Testing and calibrating a landscape evolution model using a natural experiment in post-glacial landscape evolution

Gregory E. Tucker (CIRES and Department of Geological Sciences, University of Colorado, Boulder, CO, USA)


Landscape evolution models have shown great promise for generating insight into potential geomorphic responses to climate change. However, an important weakness of such models is the difficulty in testing and calibrating them over time and space scales relevant to landform change. Here we report results of a model testing and calibration exercise that takes advantage of a natural experiment in rapid post-glacial landscape change in a ~100km2 drainage basin in western New York state. The unique terrain, geology, and glacial history of the study site make it possible to reconstruct the topography as it existed shortly after glacial retreat. The retreating ice left behind a valley network filled with a thick sequence of till and inter-stadial sedimentary units. The tops of these deposits form a low-relief plateau surface along the axis of the main valley. Following glacial retreat, this surface was incised to form a network of gullies and canyons up to several tens of meters deep. Optically stimulated luminescence (OSL) dates on fluvial sediments near the top of the glacial plateau and on strath terraces bracket the onset of incision, and provide an indication of post-glacial incision rates. These constraints make it possible to test the ability of a landscape evolution model to reproduce the modern topography given the paleo-topography as an initial condition. We used this approach to test and calibrate the Channel-Hillslope Integrated Landscape Development (CHILD) model, a physically based numerical model of fluvial landform evolution. The model was driven with a stationary climate and a range of baselevel histories consistent with OSL dates. To calibrate the model, a series of Monte Carlo simulations were computed, and the results were compared with the mainstream longitudinal profile, the positions of two well-dated strath terraces, and several statistical measures of terrain and drainage geometry. Best-fitting simulations provided a good match with the modern topography in terms of incision depth, long profile shape, and drainage patterns. As expected, the calibration process revealed tradeoffs among parameters, such that several alternative parameter sets provide essentially equal fits to the data. The results demonstrate that the model's process laws are sufficient to account for the observed patterns of post-glacial landform change. The results also illustrate the potential for using natural experiments like this one to test models of landscape evolution and to determine the necessary and sufficient combination of process laws, boundary conditions, and parameters required to explain millennial-scale patterns of landform development.