In this paper, we present a study of temperature effects on BGO calorimeters using proton MIPs collected in the first year of operation of DAMPE. By directly comparing MIP calibration constants used by the DAMPE data production pipe line, we find an experimental relation between the temperature and signal amplitudes of each BGO bar: a general deviation of -1.162%/degrees C, and -0.47%/degrees C to -1.60%/degrees C statistically for each detector element. During 2016, DAMPE's temperature changed by similar to 8 degrees C due to solar elevation angle, and the corresponding energy scale bias is about 9%. By frequent MIP calibration operation, this kind of bias is eliminated to an acceptable value.