Predicted PM10 Concentrations Research Articles

Интегрирование химической транспортной модели и искусственной нейронной сети для прогноза концентрации PM10

A hybrid model for forecasting PM10 surface concentrations in the Moscow region, consisting of a chemical transport model (CTM) and an artificial neural network (ANN), with the use of PM10 automatic measurements data was developed and tested. The ANN was trained to predict PM10 concentrations based on the forecasts of CTM concentrations and meteorological parameters on a 2 km grid. The results of testing the ANN based on independent samples, including the episodes of high PM10 pollution, are discussed.

Modeling of PM10 Particulate Matter in Ahvaz City Using Remote Sensing and Meteorological Parameters

Introduction: In recent years, remote sensing (RS) products have emerged as effective tools for monitoring air pollution. This study aims to predict the concentrations of particulate matter with a diameter smaller than 10μm (PM10) using a multivariate linear regression (MLR) model, incorporating both Aerosol Optical Depth (AOD) products and meteorological parameters. Material and Methods: In this study, data on PM10 concentrations, Aerosol Optical Depth (AOD), and meteorological parameters (wind speed, temperature, humidity, and horizontal visibility) were used. The study focused on the time 15:00 each day, as this time was identified as having significant data relevance. The methodology section also consisted of three steps: 1) pairwise correlation analysis: The relationship between meteorological parameters, AOD, and PM10 was assessed using the pairwise correlation method. 2) Model development: A MLR model was developed to predict PM10 concentrations. 3) Validation: The model was validated using a separate dataset, ensuring that 70% of the data was used for training, and 30% for testing and validation. Results: The pairwise correlation analysis revealed a strong correlation (0.86) between AOD remote sensing index and PM10. The highest correlation (0.9) was observed during the spring season. The five developed equations to estimate the PM10 index yielded correlation coefficients ranging from 0.86 to 0.90. Notably, the highest correlation was achieved when AOD data and all the meteorological parameters were u tilized simultaneously. These results highlighted the utility of remote sensing products and meteorological data in air quality monitoring and prediction. Conclusion: This study demonstrates that a MLR model incorporating AOD and meteorological parameters can effectively predict PM10 concentrations in Ahvaz City, particularly during dust storms in hot seasons. These findings can aid policymakers and public health officials in developing strategies to mitigate the adverse effects of dust storms on air quality and public health.

Changes in urban residential and agricultural sources induce a decrease in PM10 levels in cold clean area: A thirteen-year monitoring at the Longfengshan WMO/GAW regional background station in Northeast China

Atmospheric background monitoring provides insights into the long-term changes in atmospheric composition resulting from human activities on both global and regional scales. It is important to understand the local characteristics of key air pollutants to evaluate and manage air quality control policies effectively. This study analyzed 13 years of continuous data of PM10 (particulate matter with aerodynamic diameter ≤10 μm) concentration from the only atmospheric background station in Northeastern China, the Longfengshan (LFS) World Meteorological Organization (WMO)/Global Atmosphere Watch (GAW) regional station. The results revealed a significant decrease of 58.02% in PM10 concentrations in the LFS from 2007 to 2019, which was primarily influenced by transport from Changchun and Harbin. Over periods exceeding eight months, the correlation between PM10 levels at the LFS atmospheric background station and those in Changchun and Harbin became more pronounced. The observed decline in PM10 concentrations at the LFS atmospheric background station was largely attributed to reduced emissions from urban residential and agricultural sources. Emissions of carbon monoxide and ammonia from these sources, along with temperature and relative humidity (RH), significantly affected the PM10 concentrations. Using machine learning (ML) methods and PM10 data from the LFS atmospheric background station, this study effectively predicted PM10 concentrations in neighboring cities, demonstrating that the LFS atmospheric background station can effectively assess the trend of PM10 changes in Northeast China. This study quantified the influence of urban areas on atmospheric background stations, highlighting the effectiveness of existing environmental policies, and offering a reference for similar studies and forecasts in other regions.

