AI improves monsoon rainfall forecasting

Posted by

A simplified diagram of a NoC algorithm with two dynamic combination members for simplicity. Here, the member of the second (violet) group will receive more weight, because it is closer to the MISO prediction in the subspace. Note that in the actual implementation, we reduce the dynamics in the MISO subspace to the first two principal components of the MISO mode. Credit: Proceedings of the National Academy of Science (2024). DOI: 10.1073/pnas.2312573121

Every year, the South Asian monsoon season brings heavy rains to more than a billion people across the Indian subcontinent between June and September. Precipitation occurs in oscillations: some weeks see 1 to 4 inches of rain, while other weeks are mostly dry. It is important for agriculture and urban planning to predict when dry and wet periods will occur, helping farmers know when to harvest crops and helping city officials prepare for floods. However, while weather predictions are mostly accurate to within a day or two, it is very difficult to accurately predict the weather over a week or month.

Now, a new machine-learning-based forecast more accurately predicts South Asian monsoon rainfall 10 to 30 days in advance, a significant improvement over current state-of-the-art forecasts that use numerical modeling instead of artificial intelligence. To make predictions. Understanding the behavior of monsoons is also important because this type of rainfall is a major atmospheric feature in the global climate.

The research was led by Aviator Bach, the Foster and Coco Stanback Postdoctoral Scholar Research Associate in Environmental Science and Engineering, who works in the laboratories of Tapio Schneider, Theodore Y. Wu Professor of Environmental Science and Engineering and senior research scientist at JPL; and Andrew Stuart, Bren Professor of Computing and Mathematical Sciences.

a paper Description of new method appears Proceedings of the National Academy of Science,

“There is a lot of concern about how climate change will impact monsoons and other weather events like hurricanes, heat waves,” Bach says. “Improving predictions on shorter time scales is an important part of responding to climate change because we need to be able to improve preparedness for these events.”






A model of how monsoon rainfall varies over the Indian subcontinent within a single season, called the “monsoon inter-seasonal oscillation”. Credit: E. Bach

Weather is difficult to predict because there are many instabilities in the atmosphere – for example, the atmosphere is constantly heated from beneath the Earth, causing cold, dense air above hotter, less dense air above – as well as uneven heating of the Earth and the Earth. Instability occurs due to rotation. These instabilities lead to a chaotic situation in which errors and uncertainties in modeling the behavior of the atmosphere grow exponentially, making it almost impossible to predict further into the future.

Current state-of-the-art models use numerical modeling, which are computer simulations of the atmosphere based on physics equations describing the motion of fluids. Due to the chaos, the maximum predictable time for large-scale weather is usually about 10 days. It is also possible to predict the long-term average behavior of the atmosphere – that is, climate – but predicting weather over time intervals ranging from two weeks to several months has been a challenge with numerical models.

With the South Asian monsoon, rainfall occurs in cycles of intense bursts followed by dry spells. These cycles are known as Monsoon Intraseasonal Oscillations (MISOs). In the new research, Bach and his colleagues added a machine-learning component to current state-of-the-art numerical models. This allowed researchers to gather data about MISO and better predict rainfall on the elusive timescale of two to four weeks. The resulting model was able to improve the correlations of predictions with observations by up to 70%.

“Over the past few years, there has been a growing interest in using machine learning to predict weather,” says Bach. “Our work shows that combining machine learning and more traditional numerical modeling can produce accurate results.”

The paper is titled “Improved subseasonal prediction of South Asian monsoon rainfall using data-driven forecasts of oscillatory modes.” In addition to Bach, the co-authors are V. Krishnamurthy and Jagdish Shukla of George Mason University; Safa Mote of Portland State University; A. Surjalal Sharma and Eugenia Kalne of the University of Maryland; and Michael Ghil of the École Normale Supérieure in Paris, UCLA and Imperial College London.

more information:
Aviator Bach et al, Improved subseasonal prediction of South Asian monsoon rainfall using data-driven forecasts of oscillatory modes, Proceedings of the National Academy of Science (2024). DOI: 10.1073/pnas.2312573121

Provided by California Institute of Technology


Citation: AI improves monsoon rainfall forecast (2024, 1 April) Retrieved on 1 April 2024 from https://phys.org/news/2024-04-ai-monsoon-rainfall.html.

This document is subject to copyright. No part may be reproduced without written permission, except in any fair dealing for the purpose of private study or research. The content is provided for information purposes only.


#improves #monsoon #rainfall #forecasting

Leave a Reply

Your email address will not be published. Required fields are marked *