This paper presents a study to test the ability of the current generation of landscape evolution models to correctly predict landscape form from measured erosion processes. Landscapes generated by the SIBERIA landscape evolution model were compared to experimental model landscapes in declining equilibrium. The model simulations were compared using geomorphologically and hydrologically significant parameters. Comparisons of the simulated landscapes with the experimental landscapes were carried out using the hypsometric curve, width function, cumulative area distribution, and area-slope relationship. These comparisons demonstrate that SIBERIA can correctly simulate the experimental model landscape at declining equilibrium. The simulation showed sensitivity to the spatial distribution of the rainfall, particularly with respect to hypsometry. Using the correct measured distribution of rainfall was necessary rather than using a spatially uniform rainfall distribution. The results also highlighted the importance of digital terrain map (DTM) error in deriving geomorphic statistics. The observed landform had a consistently larger width function than the simulation. Only when the simulations were corrupted with elevation errors with statistical properties of the errors in the experimental landscape DTM, as measured by photogrammetry, did the observed and simulated width functions match. ¿ 2001 American Geophysical Union