Abstract: The qualities of weather forecast in Guangzhou city of five information from April to December in 2001 are tested, the time span of forecast is 24, 38, 72 hours ahead, and the tested elements are precipitation and temperature. For precipitation, the common weather of line or rain, and weather of heavy rain have been tested respectively by using the methods of operational appraisal and Ts valuation. As for temperature forecast, the variance and graded statistics have been computed. The analyses show that for the common weather forecast of fine or rain, the forecast qualities of HRMM are generally better than others, and having good stability. For 24 hours ahead, the forecast qualities of forecasters are equal to the model, but for 24~28 hours ahead, generally the model prediction is superior to other methods. All methods have poor abilities to predict the heavy rain. The Ts valuation reveals that the model has better ability to predict rain or not. For temperature forecast, generally the forecast qualities of forecaster are the best, but the HRMM has systematical error, the model forecasts are commonly lower than those of realities. For the forecasts of precipitation and temperature, the qualities of all methods are tending down with the growth of prediction time, and the errors of forecast quality of maximum temperature are always larger than that of minimum temperature.