Next-Generation Satellite Architecture Planning-Scientific Arguments for Fusing Hyperspectral Microwave/Infrared Data Using Machine Learning

This study explores the scientific arguments for fusing hyperspectral microwave and hyperspectral infrared satellite data using machine learning (ML) and artificial intelligence (AI), with a particular focus on enhancing the planning of next-generation satellite architectures. Through simulation experiments and assessments of the information content from both program-of-record infrared and microwave sounder observations and notional hyperspectral microwave sensors, we demonstrate that combining these data types significantly improves our understanding of the thermodynamic atmospheric state, from the surface to the top of the atmosphere, under all-surface and all-weather conditions. By leveraging the unique advantages of both microwave and infrared hyperspectral observations, we show how this fusion enhances the retrieval of key meteorological parameters, such as severe weather convective indices, planetary boundary layer (PBL) heights, surface wind speed, and surface pressure—critical parameters for accurate weather forecasting. The study highlights the value of exploring the microwave window regions between 5 GHz and 320 GHz, which provides unprecedented insights into PBL height, a crucial component in understanding atmospheric dynamics. The study also emphasizes the benefits of deploying a constellation of hyperspectral infrared and microwave sensors to improve global coverage and data availability, enabling more accurate and timely weather forecasting. The fusion of hyperspectral infrared and microwave data from a constellation of low-Earth orbit (LEO) satellites offers a significant opportunity to improve in temporal, spatial and spectral coverage compared to current state-of-the-art satellite sounders. This improvement is particularly valuable for enhancing the real-time monitoring of severe meteorological events, such as tropical cyclones, by offering more frequent and detailed data on atmospheric conditions. Using ML/AI techniques, we show that these complex data sets can be effectively and efficiently fused to generate comprehensive geophysical parameter retrievals, providing critical insights into the atmosphere. The conclusions drawn from this work offer valuable recommendations for the future development of Earth observing systems, underscoring the transformative potential of advanced satellite architectures in improving the precision and reliability of weather predictions.