Microsoft Starts's new AI models could make 30-day weather forecasting more accurate

The research team at Weather from Microsoft Start is developing AI-powered weather models. These models will be able to predict 30-day weather forecasts more accurately. While Microsoft Start has been lauded for their work in the past, they seem to keep up the good work.

Notably, Weather from Microsoft Start was honored with the title of the World’s most accurate global forecast provider. ForecastWatch, a US-based weather forecast monitoring company honored the team with the title back in March last year.

On 3rd May 2024, Microsoft took to Bing blogs to inform that the Start team published an article on the ArXiv website regarding the new medium-range weather forecast model.

Microsoft’s new AI weather modeling is better than the current ones used by the ECMWF

The published article explains how AI weather models compare to the latest state-of-the-art European Centre for Medium-range Weather Forecasts (ECMWF).

Microsoft in its recent Bing blogs says that instead of using a single AI model, it combined five different trained models. This includes three other deep learning architecture together which ensures weather forecasting of one month in advance.

The company further adds that these AI models can learn from decades of weather to understand the pattern to accurately predict the weather in advance. Talking more on the same, Microsoft adds:

They operate in much the same way as an NWP model, though: given the current state of the atmosphere on a 3-D globe (latitude, longitude, and height), predict the state of the atmosphere for some future time, say one hour later. They then feed this prediction back into the model to predict two hours later, and so on.

The new AI models can predict the weather much faster and can only take minutes using a GPU. This allows models to run much more frequently ensuring better weather forecasting.

The new weather forecasting AI model developed by the Start team can measure temperature errors more effectively. Microsoft says:

The results are quite encouraging: when measuring temperature errors using the Continuous Ranked Probability Score (CRPS) metric, our out-of-the-box AI ensemble outperforms the ECMWF model by 17% for one-week forecasts and 4% for four-week forecasts (Figure 1).

Overall, the latest AI weather models could potentially bring major improvements to weather forecasting of above ten days. Hence, it is adding the 30-day weather forecasts to its ever-growing Microsoft Start weather modeling.

You can check out more details here.