Compressing multidimensional weather and climate data into neural networks

TL;DR

This paper introduces a neural network-based compression method for high-resolution weather and climate data, achieving over 300x compression while maintaining data fidelity and enabling easier access for research.

Contribution

The authors propose a coordinate-based neural network approach that significantly outperforms existing compressors in data compression ratios and preserves essential climate structures.

Findings

Achieves 300x to 3000x compression ratios.

Outperforms SZ3 in weighted RMSE and MAE.

Less than 2% increase in forecasting RMSE using compressed data.

Abstract

Weather and climate simulations produce petabytes of high-resolution data that are later analyzed by researchers in order to understand climate change or severe weather. We propose a new method of compressing this multidimensional weather and climate data: a coordinate-based neural network is trained to overfit the data, and the resulting parameters are taken as a compact representation of the original grid-based data. While compression ratios range from 300x to more than 3,000x, our method outperforms the state-of-the-art compressor SZ3 in terms of weighted RMSE, MAE. It can faithfully preserve important large scale atmosphere structures and does not introduce artifacts. When using the resulting neural network as a 790x compressed dataloader to train the WeatherBench forecasting model, its RMSE increases by less than 2%. The three orders of magnitude compression democratizes access to high-resolution climate data and enables numerous new research directions.