Prospects for an improved understanding of the low frequency variability of tropical cyclones from reanalysis

Studies of the role of tropical cyclones (TC) in the large-scale circulation and TC interannual variability have been limited by the availability of a long time series of high quality tropical analyses. Reanalysis, in principle, can provide such a data set and given 40 years analyses and once daily 72-h forecasts planned in the next ECMWF reanalysis (ERA-40), it will be possible to quantify the suitability of reanalysis for TC research.

As the only high-amplitude synoptic-scale feature of the tropical atmosphere, the reanalyzed tropical cyclone and its motion forecast is strongly dependent on many aspects of the model and data assimilation system. Thus, detection of tropical cyclones in the reanalysis and track forecast skill of the "reforecasts" is a convenient measure of the quality of the tropical analyses.

This paper first reviews the ability of the NCEP and ERA-15 (the first ECMWF reanalysis) to detect TCs in the reanalyses. Both systems consistently achieved detection rates of nearly 90% in most basins during the common period of 1979-93. In contrast, the operational ECMWF analyses (with a comparable resolution model) showed very low detection in the mid 80's (30%) implying that improved modeling and assimilation are dominant factors in detection. We then compare the performance of two very different operational systems (the T159L24 U.S. Navy model with synthetic TC observations, and the T319L31 ECMWF model with no synthetic obs) during the 1998-99 season where we find that long-term (72-120 h) track skill is not dependent on analysis error (position and intensity) near the tropical cyclone. Finally, we show results from an adaptation of the Navy synthetic TC observation scheme to the perspective ERA-40 system (T159L60). In conventional data only experiments (a proxy to the pre-satellite period before 1979), we find poor detection rates in the Atlantic and eastern North Pacific basins (< 50%) and better rates in the relatively observation-rich western North Pacific. Use of synthetic TC observations greatly improves detection (consistently > 90%), in both the analysis and forecast, and gives much better track forecasts particularly in the 0-72 h time period. These results suggests that a combination of high quality TC position, intensity and motion data and the high resolution model/data assimilation in ERA-40 will yield high quality tropical analysis that will enable new and more insightful studies of tropical cyclones and the general circulation.