Classification of PM10 distributions in Taiwan

Abstract Daily average PM10 concentrations of 71 stations in Taiwan in wintertime (October to March) and summertime periods (April to September) were fitted individually by a lognormal distribution for a 2 yr period (2001 to 2002). The distribution parameters (geometric mean and geometric standard deviation) in wintertime were used to determine the air-quality basins for PM10 by utilizing three clustering techniques, viz. of hierarchical clustering (Ward's method), non-hierarchical clustering (K-means) and two-level approach (self-organizing maps neural network, then K-means clustering). All three techniques suggested that 71 air-monitoring stations in Taiwan can be divided into five air-quality basins which are located in northern, central, eastern, southwestern and southern Taiwan, respectively. The sequence of PM10 pollution levels in the five basins is southern Taiwan>southwestern Taiwan>central Taiwan>northern Taiwan>eastern Taiwan. Geometric means and geometric standard deviations in each of the five air-quality basins were significantly different from each other for the two-level approach method by the Waller–Duncan k-ratio t-test ( k = 1 0 0 , P = 0.0 5 ), suggesting that the two-level approach method is best among the three clustering methods. The clustering results of five air-quality basins in Taiwan are useful to decide the corresponding control strategy at different air-quality basins.

[1]  Kuang-Ling Yang,et al.  Spatial and seasonal variation of PM10 mass concentrations in Taiwan , 2002 .

[2]  David West,et al.  A comparison of SOM neural network and hierarchical clustering methods , 1996 .

[3]  Gregory R. Madey,et al.  Heuristic and optimization approaches to extending the Kohonen self organizing algorithm , 1996 .

[4]  Joel Schwartz,et al.  REVIEW OF EPIDEMIOLOGICAL EVIDENCE OF HEALTH EFFECTS OF PARTICULATE AIR POLLUTION , 1995 .

[5]  Perry J. Samson,et al.  Use of Cluster Analysis to Define Periods of Similar Meteorology and Precipitation Chemistry in Eastern North America. Part I: Transport Patterns , 1990 .

[6]  Qing Yang,et al.  Modeling the effects of meteorology on ozone in Houston using cluster analysis and generalized additive models , 1998 .

[7]  David M. Holland,et al.  Fitting statistical distributions to air quality data by the maximum likelihood method , 1982 .

[8]  F. L. Ludwig,et al.  Classification of ozone and weather patterns associated with high ozone concentrations in the san francisco and monterey bay areas , 1995 .

[9]  Esa Alhoniemi,et al.  Clustering of the self-organizing map , 2000, IEEE Trans. Neural Networks Learn. Syst..

[10]  Anthony J. Jakeman,et al.  Modeling distributions of air pollutant concentrations—II. Estimation of one and two parameter statistical distributions , 1986 .

[11]  Sovan Lek,et al.  A comparison of self-organizing map algorithm and some conventional statistical methods for ecological community ordination , 2001 .

[12]  V. Lo Brano,et al.  Forecasting daily urban electric load profiles using artificial neural networks , 2004 .

[13]  Hsin-Chung Lu,et al.  The statistical characters of PM10 concentration in Taiwan area , 2002 .

[14]  Subhash Sharma.  Applied multivariate techniques , 1995 .

[15]  Tai-Yi Yu,et al.  Delineation of air-quality basins utilizing multivariate statistical methods in Taiwan , 2001 .

[16]  A S Kao,et al.  Frequency Distributions of PM10 Chemical Components and Their Sources. , 1995, Environmental science & technology.