Climate warming and urbanisation compound the public health risk posed by heat. Heat can be mitigated at local scales through urban greening, which provides shade and reduces surface and air temperatures. Yet, the relative effectiveness of different greening options on human thermal comfort based on physiology-based indices is understudied. We installed microclimate stations at 17 locations covering a gradient of tree canopy cover and perviousness in the city of Ghent, Belgium, and monitored the modified Physiologically Equivalent Temperature (mPET) during 195 days over Spring and Summer. We assessed the canopy cover, pervious surface fraction and building sky fraction based on field measures and hemispherical pictures. Unpaved locations with trees experienced a 2.4-fold reduction in the number of days with strong heat stress (mPET > 35 °C) compared to paved, treeless locations. Based on mixed models and our selected environmental variables, cooling effects were predominantly driven by tree canopy cover, where locations with 100% canopy cover had temperature maxima 5.5 °C mPET lower than treeless locations throughout the monitoring period. When air temperatures rose to 40 °C, cooling by tree canopies increased to 8.8 °C mPET. The pervious surface fraction and building view factor were less influential, generating variation of at most 1.7 °C and 1.1 °C mPET, respectively. In contrast, night-time temperatures were rather determined by the regional-scale urban heat island effect than by aforementioned local factors. Still, tree canopies slightly cooled the warmest nights only, whereas the vicinity of buildings led up to 1.2 °C mPET warming on average. Expanding the urban tree cover may therefore be the best solution for improving local thermal comfort levels when daytime heat peaks, but will provide little relief at night.