Unlocking the power of dual-angle satellite observations for synergistic retrieval of aerosol size information

Aerosol particle size distribution is vital for distinguishing natural from anthropogenic emissions and understanding fine particulate matter formation. Most existing deep learning methods rely on single-view satellite data and overlook significant links between aerosol parameters. Here, we propose VISTA-Net (View-Integrated Synergistic Task-Aware Network), a novel deep learning model that integrates dual-angle GOSAT-2 observations for simultaneous retrieval of fine-mode aerosol optical depth (fAOD) and fine mode fraction (FMF). Validated against ground-based AERONET measurements, VISTA-Net achieves high correlation coefficients of 0.86 for fAOD and 0.57 for FMF over six major global regions (the United States, Europe, the Middle East, Japan, South Asia, and Australia). This approach demonstrates the value of multi-angle fusion and multi-task learning in aerosol remote sensing, offering precise aerosol size information datasets to support climate impact assessments.