Abstract: In order to overcome the shortcomings of using ships to measure sea-depth with low efficiency and high cost, the authors selected the Mariana Trench area formed by the Western Pacific plate dive to the Philippine plate as the experimental area, and used gravity anomalies data to inverse the sea-depth according to the gravity-geologic method. The density difference constant of the GGM (gravity-geologic method) model was determined to be by using the remove-restore technique. The differences between the GGM sea depth model, the ETOPO1 model and the model 1 which directly gridded dispersed ship data were analyzed and compared, and the landforms of the "challenger abyss" trench were studied based on the inversion results. The result shows that the seafloor topography computed by GGM model is superior to that by the ETOPO1 model, and even better than the results of the directly gridding of the ship survey data. The relative error of the model is not related to sea depth, but affected obviously by the change of seafloor topography. The sea-depth differences between the models are large in the area where the ship data are missing or too sparse. The landforms on the both sides of the "challenger abyss" trench are different. In the south side, when the water depth is less than 5 000 m, the gentle slope is about 2°-5°; while the water depth is greater than 5 000 m, the slope is increased significantly to about 10°-15°. In the north side, the slope is about 10°-15°. There is a buffer zone near latitude.

Key words: gravity-geologic method, density difference contrast, downward continuation, remove-restore technique, Mariana trench