Abstract: A method for constructing an adaptive grid digital depth model based on mean vertical uncertainty of area is proposed. Depth uncertainty and its propagation mechanism is used for estimating the mean uncertainty of depths in the modeled area, levels of grid partition are determined automatically and depths are organized effectively; constraint conditions at common nodes of adjacent areas are added for making the depth model continuous. Experimental results demonstrate: (1) in the area where the depth change trend is not consistent, the method based on mean vertical uncertainty of area has the higher quality than the method with equal spacing and the method considering depth complexity; (2) The method based on mean vertical uncertainty of area is able to determine the grid spacing automatically and make the model quality meet the design requirements while the regular grid method is not; The method based on mean vertical uncertainty of area is more adaptive than the method considering depth complexity. (3) The method based on mean vertical uncertainty of area is more automatic and efficient than the method considering depth complexity.