Physics-informed Machine Learning (Φ-ML) Meets Engineering Webinar

Thursday 12 Jan 2023, 13:00 → 14:00 UTC

FloodSENS – Smart Sensing of Floods

FloodSENS seeks to exploit optical satellite data for rapid flood mapping. While SAR (Synthetic Aperture Radar) penetrates clouds and is a prime candidate for flood mapping it faces performance limitations in urban areas and areas with dense vegetation. FloodSENS is intended to operate as a complementary algorithm to address the limitations of SAR based flood mapping.
FloodSENS uses Machine Learning and formulates flood mapping as a supervised segmentation problem. Alongside the optical data from Sentinel-2 auxiliary data is introduced to guide the learning process. Auxiliary data includes simple quantities such as DEM (Digital Elevation Model) and slope and advanced derived hydrologic indices such as TWI (Topographic Wetness Index).
A multitude of interesting challenges were uncovered during the development of FloodSENS. Transferability between different geographic areas is a general challenge in Earth Observation. For Machine Learning approaches it is hard to achieve good generalization for various topographies, climates and biomes that exist.
A second challenge is the frequent presence of substantial cloud cover. FloodSENS targets partial reconstruction under cloud cover to maximize the information provided to end users. This activity is being developed by the Luxembourg-based company RSS-Hydro and in partnership with ESA’s InCubed program. The main objective is to offer a better flood disaster response. Possible applications target the humanitarian and disaster relief sector, and the global (re)insurance market.

Speaker: Dr. Guy Schumann (CEO RSS-Hydro), Ben Gaffinet Applied Physicist RSS-Hydro)

ZOOM link:
https://turing-uk.zoom.us/j/99066088699?pwd=OXZFamJ4eW1XOWtXelZWQzNJRWkyZz09 

meeting ID: 990 6608 8699 / passcode: 484477

More Information / Subscribe / Past Seminars
 https://www.notion.so/ML-meets-Engineering-fa48aefc1a7f40e4b98cb6f861f766cd