General Considerations:Grid Size
Distributed models are used over a very wide range of grid sizes, from 10 to 1000 m. The selection of an appropriate grid size for a GSSHA model requires consideration of both the available data and computational effort required. Typical grid sizes range from 10 to 250 m. The selection of the grid size for a given watershed determines the total number of grid cells used to describe the watershed, setting the computational effort and memory required. Note that if the grid size is halved, the memory required and computational time increase by a factor of 4. In general, smaller grids are less sensitive to sub-grid variability for Hortonian runoff (Ogden and Julien, 1993). Therefore, smaller grids are generally “better”, if the data exists to assign relevant watershed characteristics to each grid cell. However, smaller grid-sizes do not guarantee superior model performance. As with the time step, the most appropriate grid size can be determined with a convergence study, where the effects of increasing or decreasing the grid size can be observed on model output.
- 4 General