Closure on the single scattering albedo in the WRF-Chem framework using data from the MILAGRO campaign

Abstract

Data from the MILAGRO field campaign, which took place in the Mexico City Metropolitan Area (MCMA) during March 2006, is used to perform a closure experiment between aerosol chemical properties and aerosol optical properties. Measured aerosol chemical properties, obtained from the MILAGRO T1 site, are fed to two different "chemical to optical properties" modules. One module uses a sectional approach and is identical to that used in the WRF-Chem model, while the other is based on a modal approach. This modal code is employed as an independent check on the WRF-Chem module. Both modules compute aerosol optical properties and, in particular, the single-scattering albedo, ϖ 0, as a function of time. The single-scattering albedos are compared to independent measurements obtained from a photoacoustic spectrometer (PAS). Because chemical measurements of the aerosol coarse mode were not available, and the inlet of the PAS could not ingest aerosols larger than about 2 to 3 μm, we focus here on the fine-mode ϖ 0. At 870 nm, the wavelength of the PAS measurements, the agreement between the computed (modal and WRF-Chem) and observed fine-mode ϖ 0, averaged over the course of the campaign, is reasonably good. The observed ϖ 0 value is 0.77, while for both modules, the calculated value was 0.75 resulting in a difference of 0.02 between observations and both computational approaches. This difference is less than the uncertainty of the observed ϖ 0 values (6%, or 0.05), and therefore "closure" is achieved, at least for mean values. After adjusting some properties of black carbon absorption and mass concentration within plausible uncertainty limits, the two modules simulate well the diurnal variation of ϖ 0, and the absorption coefficient, B abs, but are less successful in calculating the variation of the scattering coefficient, B scat. This difficulty is probably caused by the presence of larger particles during the day when windblown dust is ubiquitous; this dust likely increases the proportion of large particles introduced into the PAS. The dust also contributes to a very large aerosol mass loading in the coarse mode, and neglect of the coarse mode may cause significant errors, estimated to be as large as 0.07, in the calculation and measurement of ambient ϖ 0. Finally, the observed ϖ 0 is compared to the ϖ 0 computed by the full WRF-Chem model, which includes prognostic aerosol chemistry. Unlike the results discussed above, a comparison between observed and simulated ϖ 0 values reveals major differences. This large discrepancy is probably due, in part, to poor characterization of emissions near the T1 site, particularly black carbon emissions.