Toward Sub-Kilometer-Scale WRF-UCM Modeling of Winter Urban Climate: A Case Study of Ulaanbaatar, Mongolia, on Extreme Local Climate and Thermal Environments

Cities create their own local climate, and numerical weather models can reproduce it if given an accurate picture of the urban surface. The local climate zone (LCZ) framework classifies neighborhoods by building height, density, and materials and provides this information to fine-scale weather models. This approach has mostly been evaluated in warm seasons, leaving open how it performs in the cold, air-stagnant winters of high-latitude cities like Ulaanbaatar, Mongolia. We ran nine model versions over one cold week (22–29 February 2024) across three nested domains (12.5, 2.5, and 0.5 km) with LCZ data resolved to 100 m. Run 8 achieved the highest aggregate validation skill, whereas Run 9 was retained as the configuration most suitable for the LCZ-based analysis rather than as the best model overall. Run 9 reproduced near-surface temperature at the urban Bayanzurkh station (R = 0.89) and gave the smallest wind-speed error (1.5 m s−1), while uniquely resolving the inter-class morphological contrasts required here. The dense urban core proved warmer than the non-urban area by 4.1 °C on average and up to 9.3 °C at night in the compact high-rise zone. Yet this warming barely relieves cold stress: the Universal Thermal Climate Index (UTCI) averaged −15.5 °C across the LCZ classes, within the strong-cold-stress range, with only the compact high-rise zone reaching a milder category. In the low-rise “ger districts”, wind speed rather than air temperature governs perceived cold, and compact high-rise form offers the strongest wind shelter. These findings provide a baseline for future scenario testing of winter thermal exposure in Ulaanbaatar.