Information on situation of air pollution is critically needed as input in four disciplines of research including risk management, risk evaluation, environmental epidemiology, as well as for status and trend analysis. Two normal practices were identified to evaluate daily air pollution situation; first, pollution magnitude has been treated as the common indicator, and second, the analysis was often conducted based on hourly average data. However, the information on the magnitude level alone to represent the pollution condition based on a rigid point data such as the average was seen as insufficient. Thus, to fill the gap, this study was conducted based on continuously measured data in the form of curves, which is also known as functional data, whereby pollution duration is emphasised. A statistical method based on curve ranking was used in the investigation. The application of the method at Klang, Petaling Jaya and Shah Alam air quality monitoring stations located in the Klang Valley, Malaysia, has shown that pollution duration decreases as the magnitude increases. Shah Alam has the longest pollution duration at low and medium magnitude levels. Meanwhile, all the three stations experienced quite a similar length of average pollution duration for the high magnitude level, that is, about 2.5 days. It was also shown that the occurrence of PM10 pollution at the area is significantly not random.
Spectral data is often required to be pre-processed prior to applying a multivariate modelling technique. Baseline correction of spectral data is one of the most important and frequently applied pre-processing procedures. This preliminary study aims to investigate the impacts of six types of baseline correction algorithms on classifying 150 infrared spectral data of three varieties of paper. The algorithms investigated were Iterative Restricted Least Squares, Asymmetric Least Squares (ALS), Low-pass FFT Filter, Median Window (MW), Fill Peaks and Modified Polynomial Fitting. Processed spectral data were then analysed using Principal Component Analysis (PCA) to visually examine the clustering among the three varieties of paper. Results show that separation among the three varieties of paper is greatly improved after baseline correction via ALS, FP and MW algorithms.