Geostatistical assessment of long term human exposure to air pollution

Nicolas Jeannée (1), Vincent Nedellec (2), Souad Bouallala (3), Jacques Deraisme (1), Hélène Desqueyroux (3)

(1) Geovariances, 49bis av. Franklin Roosevelt,77212 Avon, France
(2) Vincent Nedellec Consultants, 15 rue Firmin Gillot, 75015 Paris, France
(3) Ademe, Direction de l’Air et des Transports, Département Air, 27 rue Louis Vicat, 75737 Paris Cedex 15, France

Abstract

Keywords: air pollution, particulate matter PM10, human exposure assessment, multivariate geostatistics, conditional co-simulations.

The paper aims at illustrating the efficiency of geostatistics to provide the basic figures, based on national air quality monitoring network data, to perform Health Impact Assessment (HIA). The exposure and health assessment case studies are part of the French research effort contributing to the UNECE-WHO Pan European Program for Transport, Health and Environment (THE Pep Project): “Transport-related health impacts and their costs and benefits with a particular focus on children”. The aim of the international project is to provide environment’s European Minister with sound scientific information on social cost and impact of road traffic and recommend action or regulation that can decrease external cost or protect population health.

Health Impact Assessment (HIA) of ambient air pollution requires to accurately assess the population exposure to air pollutants. Specific HIA on transport-related air pollution implies the ability to separate traffic-related air pollution from other air pollution sources (building, industry, energy, etc.). The paper focus on a preliminary step of this HIA: the assessment, from the French air monitoring network, of ambient air concentration of particulate matter PM10 (particulate matter with an aerodynamic diameter less than 10 micron), considered as a well defined (i.e.: exposure response function between PM10 air concentration and increase frequency in many health outcomes are based on numerous concordant epidemiological study results) and health relevant indicator of transport-related emissions. As the interest is put on long-term exposure, the study is based on average annual PM10 concentrations from each measuring station. Therefore this exposure assessment ignores the day to day variability of the ambient PM10 air pollution. Special attention is paid to the heterogeneous nature of the data (rural, traffic, industrial and urban stations), excluding in particular proximity stations because of their lack of representativity.

Due to the poor efficiency of linear estimation techniques based on kriging/cokriging to address exposure to ambient air pollution issues, conditional simulations of PM10 are performed. A significant increase in the reliability of the results is obtained by taking into account the existence of: (i) a correlation between PM10 concentrations and more densely acquired NO2 data, and (ii) more recent PM10 data that allowed to supplement the PM10 monitoring network in otherwise entirely not sampled areas.

Coupling air concentration results with geo-data from the last national census (1999), the exposed population, stratified in 5 years age classes, to different levels of average annual concentrations is calculated for each studied scenario: whole territory, urban areas only, all air emission sources, only road traffic emissions. Statistical parameters from the resulting distributions are derived in the perspective of carrying out an health impact assessment study on transport related air emissions.

GeoEnv 2004