Impacts of urban forests and landscape characteristics on land surface temperature in two urban agglomeration areas of China.

• An IEU-Net model and high-resolution images are reliable for urban forest mapping. • Afforestation and deforestation exerted a net cooling and warming. • Two landscape-type conversions displayed greater cooling and weaker warming. • A robust and accurate land surface temperature feedback framework was proposed. Accurate characterization of urban forest change can help quantify its impact on the urban thermal environment. Taking Hangzhou City and Zhaoqing City, China as two cases, we proposed a deep learning-based approach to extract forest/non-forest types using China's Gaofen data and the labels it generates. We then explored the impacts of urban forest patterns on land surface temperature (LST). Our results indicate that Gaofen-1/6 images performed well for urban forest/non-forest mapping with an accuracy of over 90%. We found that afforestation and deforestation exerted a net cooling of -0.62±0.30°C and -1.23±0.26°C and warming of 1.71±0.16°C and 1.29±0.14°C in Hangzhou and Zhaoqing, respectively. Balancing the LST changes of both afforestation and deforestation found total warming due to deforestation, especially in Hangzhou. Latitudinal reduced cooling and enhanced warming were found from Zhaoqing to Hangzhou, which can be explained by strong evaporative effects and inconsistent albedo effects. Moreover, a more pronounced cooling when other types converted to landscape core type and a weaker warming when landscape core type converted to other types were detected, especially in Zhaoqing. This study highlights the importance of designing rational afforestation plans, avoiding deforestation activities, and reducing forest fragmentation in delineating effective climate mitigation strategies. [ABSTRACT FROM AUTHOR]