Cities adjacent to large water bodies are influenced by intricate interactions between urban-scale processes and land-water dynamics, resulting in significant impacts on climate and its intra-urban variability. In this study, an ensemble of simulations using the Weather Research and Forecasting (WRF) model, incorporating urban canopy representations, is employed to isolate the interplay between urbanization and Lake Michigan on the overall climate of Chicago with a focus on heat exposure and its disparities across the city’s community areas. While prior similar investigations predominantly focused on air and occasionally skin temperatures, we also consider various indices and commonly used proxies of heat stress, including wet-bulb and wet-bulb globe temperature, to understand the influence of moisture and wind flow from the adjacent water body on urban heat exposure. Our results show substantial mitigation of daytime heat exposure and heat stress due to the presence of Lake Michigan. Conversely, urbanization increases heat stress at night. Furthermore, the influences of urbanization and the lake on temperature, particularly skin temperature, including its extreme values and associated gradients from the lake to the land, are much higher than their corresponding effects on heat stress, partially attributed to moisture feedback mechanisms. Our analysis also reveals a marked difference in environmental conditions depending on variable chosen. For instance, skin temperature across community areas shows a gradient of -1.29 °C per $10,000 increment in median income per capita. In contrast, disparities in wet-bulb globe temperature are relatively milder at -0.23 for each $10,000 increase in income. These results emphasize the significance of adopting physiologically relevant heat exposure metrics when aiming to accurately capture the implications of urbanization on public health. Finally, we use several different model configurations, including various number of nested domains, multiple boundary-layer schemes, and different resolutions, to check the consistency of the results. We find consistent results during daytime, but higher variability between configurations at night. This poorer and less consistent performance being also reflected in evaluations against satellite and crowdsourced observations suggest more improvements are needed to capture realistic urban-scale processes over this region at night.