Temperature is directly affected by the earth’s energy process and is an important indicator of climate change, the accurate measurement and the grasp of spatiotemporal distribution of temperature is of great significance for the study of energy process and climate change. Spatial interpolation is an important method to obtain temperature spatial distribution; however, terrain fluctuations affect the interpolation results greatly, which restricts the accuracy of estimated data of spatial temperature. To address such challenge, we performed a case study to test temperature interpolation with and without considering terrain fluctuations in the Loess Plateau with huge elevation variations. Based on the temperature observations from 33 national meteorological stations and Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) of 1 km, differences between temperature values interpolated with and without considering terrain fluctuations were systematically studied at hourly, daily, and monthly scales, respectively. The results show that at these three scales, the interpolation results vary greatly before and after terrain correction, and the difference increases with the increase of elevation. When the elevation is below 1, 000 m, the temperature difference before and after terrain correction is relatively small (the mean value is less than 0.5 ℃), whereas it can reach 3 ℃ when the elevation exceeds 2, 000 m, and 6 ℃ when the elevation is greater than 3, 000 m at monthly scale. Moreover, the influence of elevation fluctuations on interpolation results tends to weaken when the temporal scale changes from hourly to daily and to monthly.