KWKyung-Hui Wang

Kyung-Hui Wang

*This is an algorithmically generated author record based on publicly available information.

Kyung-Hui Wang has a total of 26 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Sensitivity of Transformer-Based PM2.5 Forecasting to Input Sequence Length

    by Kwon Jang, Seung-Hee Han, Kyung-Hui Wang, Hui-Young Yun

    Transformer-based time series models are already widely used in PM2.5 prediction studies due to their ability to learn long-term dependencies. However, despite input sequence length being a key design factor governing prediction performance, systematic evaluations that independently control this factor and examine how its effects vary across forecast lead times and atmospheric conditions remain limited. In particular, quantitative evidence is lacking on whether extending the input sequence length consistently improves long-term forecasting skill, or whether, under certain conditions, excessive historical information can instead degrade forecast stability.In this study, Seoul—characterized by frequent high-pollution episodes and pronounced seasonal variability—is selected as a case study region. Using hourly observational data from 2018 to 2024, we quantitatively analyze the effects of input sequence length (3, 7, and 15 days) on Transformer-based PM2.5 prediction performance. Vanilla Transformer, Informer, and Autoformer models are evaluated under identical data partitioning, preprocessing, input variable configuration, training strategies, and output structures, allowing the effects of input sequence length to be isolated from other modeling factors. Prediction performance is assessed for short-term (24 h) and long-term (72 h) forecast horizons using MAE and RMSE, enabling joint analysis of error reduction, error accumulation, and forecast stability as input sequence length increases.The results show that extending the input sequence length from 3 to 7 days leads to reduced short-term prediction errors and improved stability in long-term forecasts across all models. However, further extension to 15 days yields diminishing returns and, in some cases, increased errors for long-term forecasts. In particular, differences in MAE associated with input sequence length reach up to approximately 10–15% for 72 h predictions, indicating that longer input sequences can introduce not only useful long-term dependencies but also redundant or irrelevant historical patterns. Seasonal analyses further reveal that sensitivity to input sequence length is amplified during wintertime conditions with frequent high-pollution events, suggesting that the utilization of accumulated historical information plays a more critical role under stagnant atmospheric regimes.This study demonstrates that longer input sequences are not universally optimal across all forecast horizons and highlights the need to tailor input sequence length according to forecast lead time and environmental context, even within the same Transformer architecture. By reframing input sequence length as a purpose-driven design parameter rather than a fixed hyperparameter, this work provides empirical guidance for the development and application of Transformer-based PM2.5 forecasting models.Acknowledgments"This research was supported by Particulate Matter Management Specialized Graduate Program through the Korea Environmental Industry & Technology Institute(KEITI) funded by the Ministry of Environment(MOE)“

  • Open Access

    Assessing the added value of high-resolution PM2.5 mapping for population exposure estimates in Seoul, South Korea

    by Kyung-Hui Wang, Seung-Hee Han, Kwon Jang, Hui-Young Yun

    Accurate representation of fine particulate matter (PM2.5) at urban scales remains a major challenge in air pollution modelling, as conventional chemical transport models (CTMs) are typically operated at kilometer-scale resolution, which may smooth out strong local concentration gradients. Such limitations can influence not only the interpretation of model results but also the application of model-derived concentration fields to exposure-related analyses.In this study, we assess the added value of high-resolution PM2.5 concentration fields for urban-scale air pollution modelling applications by comparing population-weighted exposure (PWE) estimates derived from multiple PM₂.₅ datasets. Daily PM2.5 fields from CMAQ simulations at 9 km resolution, observation-based gridded PM2.5, and a deep-learning-based super-resolution PM2.5 product at 100 m resolution were harmonized on a common analysis grid over the Seoul metropolitan area. These concentration fields were combined with 100 m gridded population data to calculate district-level PWEs for 25 administrative districts.The results show that CMAQ reproduces the broad spatial patterns of PM2.5 across Seoul, while the high-resolution PM2.5 product reveals localized variability that is not captured at coarser resolution, particularly in densely populated districts. Comparison with observation-based PWEs indicates that exposure estimates derived from the high-resolution PM2.5 fields are often closer to observations than those based on the original CMAQ outputs, although the magnitude of improvement varies by district. Population distribution maps further highlight that spatial heterogeneity in population density plays a key role in shaping district-level exposure patterns.Overall, this study demonstrates that enhancing the spatial representation of PM2.5 concentration fields can provide additional insight when air pollution model outputs are applied to population exposure assessment in urban environments. The proposed framework illustrates a practical approach for evaluating the application-level benefits of high-resolution air quality products within air pollution modelling studies. AcknowledgementThis research was supported by the Korea National Institute of Health (KNIH) research project(Project No. 2024-ER0606-01) and the Particulate Matter Management Specialized Graduate Program through the Korea Environmental Industry & Technology Institute (KEITI), funded by the Ministry of Environment (MOE).

