A Novel Transformer Network with Shifted Window Cross-Attention for Spatiotemporal Weather Forecasting

TL;DR

This paper introduces a novel video transformer architecture with shifted window cross-attention for efficient and accurate short-term weather forecasting, addressing computational and data challenges in vision transformer models.

Contribution

It proposes a new transformer-based model with a specialized augmentation scheme and cross-attention mechanism tailored for spatiotemporal weather prediction tasks.

Findings

Achieved an MSE of 0.4750 on training data

Reduced MSE to 0.4420 with transfer learning

Demonstrated effectiveness on Weather4Cast2021 dataset

Abstract

Earth Observatory is a growing research area that can capitalize on the powers of AI for short time forecasting, a Now-casting scenario. In this work, we tackle the challenge of weather forecasting using a video transformer network. Vision transformer architectures have been explored in various applications, with major constraints being the computational complexity of Attention and the data hungry training. To address these issues, we propose the use of Video Swin-Transformer, coupled with a dedicated augmentation scheme. Moreover, we employ gradual spatial reduction on the encoder side and cross-attention on the decoder. The proposed approach is tested on the Weather4Cast2021 weather forecasting challenge data, which requires the prediction of 8 hours ahead future frames (4 per hour) from an hourly weather product sequence. The dataset was normalized to 0-1 to facilitate using the evaluation metrics across different datasets. The model results in an MSE score of 0.4750 when provided with training data, and 0.4420 during transfer learning without using training data, respectively.