Evaluation of Model Simulated Atmospheric Constituents with Observations in the Factor Projected Space: CMAQ Simulations of SEARCH Measurements

Two-year CMAQ simulations of gases and aerosols over the southeast are evaluated using SEARCH observations for 2000 and 2001, both by direct comparison to observations and by projecting both datasets to the factor space using the Positive Matrix Factorization (PMF) model. Model performance for secondary species (sulfate, ozone) is generally better than for primary species (EC, CO). Nitrate concentrations are overestimated, mainly due to wintertime over-partitioning to the particulate phase. Projecting both observed and simulated constituents to the factor space using PMF, four common factors are resolved for each surface site (two urban sites and two rural sites). The resolved factors include (1) secondary sulfate, (2) secondary nitrate, (3) a fresh motor vehicle factor characterized by EC, OC, CO, NO and NOy, and (4) a mixed factor characterized by EC, OC, and CO. Performance for the sulfate and nitrate factors follow that of the corresponding driving species, while the motor vehicle and ‘‘mixed’’ factors exhibit performance corresponding to that of primary species. Comparing observations and CMAQ simulations in the projected space allow for an evaluation of the co-variability between species, an indicator of source impacts. The fact that similar factors were resolved by PMF from both the observations and the CMAQ simulations suggests that temporal processes related to emissions from specific source categories, as well as the subsequent dispersion and reactivity, are well captured by the CMAQ model. The ability to identify additional factors can be enhanced by adding tracer species in CMAQ simulations. 2008 Elsevier Ltd. All rights reserved.