The Fidelity of Tropical Cyclone Representation in Atmospheric Reanalysis Datasets

Atmospheric reanalysis datasets have been notorious for under representing tropical cyclone (TC) structure due to their coarse spatial resolution. With regards to studies utilizing these datasets in studying TCs, reanalyses have only been practical for evaluating the large scale effects of TCs. With the release of several new reanalyses including the JMA's JRA (Onogi and coauthors 2005) and NASA's MERRA (Bosilovich et al. 2006) and the impending release of the ECMWF's ERA Interim (Simmons and coauthors 2006) and NCEP's CFSRR (Saha and Pan 2008) comes the hope of a more truthful depiction of TCs. The following work will determine the extent to which TCs are represented realistically in reanalyses.

Using the methodology of Manning and Hart (2007), this study will evaluate the structure of TCs within three atmospheric datasets: the ERA40 (Uppala and coauthors 2005), the JRA, and the MERRA. TCs from the period 1982-2001 are manually tracked from three basins: the North Western Pacific, North Eastern Pacific, and North Atlantic. Absolute errors in track in comparison to the NHC best track (Jarvinen 1984) are made. Metrics used for evaluating TC structure are drawn from the cyclone phase space (Hart 2003) in addition to the use of more traditional fields such as minimum sea level pressure and 10 m surface winds. Additionally, Student's t-tests are used to determine whether the differences between TCs within basins, between basins, and amid the datasets are statistically significant.