Air temperature is the primary indicator of climate change. Reanalysis temperature products are important datasets for temperature estimates over high-elevation areas with few meteorological stations. However, they contain biases in observations, so a bias correction is required to enhance the accuracy of modeling predictions. In this study, we used the temperature lapse-rate method to correct ERA-Interim reanalysis-temperature data in the Qilian Mountains of China from 1979 to 2017. These temperature lapse rates were based on observations (ΓObs) and on model internal vertical lapse rates derived from different ERA-Interim pressure levels (ΓERA). The results showed that the temperature lapse rates in warm periods were larger than those in cold periods. Both the original and corrected ERA-Interim temperature can significantly capture the warming trend exhibited by observations. In general, the temperature lapse rate method was reliable for correcting ERA-interim reanalysis-temperature data. Although ΓObs performed best in bias correction, it depends heavily on the density of ground observation stations and is not appropriate for remote areas with a low data coverage. Correction methods based on ΓERA were shown to be reliable for bias correction, and will be especially applicable to mountainous areas with few observation stations. Our results contribute to the improvement of quality of data products and enhance the accuracy of modeling of climate change effects and risks to the environment and human health.