Integrating monitoring networks to obtain estimates of ground-level ozone concentrations -- A proof of concept in Tuscany (central Italy)

Prior to 2000 a network of conventional ozone (O₃) analysers existed in the Province of Firenze (Tuscany, central Italy). Between 2000 and 2004 the network was extended to incorporate a newly designed bioindicator network of tobacco plants (Nicotiana tabacum Bel W3). The objective was to set-up an integrated monitoring system to obtain estimates of ground-level O₃ concentrations over the whole study area (3513 km²) in order to fill data gaps and cover reporting requirements. The existing conventional monitors were purposefully located mainly in urban areas. A total of 45 biomonitoring sites were selected using a systematic design to cover the target area. Two to five additional biomonitoring sites were co-located with conventional O₃ analysers for calibration purposes, and five more sites for independent validation of modelled O₃ concentrations. Visible Leaf Injury Index (LII) on the tobacco plants was significantly correlated (P: 0.018[divide sign]0.0014) with a series of O₃ exposure variables (mean of weekly 1-hour maxima, M1; mean of 7-hour means, M7; 24-hour mean, M24; and weekly AOT40). LII was found to be a significant predictor of weekly means of the O₃ exposure variables with a standard error of estimates between 13.6 and 24.3 μg m⁻³ (absolute values). LII was mapped with an ad-hoc spatial model over the study area at a 2[low asterisk]2 km grid resolution, and mapped values were used to predict O₃ concentrations by means of a first order linear model. Results showed that high estimates of O₃ (up to 188 μg m⁻³ as mean of weekly maxima, M1) occurred more frequently in hilly and mountainous areas, with a spatial pattern changing on an annual basis. Predicted O₃ concentrations were not significantly different from the measured concentrations (P: 0.34), although marked differences were observed for individual sites and years. The study provided evidence that integration of monitoring networks using different methods can be a viable option to obtain estimates of O₃ concentrations over large areas.