The planned 2003 Mars Exploration Rover (MER) Mission and follow-on surface activities associated with landed missions will focus on long distance roving and sample return, which require detailed knowledge of vehicle locations in both local and global reference systems. In this paper we argue that this rover localization should be known to within 0.1 of the distance traversed for local coordinate systems. To test the ability to meet this goal using only descent and rover-based data, we conducted experiments with simulated descent images and Field Integrated Design and Operations Rover data collected during field tests at Silver Lake, California, in April 1999. Specifically, an integrated bundle adjustment system incorporating both descent and rover-based images was developed and used to localize the rover positions. On the basis of surveyed ground control points it is demonstrated that the joint analysis produces RMS errors of 0.24, 0.15, and 0.38 m in x, y, and z% directions in a local coordinate system, respectively, for ground points within 500 m from the landing point and 0.23, 0.21, and 0.46 m within a distance of 1.5 km. Results show that it is possible to meet the 0.1 goal using descent and rover-based data only.