FLOOD-WATER LEVEL ESTIMATION FROM SOCIAL MEDIA IMAGES

We introduce a new deep-learning model that addresses limitations of existing urban flood models in handling varied terrains and rainfall events. Our model subdivides the city into small patches and presents a novel approach to incorporate broader spatial information. It accurately predicts high-resolution flood maps across diverse rainfall events and cities (on a minutes and meters scale) that haven’t been seen by the model, which offers valuable insights for urban flood mitigation strategies.

Floods in urban areas are one of the most common natural hazards. Due to climate change enhancing extreme rainfall and cities becoming larger and denser, the impacts of these events are expected to increase. A fast and reliable flood warning system should thus be implemented in flood-prone cities to warn the public of upcoming floods. The purpose of this brief communication is to discuss the potential implementation of low-cost acoustic rainfall sensors in short-term flood warning systems.

Grain size analysis is the key to understanding the sediment dynamics of river systems and is an important indicator for mitigating flood risk and preserving biodiversity in aquatic habitats. We propose GRAINet, a data-driven approach based on deep learning, to regress grain size distributions from georeferenced UAV images. This allows for a holistic analysis of entire gravel bars, resulting in robust grading curves and high-resolution maps of spatial grain size distribution at large scale.

This work demonstrates a new approach to obtain flood level trend information from surveillance footage with minimal prior information. A neural network trained to detect flood water is applied to video frames to create a qualitative flooding metric (namely, SOFI). The correlation between the real water trend and SOFI was found to be 75 % on average (based on six videos of flooding under various circumstances). SOFI could be used for flood model calibration, to increase model reliability.