: Accuracy assessment of tide models in polar ocean has to be performed to accurately analyze tidal response of glaciers by using Double-Differential Interferometric SAR (DDInSAR) technique. In this study, we used 120 DDInSAR images generated from 16 one-day tandem COSMOSkyMed DInSAR pairs obtained for 2 years and in situ tide height for 11 days measured by a pressure type wave recorder to assess the accuracy of tide models such as TPXO7.1, FES2004, CATS2008a and Ross_Inv in Terra Nova Bay, East Antarctica. Firstly, we compared the double-differential tide height (Δ . T) for Campbell Glacier Tongue extracted from the DDInSAR images with that predicted by the tide models. Tide height (T) from in situ measurement was compared to that of the tide models. We also compared 24-hours difference of tide height ( . T) from in situ tide height with that from the tide models. The root mean square error (RMSE) of Δ .
T, T and .T decreased after the inverse barometer effect (IBE)-correction of the tide models, from which we confirmed that the IBE of tide models should be corrected requisitely. The RMSE of .T and Δ .T were smaller than that of T. This was because . T is the difference of tide height during temporal baseline of the DInSAR pairs (24 hours), in which the errors from mean sea level of the tide models and in situ tide, and the tide constituents of S2, K2, K1 and P1 used in the tide models were canceled. This confirmed that .T and Δ .T predicted by the IBE-corrected tide models can be used in DDInSAR technique. It was difficult to select an optimum tide model for DDInSAR in Terra Nova Bay by using in situ tide height measured in a short period. However, we could confirm that Ross_Inv is the optimum tide model as it showed the smallest RMSE of 4.1 cm by accuracy assessment using the DDInSAR images.