The fiber optic gyroscope has became to one of the most important sensors in developing due to light in quality, high accuracy, compact in dimension and long life and has played a very important role in both military and civil use. The fiber coil, as one of the most critical components in FOG, is extremely sensitive to changes in temperature. In this paper, at first, by studying the thermal stress of fiber optic gyros, the element model of the fiber coil was built based on the discrete mathematics formulae of Shupe error in FOG. Then based on the temperature distribution model mentioned above, the effects of the Shupe-like bias caused by thermal stress and the Shupe bias caused by temperature gradient are simulated. A turn-by-turn quantization bias error model is established. Theoretical analysis and experimental results show the Shupe-like bias caused by thermal stress and the Shupe bias caused by temperature gradient had seriously affected the temperature performance of FOG. By optimizing the winding method of fiber coil, the Shupe error of fiber coils can be reduced. At the same time, Shupe-like bias caused by thermal stress can be reduced too.