PulseAugur
EN
LIVE 13:16:11

Machine learning forecasts weather beyond 30 days

Researchers have demonstrated that machine learning models can extend deterministic weather forecast skill beyond the traditional two-week limit. By optimizing initial conditions for the GraphCast model, they achieved an 86% error reduction at ten days, with useful skill extending past 30 days. This method revealed large-scale atmospheric corrections, and when applied to the Pangu-Weather model, it showed a 21% error reduction. The findings suggest that skillful deterministic forecasts far beyond two weeks are possible, though real-time application for operational forecasts requires further research. AI

IMPACT Extends the potential for accurate long-range weather prediction, impacting sectors reliant on climate and weather data.

RANK_REASON This is a research paper detailing a new finding in machine learning applied to weather forecasting. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

COVERAGE [1]

  1. arXiv cs.LG TIER_1 English(EN) · P. Trent Vonich, Gregory J. Hakim ·

    Atmospheric Predictability Beyond 30 Days with Machine Learning

    arXiv:2504.20238v2 Announce Type: replace-cross Abstract: Atmospheric predictability research has long held that rapid error growth at small spatial scales imposes an intrinsic limit of roughly two weeks on deterministic weather forecast skill. We challenge this limit using Graph…