Integrating meteorological and hydrological modeling in KI-HopE-DE: AI-basedflood prediction in small river basin in Germany using ICON-FORCE reanalysis

Accurate flood forecasting in small and medium-sized catchments remains a major
challenge due to rainfall prediction uncertainties, limited hydrological data at that
scale as well as short warning times related to rapid response times of such systems.
The joint German research project KI-HopE-DE (KI-gestützte Hochwasserprognose
für kleine Einzugsgebiete in Deutschland) aims to improve flood prediction in
Germany by testing new machine learning-based approaches and bringing together
meteorologists and hydrologists, both from academia and from operational services.
Within this framework, the German Weather Service (DWD) contributes by providing
high-resolution meteorological datasets to support data-driven modelling, as well as
by testing training strategies that stronger account for weather model particularities.
KI-HopE-De develops a regionally-trained Long Short-Term Memory (LSTM) using
data from 1,626 catchments across Germany. Besides the classical training based on
meteorological observations, in our study we explore the use of the novel ICON-
FORCE (Fine-scale Observation-based Reanalysis for Central Europe) reanalysis
dataset for optimizing the LSTM. This approach attempts to leverage the proximity of
the ICON-FORCE reanalysis data and the ICON-D2 forecast model later used for
inference and making forecasts. Meteorological variables derived from ICON-FORCE
reanalysis are used as input features to capture spatiotemporal dependencies and
interactions relevant for runoff generation. The study is designed as a large-sample
experiment to systematically assess the added value of high-resolution
meteorological reanalysis for data-driven flood prediction, with a particular focus on
model robustness and applicability across divers catchments.
By focusing on the interface between meteorology and hydrology, this work
contributes to ongoing efforts to better integrate atmospheric and hydrological
information in flood forecasting models.