The European Centre for Medium-range Weather Forecasts (ECMWF) is now in the production phase of a 40-year global reanalysis project (ERA-40) for the period 1957-present using a high resolution global model (T159 or 80 km and 60 layers to 0.1 hPa).
We first compare TC detection rates between ERA-40 and the first National Centers for Environmental Prediction (NCEP) reanalysis (NCEP/R1) for the years 1957-, 1973- and 1989-94. Despite large variation in the observing system, both reanalyses achieve detection rates of 80% with ERA-40 substantially higher at 90%. However, initial position and maximum wind speed error are considerably higher than in the current operational models implying a large analysis error in the vicinity of TCs.
Comparisons of once-daily, 5-day TC track forecasts from the ERA-40 global model are made against the no-skill aid CLImatology and PERsistence (CLIPER) and it is shown that the 3-day track forecast errors are only slightly better than CLIPER. In contrast, results (to be presented) for the operational global models in 2001, with comparable horizontal resolution as ERA-40, show improvements over CLIPER at 72 h of nearly 40% in the northern Hemisphere. Thus, the ERA-40 reanalyses, from a TC forecast perspective, are not as good as had been hoped. Nonetheless, 40 years of dynamical TC track forecasts is of historical interest and results from classical TCs will be given.
TC observations (location, intensity and motion) were not assimilated in ERA-40 or NCEP/R1. Experiments with the assimilation of near-center wind profiles retrieved from they TC observations demonstrates that much of the TC analysis error could have been reduced and that there was some feedback of TC analysis error with the lower frequencies (e.g., monthly means).
The paper will conclude with an overall assessment of the pros and cons of using reanalysis for TC studies and some suggestions for future research.
Supplementary URL: