Modelling catchment water flows at the right scale

Posted on October 15, 2015

Large-scale patterns of catchment water flow are most accurately represented by integrating small-scale processes across space and time. Typically, contemporary hydrological models combine remote sensing data on land cover, land use and topography with soil information, but decisions have to be made about the resolution of the data to be used in model simulations. To investigate how the use of different scales affects the outputs of hydrological modelling, data on soil, surface water and groundwater flows from a small (7.3 km2) catchment in the Susquehanna river basin in Pennsylvania, U.S.A. were combined at a range of grid scales (15-200 metres) and time intervals (5-60 minutes). The data originated from two summer storms, both of which lasted for four hours. Simulations carried out at the smallest resolution produced patterns of flow very close to those observed during the original storms. However, due to the smoothing effect of larger scales, the use of large grids or time intervals had the effect of underestimating the amount of runoff – for example, peak runoff was reduced by about 35% when the grid scale was increased from 50 to 200 m, and by about 19% when the sampling time interval was increased from 15 to 60 min. Also, with large grid scales the runoff peak occurred earlier in a storm event. While these results convincingly highlight the importance of collecting hydrographic information at small scales, the investigators also note that there’s a point beyond which the advantages of further scale reduction become insignificant and computational demands become unnecessarily high.

Reference: Yu, Z. et al. 2014. Spatial and temporal scale effect in simulating hydrologic processes in a watershed. Journal of Hydrologic Engineering
19 (1), January 1, 2014.