  • Open Access

    Comparison of Models for Missing Data Imputation in Environmental Data: A Case Study of PM-2.5 in Seoul

    by Ju-Yong Lee, Seung-Hee Han, Kwon Jang, Kyung-Hui Wang, Hui-Young Yun, Dae-Ryun Choi

    PM-2.5 is a critical pollutant for air quality evaluation and public health policymaking, necessitating accurate data for reliable analysis. However, environmental data often contain missing values due to equipment malfunctions or extreme weather conditions, which undermine the credibility of analysis and predictions. In particular, the frequent fluctuations of PM-2.5 levels in Seoul highlight the importance of addressing missing data issues.This study systematically compares the performance of various missing data imputation methods for PM-2.5 data in Seoul, aiming to identify the optimal approach for medium- and long-term predictions. By generating and evaluating missing data during high- and low-concentration periods, this research differentiates itself from prior studies and enhances practical applicability.A range of statistical and machine learning-based methods, including FFILL, KNN, MICE, SARIMAX, DNN, and LSTM, were applied to impute missing data. The performance of each method was evaluated over 6-hour, 12-hour, and 24-hour intervals using metrics such as RMSE, MAE, and correlation coefficients. The experimental design incorporated real-world air quality conditions by selecting data from periods of significant PM-2.5 variation.KNN demonstrated balanced performance across all time intervals and yielded the best results for medium- and long-term predictions. FFILL showed excellent accuracy over short time intervals but exhibited declining performance as the interval length increased. Conversely, deep learning-based models, such as DNN and LSTM, showed relatively poor performance, indicating the need for further optimization to account for the characteristics of time-series data.This study confirms that KNN is the most suitable method for PM-2.5 missing data imputation due to its simplicity and computational efficiency. These findings enhance the reliability of air quality data analysis and provide a valuable foundation for effective air quality management and policymaking. Furthermore, the results underscore the importance of selecting appropriate imputation methods to improve predictive accuracy and analytical reliability. "This research was supported by Particulate Matter Management Specialized Graduate Program through the Korea Environmental Industry & Technology Institute(KEITI) funded by the Ministry of Environment(MOE)“


  • Open Access

    Application of a Super-Resolution Algorithm to Improve the Spatial Resolution of Air Pollutant Concentrations in the Seoul Area

    by Kyung-Hui Wang, Min-Woo Jung, Seung-Hee Han, Ju-Yong Lee, Kwon Jang, Dae-Ryun Choi, Hui-Young Yun

    Air pollution not only poses harmful effects on human health but also causes various diseases, leading to severe issues such as increased premature mortality. To accurately assess the health impacts and exposure levels of air pollution, high-resolution spatiotemporal concentration data is essential.In previous studies, Hybrid Modeling combining CMAQ and CALPUFF was applied to estimate air pollutant concentrations at a spatial resolution of 100m. However, the Hybrid Model has limitations in that each modeling process must be conducted independently, requiring significant time and computational resources.This study aims to improve computational efficiency and simplify the modeling process by applying a Super-Resolution Convolutional Neural Network (SRCNN) algorithm. SRCNN uses low-resolution (9km) CMAQ data as input to produce spatial distributions similar to those generated by the Hybrid Model at a high resolution of 100m. The target pollutant is PM2.5 and NO2 in Seoul, with a training period from 2015 to 2021 and a test period in 2022. Model evaluation results show that SRCNN outperformed CMAQ in terms of PSNR, SSIM, and Spatial RMSE metrics. This demonstrates the potential of SRCNN to efficiently generate high-resolution air pollution concentration data, contributing to more precise exposure assessments and health impact analyses. AcknowledgementThis research was supported by the Korea National Institute of Health (KNIH) research project (Project No.2024-ER0606-00) and Particulate Matter Management Specialized Graduate Program through the Korea Environmental Industry & Technology Institute (KEITI) funded by the Ministry of Environment (MOE)

