Google Announces Generative AI Model to Mitigate the Uncertainty in Weather Forecasting


Google Research on Friday announced a new generative artificial intelligence (AI) model that could help reduce uncertainty and inaccuracies in weather forecasting. The AI ​​model is called Scalable Ensemble Envelope Diffusion Sampler (SEEDS), and instead of following traditional probabilistic models of weather forecasting, the AI ​​model is based on denoising diffusion probabilistic models. This is not the first weather forecasting model that the tech giant is working on, as it has previously unveiled GraphQL, a model that can predict weather up to 10 days in advance, and MetNet-3, a high-resolution for 24- -resolution prediction model. Hour duration.

The announcement was made in a blog post by senior software engineer Lizhao Li and Google Research research scientist Rob Carver. The team has published a paper on generative AI model seeds in the journal Science Advances. According to the announcement, the AI ​​model will innovate weather forecasting in two different ways – making it more accurate and reducing the cost of predicting the weather.

Highlighting two key issues in modern weather forecasting, the newspaper said that right now models run something called “probabilistic forecasts”. Essentially, they focus on initial conditions to generate a preliminary forecast and as conditions improve and the weather model receives more data, the model corrects itself to generate a more accurate forecast. Google says this method allows for more uncertainty in long-term predictions. On cost, the research team highlighted that huge supercomputers running highly complex numerical weather models, where forecasts need to be continuously generated to achieve accurate results, can run up high costs.

According to the research paper, SEEDS works on modeling the diffusion probabilistic model, which was developed by Google Research. It was trained on skill-based metrics such as rank histogram, root-mean-square error (RMSE), and continuous rank probability score (CRPS). The paper claims that while the model runs a negligible computational cost, it also improves the accuracy of the initial forecast, thereby requiring a lower number of forecast production during a particular time period.

The research team also included examples of running an AI model to predict weather and found that it provided greater reliability than Gaussian models. Highlighting the example of a geopotential trough west of Portugal, it said, “Although the Gaussian model adequately predicts the marginal univariate distribution, it fails to capture cross-field or spatial correlations. This hinders assessment of the effects that these anomalies may have on the intrusion of warm air from North Africa, which could exacerbate heat waves in Europe. According to Google Research, SEEDS is able to take these factors into account to improve its predictions. The model is yet to be peer-reviewed, and depending on its feasibility, it may later be developed into a commercial model.

Affiliate links may be automatically generated – see our ethics statement for details.

Related Post

Leave a Reply

Your email address will not be published. Required fields are marked *