مقایسه تطبیقی رگرسیونهایخطیچندگانهو جنگل تصادفی درتخمین متوسطدمای سطح زمین: مطالعه موردی شهر تبریز.

Land surface temperature, as one of the important and fundamental parameters in climatology, indicates the relationship between the atmosphere and the Earth. Considering the environmental issues of cities, including the intensification of urban heat islands, accurately estimating LST and identifying its influencing factors play a significant role in urban thermal management and adopting adaptive strategies for heat islands. In this regard, this study compares two regression methods: multiple linear regression and random forest in order to estimate the LST. Daily nighttime MODIS images were used to extract the LST of Tabriz city during the summer. These images were processed in the Google Earth Engine platform and averaged for the period from 2018 to 2022. According to the results, the random forest showed significantly better performance with a coefficient of determination of 0.924 (RMS = 0.009) compared to multiple linear regression. The random forest was also used to determine the importance of the indices. Based on the index importance results, night lights (51/06%), sky view factor (48/01%) and frontal area index (45/27%) were the most important factor affecting the nighttime summer LST in Tabriz city, respectively. The findings of this study, in addition to revealing the strength of the random forest regression in estimating LST, also highlight the importance of various indices in the LST. In this context, the study's results will be practical for managing the thermal environment of Tabriz city and adopting mitigation strategies for its heat islands. [ABSTRACT FROM AUTHOR]