Observation-driven correction of numerical weather prediction for marine winds
Researchers have developed ORCA, a transformer-based deep learning model designed to correct errors in numerical weather predictions for marine winds. By assimilating in-situ observations, ORCA adjusts Global Forecast System (GFS) output, demonstrating significant error reduction up to 48 hours in advance. The model shows particular effectiveness along coastlines and shipping routes where observational data is more plentiful, offering a practical post-processing solution for improving forecast accuracy. AI
IMPACT This model could improve maritime safety and operational efficiency by providing more accurate wind forecasts.