HydroDiffusion: Diffusion-Based Probabilistic Streamflow Forecasting with a State Space Backbone

2025-12-13

AI Summary

In medium-to-long-term probabilistic streamflow forecasting, existing diffusion models relying on LSTMs suffer from weak long-range dependency modeling and inconsistent multi-step prediction trajectories. To address this, we propose the first diffusion framework built upon a decoder-only state space model (SSM), enabling joint denoising of entire multi-day flow sequences in a single forward pass—thereby explicitly enforcing temporal consistency and avoiding error accumulation inherent in autoregressive generation. Our method integrates SSMs’ superior sequential modeling capacity, multi-step joint denoising training, and hydro-meteorological driver conditioning. Experiments across 531 basins in the contiguous United States (CONUS) demonstrate significant improvements over LSTM-based and DRUM baselines in both deterministic skill (e.g., NSE, RMSE) and probabilistic skill (e.g., CRPS, PICP) across the full forecast horizon and nowcasting period. This work establishes a more robust and temporally coherent paradigm for hydrological probabilistic forecasting.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Computer Vision: Diffusion Models for Vision
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Recent advances have introduced diffusion models for probabilistic streamflow forecasting, demonstrating strong early flood-warning skill. However, current implementations rely on recurrent Long Short-Term Memory (LSTM) backbones and single-step training objectives, which limit their ability to capture long-range dependencies and produce coherent forecast trajectories across lead times. To address these limitations, we developed HydroDiffusion, a diffusion-based probabilistic forecasting framework with a decoder-only state space model backbone. The proposed framework jointly denoises full multi-day trajectories in a single pass, ensuring temporal coherence and mitigating error accumulation common in autoregressive prediction. HydroDiffusion is evaluated across 531 watersheds in the contiguous United States (CONUS) in the CAMELS dataset. We benchmark HydroDiffusion against two diffusion baselines with LSTM backbones, as well as the recently proposed Diffusion-based Runoff Model (DRUM). Results show that HydroDiffusion achieves strong nowcast accuracy when driven by observed meteorological forcings, and maintains consistent performance across the full simulation horizon. Moreover, HydroDiffusion delivers stronger deterministic and probabilistic forecast skill than DRUM in operational forecasting. These results establish HydroDiffusion as a robust generative modeling framework for medium-range streamflow forecasting, providing both a new modeling benchmark and a foundation for future research on probabilistic hydrologic prediction at continental scales.

Problem

• Develops HydroDiffusion for probabilistic streamflow forecasting
• Addresses limitations in capturing long-range dependencies in forecasts
• Ensures temporal coherence across multi-day forecast trajectories

Innovation

• Uses decoder-only state space model backbone
• Denoises full multi-day trajectories in single pass
• Ensures temporal coherence across forecast lead times