The temperature is the important factor that affects the ring laser gyro (RLG) bias. The bias can be divided into two parts, one is the constant bias which has no relation with the temperature and the other is the trend bias related to temperature. The constant bias can be compensated by calibration experiment. The subsection polynomial regression model for the trend bias and the temperature variation has been established based on great many of repeated temperature experiments at the full temperature range from -40 ℃ to +60 ℃. The model has been applied to compensate the high low temperature experimental data the calibration compensation has been carried out. The results showed that the gyro bias has been almost eliminated and could meet the requirements of the real time compensation in the engineering applications. Therefore, this compensation method is valuable to the practical application of RLG.