Air quality pattern assessment in Malaysia using multivariate techniques

This study aims to investigate the spatial characteristics in the pattern of air quality monitoring sites, identify the most discriminating parameters contributing to air pollution, and predict the level of air pollution index (API) in Malaysia using multivariate techniques. Five parameters observed...

وصف كامل

محفوظ في:
التفاصيل البيبلوغرافية
المؤلفون الرئيسيون: Azman, Azid, Hamza, Ahmad Isiyaka
التنسيق: مقال
اللغة:English
منشور في: Malaysian Society of Analytical Sciences 2015
الموضوعات:
الوصول للمادة أونلاين:http://eprints.unisza.edu.my/6740/1/FH02-ESERI-15-04072.jpg
http://eprints.unisza.edu.my/6740/
الوسوم: إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
الوصف
الملخص:This study aims to investigate the spatial characteristics in the pattern of air quality monitoring sites, identify the most discriminating parameters contributing to air pollution, and predict the level of air pollution index (API) in Malaysia using multivariate techniques. Five parameters observed for five years (2000-2004) were used. Hierarchical agglomerative cluster analysis classified the five air quality monitoring sites into two independent groups based on the characteristics of activities in the monitoring stations. Discriminate analysis for standard, backward stepwise and forward stepwise mode gave a correct assignation of more than 87% in the confusion matrix. This result indicates that only three parameters (PM10, SO2 and NO2) with a p<0.0001 discriminate best in polluting the air. The major possible sources of air pollution were identified using principal component analysis that account for more than 58% and 60% in the total variance. Based on the findings, anthropogenic activities (vehicular emission, industrial activities, construction sites, bush burning) have a strong influence in the source of air pollution. Furthermore, artificial neural network (ANN) was used to predict the level of air pollution index at R = 0.8493 and RMSE = 5.9184. This indicates that ANN can predict more than 84% of the API.