Assessment of human and meteorological influences on PM10 concentrations: Insights from machine learning algorithms

Air pollution prediction assists residents in planning activities and avoiding regions with elevated pollution levels. This study employed Random Forest (RF), Extreme Gradient Boosting (XGB), and Histogram-Based Gradient Boosting (HistGB) based models to evaluate PM10 under 10-year average meteorological conditions (2012–2021) for Rocklea monitoring station in Queensland, Australia. The meteorological normalisation method adjusted air pollution data was assessed by removing weather-related effects to focus on the impact of human activities. By comparing observed and meteorologically normalised PM10 values, the impact of weather and human activities was assessed. The performance of these models was evaluated through k-Fold cross-validation, with a particular emphasis on their ability to predict PM10 concentrations accurately. Contrary to initial expectations, the RF model exhibited superior performance compared to XGB and HistGB, demonstrating the highest accuracy and consistency in predicting PM10 levels. This result underscores the effectiveness of the RF model in environmental air quality prediction. A key insight from our analysis was the role of weather-related factors in influencing PM10 levels. While the observed PM10 values showed a general decreasing trend from 2012 to 2021, except for an anomaly in 2019, the meteorologically normalised values remained relatively stable. This finding suggests that weather variations, rather than human activities, have been the more significant influence on the observed trend in PM10 levels over the past decade. The consistent meteorologically normalised values further reinforce the importance of considering natural meteorological variability in air quality studies.

A Novel Hybrid Model Combining the Support Vector Machine (SVM) and Boosted Regression Trees (BRT) Technique in Predicting PM10 Concentration

The PM10 concentration is subject to significant changes brought on by both gaseous and meteorological variables. The aim of this research was to explore the performance of a hybrid model combining the support vector machine (SVM) and the boosted regression trees (BRT) technique in predicting the PM10 concentration for 3 consecutive days. The BRT model was trained by utilizing maximum daily data in the cities of Alor Setar, Klang, and Kuching from the years 2002 to 2017. The SVM–BRT model can optimize the number of predictors and predict PM10 concentration; it was shown to be capable of predicting air pollution based on the models’ performance with NAE (0.15–0.33), RMSE (10.46–32.60), R2 (0.33–0.70), IA (0.59–0.91), and PA (0.50–0.84). This was accomplished while saving training time by reducing the feature size given in the data representation and preventing learning from noise (overfitting) to improve accuracy. This knowledge establishes the foundation for the development of efficient methods to prevent and/or minimize the health effects of PM10 exposure on one’s health.

Evaluating the Performance of Random Forest and Multiple Linear Regression for Higher Observed PM10 Concentrations

Background: Air pollution is notable for its direct impact on human health. Hence, the ability to accurately predict air pollution concentrations is vital to raising public awareness of this issue and for better understanding of air quality management. Aim: Therefore, the aim of this research is to predict PM10 concentrations in Malaysia, specifically on Langkawi Island using random forest and multiple linear regression. Method: The predictive analytics were based on air pollution hourly data from 2003 until 2017. The eight parameters chosen in this study were PM10, NO2, O3, CO, SO2, Relative Humidity (RH), Temperature (T), and Wind Speed (WS). The findings revealed that PM10, SO2, NO2, CO, and O3 hourly trends at Langkawi Island were below the recommended Malaysian Ambient Air Quality Guidelines (MAAQG) standard. Multiple linear regression (MLR) and random forest (RF) were used for modelling and compared based on their prediction accuracy. Result: The values of RMSE, NAE, IA, PA and R2 for MLR were 8.0698, 0.1368, 0.8584, 0.7737 and 0.5984 respectively while the values of RMSE, NAE, IA, PA and R2 for RF were 6.674038, 0.107664, 0.911974, 0.852570 and 0.726681 correspondingly. From the results, the RF method was chosen as a better model than MLR since both; the error measures and the accuracy measures results are close to 1. Nevertheless, the PM10 models (RF and MLR) are unable to take into account the higher observed concentrations.

