Self-supervised Spatial-Temporal Learner for Precipitation Nowcasting

TL;DR

This paper introduces SpaT-SparK, a self-supervised spatial-temporal model for precipitation nowcasting that outperforms supervised models by leveraging masked image modeling and temporal translation networks.

Contribution

The paper presents a novel self-supervised learning framework combining CNN-based encoding with temporal translation for improved precipitation nowcasting.

Findings

SpaT-SparK outperforms baseline models like SmaAt-UNet.

Self-supervised pretraining enhances spatial-temporal weather prediction.

The model achieves higher accuracy on the NL-50 dataset.

Abstract

Nowcasting, the short-term prediction of weather, is essential for making timely and weather-dependent decisions. Specifically, precipitation nowcasting aims to predict precipitation at a local level within a 6-hour time frame. This task can be framed as a spatial-temporal sequence forecasting problem, where deep learning methods have been particularly effective. However, despite advancements in self-supervised learning, most successful methods for nowcasting remain fully supervised. Self-supervised learning is advantageous for pretraining models to learn representations without requiring extensive labeled data. In this work, we leverage the benefits of self-supervised learning and integrate it with spatial-temporal learning to develop a novel model, SpaT-SparK. SpaT-SparK comprises a CNN-based encoder-decoder structure pretrained with a masked image modeling (MIM) task and a translation network that captures temporal relationships among past and future precipitation maps in downstream tasks. We conducted experiments on the NL-50 dataset to evaluate the performance of SpaT-SparK. The results demonstrate that SpaT-SparK outperforms existing baseline supervised models, such as SmaAt-UNet, providing more accurate nowcasting predictions.