The predictability of above roof wind resource (three-dimensional mean wind speed) in the roughness and inertial sublayers over idealized urban arrays has been studied using data previously obtained from wind tunnel experiments and a validated large eddy simulation. Both uniform arrays and arrays of heterogeneous heights (representing suburban and more complex urban areas, respectively) were considered. Firstly, the spatial variation of the wind resource within the roughness sublayers over the more complex heterogeneous array was assessed, and suggestions for ideal rooftop turbine placement in this type of geometry were made. It was found that the spatially averaged wind profile can be considered to be a lower bound for the available wind resource at the most viable turbine locations. Secondly, available methods of estimating spatially averaged wind profiles above urban-like surfaces were discussed, and bearing in mind the uncertainties inherent in these methods, a typical wind resource prediction methodology was followed. Significant uncertainties were found to occur at each stage of the prediction, but it was found that the errors occurring because of the typical methods used to parameterize surface aerodynamics are potentially the most significant. Therefore, to increase the accuracy of these prediction methodologies, it is necessary to improve the methods of parameterizing surface aerodynamics and estimating roughness sublayers spatial averages for urban-like surfaces. Copyright © 2011 John Wiley & Sons, Ltd.