Skip to content

EHP logoISEE logo

2016 Conference

Abstract Number: E-01 | ID: 3674

Fine Spatio-Temporal Resolution of PM10 and PM2.5 Concentrations in Italy (2006-2012) using Satellite Data and Several Land Use Variables

Massimo Stafoggia*, Department of Epidemiology, Lazio Region Health Service, Italy, m.stafoggia@deplazio.it; Itai Kloog, Department of Geography & Environmental Development, Ben-Gurion University of the Negev, Israel, ikloog@bgu.ac.il; Chiara Badaloni, Department of Epidemiology, Lazio Region Health Service, Italy, c.badaloni@deplazio.it; Giorgio Cattani, Italian National Institute for Environmental Protection and Research, Italy, giorgio.cattani@isprambiente.it; Alessandra Gaeta, Italian National Institute for Environmental Protection and Research, Italy, alessandra.gaeta@isprambiente.it; Francesco Forastiere, Department of Epidemiology, Lazio Region Health Service, Italy, f.forastiere@deplazio.it; Gianluca Leone, Italian National Institute for Environmental Protection and Research, Italy, gianluca.leone@isprambiente.it; Joel Schwartz, Department of Environmental Health, Harvard T. H. Chan School of Public Health, United States, jschwrtz@hsph.harvard.edu;
Introduction: Health effects of particulate matter (PM) have been widely investigated. However, studies in industrial, small cities or rural settings are rare due to the lack of PM data. We aimed to estimate daily PM10 and PM2.5 concentrations at 1x1km scale in Italy over 2006-2012 matching satellite data with meteorology, land-use, emission and population variables.
Methods: Satellite-based Aerosol Optical Depth (AOD), modelled Planetary Boundary Layer (PBL) and meteorological data were collected alongside PM data from monitoring stations, for each day in the study period. Spatial predictors were built for each 1x1km cell of Italy, including: land-use, elevation, roads and population density. A random-effects 3-step statistical model was performed. In stage 1, calibration between PM and AOD was defined for each year. Stage 1 was used to predict PM in grid cells/days without monitors but with available AOD (stage 2). Finally, we estimated PM10/PM2.5 concentrations in cells/days with no AOD by taking advantage of the association of AOD values with PM10/PM2.5 monitoring located elsewhere, and the association with available AOD values in neighboring grid cells (stage 3).
Results: Cross-Validated (CV) R2 and root mean squared errors (RMSE) were in the range 0.61-0.67 and 9-12 ug/m3, respectively, with little bias (slopes of predicted VS observed PM between 0.94-0.97). CV results of the spatial (yearly average) and temporal (daily variability) components displayed good fitting (median CV-R2=0.50 and 0.65 with RMSE < 6 and 10ug/m3, respectively). PM showed a downward trend over the study period, with annual PM10 averages decreasing from 25 to 20 ug/m3 across the country, but the rates of change varied substantially by areas.
Conclusions: High resolution PM daily and annual exposure will allow joint estimation of long- and short-term health effects for the whole Italy in urban, rural and industrial settings during a period of several years including the economic crisis.