Thanks to artificial intelligence, a new age in weather forecasting has begun.
Google DeepMind, the company’s AI research department, is working on a machine learning model that it claims can surpass 90% of the benchmarks used by the most advanced weather prediction systems in the world and predict weather properly in seconds rather than hours.
For weather forecasting and decision-making, the AI represents a “turning point,” according to Google DeepMind experts.
What is the process of AI weather forecasting?
To increase its accuracy, Google DeepMind’s AI weather forecasting model, known as GraphCast, has been trained on nearly 40 years’ worth of historical weather data, according to the publication, Science.
It took the AI four weeks and thirty-two computers to train.
On a single desktop computer, however, the algorithm this generated can forecast weather up to ten days ahead of time in less than a minute.
GraphCast outperforms existing weather systems on 90% of 1,380 parameters with much better accuracy.
The Google DeepMind scientists explain in their Science journal publication that the AI is also better at forecasting catastrophic weather occurrences, such as excessive temperatures and the tracking of tropical storms.
What is the process of a traditional weather forecast?
Supercomputers, or high performance computers, are used in weather forecasting today to do intricate computations based on data from buoys, satellites, and weather stations.
This procedure is expensive and time-consuming. In particular, Google DeepMind experts explain that the world’s most accurate weather forecasts are produced by the European Centre for Medium-Range Weather Forecasts in Italy every six hours.
This process is repeated every six hours, or four times a day on average.
The weather regulating equations must be cracked in order for the supercomputers to employ a method known as “numerical weather prediction,” which is a laborious procedure.
Why does AI have superior weather prediction skills?
Investing in more expensive computer power is necessary to increase the accuracy of this conventional weather forecasting method.
However, GraphCast can provide more accurate and affordable weather predictions using previous meteorological data, according to the Google DeepMind experts.
Patterns that are difficult to perceive in equations can be found in the data by the AI. It can then make better weather forecasts by utilizing these discoveries.
According to the Financial Times, GraphCast’s energy efficiency is around 1,000 times less expensive than that of traditional weather forecasting techniques.
Does AI weather forecasting aid in the fight against global warming?
In order to assure the appropriate development and implementation of the technology, Google is a partner in the World Economic Forum’s new effort, the AI Governance Alliance.
The Google DeepMind experts note that GraphCast itself may be used to foresee issues related to “climate and ecology, energy, agriculture, and human and biological activity” among other issues.
Since the climate problem is one of the top three global dangers for the next two to ten years, according to the Forum’s Global dangers Report 2023, technology may be able to help predict the extreme weather events brought on by it.
