This work demonstrates that using thresholds of acceptable error and bias to evaluate meteorological models for air quality modeling is insufficient in a complex topography region. A new meteorological model evaluation technique that quantifies and uses air quality driving features for evaluation is therefore proposed. A Thresholdbased Criterion (TC) evaluation fails frequently in a region of complex topography, Bogota, Colombia. However, a Feature-based Criterion (FC) evaluation reveals that a model can more frequently reproduce air quality driving meteorological features and accurately represent the distribution of these features.

An air quality simulation using the modeled meteorology then demonstrates that air quality in Bogota could be well represented. This is true if local emissions developed for Bogota account for natural mitigation to the re-suspended particulate matter (PM) emissions. Natural mitigation is the suppression of emissions by precipitation, moisture, or dry deposition on nearby surfaces. Natural mitigation leads to a reduction in re-suspended PM to an average of 43 percent of the original emissions estimates.

The TC and FC techniques were then evaluated for their ability to forecast air quality model performance. It was found that there were too few days meeting TC to consider it an indicator of air quality model skill. On the other hand, it was found that accurately predicting meteorological features meaningfully drives modeled air quality conditions. This implies that in regions of complex topography like Bogota, an FC is a more applicable evaluation of meteorological models for air quality purposes.