Simulation of mesoscale uban transport processes with numerical models: guidelines to accurate simulations

Accurate numerical simulations of mesoscale urban dispersion processes requires an understanding of the important urban PBL processes to be modeled, as well as of the mesoscale numerical code and its inputs. Based on 35 years of experience with such models, e.g., URBMET, TVM, MM5, and uMM5, this paper presents suggestions on how such models can produce accurate results. The first suggestions involve model urbanization techniques, while the second involves the order in which to correct model formulations and inputs. Urbanization began with specification of urban surface temperatures to reproduce observed UHIs and now involves modifications of sub-surface, surface, SBL, and PBL formulations. As PBL flows are combinations of synoptic scale forcing and mesoscale thermal circulations, the first output analysis step is determination if synoptic IC/BC capture key observed pressure centers and their temporal changes and if they provide mesoscale features not in the NWS synoptic charts. Model biases in nocturnal min surface temperature are frequently due to inaccurate model diagnostic-formulations of BC deep soil temperatures, while biases in daytime max values are frequently due to inaccurate IC soil moisture profiles. As mesoscale thermal circulations arise from horizontal temperature gradients, it is next necessary to see if IC/BC SST's are correctly specified. If wind speed biases exist even with correct horizontal temperature gradients, input surface roughness length values should be checked. Remaining wind direction biases could be due to horizontal grids too large to resolve complex terrain features. All comparisons should be done with model output and observations correctly adjusted to the same heights. Illustrations will be provided from simulations in wide variety of cities, e.g., San Francisco, Los Angeles, Houston, Atlanta, Jerusalem, NYC, and Phoenix.