Since anthropogenic and natural aerosols accumulate in the PBL, the knowledge of Planetary Boundary Layer height (zPBL) evolution is very useful for prediction of pollutant concentration [1]. In our work, we considered two techniques to evaluate the zPBL: one based on the vertical gradient in the Lidar (Light Detection and Ranging) signal, the second based on the thermal inversion estimated by means of Radio Sounding measurements. Since a homogeneous distribution of observational points (such as Lidar and Radiosoundings stations) is not possible, it is very useful to evaluate zPBL by Numerical Weather Prediction (NWP) models. The goal of the proposed work is to compare the PBL height predicted by means of the WRF (Weather Research Forecast) model with the PBL height indirectly evaluated by means of the two aforementioned techniques. Different WRF settings have been considered, associated with different PBL model Physics. In particular, we ran five different combinations of Planetary Boundary Layer schemes and Surface Layer Schemes. To evaluate the WRF output performances, we considered two-fixed location [2]. In particular, we compared the WRF output with the Lidar measures in the city of Taranto in the morning hours and with the radiosounding measures in the city of Brindisi at midday. Both cities are located in Apulia, the southeastern Italian region. Results show that predicted values from all five different WRF parameterizations are comparable to each other and the simulated zPBL is on average less than the measured PBL height. We suppose that this behavior is due to the complexity of the Apulia morphology, a narrow peninsula between two seas, with particular reference to the effect of the land-sea discontinuity on the vertical structure of the PBL. Since no WRF model implementation showed better performances, we used different methods of post processing, the first based on the Artificial Neural Network algorithm and the second on the Kalman Filter to improve our results. Both methods show a Mean Bias lower than each single forecast.
Multi-physics ensemble using different planetary boundary layer scheme in WRF model for PBL height prediction over Apulia region
2018
Abstract
Since anthropogenic and natural aerosols accumulate in the PBL, the knowledge of Planetary Boundary Layer height (zPBL) evolution is very useful for prediction of pollutant concentration [1]. In our work, we considered two techniques to evaluate the zPBL: one based on the vertical gradient in the Lidar (Light Detection and Ranging) signal, the second based on the thermal inversion estimated by means of Radio Sounding measurements. Since a homogeneous distribution of observational points (such as Lidar and Radiosoundings stations) is not possible, it is very useful to evaluate zPBL by Numerical Weather Prediction (NWP) models. The goal of the proposed work is to compare the PBL height predicted by means of the WRF (Weather Research Forecast) model with the PBL height indirectly evaluated by means of the two aforementioned techniques. Different WRF settings have been considered, associated with different PBL model Physics. In particular, we ran five different combinations of Planetary Boundary Layer schemes and Surface Layer Schemes. To evaluate the WRF output performances, we considered two-fixed location [2]. In particular, we compared the WRF output with the Lidar measures in the city of Taranto in the morning hours and with the radiosounding measures in the city of Brindisi at midday. Both cities are located in Apulia, the southeastern Italian region. Results show that predicted values from all five different WRF parameterizations are comparable to each other and the simulated zPBL is on average less than the measured PBL height. We suppose that this behavior is due to the complexity of the Apulia morphology, a narrow peninsula between two seas, with particular reference to the effect of the land-sea discontinuity on the vertical structure of the PBL. Since no WRF model implementation showed better performances, we used different methods of post processing, the first based on the Artificial Neural Network algorithm and the second on the Kalman Filter to improve our results. Both methods show a Mean Bias lower than each single forecast.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
