Next Generation Air Quality Models: Dynamical Mesh, New Insights into Mechanism, Datasets and Applications

Abstract

Purpose of Review

Air quality modelling and forecasting have been well recognised to play important roles in environmental research as well as government policy assessments and management strategies. To address the recent progresses in air quality modelling, we conduct a literature review focusing on air quality forecasting models and reanalysis datasets.

Recent Findings

First of all, the implementation of three-dimensional adaptive meshes/horizontal resolution-variable grids in air quality models offers a crucial insight on multi-scale simulations down to the hectometre level. These models balance high accuracy with computational efficiency. Secondly, current reanalysis still has limitations in its horizontal resolution (dozens of kilometres) that are insufficient to support the analysis and management of air pollution at factory levels or neighbourhood scales. The development of adaptive mesh resolution method provides a promising way to deal with this issue and allows the construction of the chemistry reanalysis at ultra-high resolutions (< 1 km). However, the use of adaptive mesh method in data assimilation is currently still restricted to the column-based one-dimensional models. Thirdly, the application of graphics processing units to air quality predictions enables more optimised resource usage and enhances model performance through hardware acceleration effects, while machine learning methods can both maintain the consistency with numerical solutions and increase the accuracy of air quality predictions for specific chemical species. Furthermore, parameters that describe more complicated processes and mechanisms have been added into pre-existing physical and chemical parameterisations to enable more accurate representation of various small-scale features, such as the parameterisation of inorganic chemistry on the surface of aerosols, as well as various photolysis schemes.

Summary

The increase of resolution brings computational burdens and shifts the boundary of resolved and sub-grid phenomena in air quality prediction, which in turn stimulates the development and usage of new technologies (e.g. adaptive mesh techniques, graphics processing unit acceleration, machine learning methods). They are conducive to the improvement of prediction accuracies and the acquisition of new insights on atmospheric physical and chemical mechanisms. However, new challenges also ensued, including the selection criteria for mesh refinement, the acquisition of high-resolution observational data and the integration of artificial intelligence-hybrid air quality models. More efforts are required to develop the adaptive irregular mesh grid data assimilation method to overcome the resolution problems of current chemical reanalysis.