Rising urban temperatures and heat extremes pose an urgent global challenge, yet the potential for mitigating excessive urban heat-particularly at the global scale-remains unclear. Here, we quantify the cooling potential across 2,265 cities worldwide by the 2050s using validated urban climate simulations. Cooling effects are quantified as the reduction in the summer average wet-bulb globe temperature (WBGT) and heat danger hours (HDH; WBGT > 31.4 °C) under the combined implementation of reflective surfaces, green transformation, and anthropogenic heat reduction. We show a distinct spatial asymmetry: while the cooling potential increases with latitude, primarily due to greater cooling from reflective surfaces, the highest heat risk is concentrated in low- to mid-latitude regions (10°N-40°N). In these high-risk regions, combined mitigation is more effective at night, reducing HDH by an average of 37%, whereas daytime heat is mitigated to a lesser extent (11%). These asymmetries underscore the need for context-specific strategies-particularly accelerated action and localized innovation for low-latitude humid regions-as well as the integration of city-scale planning with targeted daytime heat risk interventions.

This study presents a comparative analysis of two urban microclimate models for assessing pedestrian thermal comfort: the simplified model SOLWEIG and the high-fidelity CFD-based urbanMicroclimateFoam (uMFoam). Thermal comfort is evaluated during a heatwave in a neighborhood in Geneva using the Universal Thermal Climate Index (UTCI). Significant discrepancies are found between the models, reaching up to 30 °C in mean radiant temperature (MRT) and 10 °C in UTCI. To systematically identify the sources of these differences, a stepwise methodology is developed to isolate the effects of static data, environmental forcing, and MRT calculation method. First, SOLWEIG’s point-cloud-based static inputs are replaced with vector-based data consistent with uMFoam, reducing differences in vegetated areas and highlighting the role of accurate tree representation, particularly its influence on shading. Next, SOLWEIG’s environmental inputs (air temperature, relative humidity, and wind speed) from WRF are replaced with high-resolution outputs from uMFoam, reducing UTCI discrepancies and emphasizing the importance of accurately resolving environmental variables. Nevertheless, substantial differences persist, primarily due to simplifications in the MRT calculation linked to SOLWEIG’s surface temperature model. A novel consensus clustering approach is introduced to identify coherent temporal behavior and quantify spatial agreement between the models. Overall, while SOLWEIG enables efficient simulations over large urban domains at lower accuracy, uMFoam provides more physically consistent and accurate predictions but at higher computational cost. The results highlight the importance of resolving airflow, heat, and moisture transport for reliable thermal comfort assessment, while emphasizing the trade-off between computational efficiency and physical fidelity.