In order to improve TC track and landfall forecast and to reduce the loss caused by TC, some nonlinear mathematic models of tropical cyclones are built. The results of research and experiments of more than 5 years show that TC track forecast with the exponential curve model has higher accuracy. The 24-hour average error of TC track forecast for 199 forecasts is about 123 km, which reaches an advanced level in TC track forecasting in China. Nonlinear forecasts such as multinomial curve models have been applied for 3 years in forecasting the number of TC in Guangdong Province. The accuracy can be as high as 70% to 90%. By means of the nonlinear model to analyze the long-term forecasts of day-to-day pressure and precipitation fields to make the landfall time and location forecast of TC, as well as the time forecast of TC appearing in South China Sea. The accuracy can be as high as 70% to 80%. The way of integrating medium and long term, numerical and statistical forecasting methods to forecast the TC track and landing time and location has been proved successful in 2002 TC forecasting practice.