ML for Flood Forecasting at Scale

TL;DR

This paper explores the application of machine learning techniques to improve large-scale riverine flood forecasting, aiming to overcome data limitations, reduce reliance on human calibration, and enhance global prediction accuracy.

Contribution

It proposes a novel ML framework leveraging transfer and multitask learning to improve flood prediction accuracy at continental and global scales.

Findings

ML models outperform human experts in complex flood scenarios

Transfer learning enhances local and global flood prediction accuracy

Proposed system enables timely and accurate flood forecasts at scale

Abstract

Effective riverine flood forecasting at scale is hindered by a multitude of factors, most notably the need to rely on human calibration in current methodology, the limited amount of data for a specific location, and the computational difficulty of building continent/global level models that are sufficiently accurate. Machine learning (ML) is primed to be useful in this scenario: learned models often surpass human experts in complex high-dimensional scenarios, and the framework of transfer or multitask learning is an appealing solution for leveraging local signals to achieve improved global performance. We propose to build on these strengths and develop ML systems for timely and accurate riverine flood prediction.