Particle.news
Get it on Google Play
Download on the App Store
5 ARTICLES
·
6h ago

Machine-Learning Wrapper Doubles Six‑Week Forecast Skill, Study Finds

Farmers, utilities, and emergency planners could gain more lead time because the tool corrects systematic model biases using a CPU‑efficient machine‑learning wrapper.

Image
© Scott Olson/Getty Images
A new technique significantly improves weather forecasting beyond two weeks. The researchers developed a machine-learning model that learns to correct errors in existing forecasting models to improve their output over the two-to-six-week time frame, including for extreme weather events.

Overview

  • The Nature paper published Wednesday reports that Probabilistic Bias Correction (PBC) doubled subseasonal (2–6 week) accuracy for an AI system and raised skill for a leading numerical model across most targets.
  • PBC is a model‑agnostic postprocessing layer that uses machine learning to identify and correct systematic errors that build up in both physics‑based and AI forecasts, such as biases from sparse data and changing climate.
  • Retrospective tests show large gains for extreme events and floods, with flood‑prediction skill improving roughly 76%–92% and severe‑flood skill improving about 86%–150% in the authors’ analyses.
  • The authors released PBC as open‑source software that can run on ordinary CPUs, and early pilots with the Kenya Meteorological Service and Rhiza Research are already under way ahead of a strong El Niño season.
  • Experts say the next task is operational integration: centers like ECMWF and national services must embed PBC into real‑time pipelines and validate its performance across regions before its societal benefits are realized.