Can accurately forecast weather conditions in the next 30 days
The model is based on the latest research results of the Microsoft Start team, combines five different artificial intelligence models and three deep learning architectures, and uses weather data from the past decades for training, and can accurately predict the weather forecast within 30 days.
In terms of predicting weather forecasts for the next week, the accuracy of temperature measured by the new AI model is 17% higher than that of the European Center for Medium-Range Weather Forecasts (ECMWF), and the accuracy of weather forecasts for the next four weeks is 4% higher.
Compared to the currently commonly used NWP / ECMWF model that requires a supercomputer to run, this model can produce 100 30-day predictions in less than 2 hours on a single NVIDIA V100 graphics processing unit.
Extremely efficient!
A run model ensemble weather forecast system is proposed that uses a hybrid driven weather forecast model combined with the European Center for Range Weather Forecasts (F) ocean model to
Predicts global weather with a 1-degree resolution and a 4-week delivery time. For predictions of 2-meter temperatures, the set is on average 4-17% higher than the original ECMWF extended range set, depending on delivery time.
However, after applying statistical bias correction, the overall ECMWF performance improved by approximately 3% at 4 weeks. For other surface parameters, our ensemble also differs by a few percentage points from the ECMWF ensemble.
Using a multi-model integration approach and data-driven weather forecast model, near-state-of-the-art sub-season-to-season forecast can be achieved.
If you want to learn more, you can click on the link below the video.
Thank you for watching this video. If you like it, please subscribe and like it. thank
Thesis:https://arxiv.org/abs/2403.15598
Video: