Storm Classification and Dynamic Targeting for a Smart Ice Cloud Sensing Satellite

Smart Ice Cloud Sensing (SMICES) is a small-sat concept in which a radar intelligently targets ice storms based on information collected by a look-ahead radiometer. Often, space observations are performed by continuously collecting data from an instrument aimed at nadir (e.g., directly below the space platform). However, if the platform has the ability to assess science utility of features that will be overflown, an intelligent measurement scheme can improve science return. In the case of SMICES, power constraints and the rarity of storms mean that with blind nadir targeting SMICES would collect a limited amount of ice-storm radar data. The work proposed acquires measurements to maximize acquired high-interest storms while concurrently collecting a background sampling of all features. This is accomplished through two steps: storm classification and dynamic targeting. For classification, we describe a multistep use of machine learning and digital twin of Earth’s atmosphere to create a classifier. We discuss an autonomous data labeling pipeline used to train five different models to identify storms in tropical and nontropical regions and assess the results and impact of expected noise. For dynamic targeting, six algorithms, ranging from “blind” to more selective, are described and evaluated on their improvement over “blind” targeting.