Response of simulated sulfur in Models-3/CMAQ to alternate cloud parameterizations

Models-3/CMAQ—the air quality modeling system developed by the U. S. Environmental Protection Agency—simulates atmospheric sulfur chemistry using both homogeneous and heterogeneous reactions. Homogeneous gas-phase oxidation of SO₂ proceeds in response to available sunlight. Heterogeneous reactions in clouds oxidize SO₂ only when clouds and certain oxidants and/or catalysts are present. The atmospheric balance between SO₂ and sulfate affects the lifetimes of both species and influences source-receptor relationships. Tests of CMAQ on a set of four periods covering 32 days revealed that the model possessed a serious bias in cloud cover including large underestimates of afternoon clouds. The mean observed cloud cover fraction, f_c, for clouds below 3.7 km was 0.44. Average CMAQ estimates of f_c were 0.16 for the default 7-layer configuration and 0.35 for the default 29-layer configuration. This bias occurred despite relatively good cloud performance from the meteorological driver (MM5) and in the absence of serious bias in simulated surface variables such as temperature, water vapor mixing ratio and wind speed. Alternate parameterizations and grid structures were tested to determine the extent to which they influence both simulated cloud cover and the sulfate-SO₂ balance. Tests included different grid cell sizes (36 and 12 km), different numbers of vertical layers (7 and 29), alternate limits on cloud base heights, alternate convective cloud initiation options, alternate cloud fraction diagnostic equations, alternate thresholds for diagnosing “resolved” clouds, and a scheme for simulating the effect of subgrid convective moisture enhancement on localized convection. Whereas simulated cloud cover was insensitive to grid cell size, relatively large responses in both modeled cloud cover and sulfur balance were produced from changes in vertical layer structure, changes in cloud base height limits, and changes in the cloud diagnostic scheme. A variety of alternate parameterizations were needed to produce simulation results with only small biases in cloud cover (mean f_c=0.44 for all simulated periods). Modeled sulfate-SO₂ balance proved to be sensitive to simulated cloud cover, with an average ratio of sulfate sulfur to total sulfur of 0.39 (observed = 0.49) using the basic model options and 29 layers versus a ratio of 0.48 following changes in the cloud scheme.