Content area
Full text
Introduction
Floods are the deadliest natural hazards, striking numerous regions in the world each year1. Floods cause 40 billion dollars (2015 USD) in damages annually2 and affected 2.5 billion people between 1994 and 20143. Furthermore, the population of people living in flood-prone areas is increasing due to migration and population growth4,5. All of these impacts are expected to become even more severe with climate change6.
A key factor in understanding and mitigating impacts from flooding is knowing where flooding occurs on a regular basis. Accurate flood extent mapping provides essential data for various purposes. It helps urban planners design resilient infrastructure, aids in developing early warning systems, and supports insurance companies and policymakers in assessing risks and allocating resources. By understanding past flooding events, communities can better prepare for future occurrences, leading to safer and more resilient living environments. While mapping flood extent is important, doing this via on-the-ground efforts is often challenging, especially in developing countries, where resources for this time-intensive work are scarce. In addition, ground-based assessments are often for small areas.
Satellite data offers a powerful solution for mapping flood extent at scale. Two primary types of sensors are commonly used for this purpose: optical/infrared and Synthetic Aperture Radar (SAR). Optical/infrared sensors passively capture reflected light, yielding familiar photographic images with benefits like wide availability and frequent observations at various resolutions. However, their effectiveness is limited by cloud cover and a dependence on daylight. In contrast, SAR actively emits microwave signals and records their reflections, offering advantages such as being able to penetrate through cloud cover and operate in both day and night conditions. While SAR is often described as all-weather, its signal may be affected during extremely heavy rainfall7,8. However, SAR satellites like Sentinel-1 typically provide observations every 6–12 days for a given location, whereas optical satellite constellations have revisit times ranging from daily to several days, often enabling more frequent coverage than SAR. This lower temporal frequency can affect our ability to capture the full dynamics of flood events with SAR, particularly flash floods and peak flood extent that may occur between SAR observations. Additionally, both technologies face challenges in complex terrain such as narrow valleys, which are common...