Abstract:A conventional way to monitor severe convective weather is using the composite reflectivity of radar as an indicator. For oceanic areas without radar deployment, reconstruction from satellite data is useful. However, those reconstruction models built on a land dataset are not directly applicable to the ocean due to different underlying surfaces. In this study, we built reconstruction models based on U-Net (named STR-UNet) for different underlying surfaces (land, coast, offshore, and sea), and evaluated their applicability to the ocean. Our results suggest that the comprehensive use of land, coast, and offshore datasets should be more suitable for reconstruction in the ocean than using the sea dataset. The comprehensive performances (in terms of RMSE, MAE, POD, CSI, FAR, and BIAS) of the Land-Model, Coast-Model, and Offshore-Model in the ocean are superior to those of the Sea-Model, e.g., with RMSE being 5.61, 6.08, 5.06, and 7.73 in the oceanic area (Region B), respectively. We then analyzed the importance of different types of features on different underlying surfaces for reconstruction by using interpretability methods combined with physical meaning. Overall, satellite cloud-related features are most important, followed by satellite water-related features and satellite temperature-related features. For the transition of the model from land to coast, then offshore, the importance of satellite water-related features gradually increases, while the importance of satellite cloud-related features and satellite temperature-related features gradually decreases. It is worth mentioning that in the offshore region, the importance of satellite water-related features slightly exceeds the importance of satellite cloud-related features. Finally, based on the performance of the case, the results show that the STR-UNet reconstruction models we established can accurately reconstruct the shape, location, intensity, and range of the convective center, achieving the goal of detecting severe convective weather where a radar is not present.

Context Constraints-Guided Reconstruction Robust Network for Cloud-Induced Extensive Missing Area in Sea Surface Temperature

Deep Learning for Sea Surface Temperature Reconstruction under Cloud Occlusion

Missing Data Reconstruction in Remote Sensing image with a Unified Spatial-Temporal-Spectral Deep Convolutional Neural Network

A deep-learning reconstruction method for remote sensing images with large thick cloud cover

A Daily High-Resolution Sea Surface Temperature Reconstruction Using an I-DINCAE and DNN Model Based on FY-3C Thermal Infrared Data

Context-Based Multiscale Unified Network for Missing Data Reconstruction in Remote Sensing Images

Reconstruction of Incomplete Satellite Sst Data Sets Based on Eof Method

A Correlation Context-Driven Method for Sea Fog Detection in Meteorological Satellite Imagery

Radar Echo Reconstruction in Oceanic Area via Deep Learning of Satellite Data

Transforming Observations of Ocean Temperature with a Deep Convolutional Residual Regressive Neural Network

Boundary-Guided Semantic Context Network for Water Body Extraction from Remote Sensing Images

Thick Cloud Removal in Optical Remote Sensing Images Using a Texture Complexity Guided Self-Paced Learning Method

Data driven three-dimensional temperature and salinity anomaly reconstruction of the northwest Pacific Ocean

Reconstructing three-dimensional salinity field of the South China Sea from satellite observations

Learning to Aggregate Multi-Scale Context for Instance Segmentation in Remote Sensing Images

Reconstruction of Sea Surface Temperature under Clouds Using Masked Autoencoders

An Operational Global Near-Real-Time High-Resolution Seamless Sea Surface Temperature Products From Satellite-Based Thermal Infrared Measurements

High-Resolution Seamless Daily Sea Surface Temperature Based on Satellite Data Fusion and Machine Learning over Kuroshio Extension

Spatial-spectral-temporal connective tensor network decomposition for thick cloud removal

A Non-Uniform Grid Graph Convolutional Network for Sea Surface Temperature Prediction

CCT: Conditional Co-Training for Truly Unsupervised Remote Sensing Image Segmentation in Coastal Areas