This paper presents a method to evaluate the mapping error present in point cloud terrain maps created using robotic vehicles and range sensors. This work focuses on mapping environments where no a priori ground truth is available and self consistency is the only available check against false artifacts and errors. The proposed error measure is based on a disparity measurement between common sections of the environment that have been imaged multiple times. This disparity measure highlights inconsistency in the terrain map by showing regions where multiple overlapping point clouds do not fit together well. This error measure provides the map interpreter with a localized error measurement to help judge the validity of the final point cloud or gridded surface. It is shown that the proposed method highlights mapping errors more clearly than a principle component based measure used in the 3D modelling community. Results are presented for bathymetric mapping over natural terrain in the deep ocean using a remotely operated vehicle (ROV) outfitted with navigation sensors and a multibeam sonar system.