As the impacts of climate change become more frequent and intense, emergency management services require increased accuracy for monitoring flood impacts. Remote sensing offers the opportunity to address these concerns in a cost-effective manner; in particular, synthetic aperture radar (SAR) sensors are often utilized to accomplish this as a result of their penetration of cloud cover and ability to image during both day and night. The Red River of the North, located in north central United States, is a region that faces annual flooding due to spring snowmelt conditions, and thus offers an ideal location for evaluation of flood product accuracy across different satellite-derived flood inundation extent maps. In this study, we offer a time series comparison of flood products derived from both high-resolution optical data and SAR images of the Red River region. The utilization of a new Self-Supervised Water Detection (SWD) method for flood mapping optical imagery shows promising results in validating SAR-derived flood inundation mapping (FIM).

Currently, more than two billion people live in or near-coastal zones at the ocean-land interface with almost a billion more living in adjacent low-lying coastal areas. These areas and populations are at risk from increasingly severe storms and longer-term sea level rise resulting in coastal erosion, water pollution, urban and agricultural inundation and ecosystems degradation. Earth Science Digital Twins, that is, the combination of data, models, and AI/ML technologies to simulate Earth system processes and enables short- and long-term forecasts, provide understanding and actionable information to reduce risks to humans, infrastructure and ecosystems. Satellite and in situ data are critical components of a coastal zone digital twin (CZDT) to provide timely and spatially relevant input data for models and serve as a check or validation of digital twin performance.The goal of this paper is to introduce a CZDT concept being developed as part the Satellite Climate Observatory with joint participation of NASA, NOAA and CNES and discuss initial use cases, satellite data, modeling, and advanced technology components.

During the Asian summer monsoon rainy season, heavy and continuous rainfall usually causes widespread floods in many Asian countries, including the two most populated countries China and India. Flood mapping datasets from moderate-resolution satellites, such as the JPSS (Joint Polar Satellite System) series including Suomi-NPP (Suomi National Polar-orbiting Partnership) and NOAA-20, can be invaluable for monitoring floods and assessing the affected population. The flood maps derived from the VIIRS (Visible Infrared Imaging Radiometer Suite) imagery can provide a big picture of what’s on the ground over the entire Asia flooding region. The VIIRS 5-day composite flood maps, along with a population density dataset, can be combined to estimate the population potentially exposed (PPE) to flooding. The VIIRS flood maps demonstrate floods in China primarily occurred along the Yangtze River Basin and its tributaries. The VIIRS 5-day composite flood maps, along with a population density dataset, were combined to estimate the population potentially exposed to flooding. Here we use the summer of 2020 as an example, based on the flood extent along with the population density map, approximately over 50 million people in the entire China might have been affected by the floodwaters. In addition to China, several other countries including India, Bangladesh, and Myanmar were also affected. In India, the worst inundation and the affected population were located mainly in the northern states of Bihar, Assam, and West Bengal.

Remote-sensed flood monitoring is rapidly advanced by the growing abundance of satellite data. This study presents the progress in building an operational system that harnesses the power of spatially high-resolution (HR), multi-source remote sensing data for event-based flood extent and depth mapping. By integrating pioneering extent retrieval—Radar Produced Inundation Diary (RAPID), Self-supervised Waterbody Detection (SWD), and depth retrieval processors, Global-LOCAL Solvers Integration Algorithm (GLOCAL), and the Emulated Flood Recession Algorithm (EFRA)—our proposed framework marks a significant leap in flood monitoring capabilities. The upgraded RAPID addresses the complexities of Synthetic Aperture Radar (SAR) flood mapping in diverse environments, including snow-covered and arid regions, enhancing adaptability and consistency across multiple SAR satellites such as ESA Sentinel-1, CSA RADARSAT Constellation Mission (RCM), MDA RADARSAT-2 (RS2), and Capella. Meanwhile, SWD brings a method to automatically map flood extents from high-resolution optical images, including Planet, Sentinel-2, and Landsat, during clear weather conditions. GLOCAL and EFRA are tailored for depth estimation from the generated HR remotely sensed flood extents, and HR or VHR topography. Both algorithms tolerate the error from extent and topography, with EFRA physically considering the influence of micro-topography. The proposed framework has proven to produce consistent flood extent results across multi-source, high-resolution satellite images and demonstrates robust depth estimation with various input DEMs. Validation against USGS high water masks yields 1.01 m RMSE when 1 m resolution DEM is used. With continuous development and integration into NOAA's near real-time operational platform, these approaches do not only improve the resolution and dependability of operational flood monitoring globally but also significantly bolster emergency response strategies.

Long-term flood mapping datasets can be invaluable for historic flood investigation, flood potential or probability estimate, time series analysis and modelling, and climate change studies. With the developed flood detection algorithm and software for JPSS/VIIRS (Visible Infrared Imaging Radiometer Suite) (Li et al., 2017), in this study, VIIRS historic data since 2012 has been reprocessed from JPSS (Joint Polar Satellite System) series including Suomi-NPP (Suomi National Polar-orbiting Partnership) and NOAA-20. VIIRS global flood time series datasets have been generated and distributed by NOAA through Amazon Web Services (AWS). The derived dataset includes granule flood product, daily and 5-day composited flood products in netCDF4, geotiff and shapefile formats from 2012 to 2020. The dataset not only provides data records of historic flood events, but also shows potential in flood analysis and modelling. With the dataset, a simple application using annual composition is performed to analyze the annual change of flood extent globally and in each continent. The analysis has shown a slightly increasing trend in flood extent at a global scale, but varying in different regions.

One remaining challenge in synthetic aperture radar (SAR) based flood mapping is the flood detection over urban areas, particularly around buildings. To address it, we propose an unsupervised change detection method that decomposes SAR images into noise and ground condition information, and to detect flood induced changes using out-of-distribution detection. The proposed method is based on adversarial generative learning, specifically using the multimodal unsupervised image translation model. It involves training the model by reconstructing SAR images under different combinations of noise models and ground conditions. The method is tested using Sentinel-1 dual-polarization and ancillary data in Hurricane Harvey 2017 over the urban area of Houston. Subsequent research endeavors will be dedicated to enhancing the method's stability, conducting additional validation, and extending its application to diverse SAR sensors.

This paper presents a new approach to operational snow depth analysis. The analysis uses 2-dimensional optimal interpolation to blend various snow depth data weighted against their errors relative to a first guess and spatial correlations. The blended analysis is applied operationally within the National Oceanic and Atmospheric Administration (NOAA)'s Interactive Multi-Sensor Snow and Ice Mapping System (IMS) at its snow cover-classified grid points over the Northern Hemisphere. The 4-km snow depth estimates are blended from satellite-derived estimates of the Advanced Microwave Scanning Radiometer 2 (AMSR2) or the Advanced Technology Microwave Sounder (ATMS), in-situ surface reports and analyst estimates. Unique to the production is the application of snow depth estimates with the associated confidence values generated interactively from the analyst that are also ingested into the objective analysis and fully consistent with optimal interpolation method.