Last Updated on December 26, 2023 by SPN Editor
The Indian Meteorological Department (IMD) is exploring the use of AI in weather forecasting. The IMD is testing AI for creating climate models to forecast severe weather conditions. Artificial Intelligence has demonstrated the potential to enhance the speed and efficiency of weather forecasting, and in some instances, it may also improve accuracy.
K.S. Hosalikar, head of climate research and services at the IMD, believes that AI-based climate models and advisories will use AI in weather forecasting, especially for severe weather conditions like floods and droughts.
Currently, the IMD uses mathematical models and supercomputers for forecasting. The incorporation of AI could potentially generate higher-quality and less expensive weather data. AI is already being used to create public alerts for extreme weather conditions like heat waves and disease outbreaks like malaria.
The IMD plans to increase the number of weather observatories down to a village level. This is aimed at obtaining higher-resolution forecasting data. The government has set up a center to test the idea of incorporating AI in weather forecasting. They also plan to conduct workshops and conferences on this topic.
Accurate weather forecasting is critical in a country like India, which has diverse weather patterns. Its 1.4 billion inhabitants will be increasingly affected by worsening droughts, heat waves, and intense flooding. India is the second-largest producer of rice, wheat, and sugar, making accurate weather forecasting crucial for agricultural production.
The weather forecasting scene is already utilizing emerging technologies like blockchain and crypto to optimize weather data collection. For instance, the startup WeatherXM has deployed hundreds of decentralized weather stations around the world to harvest local data and provide station owners with utility tokens.
At the latest Wired Impact Conference in London, Google’s DeepMind climate action lead Sims Witherspoon suggested a new strategy called the “Understand, Optimize, Accelerate” framework. This outlines steps to use AI to tackle climate change.
Accuracy of Using AI in Weather Forecasting
The precision of AI in weather forecasting is influenced by several factors, such as the quality and volume of the data used, the complexity of the AI model, and the specific type of weather event being forecasted.
For example, Google’s DeepMind has developed an AI model called GraphCast that can provide weather forecasts for the next 10 days with remarkable accuracy in less than a minute. This model can predict weather conditions up to 10 days ahead with greater accuracy and speed than the industry’s standard weather simulation system, the High Resolution Forecast (HRES), which is produced by the European Centre for Medium-Range Weather Forecasts (ECMWF).
GraphCast’s open-source design revolutionizes access to state-of-the-art forecasting technology. By democratizing this powerful tool, it extends its reach to diverse users, spanning from remote small-scale farmers to major meteorological institutions.
This accessibility fuels immense innovation potential, fostering tailored weather solutions that cater to specific localities. The broadened availability of this technology ignites a wave of creativity and problem-solving using AI in weather forecasting, benefitting communities worldwide.
While AI models can be quicker and more efficient, they may not always surpass the accuracy of traditional forecasting models. In fact, a newly developed AI model was found to be less precise than the leading traditional forecasting models of today, even though it required about 7,000 times less computational power to generate forecasts for the same number of global points.
AI is Key to Better Climate Models
The concept of the butterfly effect beautifully illustrates the challenge of forecasting chaotic systems like weather, where small actions can lead to significant consequences. While short-term weather predictions of up to 10 days are now quite reliable, long-term climate forecasting remains intricate.
Accurate climate models are crucial for decision-making across various sectors, but currently, they lack the computational power to simulate all global climate aspects, especially small-scale processes. A proposed high-resolution modeling project may be impractical due to its immense computing demands and exclusive accessibility.
In response, a group of scientists suggests an alternative: moderately higher-resolution models coupled with machine learning to fill gaps. This innovative approach aims to create robust models without the need for massive computing centers, thereby using AI in weather forecasting.
Andrew Stuart’s work on reducing computational costs and leveraging machine learning in climate modeling holds promise. These tools, akin to image generators, seek to mitigate uncertainties and generate data, improving the accuracy of climate predictions despite existing computational limitations.
Artificial Intelligence, particularly machine learning, has revolutionized weather forecasting by extensively utilizing data to enhance accuracy. It has improved weather forecasts significantly over the years by harnessing vast amounts of information about Earth.