Testing the sensitivity of the CAESAR-Lisflood landscape evolution model to grid cell size

Abstract

Landscape evolution models (LEMs) are useful for understanding how large-scale processes and perturbations influence the development of the surface of the Earth and other planets. With their increasing sophistication and improvements in computational power, they are finding greater uptake in analyses at finer spatial and temporal scales. For many LEMs, the land surface is represented by a grid of regularly spaced and sized grid cells, or pixels, referred to as a digital elevation model (DEM), yet despite the importance of the DEM to LEM studies, there has been little work to understand the influence of grid cell size (i.e. resolution) on model behaviour. This is despite the choice of grid cell size being arbitrary for many studies, with users needing to balance detail with computational efficiency. Using the Morris method (MM) for global sensitivity analysis, the sensitivity of the CAESAR-Lisflood LEM to the grid cell size is evaluated relative to a set of influential user-defined parameters, showing that it had a similar level of influence as a key hydrological parameter and the choice of sediment transport law. Outputs relating to discharge and sediment yields remained stable across different grid cell sizes until the cells became so large that the representation of the hydrological network degraded. Although total sediment yields remained steady when changing the grid cell sizes, closer analysis revealed that using a coarser grid resulted in it being built up from fewer yet more geomorphically active events, risking outputs that are “the right answer but for the wrong reasons”. These results are important considerations for modellers using LEMs and the methodologies detailed provide solutions to understanding the impacts of modelling choices on outputs.