Improving Air Pollution Prediction Modelling Using Wrapper Feature Selection

Feature selection is considered as one of the essential steps in data pre-processing. However, all of the previous studies on predicting PM10 concentration in Malaysia have been limited to statistical method feature selection, and none of these studies used machine-learning approaches. Therefore, the objective of this research is to investigate the influence variables of the PM10 prediction model by using wrapper feature selection to compare the prediction model performance of different wrapper feature selection and to predict the concentration of PM10 for the next day. This research uses 10 years of daily data on pollutant concentrations from two stations (Klang and Shah Alam) obtained from the Department of Environment Malaysia (DOE) from 2009 until 2018. Six wrapper methods (forward selection, backward elimination, stepwise, brute-force, weight-guided and genetic algorithm evolution and the predictive analytics multiple linear regression (MLR) and artificial neural network (ANN)) were implemented in this study. This study found that brute-force is the dominant wrapper method in most of the best models in selecting important features for MLR. Moreover, compared to MLR, ANN provides more advantages regarding model accuracy and permits feature selection in predicting PM10. The overall results revealed that the RMSE value for next day prediction in Klang is 20.728, while the AE value is 15.69. Furthermore, the RMSE value for next day prediction in Shah Alam is 10.004, while the AE value is 7.982. Finally, all of the predicted models in Klang and Shah Alam can be used to predict the PM10 concentrations. This proposed model can be used as a tool for an early warning system in giving air quality information to local authorities in order to formulate air-quality-improvement strategies.

A Suitable Model for Spatiotemporal Particulate Matter Concentration Prediction in Rural and Urban Landscapes, Thailand

Spatiotemporal particulate matter (PM) concentration prediction using MODIS AOD with significant PM factors in rural and urban landscapes in Thailand is necessary for public health and has been complicated by the limitations of PM monitoring stations. The research objectives were (1) to identify significant factors affecting PM10 concentrations in rural landscapes and PM2.5 in urban landscapes; (2) to predict spatiotemporal PM10 and PM2.5 concentrations using geographically weighted regression (GWR) and mixed-effect model (MEM), and (3) to evaluate a suitable spatiotemporal model for PM10 and PM2.5 concentration prediction and validation. The research methodology consisted of four stages: data collection and preparation, the identification of significant spatiotemporal factors affecting PM concentrations, the prediction of spatiotemporal PM concentrations, and a suitable spatiotemporal model for PM concentration prediction and validation. As a result, the predicted PM10 concentrations using the GWR model varied from 50.53 to 85.79 µg/m3 and from 36.92 to 51.32 µg/m3 in winter and summer, while the predicted PM10 concentrations using the MEM model varied from 50.68 to 84.59 µg/m3 and from 37.08 to 50.81 µg/m3 in both seasons. Likewise, the PM2.5 concentration prediction using the GWR model varied from 25.33 to 44.37 µg/m3 and from 16.69 to 24.04 µg/m3 in winter and summer, and the PM2.5 concentration prediction using the MEM model varied from 25.45 to 44.36 µg/m3 and from 16.68 and 23.75 µg/m3 during the two seasons. Meanwhile, according to Thailand and U.S. EPA standards, the monthly air quality index (AQI) classifications of the GWR and MEM were similar. Nevertheless, the derived average corrected Akaike Information Criterion (AICc) values of the GWR model for PM10 and PM2.5 predictions during both seasons were lower than that of the MEM model. Therefore, the GWR model was chosen as a suitable model for spatiotemporal PM10 and PM2.5 concentration predictions. Furthermore, the result of spatial correlation analysis for GWR model validation based on a new dataset provided average correlation coefficient values for PM10 and PM2.5 concentration predictions with a higher than the expected value of 0.5. Subsequently, the GWR model with significant monthly and seasonal factors could predict spatiotemporal PM 10 and PM2.5 concentrations in rural and urban landscapes in Thailand.

Impact of limited residential address on health effect analysis of predicted air pollution in a simulation study

