Functional kriging prediction of atmospheric particulate matter concentrations in Madrid, Spain: Is the new monitoring system masking potential public health problems?

Abstract

Prediction of particulate matter concentrations is of particular interest in the field of air pollution control. We focus on the spatio-temporal geostatistical approach to predicting particulate matter in large urban areas. However, due to both the poor performance and the prohibitive computational burden of traditional spatio-temporal kriging when using massive databases (the “big n problem”), we use functional kriging prediction, a promising strategy in such situations. Since it deals with functional data representing the observations recorded at each observed location, spatial-only kriging based on such functional observations and also a trace semivariogram enables prediction of functional data, thus overcoming the “big n problem”. This approach has been applied in Madrid (Spain), where, in 2010, the city's atmospheric pollution monitoring system was reorganized and ecologist associations suspected that the Municipality of Madrid removed stations from the sites that were potentially the most polluted. To our knowledge, this is the first study that uses functional kriging models to predict air pollution in Spain. Our results indicate that such suspicions are unfounded because predicted particulate matter concentrations at the sites where the removed stations were placed are in line with the concentrations levels measured by the new stations. This methodology can be applied to other air pollutants as well as to other types of pollution (water pollution, noise pollution, odour pollution, etc.).