Over the past decade, the building energy research community has increasingly focused on urban-scale models. The shortcomings of analyzing isolated buildings in an urban context are well-known and far from negligible, mainly due to the inability to account for Urban Heat Island (UHI), shading from neighboring obstructions, and obstructed wind flow. The aim of this paper is to evaluate the impact of the urban context via urban-scale modelling and inverse parameter estimation (calibration) using metered building energy consumption. We describe an automated calibration process for modelling 56 buildings in a representative district in downtown Abu Dhabi (UAE), where a detailed energy audit was conducted with data from 2008 to 2010. Since the urban ambient air temperature could differ significantly from the reference rural air temperature used in most building simulations, the calibration procedure will also consider this UHI effect. Two main approaches of district-wide energy model calibration are proposed using a genetic algorithm and compared to a baseline case where UHI is not considered. The first approach estimates seven building-related parameters together with four microclimate-related variables (describing annual average and seasonal variation of the UHI effect on both air temperature and humidity). The second one uses the Urban Weather Generator (UWG) to pre-process the urban EPW file, thereby reducing the number of the parameters to be estimated. In addition, two approaches are investigated for the calculation of the ASHRAE Guideline 14 calibration error metrics (CvRMSE and NMBE). One approach is to look at the whole district as one aggregate building, while the other, introduced for the first time herein, consists in deriving the weighted-average of the error of each building. The main contribution of this study is to provide simultaneous calibration for multiple buildings in the same district-and subject to the same UHI intensity. Hence, the UHI intensity (urban-rural temperature/humidity differential) is estimated alongside other calibration parameters. For the weighted average approach, CvRMSE found is between 19.09% and 19.40%, while NMBE is between 16.24% and 16.39%. For the aggregated building, CvRMSE is between 2.71% and 4.04%, while NMBE is between 1.95% and 2.35%.