A novel hybrid clustering model of region segmentation to fuse CMAQ simulations with observations

Abstract

The Community Multiscale Air Quality (CMAQ) model can predict air pollutant concentrations at every pre-determined location but usually suffers from considerable systematic bias. By post-process the CAMQ simulations with ground observations, K-means-cluster-Observation-examiner-Boundary-denoiser (KOB) model was developed as an unsupervised hybrid model, to segment regions by capturing spatial dependencies of CMAQ time-series and systematically reduce the bias of CMAQ simulations. The KOB model was then applied in 2019 to improve the CMAQ predictions of respirable suspended particulates (RSP) concentrations in the Pearl River Delta (PRD) region of southern China. Compared to the CMAQ simulations, the KOB model is implemented with hourly CMAQ simulations and significantly improves the 1-h predictions of RSP at all stations, with the index of agreement (IOA) increasing from 0.70 to 0.86 and average root mean square error (RMSE) reducing from 28.4 to 20.0 μg/m3. A better agreement was also achieved between the monthly average RSP from predictions using the KOB model and ground observations, with IOA increased from 0.87 to 0.93 and RMSE decreased from 12.4 to 8.2 μg/m3. In addition, the capability of the KOB model to predict extreme pollution was greatly improved, with the RMSE reduced from 76.3 to 50.8 μg/m3. Furthermore, the improvement was identified not only for the next 1-h predictions, but also for the predictions in the next 6 h and after 6 h the SBC method outperforms the KOB model. Through data fusion, our hybrid clustering model can achieve robust real-time air quality forecasting.