Communities are increasingly affected by excessive heat. The likelihood of extreme heat events is predicted to increase in the Midwest region of the United States. By mid-century (2036–2065), one year out of 10 is projected to have a 5-day period that is 13°F warmer than a comparable earlier period (1976–2005). The frequency of high humidity/dew point days (“extra moist tropical air mass days,” MT++ synoptic climate classification system) has also increased significantly during a similar period (1975–2010) and between 2010 and 2014 included 8 of 26 heat events. This impact is exacerbated by the fact that many residences in low-income neighbourhoods in the US do not have central air-conditioning systems (e.g., up to 50% of low-income homes in Polk County, the location of our study in the US Midwest). Modifications to urban landscapes by the addition of trees can modify temperatures in the nearby environment, which is important for reducing summer heat loads on building surfaces. Trees can reduce energy use and improve indoor and outdoor comfort for cooling in summer by casting shade and providing evapotranspirational (ET) cooling. This paper presents a methodology to combine spatially explicit three-dimensional tree morphology and estimates of ET rates with building location and wall characteristic data to test their relative contribution to building energy consumption. Based on a comprehensive tree inventory for our Midwestern study neighbourhood, tree morphology and building data have been integrated in a three-dimensional array in the “Urban Modeling Interface” (umi) to estimate cooling due to interception of sunlight. We then perform a series of parametric computational fluid dynamics (CFD) studies to simulate ET cooling for various tree morphologies and relative locations to walls. We resolve conventional mesh generation challenges associated with CFD by introducing a novel, immersed boundary framework based on adaptive octree meshes. This approach can seamlessly include trees and buildings at arbitrary locations with minimal human effort. This model was run with and without treesto quantify the relative impact of that process in the