Abstract: To examine the influence of landscape pattern on the urban heat island effect for Xi’an, China, landscape classification and temperature inversion were performed using 2000, 2006, and 2015 remote sensing datasets, and the regression model was established. The results showed that correlations between landscape indices and temperature differed according to spatial scale. On the landscape scale, the landscape shape index (LSI), landscape division index (DIVISION), and Shannon’s diversity index (SHDI) were significantly correlated with temperature. On the class scale, CA₁, PD₁, LSI₁, AI₁, LPI₂, AI₂, CA₃, DIVISION₃, LPI₃, AI₄(where “1” indicated built-up land, “2” indicated cropland, “3” indicated forestland, and “4” indicated grassland; where CA referred to class area, PD referred to patch density, LSI referred to landscape shape index, AI referred to aggregation index, DIVISION referred to landscape division index, and LPI referred to largest patch index) were significantly correlated with temperature. The landscape pattern was the main factor affecting the urban heat island, and there was a strong response for this effect. Lots of the landscape indices could represent surface temperature, and modifying the configuration of landscapes could have significant long-term effect on reducing the urban heat island effect. The prediction of multiple linear regression models established on different scales showed that the heat island effect was lower in 2030 than in 2015, but the decrease was not significant, and the effect was still stronger than that in 2006. The heat island effect first spread outward from the main city by point-to-area expansion, then it was weakened in the main city and increased in the surrounding area (counties).