As the numerical weather prediction community seeks deeper understanding of multi-scale interactions among the atmosphere, human systems and the overall earth system, more explicit representation of surface terrain in these models has become necessary. While a great body of work has examined the differences in error and uncertainty of simulations at various horizontal grid resolutions, no studies have been performed that compare the results of running the models at the same horizontal grid resolution but with different resolutions of embedded urban neighborhood morphology. We examine the differences in meteorological output from the Weather Research and Forecasting (WRF) model run at 270m horizontal resolution using 10m, 100m and 1km resolution neighborhood morphological inputs and with no morphological inputs. We find that horizontal resolution differences in urban morphological inputs to numerical weather models result in model output differences, especially in the spatial variability of meteorological parameters. We show that these results affect how we determine the heterogeneity of human vulnerability to a heat wave across an urban area.