Meteorological Satellite Images Prediction Based on Deep Multi-scales Extrapolation Fusion

Fang Huang,Wencong Cheng,PanFeng Wang,ZhiGang Wang,HongHong He

DOI: https://doi.org/10.48550/arXiv.2209.11682

2022-09-20

Abstract:Meteorological satellite imagery is critical for meteorologists. The data have played an important role in monitoring and analyzing weather and climate changes. However, satellite imagery is a kind of observation data and exists a significant time delay when transmitting the data back to Earth. It is important to make accurate predictions for meteorological satellite images, especially the nowcasting prediction up to 2 hours ahead. In recent years, there has been growing interest in the research of nowcasting prediction applications of weather radar images based on deep learning. Compared to the weather radar images prediction problem, the main challenge for meteorological satellite images prediction is the large-scale observation areas and therefore the large sizes of the observation products. Here we present a deep multi-scales extrapolation fusion method, to address the challenge of the meteorological satellite images nowcasting prediction. First, we downsample the original satellite images dataset with large size to several images datasets with smaller resolutions, then we use a deep spatiotemporal sequences prediction method to generate the multi-scales prediction images with different resolutions separately. Second, we fuse the multi-scales prediction results to the targeting prediction images with the original size by a conditional generative adversarial network. The experiments based on the FY-4A meteorological satellite data show that the proposed method can generate realistic prediction images that effectively capture the evolutions of the weather systems in detail. We believe that the general idea of this work can be potentially applied to other spatiotemporal sequence prediction tasks with a large size.

Computer Vision and Pattern Recognition,Image and Video Processing

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the nowcasting prediction of meteorological satellite images, especially predicting meteorological satellite images within the next 2 hours. Due to the large spatio - temporal scale of meteorological satellite images and the significant time delay during transmission, how to accurately predict these images has become the focus of research. Existing deep - learning methods face computational resource limitations when dealing with large - scale images, resulting in blurred prediction results and discontinuity between adjacent blocks. For this reason, the paper proposes a deep - learning method based on multi - scale extrapolation fusion (MEF - GAN) to improve prediction accuracy and image quality.

Specifically, the paper solves the problem through the following steps:
1. **Multi - scale down - sampling**: Down - sample the original high - resolution satellite image into multiple image sequences with smaller resolutions for processing under limited computational resources.
2. **Multi - scale prediction**: Use a deep spatio - temporal sequence prediction model (such as the MIM model) to predict image sequences with different resolutions respectively.
3. **Result fusion**: Use a conditional generative adversarial network (CGAN) to fuse multi - scale prediction results into a prediction image with the same resolution as the original one, thereby generating a high - quality prediction image.

This method not only solves the problems of blurred and discontinuous prediction results in existing methods but also improves the accuracy and detail clarity of prediction. Experimental results show that this method is superior to traditional optical flow methods and single - scale prediction methods in both 1 - hour and 2 - hour predictions.