Urban green spaces, closely tied to local climates, significantly affect human comfort levels, yet existing assessment methods vary in applicability across different contexts and regions. Here, we determined the applicability of two commonly used indices to evaluate human comfort in urban green space types in Hohhot City in China, which is in an arid and semi-arid area. We established sites in four different urban green space types (S1–S4) and a control area (CK) through field-based assessment, and collected meteorological data over 10 days in each season from 2020 to 2021. Specifically, air temperature, relative humidity, and average wind speed were observed from 7:00 to 19:00. Air temperature was highest in summer and lowest in winter. Throughout the day, air temperature first increased and then decreased, with the maximum temperature occurring later in winter than in other seasons. Relative humidity showed an opposite diurnal trend to temperature, and there were no significant differences between urban green space types and CK. The average wind speed of CK was significantly higher than that of the urban green space types. HCI_Lu classifies thermal comfort levels across urban green space types and seasons into four distinct categories as uncomfortable, comfortable to less comfortable, less comfortable, and extremely uncomfortable. HCI_CMA further stratifies thermal conditions at urban green space types by season into cool and refreshing, most comfortable, most comfortable to slightly cool, cold, and uncomfortable. The HCI_Lu ranged from 2.3 to 25.1, and tended to first decrease and then increase on a daily basis. Conversely, HCI_CMA fluctuated throughout the day and ranged from 18.6 to 78.0. According to HCI_Lu, the urban green space types were comfortable for 45% of the observation time, and were comfortable for a greater proportion of time compared to if the comfort was calculated using HCI_CMA. HCI_CMA was strongly correlated with air temperature and average wind speed. According to receiver operating characteristic (ROC) curve analysis, the area under the curve (AUC) for HCI_CMA was 0.59–0.91, and was higher than that of HCI_Lu in each season, indicating greater suitability for the study site.