5C.5 Toward a Homogeneous Global Tropical Cyclone Intensity Record

Tuesday, 1 April 2014: 9:00 AM
Pacific Ballroom (Town and Country Resort )
Christopher C. Hennon, Univ. of North Carolina at Asheville, Asheville, NC; and K. R. Knapp, C. J. Schreck III, S. E. Stevens, J. P. Kossin, P. W. Thorne, P. A. Hennon, M. C. Kruk, and J. Rennie

Cyclone Center is a crowd sourcing project that engages the general public to produce a refined global tropical cyclone intensity record. A website presents users with homogenized infrared tropical cyclone (TC) centric imagery, asks questions about the images, and records responses. A consensus intensity is then determined by processing user responses through an algorithm that closely follows the Dvorak Enhanced Infrared (EIR) technique.

The project will help to reconcile uncertainties in the global TC intensity record; these uncertainties arise primarily because almost all TCs are not directly measured. Heterogeneities are introduced into the historical record with the evolution of operational procedures, personnel, and observing platforms. Furthermore, the application of the EIR procedure can be subjective. The disagreements in best track data impede our ability to identify the relationship between TC intensities and, for example, recent climate change. Differences in intensity for particular storms can be especially large in the western Pacific and Indian Ocean basins, where multiple TC forecast agencies produce best track data for a TC. Research on recent intensity changes in the area has been contradictory, and there are a number of examples of large intensity differences (> 70 kt in some cases) in storm best track data.

A global reanalysis of TC intensity using humans is impractical because of the sheer number of storms. Crowd sourcing provides us with a promising method to acquire multiple classifications for each of the nearly 300,000 storm images available in 32 years (1978-2009) of HURSAT B1 data. Focusing on the storms with the highest degree of agency disagreement first, we have recorded over 265,000 individual classifications from several thousand citizen scientists with little or no meteorological background. Simply designed questions, inline help, tutorials, discussion forums, blogs, and a Facebook page have enabled our users to provide us with all of the data necessary to address the discrepancies in these storms and further efforts toward a future global reanalysis. In addition, Cyclone Center also serves as a global TC morphology survey; information on storm characteristics such as cloud pattern distribution, eye size, and intensity trends are automatically recorded.

This presentation will emphasize the conceptual design and implementation of the Cyclone Center interface and algorithm. Preliminary results will be shown for two TC case studies (Ivan (1997) and Yvette (1992)); these suggest that the performance of the human classifiers can meet and sometimes exceed that of the Advanced Dvorak Technique (ADT) when applied to the same data, especially for times when the storm is transitioning into a hurricane.

Supplementary URL: http://cyclonecenter.org

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner