Leading meteorological centers in US and Europe have performed "reanalysis" projects since early 1990's. In Japan, recent urgent needs for more accurate environmental and climate diagnostic information boosted the plan to produce another reanalysis dataset by its own system based on the operational models of Japan Meteorological Agency (JMA). Thus, we started the Japanese long-term (25 years) reanalysis project (JRA-25) in 2001 under a collaborative research between JMA and Central Research Institute of Electric Power Industry (CRIEPI).

Our global data assimilation system consists of a spectral forecast model with a horizontal resolution of T106, various quality control processes featuring a unique database format, and a 3D-VAR assimilation method. We're using conventional and satellite observation data from various sources, mostly provided from the European Centre for Medium-Range Weather Forecast, and boundary data of sea surface temperature and sea-ice distribution from the latest assimilation system of JMA. One of the aspects that distinguish our product from existing reanalysis datasets is use of wind retrieval data derived from tropical cyclone (TC) best track data. This TC wind retrieval data (TCR) is supposed to improve the quality of tropical analysis regarding TC positions and associated wind and moisture fields. In this study, we examine the effectiveness of TCR data from some data assimilation and forecast experiments, in particular, focused on TCs generated in a period of September 1990.

In the analysis, using TCR and conventional dataset, we confirmed that the observed TCs were realistically represented even over data sparse regions such as the eastern North Pacific. Moisture fields around TCs were also corrected by modified wind there. We found that TCR had positive impacts on TC motion forecasts especially during recurving process. By using TCR, TC location in a three to four days forecast approached to the best track. Thus, modification of TC locations and moisture field with realistic wind field is useful not only for climate representation but also for improving a forecast skill. We're developing objective detection method for tropical cyclones from the analysis, which can be useful to validate analysis quality.