The Variability of Tropical Cyclone Structure within Atmospheric Reanalysis Datasets

Atmospheric reanalysis datasets provide one of the most dense and homogeneous sources of data for climate scale studies. The recent application of reanalyses for use in studying TCs (e.g. Sriver and Huber 2006; Hart et al. 2007) has brought renewed interest in determining the extent to which TCs are realistically depicted within these datasets. In this study, TC structure will be evaluated within 5 atmospheric reanalysis datasets (CFSR, ERA-40, ERA-Interim, JRA-25, and the MERRA) in order to provide guidance concerning the relative robustness with which TCs are depicted. Three-dimensional storm-relative composited anomalies stratified by intensity and location have been constructed to determine the extent to which the results agree with observations as well quantifying the distinctiveness of TC structure between datasets.

The results of this study show that substantial differences are found between TC representation not only between reanalyses, but between basins within each reanalysis. The North Atlantic is found to have the most robust depiction of TCs in all reanalyses which is primarily attributed to this region having the highest density and quality of observations. In contrast, composites of TCs within the North Eastern Pacific are severely muted particularly in the reanalyses which do not artificially alter TC intensity or location. These results together with those concerning TC track error may be indicative of larger issues concerning the assimilation of observations within the EPAC. With regards to inter dataset differences, the variability of representation between reanalyses appears to be dictated by whether the intensity or location of the TC was artificially altered during assimilation. Differences in resolution are also found to play an important role in creating dissimilarities between datasets. The results suggest that future generations of reanalyses will be able to properly capture the anomalous nature of TCs without the introduction of “artificial” data during the assimilation phase.