An experimental comparison of three tracking algorithms for tropical and extra-tropical cyclones

A realistic detection and tracking of tropical cyclones/extra-tropical cyclones (TC/ETC) is critical for assessing the impacts or risks that severe weather may bring.

There are a few tracking algorithms that are used by the research community. For example, TempestExtremes is an objective feature-based tracking algorithm (Ullrich and Zarzycki 2017). TempestExtremes uses sea level pressure (SLP) and 400-hPa atmospheric temperature as its feature-tracking variable to detect (E)TCs.

CycloTrack V2 (Pringle et al. 2021) is also available to track spatially filtered central SLPs by minimizing the cost functions of SLP differences between time snaps.

Recently, a multilevel robustness framework utilizes topological notions of critical points and robustness for the study of 2D time-varying vector fields.

This framework can improve the visual interpretability of climate model data in terms of TC/ETC tracking, selection, and comparison.

Such a framework has the advantage of identifying cyclonic features using only the wind vector fields, while the other tracking schemes require the dynamic and thermodynamic variables to meet specified criteria.

In this study, we will track hurricanes and compare the effectiveness and efficiency of TempestExtremes, CycloTrack V2, and the multilevel robustness framework in detecting and tracking TC/ETC within gridded data. Our ultimate goal is to either find the best tracking algorithm (if not, to quantify the uncertainty) for long-term climate data and investigate the changes in hurricane tracks in future climates.

References:

Pringle, W. J., Wang, J., Roberts, K. J., Kotamarthi, V. R. (2021). Projected Changes to Cool-Season Storm Tides in the 21st Century along the Northeastern United States Coast. Earth's Future, 9(7), e2020EF001940. https://doi.org/10.1029/2020EF001940

Ullrich, P. A., & Zarzycki, C. M. (2017). Tempestextremes v1. 0: A framework for scale-insensitive pointwise feature tracking on unstructured grids. Geoscientific Model Development, 10, 1069-1090.