← Back to all news
arXiv Machine Learning By Valerie Tsao, Nathaniel W. Chaney, Manolis Veveakis, Marta Zaniolo

Generated storm lifecycles fill the rare-event tail of short rainfall records

Read the original on arXiv Machine Learning →

The paper introduces StormGen, a generative machine‑learning model that expands the initial hours of an observed storm into full, physically plausible hourly rainfall lifecycles within minutes. Tested on a decade of southeast Texas rainfall, StormGen accurately reproduced the most severe storms, generating close counterparts for 4–5% of the top 2% of events, outperforming stochastic storm transposition. Remarkably, the model reconstructed Hurricane Harvey’s lifecycle and extreme catchment totals from a single hour of data, even though the event was not in the training set.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

Related stories

arXiv Machine Learning
Sep 4

SurgeGen: A Hybrid Generative Diffusion Framework for Storm Surge Scenario Synthesis

SurgeGen is a two‑stage generative diffusion framework that synthesizes storm surge scenarios conditioned on continuous storm parameters. The first stage produces a coarse baseline surge estimate, which then conditions a diffusion model that refines the output to capture realistic spatial patterns and variability. The method can generate diverse, realistic surge scenarios both within and beyond the training distribution.

By Shunan Zheng, John J. Hasenbein
diffusion
More like this →
arXiv Machine Learning
Oct 16

KNOU-Flood: Advancing global streamflow forecasting with lead-aware probabilistic machine learning

KNOU‑Flood extends the open AIFL recipe for global streamflow forecasting by retaining a single‑layer LSTM and two‑stage transfer learning while adding a longer 424‑day input window, lead‑aware forecast inputs, and GraphCast as an additional forecast source. The model also introduces a latent‑noise ensemble trained with the fair CRPS, achieving a median KGE′ of 0.65 at lead time 1 and 0.48 at lead time 10 across 2,099 test basins, outperforming AIFL and a 1,024‑unit baseline. The probabilistic head, GraphCast channels, longer window, and mixed fine‑tuning collectively reduce lead‑time decay, improve calibration, and enhance detection of multi‑year floods.

By Minjong Cheon
fine-tuning
More like this →
arXiv AI
Oct 27

EvoBench: An Evolving, Diagnostic Benchmark Framework for Environmental Modeling

EvoBench is a diagnostic benchmark framework designed to reveal where and when machine‑learning models for environmental variables fail, rather than relying on a single average accuracy score. It tracks model performance over time and across spatial extrapolation, highlighting issues such as non‑stationarity, poor performance at unmonitored sites, and under‑prediction of rare events like floods. In a case study on daily streamflow prediction across 9,000 U.S. basins, EvoBench shows that no model dominates all scenarios and that targeted fixes can mitigate identified weaknesses.

By Yiming Sun, Yuan Zhang, Qi Cheng, Yingda Fan, Runlong Yu, Yiqun Xie, Xiaowei Jia
benchmarks
More like this →