A high frequency deep-tow seismic survey was carried out in the Nankai Trough area in 1996. The objective of the survey was to obtain high resolution seismic sections and velocity profiles of the methane hydrate zone, inferred from the strong BSR events seen on conventional seismic data. A special feature of the survey is that both the source and the streamer cable are towed close to the seabed. This special acquisition geometry requires special data processing to handle the varying source and receiver depths. A CMP floating datum processing sequence was designed which led to high quality sections of the shallow geology. A key step in the processing was devising a residual statics technique to compensate for errors in the measured depths. The processing sequence was applied to a number of lines, totaling 200 km. The quality of the final stack sections was highly variable.There are 2D conventional lines close enough to some of the deep-tow lines to obtain a rough qualitative comparison of the data and to judge the merits and disadvantages of the deep-tow versus conventional airgun data. The main merit of the deep-tow data is a very high resolution image of the very shallow layers, within 300 msec of the seabed and the main disadvantage is that the lack of low frequencies does not allow imaging of the BSR or any deeper events. To obtain a more detailed evaluation of the deep-tow data it is necessary to obtain as precise a comparison as possible with conventional data. The best way to make a precise comparison is using the migrated volume of Tokai-Oki 3D survey, which was acquired in 2002 and processed in 2003. This 3D survey covers most of the deep-tow lines, including the highest quality lines.It was found during the deep-tow processing that the navigation SP positions gave very poor line-ties and that shifts of several hundred meters were needed to obtain good correlation ties. It was similarly found that when the track of the deep-tow lines was extracted from the Tokai-Oki 3D data volume the extracted 2D line profile did not always match the deep-tow profile very well in some places. The positioning problem must be addressed before any meaningful analysis or evaluation of the deep-tow data can be done. We took post-stack positioning analysis and correction. Post-stack positioning analysis was run by correlating the deep tow final stacks with the conventional 3D seismic volume, which was used as a fixed reference. It was anticipated that there would not give enough accuracy for repositioning on all data. In addition, it would never be possible to find absolute shot positions. In this case, we applied pre-stack positioning correction.After making corrections to the deep-tow stacking line positions, it was possible to obtain some detailed comparisons between the deep-tow and the 3D data which can be used to judge the value of deep-tow data. We found some lines produced high quality sections and others, much poorer sections with few interpretable events. Conventional seismic data in the area also shows variation in ...