The India Meteorological Department is preparing to adopt artificial intelligence and machine learning to facilitate precise weather forecasting. In two to three years, the department hopes to utilize AI/ML completely.
For this effort, IMD has partnered with IIITs in Prayagraj and Vadodara and IIT Kharagpur. In addition, the organization has teamed up with Google to produce accurate short- and long-term projections.
Senior authorities and meteorologists from IMD have formed internal sub-groups to discover the best AI and machine learning techniques for weather forecasting. The Ministry of Earth Sciences is seeking independent researchers to aid the team in this endeavour. IMD has also established a network of sister organizations, including the Indian Institute of Tropical Meteorology in Pune.
Mrutyunjay Mohapatra, the director-general of the IMD, indicated last year that the department would release “nowcasts” for real-time forecasting of severe weather occurrences. Extreme weather changes are predicted 3-6 hours ahead of time by the nowcasts. IMD has traditionally issued such forecasts using radars and satellite imagery. It is, however, a lengthy procedure. IMD had also invited several academic groups to look into how AI could improve weather predictions.
The Ministry of Earth Sciences announced projects in 2019 to improve weather forecasting using artificial intelligence. To increase understanding of weather and climate, the ministry seeks to expand the present supercomputing facility to 100 petaflops (PF).
The Ministry of Agriculture and Farmers Welfare struck a deal with IBM India the same year to develop AI and weather technology-based solutions for weather forecasting and soil moisture information at the farm level to aid farmers in crop management decisions.
AI for weather forecasting
Organizations worldwide rely mainly on satellite imagery and radar inputs to acquire weather and climate data for forecasting. Humans and even traditional computers cannot analyze and scan the data because it is too large and overwhelming.
Pattern recognition in weather and climate information is aided by AI systems, machine learning, neural networks, and deep learning.
According to research published by the American Meteorological Society (AMS), AI can improve weather forecasting by scanning large amounts of data in a short amount of time. According to the report, AI could assist foresee the effects of extreme weather and reduce energy consumption.
Weather bureaus all across the world have implemented sophisticated and intelligent monitoring systems. The National Oceanic and Atmospheric Administration (NOAA) of the United States said in 2020 that it would focus on four areas to improve the quality and timeliness of NOAA science, goods, and services: NOAA Unmanned Systems, artificial intelligence, Omics, and the cloud. NOAA’s Satellite and Information Service branch has partnered with Google to employ AI and machine learning to make satellite and environmental data more accessible.
Recent efforts with AI
The University of Washington and Microsoft Research teamed up last year to show how AI can analyze past weather trends and anticipate future events. The method proved to be more efficient and precise than current technologies. Using weather data from the previous 40 years, the team constructed a worldwide weather model to make predictions. “Machine learning is essentially a glorified version of pattern recognition,” said Jonathan Weyn, the study’s lead author. It recognizes a familiar pattern, understands how it typically evolves and determines what to do based on the examples it has seen over the last 40 years of data.”
MetNet, a neural meteorological model for precipitation forecasting, was unveiled by Google in January 2020. The deep learning network can forecast future precipitation with a resolution of 1 km at 2-minute intervals across timescales of up to 8 hours. According to a preliminary study, the model could predict the weather for the entire United States in a matter of seconds, rather than the hour it took previous technologies.
IBM purchased The Weather Company two years ago and created Deep Thunder to give hyper-local weather forecasts with a resolution of 0.2 to 12 miles.