A new approach to discovering the causal relationship between meteorological patterns and PM10 exceedances

Abstract

Air Pollution control is of major concern for the Greater Athens Area (GAA) of Greece. High concentrations of Particulate Matter with diameter less than 10μm (PM10) have been observed and often reported to exceed the currently EU legislated 24-hour limit of 50μg/m3. Efforts therefore have been placed on understanding the PM10 concentration pattern in the area so that mitigation measures can be taken accordingly. The present paper presents a statistical methodology to discover causal relationships between daily PM10 exceedances at a monitoring site, with PM10 concentrations from existing stations from the monitoring network and associated weather patterns. The proposed approach utilised a dimension reduction algorithm, Positive Matrix Factorisation (PMF) algorithm, coupled with the k-means clustering algorithm to identify distinct groups of data. Then for each resulted cluster, the Granger Causality method aided by the Pearson correlation is applied to establish the causal relationships between the meteorological patterns and the observed PM10 exceedances. The study was conducted using 6-years of daily PM10 concentration data from the monitoring network in the GAA, complemented with meteorological data available from the National Centres for Environmental Prediction (NCEP) Global Forecasting System (GFS). The analysis yielded that the PM10 exceedances in the Athens area can be classified into 6 distinct types identified with varying spatial distribution characteristics and air pollution contributors.