Using Global Space-based Inter-Calibration System (GSICS), the thermal infrared (TIR) channel was calibrated with high precision. During this procedure, a new calibration table was made, and the sensor non-linear effect was corrected. During the TIR channel image remapping and Level2 (L2) dataset generation procedure, the bilinear interpolation was widely used. Most of the L2 data was stored in D/N count with corresponding calibration table, which assumes D/N count is linear. But in the real world, the non-linear D/N count which comes from imprecise modeled A/D transformation of instrument sensor, will lead to the temperature bias on L2 dataset, even though the high precision calibration Look up Table (LUT) was regenerated. In this paper, D/N bias comes from the mapping process was diagnosed, with the consideration of the temperature difference between neighbor pixels.