Many east Asian countries are located at the northwest Pacific basin which is, on average, the most active tropical cyclone (TC) region in the world. This area is also one of the most populous regions, so the accurate prediction of tropical cyclone tracks is very important. As a consequence, active research programs are part of weather centers in many countries in the region, including Japan, PRC, South Korea and Taiwan. In this study, we present numerical results obtained for TCs which affected east Asia, based on a high-resolution limited-area assimilation and prediction system, HIRES (HIgh RESolution model). Numerical experiments for a total of over 50 TC cases have been performed by employing advanced assimilation techniques, including four-dimensional variational analysis (4D-Var) and inverse 3D-Var (I3D-Var; Kalnay et al. 2000), over a 12 hour period prior to the forecast period, followed by high resolution (10 km) predictions of tracks, intensity and precipitation. The variational assimilation schemes used all available data, including satellite and radar observations. The main focus has been on the 5 TC seasons (1999-2003) in both the Pacific and the East/Japan Sea regions, especially TCs affecting Japan, the PRC and South Korea. A mix of difficult and less difficult storms was chosen, to provide a more representative sample of TC cases. Results are presented concentrating on a number of aspects of the numerical experiments. The impact of assimilating non-conventional data using both 4D-Var and I3D-Var methods will be discussed. In addition, a comparison is made relative to control forecasts using an intermittent (6-hourly cycled OI) data assimilation scheme. Other aspects of the forecasts that are assessed include tracks, intensity and quantitative precipitation forecasting for several storms that produced flooding after nearing or making landfall. Planned future work includes the hybrid application of I3D-Var and 4D-Var techniques, given the large gains in efficiency provided by the I3D-Var approach.