RAIN, RAIN, PREDICTED AWAY: UP scientists build AI model to track cyclone rainfall

Diane Joy Galos

Scientists from the University of the Philippines Diliman announced that they created an artificial intelligence (AI) model to forecast rainfall from tropical cyclones.

The study, published on August 11 by the Royal Meteorological Society, was authored by Cris Gino Mesias and Dr. Gerry Bagtasa of UP Diliman’s Institute of Environmental Science and Meteorology.

According to the researchers, the model linked past tropical cyclone tracks with recorded rainfall to simulate future weather conditions.

“Specifically, a self-organizing map (SOM) clustered the TC tracks, which were then fed into a random forest (RF) regression model that used cyclone position, intensity, translational speed, and other parameters to predict accumulated rainfall,” the study read.

Comparable to Existing Dynamic Models

Bagtasa explained that conventional tropical cyclone rainfall predictions depend on dynamic models, which require high-performance computing facilities.

He emphasized that the newly built AI model could recognize patterns rapidly and accurately even when run on an ordinary laptop.

“When we assessed the AI model, its predictive skill was comparable to a dynamic model that we regularly use. The AI model had better skills for extreme rainfall from tropical cyclones,” Bagtasa said.

The study team tested the AI model’s ability by using data from 10 tropical cyclones that brought extreme rainfall across the Philippines between 2016 and 2020.

Useful for Disaster Preparedness

The validation results showed that the AI and machine learning system produced rainfall distribution patterns consistent with satellite-derived precipitation.

“When compared to the WRF model simulations, the AI model showed only slightly lower values in the prediction skill metrics,” the researchers said.

They added that the model recorded higher hit rate scores in predicting intense rainfall events of more than 100 millimeters, a feature that could strengthen disaster management planning.

The distance of the cyclone and its duration were the parameters that influenced the rainfall predictions the most.

“For instance, a typhoon near Batanes would not be expected to cause heavy rains in Mindanao. Slow-moving TCs that spend more time over land also tend to bring more rainfall overall,” the UPD-College of Science explained.

Broader Implications

According to the Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA), as many as 17 tropical cyclones were expected to develop or pass through the country’s area of responsibility from August to January, a forecast that the new AI tool might support.

Bagtasa admitted that the AI rainfall model they developed was not flawless, but stressed that it could supplement the set of forecast tools available to disaster managers.

“This AI model, admittedly, is not perfect. But it can add to the suite of rainfall forecast models available to equip our disaster managers with more information on impending hazards,” he said.

He clarified that the project was not a large language model (LLM), which demands significant energy and contributes to environmental harm.

“Some AI models, such as those for weather forecasting, can be useful and more efficient than conventional methods. But there are also some, like LLMs, that consume so much energy, leading to environmental impacts that are harmful to the planet,” Bagtasa said.

With the Philippines ranked among the world’s most disaster-prone nations, the researchers stressed that innovations like this AI tool are vital in strengthening early warning systems and helping communities prepare for the next inevitable storm.