Efficient spatio-temporal weather forecasting using U-Net

TL;DR

This paper presents an efficient U-Net based autoencoder for spatio-temporal weather forecasting, achieving competitive results with low computational resources in the Weather4cast 2021 Challenge.

Contribution

The paper introduces SmaAt-UNet, a lightweight deep learning model for weather prediction that balances accuracy and computational efficiency.

Findings

Achieved competitive weather forecasting accuracy

Maintained low computational resource usage

Discussed potential future improvements

Abstract

Weather forecast plays an essential role in multiple aspects of the daily life of human beings. Currently, physics based numerical weather prediction is used to predict the weather and requires enormous amount of computational resources. In recent years, deep learning based models have seen wide success in many weather-prediction related tasks. In this paper we describe our experiments for the Weather4cast 2021 Challenge, where 8 hours of spatio-temporal weather data is predicted based on an initial one hour of spatio-temporal data. We focus on SmaAt-UNet, an efficient U-Net based autoencoder. With this model we achieve competent results whilst maintaining low computational resources. Furthermore, several approaches and possible future work is discussed at the end of the paper.