Uncertainty-aware segmentation for rainfall prediction post processing
Simone Monaco, Luca Monaco, Daniele Apiletti
TL;DR
The paper tackles biases and uncertainties in high-resolution NWP rainfall forecasts by reframing post-processing as a segmentation task and developing uncertainty-aware deep learning models. It introduces and compares Monte Carlo Dropout U-Net, Deep Ensemble U-Net, and a novel SDE U-Net architecture tailored for segmentation, evaluating them on NW Italy data with typical and intense precipitation events. All DL models outperform the baseline Poor Man's Ensemble, with SDE U-Net providing the best accuracy-reliability balance, especially for intense events, and MC Dropout excelling in typical-event uncertainty sharpness. The findings support integrating uncertainty-aware post-processing into operational forecasting to improve decision-making and preparedness, while highlighting avenues for future refinement and broader validation.
Abstract
Accurate precipitation forecasts are crucial for applications such as flood management, agricultural planning, water resource allocation, and weather warnings. Despite advances in numerical weather prediction (NWP) models, they still exhibit significant biases and uncertainties, especially at high spatial and temporal resolutions. To address these limitations, we explore uncertainty-aware deep learning models for post-processing daily cumulative quantitative precipitation forecasts to obtain forecast uncertainties that lead to a better trade-off between accuracy and reliability. Our study compares different state-of-the-art models, and we propose a variant of the well-known SDE-Net, called SDE U-Net, tailored to segmentation problems like ours. We evaluate its performance for both typical and intense precipitation events. Our results show that all deep learning models significantly outperform the average baseline NWP solution, with our implementation of the SDE U-Net showing the best trade-off between accuracy and reliability. Integrating these models, which account for uncertainty, into operational forecasting systems can improve decision-making and preparedness for weather-related events.