Abstract

Tropical cyclones are highly organized, rotating storms which rank among the most costly natural disasters in the United States. The processes which drive intensification and weakening of such storms are still not fully understood, particularly when these intensity changes occur on short time sca... Show moreTropical cyclones are highly organized, rotating storms which rank among the most costly natural disasters in the United States. The processes which drive intensification and weakening of such storms are still not fully understood, particularly when these intensity changes occur on short time scales. The physical and environmental factors used in intensity prediction schemes often do not consider the spatial structure of the storm and, taken alone, are inadequate for describing the evolution of storms during rapid intensity change events. Since tropical cyclones generally form far from land-based observing networks, we often rely on satellite observations to assess these storms, particularly infrared observations that reveal cloud top temperatures as a proxy for strength of convection in these storms. In addition, advances in satellite instrumentation has continued to improve the spatial and temporal resolution of these observations. To take advantage of this information, we develop a suite of features which quantify the spatial structure of convection within the storm via infrared brightness temperature data from the Geostationary Operational Environmental Satellites. These features target the bulk morphology, core structure, and overall organization of the storm to provide a rich, interpretable description of the spatial structure of convection within the storm. This quantification provides a foundation for applying powerful but otherwise hard-to-interpret machine learning techniques to further our understanding of the physical processes behind rapid intensity change in tropical cyclones and support future improvements in forecasting such changes. Show less