Atmospheric reanalysis datasets provide a unique opportunity for studying tropical cyclones (TCs) on climate scales by providing a relatively homogeneous hindcast of the last several decades that is only affected by changes in the observing system. While an increasing number of studies have utilized these datasets for studying TCs (e.g. Sriver and Huber 2006; Hart et al. 2007), comparatively few have quantified the fidelity of their representation within reanalyses. The following study seeks to examine the position, intensity, and the intensity lifecycle of TCs among five reanalysis datasets (NCEP CFSR, ECWMF ERA-40, ECMWF ERA-Interim, JMA JRA-25, and NASA MERRA) to evaluate the degree to which these datasets are suitable for studying TCs. The primary focus of this work involves quantifying position differences, intensity differences, and differences among the life cycle of intensity between reanalyses and best-track data for TCs within the Eastern North Pacific (EPAC), North Atlantic (NATL), and Western North Pacific (WPAC).

The results of this evaluation indicate that the representation of reanalysis TC intensity is most problematic due to its gross underestimation beyond what can be attributed to the coarse resolution of reanalyses. Furthermore, the mean lifecycle of normalized reanalysis TC intensity exhibits an underestimation of pre-and post-peak TC intensification rates as well as a delayed peak in intensity relative to the best-track. These discrepancies between best-track and reanalysis TC intensity lifecycle are supported by nonphysical correlations between reanalysis TC intensity and either best-track TC age, best-track TC latitude, and extended best-track TC size that are generally stronger than those with best-track TC intensity. The differences in the magnitude and lifecycle of best-track and reanalysis TC intensity imply that a bias correction of reanalyses would not replicate the best-track intensity distribution.

With regards to TC position, the ERA-40, ERA-I, and MERRA exhibit decreases in position differences towards the most observation dense regions in the NATL and WPAC as shown by statistically significant correlations (0.27 ≤ R ≤ 0.38). In contrast, the assimilation of best-track data within the CFSR and JRA-25 results in smaller, more spatially homogeneous position differences in each basin as well as stronger reanalysis TC intensities. Position differences are also found to decrease with increasing best-track TC intensity and reanalysis TC intensity. Of the three basins, the EPAC is found to have TCs with the largest position differences, weakest reanalysis TC intensities, and weakest relationship between best-track age and reanalysis TC intensity implying substantial deficiencies in EPAC TC representation. The results presented here strongly suggests that reanalysis TCs should be tracked independently of their best-track counterparts for studies dependent on TC track. Further, caution should be exercised when using these datasets for research that relies strongly on TC intensity or the lifecycle of TC intensity.