When undertaking simulations of the urban environment, it is not possible to have advanced knowledge of the values for all parameters and state variables that are required for initialisation. Hence assumptions have to be made in order to proceed. Usually, the impacts of such assumptions are either neglected, or dismissed as being unimportant. As part of the urban model comparison experiment, the Joint UK Land Environment Simulator (JULES) has been run by a number of people with varying initial assumptions for the different parameters. Also, additional simulations have been undertaken which vary some of the state variables with long term memory (such as soil moisture and temperatures). Fluxes from this set of simulations have been compared against observations to assess the spread of results and quantify the uncertainty in the accuracy of the predictions. The results of this analysis will be presented, along with the implications for global simulations of urban areas which include a diverse spread of parameter values.