BackgroundRecent epidemiological studies of air pollution have adopted spatially-resolved prediction models to estimate air pollution concentrations at people’s homes. However, the benefit of these models was limited in many studies that used existing health data relying on incomplete addresses resulting from confidentiality concerns or lack of interest when designed.ObjectiveThis simulation study aimed to understand the impact of incomplete addresses on health effect estimation based on the association between particulate matter with diameter ≤10 µm (PM10) and low birth weight (LBW).MethodsWe generated true annual average concentrations of PM10 at 46,007 mothers’ homes and their LBW status, using the parameters obtained from our data analysis and a previous study in Seoul, Korea. Then, we hypothesized that mothers’ address information is limited to the district and compared the properties of their health effect estimates of PM10 with those using complete addresses. We performed this comparison across eight environmental scenarios that represent various spatial distributions of PM10 and nine exposure prediction methods that provide different sets of predicted PM10 concentrations of mothers.ResultsWe observed increased bias and root mean square error consistently across all environmental scenarios and prediction methods using incomplete addresses compared to complete addresses. However, the bias related to incomplete addresses decreased when we used population-representative exposures averaged to the district from predicted PM10 at census tract centroids.SignificanceOur simulation study suggested that individual exposure estimated by prediction approaches and averaged across population-representative points can provide improved accuracy in health effect estimates when complete address data are unavailable.Impact statementOur simulation study focused on a common and practical challenge of limited address information in air pollution epidemiology, and investigated its impact on health effect analysis. Cohort studies of air pollution have developed advanced exposure prediction model to allow the estimation of individual-level long-term air pollution concentrations at people’s addresses. However, it is common that address information of existing health data is available at the coarse spatial scale such as city, district, and zip code area. Our findings can help understand the possible consequences of limited address information and provide practical guidance in achieving the accuracy in health effect analysis.

An LSTM-based neural network method of particulate pollution forecast in China

Particulate pollution has become more than an environmental problem in rapidly developing economies. Large-scale, long-term and high concentration of particulate pollution occurs much more frequently, which not only affects human health but also economic production. As PM10 is one of the main pollutants, the prediction of its concentration is of great significance. In this study, we present a PM10 forecast model based on the long short-term memory (LSTM) neural network method and evaluate its performance of predicting PM10 daily concentrations at five representative cities (Beijing, Taiyuan, Shanghai, Nanjing and Guangzhou) in China. Our model shows excellent adaptability for various regions in China. The predicted PM10 concentrations have good correlations with observations (R = 0.81–0.91). We also achieve great predication accuracy (70%–80%) on predicting the next-day changing trend and the model has the best performance for heavy pollution situation (PM10 > 100 μg m−3). In addition, the comparison of LSTM-based method and other statistical/machine learning methods indicates that our model is not only robust to different pollution intensities and geographic locations, but also with great potential on pollution forecast with temporal-correlated feature.

Evaluation of Boosted Regression Tree for the Prediction of the Maximum 24-Hour Concentration of Particulate Matter

Air pollution is a considerable health danger to the environment. The objective of this study was to assess the characteristics of air quality and predict PM10 concentrations using boosted regression trees (BRTs). The maximum daily PM10 concentration data from 2002 to 2016 were obtained from the air quality monitoring station in Kuching, Sarawak. Eighty percent of the monitoring records were used for the training and twenty percent for the validation of the models. The best iteration of the BRT model was performed by optimizing the prediction performance, while the BRT algorithm model was constructed from multiple regression models. The two main parameters that were used were the learning rate (lr) and tree complexity (tc), which were fixed at 0.01 and 5, respectively. Meanwhile, the number of trees (nt) was determined by using an independent test set (test), a 5-fold cross validation (CV) and out-of-bag (OOB) estimation. The algorithm model for the BRT produced by using the CV was the best guide to be used compared with the OOB to test the predicted PM10 concentration. The performance indicators showed that the model was adequate for the next day’s prediction (PA=0.638, R2=0.427, IA=0.749, NAE=0.267, and RMSE=28.455).

Modelling daily PM10 concentrations at high spatio-temporal resolution across South Africa.

Background Air pollution is a major environmental risk to health, responsible for one in every nine deaths globally. Currently, limited spatiotemporal air pollution data is available to conduct large-scale epidemiological studies investigating the adverse health effects of air pollution in South Africa. Here we aim to model South Africa’s daily average PM10 concentrations for years 2010 to 2016 using spatial and temporal predictor variables including South Africa’s data from Moderate Resolution Imaging Spectroradiometer (MODIS) and Copernicus Atmosphere Monitoring Service (CAMS). Methods We followed a two-stage modelling strategy. In stage 1, we imputed missing Aerosol Optical Depth (AOD) data, due mainly to cloud cover, by developing Random Forest (RF) models between AOD and CAMS AOD estimates also including additional large-scale spatiotemporal predictors. In stage 2, we developed RF models to explain the measured PM10 concentration using the imputed AOD data from stage 1 and spatiotemporal predictors including land cover, road density, population density, altitude, climatological zones, and meteorological variables. Results Preliminary analysis for 2016 shows that our stage 1 RF models can explain 96 percent in variability in the AOD data. In stage 2, our models explain 79% of the variation in measured ground level PM10 concentrations with the corresponding 10-fold cross-validation R2 of 78%. The RF stage 2 model includes variables depicting some PM10 source related variables like day of the week, geographical coordinates, road density, altitude and meteorological predictor variables in the top 10 of influential predictor variables. Conclusion This is the first study showing the potential of earth observation data to develop models explaining fine scale (1x1 km) temporal (daily) variation of PM10 across South Africa. We developed models for years 2010 to 2016. The predicted PM10 concentrations will be made available to facilitate health research in South Africa.

