Topology-Based Visualization and Tracking of Hurricane Inner Core Features in HWRF and HMON Models

With finer resolution and improved data assimilation and physics representations, regional hurricane prediction models are now able to capture fine features in the inner cores of hurricanes. These fine features, such as warm cores, primary and secondary eyewalls, updrafts and downdrafts, rainbands, are critical to the development and evolution of hurricanes. Given the large number and large data size of model runs, it is hard to look at every single forecast hour or cycle in a short time to identify these features. Automated tracking and detection are therefore needed and will also help reduce the cost of diagnosis by choosing the most interesting forecast hours and cycles to investigate.

A contour tree is a topological structure that stores the nesting relationships of contours in a scalar field with the tree nodes being maxima, minima, and saddles points. Given the time varying nature of the hurricane data, we enhance a static contour tree to picture the time-varying topology so that the changes of time-varying contours are captured. The tree nodes are aligned with inner core features, since these features are often associated with minima or maxima of temperature, wind speed, precipitation, etc., and saddles which present the splits and merges of contours are the candidate points where a secondary eyewall may be branched out from a primary eyewall. Quantitative properties, such as volume and time duration of contours, are computed and labelled on the contour tree. As a result, we are able to visualize and track the evolution of the inner core features over time, providing an interesting insight into their development from a topological perspective for the first time.