The Benefits of Additional Satellite Channels for a Multimodal Deep Learning Method for Precipitation Estimation

For the purpose of improving precipitation estimation, we have developed a deep learning method for merging different modalities of measurements. In a previous paper, we proposed a method that can merge rain gauge measurements with ground-based radar composite and thermal infrared satellite imagery. In this paper, we study the benefits of also using other satellite channels that are not infrared. Because non-infrared channels are not visible during the night, we also study the difference in performance during the day and the night.