Substantial Underestimation of Fine-Mode Aerosol Loading from Wildfires and Its Radiative Effects in Current...

Wildfires generate abundant smoke primarily composed of fine-mode aerosols. However, accurately measuring the fine-mode aerosol optical depth (fAOD) is highly uncertain in most existing satellite-based aerosol products. Deep learning offers promise for inferring fAOD, but little has been done using multiangle satellite data. We developed an innovative angle-dependent deep-learning model (ADLM) that accounts for angular diversity in dual-angle observations. The model captures aerosol properties observed from dual angles in the contiguous United States and explores the potential of Greenhouse gases Observing Satellite-2's (GOSAT-2) measurements to retrieve fAOD at a 460 m spatial resolution. The ADLM demonstrates a strong performance through rigorous validation against ground-based data, revealing small biases. By comparison, the official fAOD product from the Moderate Resolution Imaging Spectroradiometer (MODIS), the Visible Infrared Imaging Radiometer Suite (VIIRS), and the Multiangle Imaging Spectroradiometer (MISR) during wildfire events is underestimated by more than 40% over western USA. This leads to significant differences in estimates of aerosol radiative forcing (ARF) from wildfires. The ADLM shows more than 20% stronger ARF than the MODIS, VIIRS, and MISR estimates, highlighting a greater impact of wildfire fAOD on Earth's energy balance.