Application of bias adjustment techniques to improve air quality forecasts

Abstract Two bias adjustment techniques, the hybrid forecast (HF) and the Kalman filter (KF), have been applied to investigate their capability to improve the accuracy of predictions supplied by an air quality forecast system (AQFS). The studied AQFS operationally predicts NO 2 , ozone, particulate matter and other pollutants concentrations for the Lazio Region (Central Italy). A thorough evaluation of the AQFS and the two techniques has been performed through calculation and analysis of statistical parameters and skill scores. The evaluation performed considering all Lazio region monitoring sites evidenced better results for KF than for HF. RMSE scores were reduced by 43.8% (33.5% HF), 25.2% (13.2% HF) and 41.6% (39.7% HF) respectively for hourly averaged NO 2 , hourly averaged O 3 and daily averaged PM10 concentrations. A further analysis performed clustering the monitoring stations per type showed a good performance of the AQFS for ozone for all the groups of stations ( r  = 0.7), while satisfactory results were obtained for PM 10 and NO 2 at rural background ( r  = 0.6) and Rome background stations ( r  = 0.7). The skill scores confirmed the capability of the adopted techniques to improve the reproduction of exceedance events.