The application of artificial intelligence (AI) in environmental science is increasingly becoming pervasive especially in the field of quantitative precipitation estimation and weather prediction of severe storms. Adubi Tunde, an award-winning researcher believes that the crossroads of accurate weather prediction and artificial intelligence would be an exciting development around the world owing to the gradual effects of climate change followed by a corresponding advancement in machine learning.

After the 2020 pandemic, Adubi arrived in the US where he began pursuing his long-held dream of studying and researching the dynamics of severe convective storms. He had wanted to investigate extreme precipitation events that could pose socio-economic risk in Nigeria after his devastating flash-flood experience in 2012. Coincidentally, few months into his program, a similar flood occurred in Nigeria in the early summer of 2022. He stated that while it was a pitiful occurrence, it further inspired him to advance his studies.

Adubi’s research involves improving the estimation and prediction of severe precipitation events using numerical and empirical formulations. In his investigations, he unravels the cause of inaccurate rainfall estimation as well as studying the type of weather instruments that could be better applicable for heavy rainfall and storm prediction. He is also exploring the use of complex AI models that could provide precise prediction even in real-time. He stated that accurate forecasting and nowcasting – a short-time prediction of extreme rain typically within few hours that could lead to flash-floods has been an issue especially in developing countries like Nigeria.

He mentioned that many Nigerians rely on accurate weather prediction for their day-to-day activity, however whenever these predictions are inaccurate, it could cause uncertainty and changes in their plan – affecting major activities of their lives. “When the quality of dataset from low-tech precipitation instruments have low-resolutions and poor models are applied for forecasting, they make the process tedious, time consuming with even less likelihood of the event occurring” Adubi added.

“Though the widely used mathematical method is largely effective in quantifying precipitation, enabling weather enthusiasts to forecast heavy rainfall associated with convection storm, on the other hand, the application of traditional prediction models that employ physical processes like cloud formation, evaporation, and condensation during the interactions of atmosphere, land and oceans could be misleading. In other words, the over simplification of these models and processes for severe weather prediction tends to introduce inaccuracies and bias” He highlighted.

Adubi suggested that like-minded researchers could further develop and enhance AI algorithms that would substitute the numerical weather prediction methods, thereby obtaining a seamlessly dependable forecast in real-time. In one of his projects, he designed an optimized convolution neural network algorithm – an AI model to detect and classify convective storms observed by advanced radar systems. He believes that the use of AI such as random forest regression or support vector regression method could be harnessed to improve the calibration of the radar systems – a major step for accurate storm quantification and forecasting.

Just like the United States, Adubi opined that the Nigerian government should take necessary measures that would support relevant weather authorities in advancing research in oceanic and atmospheric studies. “Nationwide, high school curriculums should be regulated and geared towards promoting this field of science with incentives or scholarships provided for high performing students. Perhaps, this would motivate them in college where they could become experts in science and engineering. Without qualified professionals in the field, weather agencies would hardly function to serve the critical needs of the people” Adubi stated.

“Although the use of AI models could be energy-intensive for improving rainfall estimation and prediction, emphasis should be placed on establishing robust energy and computing infrastructure in Nigeria. Similarly, advanced observing technologies including – the S-band polarimetric radars, disdrometers, wind profilers, tipping buckets and rain gauge – should be deployed in flood-prone regions. The integration of AI with high-resolution data from these instruments could provide less computational challenges while enhancing extreme storm estimation and prediction” Adubi mentioned.

He also stated that since the application of AI in this field is quite expensive, it is mandatory for developing countries to invest heavily in all aspects of severe weather prediction as well as flood disaster prevention and management. “If these steps are taken, timely and accurate precipitation information would help inform reservoir operations, flood protection and the emergency preparedness team” Adubi added.