the information on flood occurrence exists and the AOI covering potentially affected areas is provided, the method could be used in the ondemand mode while using only its change detection results for approximate delimitation of floods extent. A similar approach was described in [10] where a unique triggering mechanism was developed based on gauge stations and satellite precipitation estimations, identifying potentially flooded zones. Similarly, alerts issued by national meteorological services could be used.…”

“…They focused only on the selected flood events themselves. Other authors, such as [10], do not operate their system permanently, but only in areas and at times when their triggering mechanism reported potential floods. We consider the presented long-term tests to be an essential part of the quality assessment of our developed solution.…”

“…It is based on a normalized difference ratio comparing pre-event and post-event images, using both VV and VH (vertical transmit and horizontal receive) polarisations. Furthermore, more recently, automatic flood detection algorithms were developed from Sentinel-1 imagery based on neural networks [8,9], kernel algorithms and machine learning [10], or alternating decision trees [11].…”

“…Most of the studies mentioned above were performed only to evaluate a specific event or a limited set of events. Only study of Yang et al was broader, covering the continental part of the USA for three and a half years [10]. However, even this study was tied to US-specific conditions.…”

“…The developed system was closely linked to other existing structures in the USA, such as the United States Geological Survey, a network of gauge stations, etc. Some other solutions, e.g., [6][7][8][9][10][11]13,14,17,18,24,26], relied on the use of computational intelligence tools, indexes and thresholding, or on specific auxiliary data sets. The computational intelligence tools used, such as neural networks, must always be trained on sample data first and only afterwards can be used to evaluate the current situation.…”