Evolution and forecasting of PM10 concentration at the Port of Gijon (Spain)

The name PM10 refers to small particles with a diameter of less than 10 microns. The present research analyses different models capable of predicting PM10 concentration using the previous values of PM10, SO2, NO, NO2, CO and O3 as input variables. The information for model training uses data from January 2010 to December 2017. The models trained were autoregressive integrated moving average (ARIMA), vector autoregressive moving average (VARMA), multilayer perceptron neural networks (MLP), support vector machines as regressor (SVMR) and multivariate adaptive regression splines. Predictions were performed from 1 to 6 months in advance. The performance of the different models was measured in terms of root mean squared errors (RMSE). For forecasting 1 month ahead, the best results were obtained with the help of a SVMR model of six variables that gave a RMSE of 4.2649, but MLP results were very close, with a RMSE value of 4.3402. In the case of forecasts 6 months in advance, the best results correspond to an MLP model of six variables with a RMSE of 6.0873 followed by a SVMR also with six variables that gave an RMSE result of 6.1010. For forecasts both 1 and 6 months ahead, ARIMA outperformed VARMA models.

Air Quality Assessment by Statistical Learning-Based Regularization

PM10 can be stated as a particulate matter smaller than 10 micrometer and it can be suspended in the air. The incremental concentration of PM10 affects both human and environment drastically. In this study, an air quality assessment by exhibiting the potential relationships among the secondary indicators and PM10 has been focused. For the analyses, statistical learning-based regularization procedures such as Ridge, the Lasso and Elastic-net algorithms have been practiced. In particular, use of Elastic-net algorithm in predicting PM10 concentration includes a novelty. As a result of the computational studies, it has been recorded that all the models showed high accuracy capacities. However, the elastic-net model outperforms the other models both accuracy and robustness (stability). Considering the error measurements (MSE and MAPE), the best numerical results have been provided by the Elastic-net model. Use of machine learning-based regularization algorithms in environmental problems can provide accurate model structures as well as generality and transparency.

Prediction of source contributions to urban background PM<sub>10</sub> concentrations in European cities: a case study for an episode in December 2016 using EMEP/MSC-W rv4.15 and LOTOS-EUROS v2.0 – Part 1: The country contributions

