Search NASASearch

DOE OSTI · 2575724

Probabilistic Diffusion Models Advance Extreme Flood Forecasting

Abstract

Extreme floods pose escalating risks in a changing climate, yet forecasting remains challenging due to peak flow underestimation and high uncertainty. We introduce diffusion-based runoff model (DRUM), a probabilistic deep learning (DL) approach that advances extreme flood forecasting across representative basins in the contiguous United States. DRUM outperforms state-of-the-art benchmarks, enhancing nowcasting skill for the top 1‰ of flows in 72.3% of studied basins. Under operational scenarios, DRUM extends reliable lead times by nearly a full day for 20- and 50-year floods. When evaluated with measured precipitation, an ideal condition, recall improves by 0.3–0.4 and the early warning window extends by 2.3 days for 50-year floods. The enhancement potential varies regionally, with precipitation-driven flood zones in the eastern and northwestern US benefiting most, gaining 3–7 days in lead time. These findings highlight the transformative potential of diffusion models as a cutting-edge generative AI technique for advancing hydrology and broader Earth system sciences.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Ou, Zhigang [Southern University of Science and Technology (SUSTech), Shenzhen (China); Chinese Academy of Science, Beijing (China)], Nai, Congyi [Chinese Academy of Science, Beijing (China)], Pan, Baoxiang [Chinese Academy of Science, Beijing (China)] (ORCID:0000000234095568), Zheng, Yi [Southern University of Science and Technology (SUSTech), Shenzhen (China)] (ORCID:000000018442182X), Shen, Chaopeng [Pennsylvania State Univ., University Park, PA (United States)] (ORCID:0000000206851901), Jiang, Peishi [Pacific Northwest National Laboratory (PNNL), Richland, WA (United States)] (ORCID:0000000349684258), Liu, Xingcai [Chinese Academy of Sciences, Beijing (China)] (ORCID:0000000157267353), Tang, Qiuhong [Chinese Academy of Sciences, Beijing (China)] (ORCID:0000000208866699), Li, Wenqing [China Institute of Water Resources and Hydropower Research, Beijing (China)] (ORCID:000000026086958X), Pan, Ming [Univ. of California, San Diego, La Jolla, CA (United States)] (ORCID:0000000333508719). 2025-07-31. Probabilistic Diffusion Models Advance Extreme Flood Forecasting. https://doi.org/10.1029/2025gl115705

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related reports

Water4Energy Step-1 Band-M Ready-to-Train Samples for TVA Weeks-to-Years Prediction, Version 0

AI-ready Band-M (monthly) labelled training pack for the Water4Energy Genesis Task-1 project on weeks-to-years prediction of Tennessee Valley temperature and precipitation. The deposit includes leakage-aware issue-time samples (samples_M_v0.nc; N=486), train-only scalers, issue-time split table, supporting monthly panels, and Python generation scripts to recreate the pack from the companion Tier-1 raw observation collection (https://doi.org/10.13139/ORNLNCCS/3398576). Each sample pairs a 12-month lookback of teleconnection indices and SST box anomalies with TVA-mean ERA5 anomaly targets (t2m, tp, msl) at leads 1–3 months.

54 ENVIRONMENTAL SCIENCES

Multi-Angle Snowflake Camera, particle analysis

The c1 level data product for the Mutli-Angle Snowflake Camera contains snowflake fall speeds and particle size, among other analysis for images associated with each hydrometeor.

54 ENVIRONMENTAL SCIENCES

Multi-Angle Snowflake Camera, time bins

The c1 level data product for the Mutli-Angle Snowflake Camera contains snowflake fall speeds and particle size, among other analysis for images associated with each hydrometeor.

54 ENVIRONMENTAL SCIENCES