A Spatiotemporal Radar-Based Precipitation Model for Water Level Prediction and Flood Forecasting

Study Region: Goslar and Göttingen, Lower Saxony, Germany. Study Focus: In July 2017, the cities of Goslar and Göttingen experienced severe flood events characterized by short warning time of only 20 minutes, resulting in extensive regional flooding and significant damage. This highlights the critical need for a more reliable and timely flood forecasting system. This paper presents a comprehensive study on the impact of radar-based precipitation data on forecasting river water levels in Goslar. Additionally, the study examines how precipitation influences water level forecasts in Göttingen. The analysis integrates radar-derived spatiotemporal precipitation patterns with hydrological sensor data obtained from ground stations to evaluate the effectiveness of this approach in improving flood prediction capabilities. New Hydrological Insights for the Region: A key innovation in this paper is the use of residual-based modeling to address the non-linearity between precipitation images and water levels, leading to a Spatiotemporal Radar-based Precipitation Model with residuals (STRPMr). Unlike traditional hydrological models, our approach does not rely on upstream data, making it independent of additional hydrological inputs. This independence enhances its adaptability and allows for broader applicability in other regions with RADOLAN precipitation. The deep learning architecture integrates (2+1)D convolutional neural networks for spatial and temporal feature extraction with LSTM for timeseries forecasting. The results demonstrate the potential of the STRPMr for capturing extreme events and more accurate flood forecasting.

💡 Research Summary

The paper addresses the acute need for faster and more reliable flood forecasting in the German towns of Goslar and Göttingen, where July 2017 flood events left only a 20‑minute warning window. The authors propose a novel data‑driven model, the Spatiotemporal Radar‑based Precipitation Model with residuals (STRPMr), which relies exclusively on high‑resolution radar precipitation fields from the German Weather Service (DWD) and does not require any upstream hydrological inputs.

Three datasets are used: (1) RADOLAN radar precipitation images (5‑minute, 1 km × 1 km resolution) covering 2003‑2018, (2) 15‑minute water‑level measurements from the Sennhuette gauge in Goslar, and (3) analogous water‑level data from a gauge in Göttingen for out‑of‑sample validation. The radar data are spatially cropped to the study area, then aggregated over three consecutive frames to match the 15‑minute water‑level cadence. The water‑level series are smoothed with an 8‑point moving average to reduce noise, modestly increasing the raw precipitation‑level correlation from 0.075 to 0.0835.

The core of STRPMr is a (2+1)D convolutional neural network that separates spatial (2‑D) and temporal (1‑D) convolutions, extracting rich spatiotemporal features while keeping computational cost low. These feature maps feed into a Long Short‑Term Memory (LSTM) module that captures the sequential dynamics of the system. Crucially, the model incorporates a residual‑learning branch: after the primary prediction of water level, the residual (the difference between observed and predicted level) is also predicted, allowing the network to explicitly model the nonlinear, non‑stationary relationship between radar‑derived precipitation and river response. Training proceeds in two stages—first learning the radar‑only representation, then fine‑tuning with the combined radar‑plus‑historical level inputs.

Performance is benchmarked against the residual‑LSTM model introduced in the Goslar flood benchmark (reference