A Statistical Analysis of Surface Ozone Variability Over the Mediterranean Region During Summer

Abstract

The Mediterranean (Med) region contains the highest surface ozone (O3surf) concentrations in Europe, in particular in summer. Although atmospheric chemistry models (ACMs) are able to reproduce the regional O3surf maximum over the Med region, they generally exhibit a high bias over this region. There are many chemical and physical processes acting on different time scales that act as controlling factors of the O3surf variability and might not be simulated accurately by the models. Thus, detailed investigation of a single model is needed to understand why (or where) the model is not able to reproduce the observed O3surf variability. In this study, we used two datasets to evaluate the O3surf variability over the Med region on the intraday (ID), diurnal (DU), and synoptic (SY) time scales. Observations from the TOAR database at 76 rural and 109 urban Med stations for summers 2010, 2011, and 2012 were compared to results of a WRF-Chem model simulation at 30 km resolution. As a first step, a scale analysis was performed on both datasets to determine the relative importance of the different time scales, i.e. ID, DU, and SY, for the O3surf variability. In a second step, a model performance indicator was used to quantify the amount of the model error in simulating the O3surf variability in each time scale. In the third step, a multiple linear regression (MLR) model was established for each component of the variability spectrum to identify the relationship between O3surf and several independent variables such as AT, RH, SP, U, V, and NOx in the measurements and in the simulation. The results of the scale analysis of the measured and simulated O3surf show that the model simulation is able to capture major features of the O3surf variability at each time scale for both rural and urban sites. However, the relative contributions of each time scale to the total O3surf variability differ from those of the observations. The model performance indicator reveals that the major part of the model variance error in simulating the O3surf variability over the Med region is associated with the model's difficulties in simulating the observed DU and SY variability. From the MLR analysis, it appears that the ID, DU and SY variability of the measured O3surf over the Med region are predominantly associated with the variability of NOx, AT, and RH, respectively. However, in the WRF-Chem model, the relationship between O3surf and NOx is much stronger than in the measurements for all time scales. Moreover, the model does not capture the observed relationship between O3surf and RH well. Contrary to prior expectations, the MLR analysis did not show a strong relationship between the observed O3surf and SP, U, and V, even though it is clear that local circulation patterns can affect measurements for example at a mountain or coastal site. This lack of correlation might be reproduced by the nonlinearity of the relationship between O3surf and these variables. Therefore, misrepresentation of the true relationship between variability of the O3surf and several variables, such as NOx, RH, and AT, can be a possible reason for the model variance error.