Abstract. A large fraction of the urban population in Europe is exposed to particulate matter levels above the WHO guideline value. To make more effective mitigation strategies, it is important to understand the influence on particulate matter (PM) from pollutants emitted in different European nations. In this study, we evaluate a country source contribution forecasting system aimed at assessing the domestic and transboundary contributions to PM in major European cities for an episode in December 2016. The system is composed of two models (EMEP/MSC-W rv4.15 and LOTOS-EUROS v2.0), which allows the consideration of differences in the source attribution. We also compared the PM10 concentrations, and both models present satisfactory agreement in the 4 d forecasts of the surface concentrations, since the hourly concentrations can be highly correlated with in situ observations. The correlation coefficients reach values of up to 0.58 for LOTOS-EUROS and 0.50 for EMEP for the urban stations; the values are 0.58 for LOTOS-EUROS and 0.72 for EMEP for the rural stations. However, the models underpredict the highest hourly concentrations measured by the urban stations (mean underestimation of 36 %), which is to be expected given the relatively coarse model resolution used (0.25∘ longitude × 0.125∘ latitude). For the source attribution calculations, LOTOS-EUROS uses a labelling technique, while the EMEP/MSC-W model uses a scenario having reduced anthropogenic emissions, and then it is compared to a reference run where no changes are applied. Different percentages (5 %, 15 %, and 50 %) for the reduced emissions in the EMEP/MSC-W model were used to test the robustness of the methodology. The impact of the different ways to define the urban area for the studied cities was also investigated (i.e. one model grid cell, nine grid cells, and grid cells covering the definition given by the Global Administrative Areas – GADM). We found that the combination of a 15 % emission reduction and a larger domain (nine grid cells or GADM) helps to preserve the linearity between emission and concentrations changes. The nonlinearity, related to the emission reduction scenario used, is suggested by the nature of the mismatch between the total concentration and the sum of the concentrations from different calculated sources. Even limited, this nonlinearity is observed in the NO3-, NH4+, and H2O concentrations, which is related to gas–aerosol partitioning of the species. The use of a 15 % emission reduction and of a larger city domain also causes better agreement on the determination of the main country contributors between both country source calculations. Over the 34 European cities investigated, PM10 was dominated by domestic emissions for the studied episode (1–9 December 2016). The two models generally agree on the dominant external country contributor (68 % on an hourly basis) to PM10 concentrations. Overall, 75 % of the hourly predicted PM10 concentrations of both models have the same top five main country contributors. Better agreement on the dominant country contributor for primary (emitted) species (70 % is found for primary organic matter (POM) and 80 % for elemental carbon – EC) than for the inorganic secondary component of the aerosol (50 %), which is predictable due to the conceptual differences in the source attribution used by both models. The country contribution calculated by the scenario approach depends on the chemical regime, which largely impacts the secondary components, unlike the calculation using the labelling approach.

Mapping and Understanding Patterns of Air Quality Using Satellite Data and Machine Learning

AbstractThe quantification of factors leading to harmfully high levels of particulate matter (PM) remains challenging. This study presents a novel approach using a statistical model that is trained to predict hourly concentrations of particles smaller than 10 m (PM10) by combining satellite‐borne aerosol optical depth (AOD) with meteorological and land‐use parameters. The model is shown to accurately predict PM10 (overall R = 0.77, RMSE = 7.44 g/m ) for measurement sites in Germany. The capability of satellite observations to map and monitor surface air pollution is assessed by investigating the relationship between AOD and PM10 in the same modeling setup. Sensitivity analyses show that important drivers of modeled PM10 include multiday mean wind flow, boundary layer height (BLH), day of year (DOY), and temperature. Different mechanisms associated with elevated PM10 concentrations are identified in winter and summer. In winter, mean predictions of PM10 concentrations >35 g/m occur when BLH is below 500 m. Paired with multiday easterly wind flow, mean model predictions surpass 40 g/m of PM10. In summer, PM10 concentrations seemingly are less driven by meteorology, but by emission or chemical particle formation processes, which are not included in the model. The relationship between AOD and predicted PM10 concentrations depends to a large extent on ambient meteorological conditions. Results suggest that AOD can be used to assess air quality at ground level in a machine learning approach linking it with meteorological conditions.

Validation of linear, nonlinear, and hybrid models for predicting particulate matter concentration in Tehran, Iran

Information on particulate matter forecast is significant as it allows residents to manage its undesirable effects. For the purpose of predicting PM10 concentration in the air of Tehran, various models were used, including (i) a linear model (multiple liner regression, MLR), (ii) two hybrid models (Adaptive Neuro-Fuzzy Inference System, ANFIS as well as ensemble empirical mode decomposition and general regression neural network, EEMD-GRNN), and (iii) a nonlinear model (multi-layer perceptron, MLP). The output variable in these models was the measure of suspended particles of PM10 while the predictor variables were the information on air quality which consisted of CO, NO2, O3, PM10 of the previous day, PM2.5, and SO2 as well as meteorological data which included average atmospheric pressure (AP), average maximum temperature (Max T), average minimum temperature (Min T), daily relative humidity level of the air (RH), daily total precipitation (TP), and daily wind speed (WS) for the year 2016 in Tehran. Analysis of the data revealed that in comparison with the results of MLR and MLP, ANFIS obtained the most accurate output (R2 = 0.97, root mean square error (RMSE) = 1.0713, and mean absolute error (MAE) = 0.6111) for the training phase and (R2 = 0.89, RMSE = 3.6165, and MAE = 2.8993) the testing phase. However, the hybrid models which were used in the current study had almost similar prediction results. As it can be concluded, in comparison with linear and nonlinear models, hybrid models turn out to have higher accuracy in predicting PM10 concentration.

