Sensitivity of WRF-Chem simulations to chemical boundary conditions, domain configuration and nesting options: An evaluation for an Andean city.

Abstract:

Regional air quality studies in zones of complex topography require high-resolution simulations to capture the fine details in topography, land use and emission patterns. Furthermore, adequate meteorological and chemical boundary conditions (CBC) are essential to obtain accurate predictions. However, CBCs are usually derived from coarse-resolution global models; hence, high-resolution studies require nested simulation to downscale the global model outputs into data with suitable spatial and temporal resolution. In this work, the WRF-Chem model was used to evaluate the sensitivity of O₃ and CO predictions to: 1) Domain configurations (three nested domains with 5:1 nesting ratio vs four nested domains with 3:1 nesting ratio) 2) changes in chemical boundary conditions (default WRF-Chem CBC vs derived from CAM-Chem outputs), and nesting options (with and without feedback among nested domains). Model simulations were conducted over the Andean city of Manizales, Colombia, characterized by a complex topography. The simulation period included 28 days with low precipitation events in 2018. Predictions of O₃, CO, and meteorological variables (T, RH, WS, WD) were compared with surface measurements. Results show a strong sensitivity to the CBC options. The use of CAM-Chem CBC reduced mean bias (27%) and RMSE (11%) for CO predictions. Likewise, a reduction in mean bias (18%) for O₃ prediction was obtained. On the contrary, the model did not exhibit a strong sensitivity to domain configurations. The simulations with four nested domains allowed slightly more accurate representations of CO and some meteorological variables but at expenses of an increase in computational time of approximately 20%, compared to the three-nested-domains simulation. Finally, model predictions were highly sensitive to the nesting options, where the 2-way nesting simulation provided the best results for O₃ concentrations but higher errors in CO forecast. Overall, these results highlight the importance of using adequate boundary conditions and downscaling procedures for improving model performance. This study can guide WRF-Chem setup for future air quality simulations over areas with a complex topography and provide the baseline for future sensitivity analysis assessing the impact of chemical mechanisms and emission inventories.