A Diffusion-Based Framework for High-Resolution Precipitation Forecasting over CONUS

2025-12-09

AI Summary

Accurate nowcasting of precipitation at 0–12 hours is critical for flash flood and infrastructure risk warning. This paper proposes a diffusion-based residual learning framework that fuses MRMS radar observations with HRRR numerical forecasts to generate 1–12-hour, 1-km-resolution precipitation forecasts over the contiguous United States (CONUS). We systematically compare three paradigms—pure data-driven, HRRR bias correction, and hybrid residual modeling—for the first time, and introduce a calibration-aware uncertainty quantification method tailored to residual learning. Results show that the hybrid model achieves the best skill at 1-hour lead time, while the HRRR correction model retains high skill up to 12 hours; overall, the framework significantly outperforms the HRRR baseline across CONUS and markedly improves extreme rainfall forecasting. This work establishes a new, interpretable, reliable, and regionally generalizable paradigm for precipitation nowcasting.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Planning, Routing, and Scheduling: Model-Based Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Accurate precipitation forecasting is essential for hydrometeorological risk management, especially for anticipating extreme rainfall that can lead to flash flooding and infrastructure damage. This study introduces a diffusion-based deep learning (DL) framework that systematically compares three residual prediction strategies differing only in their input sources: (1) a fully data-driven model using only past observations from the Multi-Radar Multi-Sensor (MRMS) system, (2) a corrective model using only forecasts from the High-Resolution Rapid Refresh (HRRR) numerical weather prediction system, and (3) a hybrid model integrating both MRMS and selected HRRR forecast variables. By evaluating these approaches under a unified setup, we provide a clearer understanding of how each data source contributes to predictive skill over the Continental United States (CONUS). Forecasts are produced at 1-km spatial resolution, beginning with direct 1-hour predictions and extending to 12 hours using autoregressive rollouts. Performance is evaluated using both CONUS-wide and region-specific metrics that assess overall performance and skill at extreme rainfall thresholds. Across all lead times, our DL framework consistently outperforms the HRRR baseline in pixel-wise and spatiostatistical metrics. The hybrid model performs best at the shortest lead time, while the HRRR-corrective model outperforms others at longer lead times, maintaining high skill through 12 hours. To assess reliability, we incorporate calibrated uncertainty quantification tailored to the residual learning setup. These gains, particularly at longer lead times, are critical for emergency preparedness, where modest increases in forecast horizon can improve decision-making. This work advances DL-based precipitation forecasting by enhancing predictive skill, reliability, and applicability across regions.

Problem

• Develops a diffusion-based deep learning framework for high-resolution precipitation forecasting.
• Compares three residual prediction strategies using different data sources for accuracy.
• Evaluates performance and reliability across lead times and extreme rainfall thresholds.

Innovation

• Diffusion-based deep learning framework for precipitation forecasting
• Hybrid model integrating MRMS observations and HRRR forecasts
• Calibrated uncertainty quantification for residual learning setup