A new approach to more accurately monitor and evaluate transboundary particulate matter (PM) pollution is introduced based on aerosol optical products from Korea's Geostationary Ocean Color Imager (GOCI). The area studied is Northeast Asia (including eastern parts of China, the Korean peninsula and Japan), where GOCI has been monitoring since June 2010. The hourly multi-spectral aerosol optical data that were retrieved from GOCI sensor onboard geostationary satellite COMS (Communication, Ocean, and Meteorology Satellite) through the Yonsei aerosol retrieval algorithm were first presented and used in this study. The GOCI-retrieved aerosol optical data are integrated with estimated aerosol distributions from US EPA Models-3/CMAQ (Community Multi-scale Air Quality) v4.5.1 model simulations via data assimilation technique, thereby making the aerosol data spatially continuous and available even for cloud contamination cells. The assimilated aerosol optical data are utilized to provide quantitative estimates of transboundary PM pollution from China to the Korean peninsula and Japan. For the period of 1 April to 31 May, 2011 this analysis yields estimates that AOD as a proxy for PM_2.5 or PM₁₀ during long-range transport events increased by 117–265% compared to background average AOD (aerosol optical depth) at the four AERONET sites in Korea, and average AOD increases of 121% were found when averaged over the entire Korean peninsula. This paper demonstrates that the use of multi-spectral AOD retrievals from geostationary satellites can improve estimates of transboundary PM pollution. Such data will become more widely available later this decade when new sensors such as the GEMS (Geostationary Environment Monitoring Spectrometer) and GOCI-2 are scheduled to be launched.

A deep recurrent neural network system based on a long short-term memory (LSTM) model was developed for daily PM₁₀ and PM_2.5 predictions in South Korea. The structural and learnable parameters of the newly developed system were optimized from iterative model training. Independent variables were obtained from ground-based observations over 2.3 years. The performance of the particulate matter (PM) prediction LSTM was then evaluated by comparisons with ground PM observations and with the PM concentrations predicted from two sets of 3-D chemistry-transport model (CTM) simulations (with and without data assimilation for initial conditions). The comparisons showed, in general, better performance with the LSTM than with the 3-D CTM simulations. For example, in terms of IOAs (index of agreements), the PM prediction IOAs were enhanced from 0.36–0.78 with the 3-D CTM simulations to 0.62–0.79 with the LSTM-based model. The deep LSTM-based PM prediction system developed at observation sites is expected to be further integrated with 3-D CTM-based prediction systems in the future. In addition to this, further possible applications of the deep LSTM-based system are discussed, together with some limitations of the current system.

In this paper, we propose a new temperature prediction model based on deep learning by using real observed weather data. To this end, a huge amount of model training data is needed, but these data should not be defective. However, there is a limitation in collecting weather data since it is not possible to measure data that have been missed. Thus, the collected data are apt to be incomplete, with random or extended gaps. Therefore, the proposed temperature prediction model is used to refine missing data in order to restore missed weather data. In addition, since temperature is seasonal, the proposed model utilizes a long short-term memory (LSTM) neural network, which is a kind of recurrent neural network known to be suitable for time-series data modeling. Furthermore, different configurations of LSTMs are investigated so that the proposed LSTM-based model can reflect the time-series traits of the temperature data. In particular, when a part of the data is detected as missing, it is restored by using the proposed model’s refinement function. After all the missing data are refined, the LSTM-based model is retrained using the refined data. Finally, the proposed LSTM-based temperature prediction model can predict the temperature through three time steps: 6, 12, and 24 h. Furthermore, the model is extended to predict 7 and 14 day future temperatures. The performance of the proposed model is measured by its root-mean-squared error (RMSE) and compared with the RMSEs of a feedforward deep neural network, a conventional LSTM neural network without any refinement function, and a mathematical model currently used by the meteorological office in Korea. Consequently, it is shown that the proposed LSTM-based model employing LSTM-refinement achieves the lowest RMSEs for 6, 12, and 24 h temperature prediction as well as for 7 and 14 day temperature prediction, compared to other DNN-based and LSTM-based models with either no refinement or linear interpolation. Moreover, the prediction accuracy of the proposed model is higher than that of the Unified Model (UM) Local Data Assimilation and Prediction System (LDAPS) for 24 h temperature predictions.

To more accurately estimate direct radiative forcing (DRF) by aerosols, and better investigate particulate pollution over East Asia, precise calculations of the optical properties of aerosols, such as aerosol optical depth (AOD), single scattering albedo (SSA) and aerosol extinction coefficient (σ_ext), are of primary importance. The aerosol optical properties over East Asia were investigated in this study, based on US EPA Models-3/CMAQ v4.5.1 model simulations. The CMAQ model simulations in this study were improved in several ways compared to those in a previous study (Song et al., 2008). Although the details of the improvements were described in the manuscript, the following points should be emphasized: (1) two data assimilation techniques were employed for producing more accurate AOD products and meteorological fields over East Asia; (2) updated/upgraded emission inventories were used in the CMAQ model simulations with a fine grid resolution of 30 × 30 km²; and (3) the 4-D particulate composition calculated from the CMAQ model simulations was converted into 3-D or 4-D aerosol optical products, using the Malm and Hand (2007) algorithm with significant further modifications. The results from the CMAQ model simulations (without assimilation) showed great improvements compared to those from a previous study. For example, in terms of the regression coefficients (R), R values were increased from 0.48–0.68 (previous study) to 0.62–0.79 (this study). The monthly-averaged CMAQ-simulated single scattering albedo (SSA) also agreed well with the AERONET SSA, with the exceptions of the Hong Kong and Taipei sites, where the air qualities were strongly influenced by active biomass burning events from January to April. There were also excellent matches between the vertical profiles of the CMAQ-simulated σ_ext and LIDAR-retrieved σ_ext. It was also found that the contributions of (NH₄)₂SO₄ during summer, NH₄NO₃ during winter, sea-salt particles during winter and dust particles during spring to the total AOD were large over East Asia. In particular, the largest contribution of NH₄NO₃ to the total AOD was found over East Asia during winter. Therefore, it was suggested that this contribution of NH₄NO₃ should not be neglected. In order to produce more accurate AOD products, the CMAQ-simulated AODs were further assimilated with the MODIS-retrieved AODs. Both of the assimilated and AERONET AODs were better correlated with each other than the CMAQ-simulated and AERONET AODs. The obvious benefits from this study would be that with these improved aerosol optical properties, the particulate pollution (e.g. AOD can be served as a proxy to PM_2.5 or PM₁₀) and DRF by aerosols over East Asia can be more satisfactorily investigated in future.