HSE scientists, together with Yandex Cloud, have developed a neural network for predicting El Niño

2023: Developing a neural network to predict El Niño

HSE scientists, together with Yandex Cloud, have developed a neural network to predict El Niño. The HSE announced this on April 20, 2023.

This algorithm helps to accurately predict the change in the average temperature of ocean waters on the surface, which can cause natural disasters in certain regions of the world. For April 2023, the model already predicts El Niño 1.5 years ahead, and in the future, scientists plan to increase the forecast period to 2 years.

El Niño is a change in the temperature distribution of the surface of water in the Pacific Ocean, which affects the weather and can cause natural disasters in certain regions.

The neural network simulates the average temperature in the equatorial zone of the Pacific Ocean in perspective. Under El Niño, the equatorial part becomes warmer than usual. There is also a reverse process with a decrease in ocean temperature - La Niña. Such a shift cycle occurs every 2-7 years. These fluctuations have a significant impact on weather in many countries of the world and can increase the risk of fires, droughts, floods and crop failures.

The HSE scientific group trained neural networks on an array of thousands of temperature maps with synthetic and real data collected from 1800 to April 2023. In addition to standard machine learning methods for predicting such phenomena, ML specialists are testing the Autoformer architecture in training. Thanks to this, it is possible to qualitatively process the sequence of temperature maps. To pre-process datacets, scientists used the Yandex DataSphere ML development service, which has all the necessary tools and dynamically scalable cloud resources for the full cycle of machine learning development.

The challenges of global climate change are becoming more urgent. It is scary not so much warming itself as the inevitable "imbalance" of the climate on the planet. The El Niño effect plays a crucial role in the emergence of global weather and climate fluctuations, leading, for example, to massive crop failures, and therefore its forecasting is especially important in the current conditions of increasing climatic "imbalance," said big data computer sciences Dmitry Vetrov, research professor at the Department and Information Search Department of the HSE Faculty.

Cloud technologies help more efficiently conduct experiments in a scientific environment. In projects such as the El Niño study, rapid and flexible access to services to test different machine learning models is important. Each such test with a new architecture helps to predict the phenomenon as early and accurately as possible, "said Anna Lemyakina, director of national strategic projects at Yandex Cloud.