Distributed Machine Learning and Data Fusion for Flood Detection and Monitoring

Abstract:

Today, multiple approaches exist for flood detection and monitoring from map generation to coordinating satellite observations with river discharge measurements. The common design concept is based on a multi-temporal processing chain using a combination of optical and/or radar-based platforms. Flood monitoring is one of the main common areas of interest. This scientific focus is also an area of application for observational data generated from the upcoming Surface Water and Ocean Topography (SWOT) mission. In the framework of this mission, a joint data science project was established between the NASA Jet Propulsion Laboratory (JPL) and the Centre National d'Études Spatiales (CNES) to develop a Machine Learning driven solution for automatic flood monitoring by integrating optical/radar imaging with satellite remote sensing and river gauges measurements. An area of interest is monitoring floods in regions where water extents can be derived through optical/radar analysis. The project serves as an example of distributed analytics solution that harmonizes multivariate data to provide an environment for ongoing flood monitoring and investigation. We will discuss the current status of our project, called FloodML, and our plan in further our effort in applying data science in identifying and understanding the impact of coastal and regional flood events as the result of climate change.