Be the first to like this
This paper explores the performance of two commonly used air quality models: dispersion models and land-use regression models. Both models are widely used in air pollution epidemiological studies and in health impact assessment studies. In this work, we looked at how the choice of the validation dataset impacts the performance of air quality models and the insights gleaned from their validation. We also looked at whether the spatial resolution for the models' setup impacts the performance of air quality models and the insights gleaned from their validation. We saw that R-squared almost halved when the air quality models' estimates were made at the centroid of the 100x100m grid in which the validation point fell, instead of at the exact location of the validation point. We also saw that the different validation datasets give very different insights.
Dispersion models and land-use regression models are widely used in air pollution epidemiological studies and in health impact assessment studies. As such the performance of these air quality models have implications on the ability of epidemiological studies to pick up associations between the exposures and the health outcomes of interest, and the ability of health impact assessment studies to quantify the impacts accurately. This work demonstrated the value of validating modeled air quality data against various datasets to obtain a better understanding of the performance of models and the value of reporting these validation results. Also, the work suggested that the spatial resolution of the models’ estimates has a significant influence on the validity at the application point. These results should be considered when air quality models are used to assign human exposures and study the health effects/impacts of these exposures. Significant work is still needed to improve the performance of air quality models and their ability to pick up the variations of air pollution levels and especially the higher and more variable levels that are related to traffic. Significant work is also still needed to account for the factors that underlie this variation in epidemiological and health impact assessment studies, especially time activity patterns of the exposed populations