Abstract: With the acceleration of urbanization, urban diseases caused by the rising land surface temperature (LST) are becoming more and more serious. In order to deeply understand the influencing factors of LST in coastal cities and provide scientific data support for improving human health and ecological environment, the single window algorithm is used to inverse the LST of Ganjingzi District, Xigang District, Shahekou District and Zhongshan District in Dalian, China, and multiscale geographical weighted regression (Multiscale-GWR) model combined with normalized difference built-up index (NDBI), normalized difference vegetation index (NDVI), modified normalized difference water index (MNDWI) and normalized difference bareness index (NDBAI) is used to explore the spatial heterogeneity relationship between LST and the underlying surface index. The results show that: (1) the surface temperature of the four districts in Dalian presents a decreasing distribution trend from east to west, the surface temperature in the north of Zhongshan District, Shahekou District and Xigang District is higher than that in the south, and the surface temperature in the southwest of Ganjingzi District is lower than that in other areas. (2) The relationship between surface temperature and underlying surface index in the four districts of Dalian has no global effect, and the spatial heterogeneity is very strong. The Multiscale-GWR model can better fit the surface temperature correlation in the selected underlying surface exponential domain. (3) In terms of correlation coefficient, the impact of underlying surface index on surface temperature is as follows: NDBAI > NDVI > MNDWI > NDBI, and NDBAI, NDVI and MNDWI indexes show negative correlation effect on the whole, while NDBI shows positive correlation effect on the whole.

Key words: multiscale geographical weighted regression, single window algorithm, land surface temperature; underlying surface index, four districts of Dalian