The role of multi-model ensembles in assessing the air quality impact on crop yields and mortality

This work promotes a critical use of modelling information on air-pollution health and agriculture impacts, with the primary goal of providing more reliable estimates to decision makers and stakeholders. To date, the accuracy of air quality (AQ) models and the quantification of the uncertainty of their results have rarely been quantified explicitly in impact assessment studies, therefore without giving information on the robustness of the information used in the decision making process and undermining the confidence in the results obtained. A suite of twelve regional-scale chemistry transport AQ models produced in the third phase of the Air Quality Model Evaluation International Initiative (AQMEII) is used here to calculate the impact of PM2.5and ozone on human health and crop yields and the associated uncertainties over Europe. A novel methodology is developed and applied to remove the offsetting bias from the models, which are then combined in multi-model (MM) ensembles. The application of unbiased MM ensembles offers an unprecedented attempt to i) establish and ii) mitigate the uncertainty due to AQ modelling on impact calculations. We use the FASST (FAst Scenario Screening Tool) impact assessment tool to demonstrate that the accuracy of assessment of ozone-induced crop loss of wheat and maize and impact on human health (mortality) can improve dramatically when using accurate MM ensembles in place of single model realizations, as it is commonly assumed.