Google releases AI-based computer forecast model.
Artificial Intelligence is quickly becoming a part of our daily lives. Now, its expanding in the weather forecasting realm.
Last week Google announced the launch of its new weather model that uses artificial intelligence to help produce more accurate forecasts when it comes to the tropics.
Traditional models like the popular European model or American Global Forecast System (GFS) use intense physics and math equations.
Large supercomputers run that data and come up with a solution for forecasters to analyze.
AI is focused on using the past to predict the future.
How does it work?
The AI model uses the data that goes into the European model to create 50 possible scenarios up to 15 days out.
According to Google, the Weather Lab is trained to reconstruct past weather over the globe from millions of observations in addition to a specialized database containing information of track, intensity, size and wind radii of nearly 5,000 storms from the last 45 years.
Is it accurate?
Like traditional models this new tool is to be used as guidance, not gospel.
However, Google’s Weather Lab has shown promise in the early going.
According to Google, during the test years of 2023 and 2024, the model’s five-day storm track prediction was, on average, 86 miles closer to the true location when compared to the physics based European ensemble.
The range is comparable of the Euro Ensemble approximately 3.5 days out from landfall.
Looking back at Otis, a storm that far-exceeded expectations, the Weather Lab model hinted at a higher-end solution.
You may recall Otis in the Eastern Pacific was expected to fizzle out. In reality, the storm made landfall near Acapulco as a Category 5 hurricane.
While the Google Weather Lab model did not forecast a Category 5, a good portion of its members suggested the storm could reach major status.
Source: www.clickorlando.com
No comments:
Post a Comment