A description and evaluation of an air quality model nested within global and regional composition-climate models using MetUM
Abstract. There is a clear need for the development of modelling frameworks for both climate change and air quality to help inform policies for addressing these issues. This paper presents an initial attempt to develop a single modelling framework, by introducing a greater degree of consistency in the modelling framework by using a two-step, one-way nested configuration of models, from a global composition-climate model (GCCM) (140 km resolution) to a regional composition-climate model covering Europe (RCCM) (50 km resolution) and finally to a high (12 km) resolution model over the UK (AQUM). The latter model is used to produce routine air quality forecasts for the UK. All three models are based on the Met Office's Unified Model (MetUM). In order to better understand the impact of resolution on the downscaling of projections of future climate and air quality, we have used this nest of models to simulate a five year period using present-day emissions and under present-day climate conditions. We also consider the impact of running the higher resolution model with higher spatial resolution emissions, rather than simply regridding emissions from the RCCM. We present an evaluation of the models compared to in situ air quality observations over the UK, plus a comparison against an independent 1 km resolution gridded dataset, derived from a combination of modelling and observations. We show that using a high resolution model over the UK has some benefits in improving air quality modelling, but that the use of higher spatial resolution emissions is important to capture local variations in concentrations, particularly for primary pollutants such as nitrogen dioxide and sulphur dioxide. For secondary pollutants such as ozone and the secondary component of PM10, the benefits of a higher resolution nested model are more limited and reasons for this are discussed. This study confirms that the resolution of models is not the only factor in determining model performance - consistency between nested models is also important.