Temporal Assessment on Variation of PM10 Concentration in Kota Kinabalu using Principal Component Analysis and Fourier Analysis

PM10 (particulate matter with aerodynamic diameter below 10 microns) has always caught scientific attention due to its effect to human health. Predicting PM10 concentration is essential for early preventive measures, especially for cities such as Kota Kinabalu. Temporal data clustering may enhance accuracy of prediction model by group data in time range. However, the necessity of temporal data clustering has yet to be studied in Kota Kinabalu. OBJECTIVE. This research is conducted to compare significance of meteorological and pollutant factors for PM10 variation in clustered and unclustered data. METHODOLOGY. This study is focused in Kota Kinabalu, Sabah. The data for meteorological factors (Ws, Wd, Hum, Temp) and pollutant factors (CO2, NO2, O3, SO2, PM10) from 2003 to 2012 provided by Department of Environment are used for this research. Missing data are imputed using nearest neighbour method before it is clustered by monsoonal clustering. Unclustered and clustered datasets are analysed using principal component analysis (PCA) to check significance of factors contributing to PM10 concentration. FINDINGS. PCA results show that temporal clustering does not have noticeable effect on the variation of PM10 concentration. For all datasets, humidity and x-component wind speed have highest factor loading on PC1 and PC2 respectively. Further statistical analysis by 2-D regression shows that humidity (ρ = -0.60 ± 0.20) and temperature (ρ = 0.63 ± 0.11) have moderate to strong correlation towards PM10 concentration. This may be due to high humidity level and strong negative correlation between temperature and humidity (ρ = -0.91 ± 0.03). In contrast, both x- and y-component wind speed generally show weak correlation towards PM10, with ρ value of 0.09 ± 0.14 and 0.24 ± 0.18 respectively probably because of varying direction of particle dispersion. Fourier analysis further confirms this result by showing that human activity contributes major effect to variation of PM10 concentration.

Predicting PM10 concentration using Bayesian regression with Non-Informative Prior and Conjugate Prior Model

The aim of this study is to predict the next day PM10 concentration using Bayesian Regression with noninformative prior and conjugate prior models. The descriptive analysis of PM10, temperature, relative humidity, nitrogen dioxide (NO2), sulphur dioxide (SO2), carbon monoxide (CO) and ozone (O3) are also included. A case study used two-years of air quality monitoring data at three (3) monitoring stations to predict the future PM10 concentration with seven parameters (PM10, temperature, relative humidity, NO2, SO2, CO, and O3). The descriptive analysis showed that the highest mean PM10 concentration occurred at Klang station in 2011 (71.30 µg/m3) followed by 2012 (68.82 µg/m3). The highest mean PM10 concentration was at Nilai in 2012 (68.86 µg/m3) followed by 2011 (66.29µg/m3) respectively. The results showed that the Bayesian regression model used a conjugate prior with a normal-gamma prior which was a good model to predict the PM10 concentration for most study stations with (R2 = 0.67 at Jerantut station), (R2 = 0.61 at Nilai station) and (R2 = 0.66 at Klang station) respectively compared to a non-informative prior.

Predicting PM10 concentration in Seoul metropolitan subway stations using artificial neural network (ANN)

The indoor air quality of subway systems can significantly affect the health of passengers since these systems are widely used for short-distance transit in metropolitan urban areas in many countries. The particles generated by abrasion during subway operations and the vehicle-emitted pollutants flowing in from the street in particular affect the air quality in underground subway stations. Thus the continuous monitoring of particulate matter (PM) in underground station is important to evaluate the exposure level of PM to passengers. However, it is difficult to obtain indoor PM data because the measurement systems are expensive and difficult to install and operate for significant periods of time in spaces crowded with people. In this study, we predicted the indoor PM concentration using the information of outdoor PM, the number of subway trains running, and information on ventilation operation by the artificial neural network (ANN) model. As well, we investigated the relationship between ANN’s performance and the depth of underground subway station. ANN model showed a high correlation between the predicted and actual measured values and it was able to predict 67∼80% of PM at 6 subway station. In addition, we found that platform shape and depth influenced the model performance.