Parameterization of urban subgrid scale processes in global atmospheric chemistry models

Abstract

We have derived a parameterization consisting of a set of analytical expressions that approximate the predictions by the California Institute of Technology ‐ Carnegie‐Mellon University (CIT) Urban Airshed Model for the net export to the environment (i.e., effective emissions) of several chemical species, as functions of 14 input parameters. For each species, effective emissions are a function of actual urban emissions of this and other species and of other urban domain properties such as meteorology. Effective emissions may be “aged” emissions of primary pollutants or actual production of secondary pollutants. To develop the parameterization we have applied the probabilistic collocation method, which uses the probability density functions of the inputs to generate a set of orthogonal polynomials. These polynomials are then used as the basis for a polynomial chaos expansion that approximates the actual response of the CIT model to its inputs. We assume that seasonal variations can be represented by sinusoidal functions. The parameterization provides a computationally very efficient simulation of the actual model behavior. We have compared the outputs of the parameterization with the outputs of the CIT model, and we conclude that it gives a quite good approximation for effective emissions, at least in the regions of highest probability of the input parameters. This parameterization is applicable to detailed uncertainty and sensitivity analyses and enables computationally efficient inclusion of urban‐scale processes as subgrid scale phenomena in global‐scale models.