Towards deep learning based flood forecasting for ungauged basins
Sprache des Titels:
Englisch
Original Buchtitel:
EGU General Assembly 2020
Original Kurzfassung:
Floods are among the most destructive natural hazards in the world. To reduce flood induceddamages and casualties, streamflow forecasts should be as accurate as possible.As of today, streamflow forecasts are usually made with either conceptual or process-basedhydrological models. The problem these models usually have is that they perform best whencalibrated for a specific basin, and performance degrades drastically if the models are used inplaces without historic streamflow measurements. To make things worse, some of the mostdevastating floods occur in developing and low-income countries, where historic records ofstreamflow measurements are scarce. Therefore, a central task for enhancing flood forecasts andhelping local authorities to manage these areas is to provide high-quality streamflow forecasts inungauged rivers. Although the IAHS dedicated an entire decade (2003-2012) to advance theproblem of Prediction in Ungauged Basins the central goal remains largely a challenge.In this talk, we will present a novel approach for tackling the problem of prediction in ungaugedbasins using a data-driven approach. More concretely, we show that the Long Short-Term Memorynetwork (LSTM), which is a special type of a deep learning model, can serve as a generalizablerainfall-runoff simulation model. We will present recent results indicating that the LSTM gives onaverage better out-of-sample predictions (ungauged prediction) than e.g. the SAC-SMA in-sample(gauged) or the US National Water Model (Kratzert et al., 2019).One place where these research results are already finding their way into operation is Google?sFlood Forecasting Initiative. The goal of this initiative is to provide (enhanced) flood warnings,where needed, starting with a pilot project in India. And as mentioned above, historic streamflowrecords in those regions are scarce, which motivates new and innovative approaches for enhancedstreamflow forecasting.
Forschungs-