Keywords
data mining, climatology, pm10, exceedances
Start Date
1-7-2008 12:00 AM
Abstract
The present study presents a data mining methodology to discover weathercharacteristics that are associated with exceedances of particulate matter (PM) withdiameter less than 10 μm. The proposed approach is an eigenvector one consisting of arotated principal components algorithm coupled with the fuzzy c-means clusteringalgorithm. This study examines only those days where at least one station exhibited anexceedance of the limit of a daily average value of 50 μg/m3. The analysis was conductedusing daily data from the monitoring network of the Grater Athens Area that werecomplemented with meteorological data from the NCEP Global Forecasting System (GFS).The analysis concluded that the PM10 exceedances in the Athens Area can be classifiedinto 10 distinct types with varying spatial characteristics and weather contribution.
A data mining approach to discover weather patterns contributing to PM10 exceedances
The present study presents a data mining methodology to discover weathercharacteristics that are associated with exceedances of particulate matter (PM) withdiameter less than 10 μm. The proposed approach is an eigenvector one consisting of arotated principal components algorithm coupled with the fuzzy c-means clusteringalgorithm. This study examines only those days where at least one station exhibited anexceedance of the limit of a daily average value of 50 μg/m3. The analysis was conductedusing daily data from the monitoring network of the Grater Athens Area that werecomplemented with meteorological data from the NCEP Global Forecasting System (GFS).The analysis concluded that the PM10 exceedances in the Athens Area can be classifiedinto 10 distinct types with varying spatial characteristics and weather contribution.
