SRViT: Vision Transformers for Estimating Radar Reflectivity from Satellite Observations at Scale

Jason Stock,Kyle Hilburn,Imme Ebert-Uphoff,Charles Anderson

2024-06-29

Abstract:We introduce a transformer-based neural network to generate high-resolution (3km) synthetic radar reflectivity fields at scale from geostationary satellite imagery. This work aims to enhance short-term convective-scale forecasts of high-impact weather events and aid in data assimilation for numerical weather prediction over the United States. Compared to convolutional approaches, which have limited receptive fields, our results show improved sharpness and higher accuracy across various composite reflectivity thresholds. Additional case studies over specific atmospheric phenomena support our quantitative findings, while a novel attribution method is introduced to guide domain experts in understanding model outputs.

Signal Processing,Computer Vision and Pattern Recognition,Machine Learning

What problem does this paper attempt to address?

The main objective of this paper is to develop a Transformer-based neural network model (named SRViT, Satellite to Radar Vision Transformer) for generating high-resolution (approximately 3 kilometers) synthetic radar reflectivity fields from geostationary satellite images. Specifically, this study aims to: 1. **Enhance short-term convective-scale forecasting**: By more accurately monitoring and predicting high-impact weather events across the Continental United States (CONUS), to improve decision-making in areas such as public safety and agricultural production. 2. **Assist data assimilation**: By integrating the generated synthetic radar reflectivity data into numerical weather prediction (NWP) models, to improve the accuracy of weather forecasts. Compared to traditional convolutional methods, the SRViT model can handle longer-range dependencies, resulting in better clarity and higher accuracy under various synthetic reflectivity thresholds. Additionally, this study introduces a novel attribution method to help domain experts better understand the model outputs. In summary, this paper focuses on leveraging advanced machine learning techniques (particularly the Transformer architecture) to improve the coverage and quality of radar data, thereby enhancing the capability of short-term convective weather forecasting and supporting the data assimilation process in numerical weather prediction systems.

SRViT: Vision Transformers for Estimating Radar Reflectivity from Satellite Observations at Scale

CSwT-SR: Conv-Swin Transformer for Blind Remote Sensing Image Super-Resolution with Amplitude-Phase Learning and Structural Detail Alternating Learning

Development and Interpretation of a Neural Network-Based Synthetic Radar Reflectivity Estimator Using GOES-R Satellite Observations

Transformer-based nowcasting of radar composites from satellite images for severe weather

Vision Transformer for Multispectral Satellite Imagery: Advancing Landcover Classification*

Radar Instance Transformer: Reliable Moving Instance Segmentation in Sparse Radar Point Clouds

A Vision Transformer Approach for Efficient Near-Field Irregular SAR Super-Resolution

Intelligent Reconstruction of Radar Composite Reflectivity Based on Satellite Observations and Deep Learning

Artificial intelligence (AI)-derived 3D cloud tomography from geostationary 2D satellite data

A Spatial Resolution Enhancement Method of Microwave Radiation Imager Data Based on Data Matching and Transformer

SatensoRF: Fast Satellite Tensorial Radiance Field for Multi-date Satellite Imagery of Large Size

STC-ViT: Spatio Temporal Continuous Vision Transformer for Weather Forecasting

Advancing Plain Vision Transformer Toward Remote Sensing Foundation Model

SWHFormer: A Vision Transformer for Significant Wave Height Estimation From Nautical Radar Images

Surface Reflectance From Commercial Very High Resolution Multispectral Imagery Estimated Empirically With Synthetic Landsat (2023)

GeoViT: A Versatile Vision Transformer Architecture for Geospatial Image Analysis

DiffSR: Learning Radar Reflectivity Synthesis via Diffusion Model from Satellite Observations

An Improvement Multitask Transformer Network for Dual-Polarization Radar Extrapolation

Automated classification of remote sensing satellite images using deep learning based vision transformer

TTSR: A Transformer-Based Topography Neural Network for Digital Elevation Model Super-Resolution