Based on Bayesian theory, a basic model of the probabilistic forecast was established out of the deterministic forecast.The probability forecast was evaluated and put into use by using European Centre for Medium-Range Weather Forecast(ECMWF) ensemble forecast of 2 m temperature taken from the TIGGE data and the observed data.The results show that the probability forecast can provide a plenty of forecast products as well as better forecast average value than the original deterministic forecast.The Bayesian model averaging(BMA) approach was applied to conduct the multimodel ensemble probabilistic forecast by using China Meteorological Administration(CMA), US National Centers for Environmental Prediction(NCEP) and ECMWF ensemble forecast data.More reasonable probability distribution was acquired from the BMA forecast, the distribution average value can be used as the model forecast, and the standard deviation and the confidence interval reflect the variable range of the forecast.Therefore, the probabilistic forecast based on the Bayesian forecast model is able to provide more accurate forecast as well as more comprehensive forecast information.The BMA ensemble forecast is superior to the ensemble mean forecast in terms of its forecast accuracy, and it can quantitatively provide the uncertainty of the forecast as well.The BMA probabilistic forecast experiment was carried out by using 51 members of the ECMWF ensemble forecasts.The results show that the BMA forecast merges the description of uncertainty of each ensemble member, and it adjusts the average value of probabilistic forecast in order to make it closer to the observed value.The probability density distribution(PDF) of BMA forecast can better reflect the realistic distribution of atmosphere.