  • Open Access

    -Meter High-Resolution Modeling and Validation of PM and NO2 Concentrations in Urban Areas of South Korea

    by Hui-Young Yun, Kyung-Hui Wang, Min-Woo Jung, Seung-Hee Han, Ju-Yong Lee, Kwon Jang, Dae-Ryun Choi

    Fine particulate matter (PM) and nitrogen dioxide (NO2) are major air pollutants that significantly contribute to health risks, including cardiovascular and respiratory diseases. This study develops and validates 100-meter resolution air pollution data for South Korean cities, focusing on PM and NO2 concentrations. A hybrid modeling approach combining the Chemical Transport Model (CMAQ) and the Dispersion Model (CALPUFF) was employed to estimate the spatiotemporal distribution of these pollutants in major metropolitan areas, including Seoul, Busan, and Incheon. The CMAQ model generated baseline data at typical resolutions of 9 km and 1 km grids, which were further refined using the CALPUFF model to produce high-resolution 100 m datasets. The hybrid modeling approach integrated primary pollutant concentrations from CMAQ with CALPUFF's precise dispersion modeling to accurately reflect localized pollutant variations critical for urban health assessments. The resulting 100 m resolution data were validated by comparing them with roadside air quality monitoring measurements, demonstrating high correlation and ensuring temporal and spatial reliability. This study overcomes the limitations of traditional 1 km and 9 km resolution datasets and presents a novel approach for analyzing fine-scale pollutant distributions in urban environments. The methodology is applicable to other regions globally, particularly those facing severe air pollution challenges, and serves as a foundational tool for urban air quality improvement efforts. The generated data will facilitate research on the relationship between air pollution exposure and health outcomes and support the development of targeted air quality management policies. Future work will focus on integrating real-time air quality monitoring data to improve model accuracy and support the implementation of evidence-based air quality management policies. Acknowledgement This research was supported by the Korea National Institute of Health (KNIH) research project (Project No. 2024-ER0606-00) and the Particulate Matter Management Specialized Graduate Program through the Korea Environmental Industry & Technology Institute (KEITI), funded by the Ministry of Environment (MOE).

  • Open Access

    Comparison of LSTM and Bi-LSTM Models for Predicting PM2.5 Concentration in Seoul

    by Kwon Jang, Ju-Yong Lee, Seung-Hee Han, Kyung-Hui Wang, Hui-Young Yun

    The importance of predicting fine particulate matter (PM2.5) concentrations has grown significantly due to deteriorating urban air quality. Seoul presents unique modeling challenges due to its intensive industrial activities, high population density, significant data variability between monitoring stations, and numerous missing values. This study analyzes and compares the predictive performance of LSTM (Long Short-Term Memory) and Bi-LSTM (Bidirectional LSTM) models using hourly PM2.5 concentration data collected from 25 monitoring stations in Seoul from 2018 to 2022.In the data preprocessing phase, we employed the MICE (Multiple Imputation by Chained Equations) method to handle missing values, which effectively preserved the data's structural characteristics by considering inter-variable relationships. The predictive performance of both models was evaluated using metrics such as RMSE (Root Mean Square Error) and MAE (Mean Absolute Error). While LSTM focuses on forward learning, Bi-LSTM can capture complex time series patterns by utilizing both forward and backward information.The results demonstrated that Bi-LSTM effectively captured complex time series patterns through its bidirectional learning structure, and optimal learning rates and dropout ratios were determined through various parameter tuning experiments. These findings present the characteristics and potential applications of both models in PM2.5 concentration prediction, expected to contribute to improved air quality forecasting systems and policy decision support.Future research will focus on enhancing model accuracy through the implementation of additional algorithms and exploring applications for PM2.5 prediction in other cities. This aims to increase the spatial resolution of air pollution prediction and contribute to broader air quality improvements.Acknowledgement "This research was supported by Particulate Matter Management Specialized Graduate Program through the Korea Environmental Industry & Technology Institute(KEITI) funded by the Ministry of Environment(MOE)"

