1. Thermodynamic characterization of Mexico City Aerosol during MILAGRO 2006 Christos Fountoukis 1, Amy Sullivan 2,7, Rodney Weber 2, Timothy VanReken 3,8, Marc Fischer 4, Edith Matías 5, Mireya Moya 5, Delphine Farmer 6, Ronald Cohen 6 and Athanasios Nenes 1,2 1 School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, Atlanta, GA 2 School of Earth & Atmospheric Sciences, Georgia Institute of Technology, Atlanta, GA. 3 National Center for Atmospheric Research, Boulder, CO 4 Environmental Energy Technologies Division, Lawrence Berkeley National Laboratory, Berkeley, CA. 5 Centro de Ciencias de la Atmosfera, Universidad Nacional Autonoma de Mexico, Mexico City, México 6 Department of Chemistry, University of California Berkeley, Berkeley, CA. 7 Now at Department of Atmospheric Science, Colorado State University, CO 8 Now at Laboratory for Atmospheric Research, Department of Civil & Environmental Engineering, Washington State University, Pullman, Washington. Acknowledgments NOAA contract NMRAC000-5-04017, EPA contract X83234201, NSF ATM-0513035, NCAR Advanced Study Program, and NSF ATM-0511829. Introduction At the heart of every air quality simulation is a module for computing the equilibrium composition of aerosol. Understanding the prediction uncertainty from assumptions on phase state and composition is required for effective PM simulations. Observational Data We use fast measurements of aerosol and gas-phase constituents sampled at the T1 site during the MILAGRO 2006 campaign. Particle Into Liquid Sampler (PILS) (Orsini et al., 2003), for PM2.5 ion concentrations. Quantum-cascade laser (QCL) (Fischer et al., 2007), for NH 3(g) Thermal dissociation-laser induced fluorescence (TD-LIF, Farmer et al., 2006; Day et al., 2002), for volatile nitrate (i.e. HNO 3(g) + NH 4 NO 3 ). Ambient temperature (T), pressure and relative humidity (RH). Aerosol particles (PM 2.5 ) were also collected with a cascade micro-orifice uniform deposit impactor (MOUDI), MSP Model 100 (Marple et al., 1991). Preferred Aerosol Phase State For Mexico City aerosol, assess the: Ability of the ISORROPIA-II thermodynamic model (Fountoukis and Nenes, 2007) to predict aerosol composition. Timescale for achieving equilibrium. Importance of explicitly including crustal species in the thermodynamics. Objectives Preferred phase state, either “stable” (solids precipitate out of solution upon saturation), or, “metastable” (aerosol is an aqueous phase regardless of saturation state). ISORROPIA-II vs. Observations The data is classified into 4 “completeness factor” (CF) categories: CF=0 (51% of the data) corresponds to 5- min average measurements of all (gas + particulate phase) species. CF=1 (26% of the data) corresponds to 20- min average measurement of all (gas + particulate phase) species. CF=2 (13% of the data) corresponds to 5-min average measurements, with the PILS nitrate being larger than the TD-LIF HNO 3(g) + NH 4 NO 3. CF=3 (10% of the data) corresponds to 20-min average measurements, with the PILS nitrate being larger than the TD-LIF HNO 3(g) + NH 4 NO 3 Diurnal profile of measured nitrate, ammonium and ambient RH for 27 March 2006. Good agreement between modeling and predictions. Large excess of NH 3(g) drives most Cl, NO 3 into the aerosol phase, so: Small errors in particulate nitrate are magnified in the gas phase. HNO 3(g) exhibits large scatter (Mean Normalized Error ~ 80%), which is however less than the estimated uncertainty (~ 100%). Predicted concentrations of gas phase HCl are low (0-0.3 μg m -3 ). For RH ( 1. This suggests that the “stable” state (solids precipitate out of solution upon saturation) is preferred when [SO 4 ]/[NO 3 ] < 1 and vice versa. Equilibration Timescale Mean Normalized Error (MNE) and Mean Normalized Bias (MNB) do not depend on the CF factor, but only on the averaging timescale. The MNB becomes minimum at ~ 20 min, and suggests this is the equilibration timescale. Three treatments of crustals (Mg, Ca, K) are considered: Explicitly in the ISORROPIA-II thermodynamic calculations Treating crustal species as “equivalent sodium” (i.e., by adding [Na]=[K]+2[Ca]+2[Mg] to the input data) Treating crustals as insoluble. Importance of explicit crustal treatment Mean prediction error and bias for all 3 crustal treatments. Treating crustals as insoluble gives the largest prediction errors and biases. The water uptake is not significantly affected by the crustal treatment assumption; full thermodynamics tend to give the lowest water uptake values. Equivalent sodium differs from the full thermodynamic treatment; the latter tends to give smaller mean errors. This has important implications for the treatment of dust in large-scale models.