Development and evaluation of processes affecting simulation of diel fine particulate matter variation in the GEOS-Chem model

Abstract. The capability of chemical transport models to represent fine particulate matter (PM_2.5) over the course of a day is of vital importance for air quality simulation and assessment. In this work, we used the nested GEOS-Chem model at 0.25° × 0.3125° resolution to simulate the diel (24 h) variation in PM_2.5 mass concentrations over the United States (US) in 2016. We evaluate the simulations with in situ measurements from a national monitoring network. Our base case simulation broadly reproduces the observed morning peak, afternoon dip and evening peak of PM_2.5, matching the timings of these features within 1–3 hours. However, the simulated PM_2.5 diel amplitude in our base case was 105 % biased high relative to observations. We find that temporal resolution of emissions, differences in vertical representativeness between model and observations, as well as boundary layer mixing are the major causes for this inconsistency. We applied an hourly anthropogenic emission inventory and converted the PM_2.5 masses from model level center to the height of surface measurements by correcting for aerodynamic resistance. The biases in the PM_2.5 diel amplitude were reduced to 25 % in the improved simulation and the timing of diel variations were better captured. In addition, notable sensitivity of the simulated diel amplitude of PM_2.5 (8 %) on the boundary layer height in the driving met fields were identified. Based on the improved model, we find that the diel variation in PM_2.5 is driven by 1) building up of PM_2.5 in early morning due to increasing anthropogenic emissions into a shallow mixed layer, 2) decreasing PM_2.5 from mid-morning through afternoon associated with mixed layer growth, 3) increasing PM_2.5 from mid-afternoon though evening as emissions persist into a collapsing mixed layer, and 4) decreasing PM_2.5 overnight as emissions diminish.