There are many examples where regional atmospheric models can be useful for evaluating potential land management and urban development alternatives. Atmospheric models are particularly attractive for urban heat island studies because the influences of specific parameters such as surface albedo can be isolated. In using a mesoscale model in this type of study, there are some fundamental problems surrounding set-up and implementation that need to be addressed. For instance how does one apply a weather model in a climate study? Mesoscale models predict specific instances of weather based on the data with which they are initialized. In a climate study information is desired concerning how atmospheric variables change on the average. Other issues involve model "spin-up", dynamic initialization, and model divergence. Model spin-up is the model run-time required to remove computational features from the initialization data. Initialization data from a baseline simulation may be inappropriate for initialization of a perturbed surface case, so model run-time is necessary for the atmosphere to adjust to the surface modifications. In a technique known as dynamic initialization, observations can be incorporated throughout the initialization period to prevent the model from diverging from the desired modeling scenario. Finally the resolution of the model can have a significant effect on model results. If the finest model resolution is too large, urban features may only be resolved as one homogenous grid cell. Mesoscale models also typically only have a few land classifications of which none may properly describe the surface properties of urban areas.
Each of these issues will be discussed and a methodology for calculating spin-up time will be presented. Simulations are presented for a number of scenarios. The sensitivity of model spin-up time is examined for modifications to several surface parameters including surface temperature, albedo, terrain height, and surface roughness. Current wisdom suggests that only a few hours of spin-up times are necessary, but our results show that in some cases as many as twelve hours are required for full model spin-up. While allowing the model to spin up completely pushes towards longer model simulations, issues associated with model divergence suggest that shorter studies are better. This issue and the relationship between model spin-up and the other factors above are addressed in terms of the design of urban heat island studies.