  • Open Access

    A Transfer Learning-Based Model for PM2.5 Prediction in Seoul

    by Seung-Hee Han, Ju-Yong Lee, Kwon Jang, Kyung-Hui Wang, Hui-Young Yun

    Predicting PM2.5 concentrations using air quality data is often hindered by the presence of missing values, which can compromise accuracy and reliability. Traditional prediction models frequently suffer significant data loss during the handling of missing values, necessitating new approaches that address data quality issues while improving prediction performance.This study proposes an efficient prediction methodology leveraging transfer learning to minimize the impact of missing values. A pre-trained model was constructed using integrated data from all monitoring stations in Seoul, followed by fine-tuning for specific monitoring stations to develop a PM2.5 prediction model. Transfer learning is a machine learning technique that utilizes knowledge from previously trained models to enhance learning efficiency and performance in new tasks or domains. Unlike traditional approaches that require training from scratch, transfer learning reuses the weights and structure of pre-trained models, reducing training time and improving performance.In this study, a deep neural network (DNN) pre-trained model was built using data from all monitoring stations in Seoul, and fine-tuning was applied using specific station data. The model was trained on six-hour average PM2.5 data to predict the next six hours, effectively addressing missing values.Preliminary results indicate that the transfer learning-based model effectively handles missing values and demonstrates improved prediction accuracy compared to independently modeled traditional approaches. By leveraging domain-wide information, the model compensates for the limitations of individual monitoring station data, achieving higher accuracy and reliability.This study provides a scalable solution for addressing data gaps in air quality prediction and contributes to research on the health impacts of air pollution and urban air quality management. Future research will explore the application of this methodology to other pollutants and regions, further enhancing its generalizability and effectiveness.Acknowledgement"This research was supported by Particulate Matter Management Specialized Graduate Program through the Korea Environmental Industry & Technology Institute(KEITI) funded by the Ministry of Environment(MOE)"

  • Development of a Conv-LSTM-Based AI Model for Forecasting PM2.5 Spatial Concentration Distribution

    by Min-Woo Jung, Ju-Yong Lee, Kyung-Hui Wang, Chae-Yeon Lee, Seung-Hee Han, Hee-Jin Kim, Seung-Min Sohn, Peel-Soo Jeong, Dae-Ryun Choi, Hui-Young Yun


  • Open Access

    The Analysis of Changes in Particulate Matter Concentrations in East Asia before and after COVID-19

    by Eun-Seong Son, HyeunSoo Kim, Peel-Soo Jeong, Kyung-Hui Wang, Seung-Hee Han, Hui-Young Yun

    The COVID-19 pandemic commenced at the end of 2019, inducing a substantial decline in human activities across various sectors, encompassing industry, economy, tourism, and daily life globally. This resulted in a reduction in anthropogenic pollutant emissions.The concentration of fine particulate matter in the atmosphere incorporates influences from both artificial and natural emission sources. East Asia is also notable for experiencing frequent occurrences of yellow dust.Nonetheless, studies conducted by Bae et al. (2023) and Zhang et al. (2020) reported that PM10 concentrations in Korea and China decreased by more than 30% following the emergence of COVID-19. This phenomenon has been commonly observed worldwide.This study delves into whether air quality is deteriorating again now that the pandemic has subsided. It examines long-term changes in air quality in East Asia, covering the Pre-COVID-19 and Post-COVID-19 periods, utilizing measurement data. To enhance reliability, the data were obtained from a nationally operated atmospheric measurement station. In addition to PM10, changes in the concentration of gaseous pollutants such as NO2 and SO2 were also analyzed.In instances where changes in air quality occurred, an analysis was conducted to determine whether these changes correlated with human activities. The study explored whether yearly and seasonal weather changes acted as variables and whether air quality fluctuations in urban and rural areas exhibited disparities. Statistical methods such as regression analysis were employed.This study analyzes the impact of human activities on the concentration of fine particulate matter in East Asia and offers implications for the direction of national policies aimed at improving air quality.Reference:Minah Bae, Yoon-Hee Kang, Eunhye Kim, Segi Kim, Soontae Kim (2023), A multifaceted approach to explain short- and long-term PM2.5 concentration changes in Northeast Asia in the month of January during 2016–2021, Science of the Total Environment, Volume 880, https://doi.org/10.1016/j.scitotenv.2023.163309Zhang, J., Cu i, K., Wang, Y.F., Wu , J.L., Hu ang, W.S., Wan, S., and Xu, K. (2020), Temporal variations in the air quality index and the impact of the COVID-19 event on air quality in western China, erosol and Air Quality Research, Vol. 20, No. 7, pp. 1552-1568. https://doi.org/10.4209/aaqr.2020.06.0297Acknowledgments:This research was supported by Particulate Matter Management Specialized Graduate Program through the Korea Environmental Industry & Technology Institute (KEITI) funded by the Ministry of Environment (MOE)