Evaluation of a Eulerian and Lagrangian air quality model using perfluorocarbon tracers released in Texas for the BRAVO haze study

The Big Bend Regional Aerosol and Visibility Observational (BRAVO) study was designed to determine the sources of haze at Big Bend National Park, Texas, using a combination of source and receptor models. BRAVO included an intensive monitoring campaign from July to October 1999 that included the release of perfluorocarbon tracers from four locations at distances 230–750 km from Big Bend and measured at 24 sites. The tracer measurements near Big Bend were used to evaluate the dispersion mechanisms in the REMSAD Eulerian model and the CAPITA Monte Carlo (CMC) Lagrangian model used in BRAVO. Both models used 36 km MM5 wind fields as input. The CMC model also used a combination of routinely available 80 and 190 km wind fields from the National Weather Serviceʹs National Centers for Environmental Prediction (NCEP) as input. A modelʹs performance is limited by inherent uncertainties due to errors in the tracer concentrations and a modelʹs inability to simulate sub-resolution variability. A range in the inherent uncertainty was estimated by comparing tracer data at nearby monitoring sites. It was found that the REMSAD and CMC models, using the MM5 wind field, produced performance statistics generally within this inherent uncertainty. The CMC simulation using the NCEP wind fields could reproduce the timing of tracer impacts at Big Bend, but not the concentration values, due to a systematic underestimation. It appears that the underestimation was partly due to excessive vertical dilution from high mixing depths. The model simulations were more sensitive to the input wind fields than the models’ different dispersion mechanisms. Comparisons of REMSAD to CMC tracer simulations using the MM5 wind fields had correlations between 0.75 and 0.82, depending on the tracer, but the tracer simulations using the two wind fields in the CMC model had correlations between 0.37 and 0.5.