AI to make it rain? UAE weather summit explores latest tech
At their most recent modular meeting, AI experts from the UAE examined the use of AI in improving weather forecasting systems, as well as its possible applications in improving cloud seeding.
'It depends on the question you're trying to answer. AI models exist at various levels. Global, regional, and local, and you have to target datasets to train for the models to be of use. Foundational models can be fine-tuned but must be utilized correctly,' was said and repeated by abang Youngman, the National Weather Service in the USA, at the AI for Weather Prediction: Advances, Challenges & Future Outlook happening in Abu Dhai.
Regarding the use of AI to predict weather systems, she continued: 'The most useful state is the three-to-seven-day forecast range. AI models struggle due to their lack of Physics. In the case of extremely short-term forecasting, that is, nowcasting, there is progress. Physics based models still dominate the area of forecasting.'
Analysing real-time data with AI
One session talked about prediction of the Earth system with AI, with its emphasis on the need for high quality, varied field datasets: a multitude of weather data – real-time weather satellite data with the IoT devices and geo-spacial IoT data fusion observations and field weather data.
To improve the models and understand the behaviour of the clouds, weather algorithms are trained on large volumes of data, streamlining the processes of predicting cloud seeding timing and locations. These capabilities of AI improves the models, and at the same time, provides valuable cloud insights and summation intelligence for successful cloud systems, integration cloud, and augment moisture).
“What if there are several clouds instead of just one? Each of them has a limited amount of time before they dissipate. In that case, cloud AI might collaborate with forecasters and using weather parameters select the best possible cloud formation regions globally and guide them. These days the entire process is completed manually, but when it is integrated with AI, the operator will be able to do much more. Instead of just sifting through clouds, they will be able to specifically identify regions that the AI will highlight.”
“It is important to highly emphasize that the addition of AI technology will not solve all the problems at hand. Meteorologists, however, will more precisely predict and alter the current state of weather phenomena. That is a significant milestone on the field of climate science, not only within the UAE but globally.”
Ian Lisk, Chair of World Meteorological Organization (WMO) SERCOM, pointed out the inadequacies of the existing models.
“Some weather parameters, for example, thunderstorms and other forms of convective storms, are difficult for even traditional physics-based models,” he noted. “AI models have similar resolution issues. One of the hardest problems is the training data, particularly in a shifting climate.”
Ko Barrett, deputy secretary of the WMO, however, noted that there is a keen demand for s. tools that enhances predictive accuracy of forecasts while lowering the resource burden.
“But it’s not a one-size-fits all solution," she stated. “Many of our members possessing rudimentary skills are quite distant from the operational use of AI. Thus, it is our burden, as the WMO, to ensure that everyone is caught up to the pace of progress in AI.”
Comments