Combination techniques of monitoring data and model estimations of air pollutants concentration in the Veneto region

Air pollution is a crucial environmental parameter related to an alteration in the chemical composition of air. High levels of air pollutants can be dangerous for health and damage the environment, biological resources and ecosystems. Measuring such concentration is an important task, especially in regions where, for morphological reasons, pollutants tend to persist for more time. The Po Valley is a highly industrialized and densely populated region surrounded by mountains that do not favour enough ventilation. Here the air pollutants concentrations can reach the legal limits set by the European Union. In Veneto, ARPAV (Regional Environmental Protection Agency Veneto) is the agency responsible for air quality control. They monitor it using a network of stations distributed over the Regional territory which measure the concentration of the principal pollutants (PM10, PM2.5, NOx, O3). A deterministic eulerian Chemical Transport Model (CTM) is also used to have better estimations of the whole territory (even far from stations) and predictions of the concentrations for the upcoming days. The goal of this work is to apply and evaluate some "data fusion" methods, statistical-based, commonly used to merge model estimations and station measurements; these techniques allow to improve the estimate of pollutants concentration in the model domain and are thus of paramount importance for ARPAV. After an introduction to the context where the problem arises, the work will focus on the data and statistics used for the analysis. Firstly, an overview of the air pollutant concentrations data (PM10, NOx and O3) for the Veneto region is presented, describing how the data are gathered and organized. Afterwards, the deterministic model is also briefly described, with a particular interest in the verification with respect to the measurements. The main part of the work is an in-depth analysis of the interpolation methods used for spatial prediction to correct the model predictions using station measurements. Finally, the results of the analysis are discussed.