Comparison between the Non-Adaptive and Adaptive Bias Correction on Hourly PM Forecasts of the Deterministic Model WRF-CAMx

Abstract

This Final Year Project mainly focuses on the comparison between two bias correction approaches carried out on forecasts of Particulate Matters (PM) within Pearl River Delta Region that were simulated by the deterministic air quality forecasting model WRF-CAMx. The two approaches, namely non-adaptive (or offline) bias correction and adaptive (or online) bias correction, were constructed under the ordinary least squares principle and the Kalman filter algorithm, respectively, as implemented by MATLAB. The obtained corrected data were then fitted into line charts and scatter plots for graphical comparison; also, they were processed and shown as indicators for numerical evaluations of the correction model’s performance. The results verified the efficacy of both correction methods, while highlighted better performances of the adaptive (online) correction model, which suggests that the WRF-CAMx system may be time-varying. Data also shows that WRF-CAMx tends underestimate the PM concentration within the study period (January, 2014).