Working on new technologies, Google, in order to expand the limits of artificial intelligence -based models in weather forecasts Weather Lab established a new platform. The most up -to -date work shared on this platform, especially Tropical cyclones It aims to produce longer -term and various scenarios about its direction, size and violence. The developed model can create 50 different estimates up to 15 days. These estimates are not only produced in a computer environment, but also by the US National Hurricane Center.
This cooperation aims not to replace classical physics -based weather forecasting models, but to offer them an alternative and supportive tool. Google Deepmind And Google Research The biggest claim of the model developed by teams is that the accuracy rate of the estimates is at least the same as the current systems. However, the company states that the model is still in the trial phase and that individual users should not benefit directly from this system.
Especially North Atlantic And East Pacific In the trials in the regions, this new artificial intelligence model, the European Medium -Term Air Forecast Center (ECMWF) data is said to make an average more accurate estimates. This difference becomes evident based on five -day predictions. However, in order to make these results permanent, there is a need for long -term and different geographies.
The model is trying to estimate future hurricanes by feeding from past data
One of the main sources used in the education of the new model is European -based Era5 data archive was. This archive brings together observational data collected for decades and physical -based model results. In this way, the model can update itself not only with numerical data, but also through past patterns. Google’s previously developed Gencast The model was trained with the same data archive and Nature According to a study published in the magazine, ECMWF has left one of the powerful models of 97 percent behind.
The Weather Lab website allows users to monitor the performance of these models visually. However, it is emphasized that this system is currently open only for research purposes. Google aims to explain the accuracy level of the new model to a wider audience with visual comparisons made through this site. Nevertheless, it is not clear whether this system will become a commercial or public service.
Google does not work alone in this field. Colorado State University The Cooperative Institute of atmospheric Research continues to examine how the model performances in different climatic conditions with researchers from the United Kingdom and Japan. Such collaborations are critical for the applicability of artificial intelligence -supported prediction models on a wider scale.
The increase in the number and intensity of excessive weather events such as hurricanes causes more attention to attract such projects. However, the personnel reduction in the US federal institutions in the past years led to weakening in the weather observation infrastructure. Especially National Ocean and Atmospheric Department (Noaa) The number of teams within the body was reduced and some meteorological measurement tasks were transferred to private companies. This brings some concerns about the future of public data.
These interventions against NOAA also opened the possibility of privatization of public -open weather forecasting services. Project 2025 Within the scope of the plans, the distribution of Noaa and some services to be carried out by the private sector. Google officials answered questions on this issue, models were developed with the intention of contributing to public service, he said.
The development of new artificial intelligence -based weather forecasting models has the potential to move the limits of the existing forecasting methods forward thanks to their work with past data and their traditional systems. However, the fact that these systems become fully reliable depends on comprehensive tests in many geographical regions and in different weather conditions. Google’s consequences to the public in this stage will be decisive in terms of long -term goals of the model.