E-Paper
Sign In
UAE: How AI can save millions and deliver faster, more accurate weather forecasts
AI-driven forecasting tools promise to be more accessible and affordable, especially for developing countries
- PUBLISHED: Tue 9 Sept 2025, 4:54 PM
Predicting the weather may soon become faster, cheaper, and more accurate with the use of artificial intelligence (AI). This technology could replace costly supercomputers and transform forecasting globally, experts highlighted at a recent conference in Abu Dhabi.
Speaking exclusively to Khaleej Times on the sidelines of the AI for Weather Prediction: Advances, Challenges & Future Outlook conference, World Meteorological Organization (WMO) President, Dr Abdulla Al Mandous, said, “We are here because the UAE government, especially Sheikh Mansour Bin Zayed Al Nahyan, Vice-President, Deputy Prime Minister, and Chairman of the Presidential Court, told us that we should use AI to improve weather prediction.
Recommended For You
"Traditionally, we have physical models, which are very expensive and not every country (nations with limited resources) can afford. We investigated the accuracy of using AI and that is why this conference is important: to discuss the advancements, the challenges, and the future of weather prediction."
Stay up to date with the latest news. Follow KT on WhatsApp Channels.
AI could replace supercomputers
The National Centre of Meteorology (NCM) currently relies on a supercomputer worth millions of dollars to run traditional numerical weather prediction models.
However, AI-driven forecasting tools promise to be more accessible and affordable, especially for developing countries.
“AI should be more cost-effective compared to traditional models that require supercomputers, manpower, and electricity. For example, the NCM supercomputer costs more than $10 million. But if AI can deliver accurate numerical predictions, it will save a lot of money and give many countries the capability to forecast weather for the safety of their people.”
The expert explained that global forecasting has already advanced with powerful institutions like the European Centre for Medium-Range Weather Forecasts (ECMWF) in Europe and the National Oceanic and Atmospheric Administration (NOAA) in the United States, supported by satellite data and advanced computational systems.
Challenges remain
There are still hurdles to overcome. “When it comes to challenges, first, many countries in the world lack the ability to install weather stations or receive satellite data, so we face a big data gap globally. Second, the atmosphere is not homogeneous — disturbances can change predictions. Third, sudden severe weather events reduce forecast accuracy. And of course, climate change itself is a major challenge to producing reliable forecasts with physical models.”
Despite these impediments, AI is being piloted successfully in various countries. “Right now, we are starting with AI, but we are not fully using it yet in WMO. We have pilot projects — for example, in Norway and Malawi. Malawi, which cannot afford very expensive supercomputers or expertise, is now able to do three-day forecasts with computers costing less than $5,000. This shows how AI can give countries the capability to produce their own forecasts at lower cost.”
Looking ahead, Mandous envisions combining AI with traditional models to build stronger hybrid systems. “We want to create a hybrid between physical numerical weather models and AI. AI can capture cases that physical models miss because it already knows the history and past scenarios. We also want to use all possible sources of data…from other agencies, as inputs to the models. Importantly, we seek international cooperation through WMO, working with UN bodies, intergovernmental, and governmental organisations.”
He pointed out that AI could ultimately help improve both the accuracy and resilience of forecasts in the face of climate change and severe weather.
“AI prediction models are cheaper, more accurate, and can even capture sudden changes in weather. They are trained on such events, so their prediction capability can be much higher than physical models.”
