To improve the predictability of concentrations of atmospheric particulate matter, a data assimilation (DA) system using ensemble square root filter (EnSRF) has been developed for the Community Multiscale Air Quality (CMAQ) model. The EnSRF DA method is a deterministic variant of the ensemble Kalman filter (EnKF) method, which means that unlike the EnKF method, it does not add random noise to the observations. To compare the performances of the EnSRF with those of other DA methods, such as EnKF and 3DVAR (three-dimensional variational), these three methods were applied to the same CMAQ model simulations with identical experimental settings. This is the first attempt in the field of chemical DA to compare the EnKF and EnSRF methods. An identical set of surface fine particulate matter (PM2.5) were assimilated every 6 h by all the DA methods over a CMAQ domain of East Asia, during the period from 01 May to 11 June 2016. In parallel with 'reanalysis experiments', we also carried out '48 h prediction experiments' using the optimized initial conditions produced by the three DA methods. Detailed analyses among the three DA methods were then carried out by comparing both the reanalysis and the prediction outputs with the observed surface PM2.5 over four regions (i.e., South Korea, the Beijing-Tianjin-Hebei (BTH) region, Shandong province, and Liaoning province). The comparison results revealed that the EnSRF produced the best reanalysis and prediction fields in terms of several statistical metrics. For example, when the 3DVAR, EnKF, and EnSRF methods were used, averaged normalized mean biases (NMBs) decreased by (57.6, 85.6, and 91.8) % in reanalyses and (39.7, 87.6, and 91.5) % in first-day predictions, compared to the CMAQ control experiment (i.e., without DA) over South Korea, respectively. Also, over the three Chinese regions, the EnSRF method outperformed the EnKF and 3DVAR methods.

Data assimilation (DA) combines incomplete background values obtained via chemical transport model predictions with observational information. Several 3-Dimensional variational (3DVAR) and sequential methods (e.g., ensemble Kalman filter (EnKF)) are used to define model errors and build a background error covariance (BEC) and are important factors affecting the prediction performance of DA. The BEC determines the spatial range, where observation concentration is reflected in the model when DA is applied to an air pollution transport model. However, studies investigating the characteristics of BEC using air quality models remain lacking. In this study, horizontal length scale (HLS) and vertical length scale (VLS) analyses of a BEC were applied to EnKF and ensemble square root filter (EnSRF), respectively, and two ensemble-based DA methods were performed; the characteristics were compared with those of a BEC applied to 3DVAR. The results of 6 h PM2.5 predictions performed for 42 days were evaluated for a control run without DA (CTR), 3DVAR, EnKF, and EnSRF. HLS and VLS respectively exhibited a high correlation with the ground wind speed and with the planetary boundary layer height for diurnal and daily variations; EnKF and EnSRF exhibited superior performances among all the methods. The root mean square errors were 11.9 μg m−3 and 11.7 μg m−3 for EnKF and EnSRF, respectively, while those for 3DVAR and CTR were 12.6 μg m−3 and 18.3 μg m−3, respectively. Thus, we proposed a simple method to find a Gaussian function that best described the error correlation of the BEC based on the physical distance.

In this study, we developed a data assimilation (DA) system for chemical transport model (CTM) simulations using an ensemble Kalman filter (EnKF) technique. This DA technique is easy to implement in an existing system without seriously modifying the original CTM and can provide flow-dependent corrections based on error covariance by short-term ensemble propagations. First, the PM2.5 observations at ground stations were assimilated in this DA system every 6 h over South Korea for the period of the KORUS–AQ campaign from 1 May to 12 June 2016. The DA performances with the EnKF were then compared to a control run (CTR) without DA and a run with three-dimensional variational (3D-Var) DA. Consistent improvements owing to the initial conditions (ICs) assimilated with the EnKF were found in the DA experiments at a 6 h interval compared to the CTR run and to the run with 3D-Var. In addition, we attempted to assimilate the ground observations from China to examine the impacts of improved boundary conditions (BCs) on the PM2.5 predictability over South Korea. The contributions of the ICs and BCs to improvements in the PM2.5 predictability were also quantified. For example, the relative reductions in terms of the normalized mean bias (NMB) were found to be approximately 27.2 % for the 6 h reanalysis run. A series of 24 h PM2.5 predictions were additionally conducted each day at 00:00 UTC with the optimized ICs. The relative reduction of the NMB was 17.3 % for the 24 h prediction run when the updated ICs were applied at 00:00 UTC. This means that after the application of the updated BCs, an additional 9.0 % reduction in the NMB was achieved for 24 h PM2.5 predictions in South Korea.