Abstract: The performance of analyses and forecasts from BJ-RUC (Beijing Rapid Updated Cycling Analysis and Forecast System) are evaluated against the operational L-band radiosonde observations at Beijing Weather Observatory sounding station at 0800 BT, 1400 BT and 2000 BT during the 2-month period of August 2010 and August 2011. From the results, it can be found that in Beijing Area, the detailed structure of the boundary layer revealed by the very high vertical resolution L-band radiosonde observations during the daytime, is generally well predicted by the model. However, systematical bias is also significantly identified. For the temperature profile, WRF model is capable of forecasting the thin layer with smaller temperature lapse rate below 400 m, while the whole boundary layer is forecasted with cold bias and warm bias at 1400 BT and 2000 BT for the boundary layer below 1 km. In addition, significant wet biases are also identified from the analyses and forecasts within the boundary layer. Overall, there are much larger systematical biases for forecasts within planetary boundary layer (PBL) than those in free atmosphere.According to the L-band radiosonde observations, the prevailing PBL mountain breeze at night usually transits into the valley breeze in the morning (around 0800 BT), there after the SSW wind will be dominant in PBL below 1500 m during the afternoon (from 1400 BT to 2000 BT) in Beijing Area. The WRF model accurately predicts such kind of diurnal feature of PBL circulation, but the wind speed in the morning is under-predicted while over-predicted in the afternoon. The forecast performance for height of PBL at 1400 BT is also verified against those derived from radiosonde data. The forecasted daily variation tendency of the height fits well with the observation during the evaluation period. But more or less, WRF model will over-predict the height under the circumstances of clear sky and light-mist, which could be partially ascribed to the feature of too strong vertical mixing in Yongsei University (YSU) PBL scheme utilized in the model.