AI can help in predicting weather extremes, says WMO report
Artificial Intelligence (AI) and Machine Learning (ML) based weather forecasting can make predicting extreme weather possible, as opposed to the numerical weather prediction system presently in use.
Artificial Intelligence (AI) and Machine Learning (ML) based weather forecasting can make predicting extreme weather possible, as opposed to the numerical weather prediction system presently in use, a multi-agency report coordinated by the World Meteorological Organisation has said.
Some evaluations have shown that AI/ML models are surpassing physics-based models in predicting some weather variables and extreme or hazardous events, such as tropical cyclones, WMO said in the report titled “United in Science: Reboot Climate Action”, adding that studies have shown that AI/ML models can also predict the El Nino Southern Oscillation (ENSO) up to three years ahead.
These recommendations come at a time when science is showing increased warming trends for the next five years which will disrupt weather patterns in several parts of the world including India.
With existing policies and Nationally Determined Contributions, global temperature is expected to rise to a maximum of 3°C over pre-industrial levels. Only in the most optimistic scenario, when all conditional NDCs and net-zero pledges are fully achieved, global warming is projected to be limited to 2°C, with just a 14% chance of limiting global warming to 1.5°C, the report said.
ML is a specific subset of AI focused on using algorithms to process data, learn from it, and make decisions or predictions based on these data. This process is known as training and involves teaching computer models to perform tasks like recognizing speech, identifying images or predicting trends by learning from large amounts of data.
Experts said that interpretation of all information made available through observations will be key in addressing gaps in weather forecasting that have been seen in recent years.
There are challenges that limit realization of the full potential of AI/ML for weather forecasting, including gaps in data availability, inadequate model resolution and concerns about ethics, such as a lack of transparency and unequal access, WMO said.
Instead of relying on costly physics-based numerical models, AI/ML models are trained on reanalysis and observational datasets, making weather forecasting faster and cheaper.
In recent years, there have been some concerns with forecasting single day rain events because often the event is far more extreme than the forecast or far milder than what is expected.
Last week, the Union Cabinet approved “Mission Mausam” to improve accuracy in forecasts and nowcasts. The mission aims to improve weather models, and use ensemble models along with artificial intelligence and machine learning. But the final interpretation of the weather information will have to done by people, or by human intelligence, said M Ravichandran, secretary, ministry of earth sciences.
On that front also, the department will try to build capacity, he added.
“Tropical weather pattern is very chaotic. There is high variability and we are trying to understand mesoscale events and forecast them better,” said Ravichandran.
With improved observations, Ravichandran said nowcast warnings which are now issued every six hours, can be issued every hour to improve preparedness. The resolution of nowcast warnings can be improved from say 25 km to 10 km radius and improve localised warnings.
As part of the project, the department plans to install 100 radars totally (39 are there); 15 wind profilers; 15 radiometers and 1 cloud chamber lab in India.
“Forecasters should not take model raw output for any decision process. The model may have systematic biases and errors. Use of AI ML methods could be used to interpret the model output more efficiently. By simply looking at model output we may mistake mistakes,” said M Rajeevan, former secretary, ministry of earth sciences.