Abstract
Kinematic wave models are widely used for rainfall-runoff simulations due to their simplicity and satisfactory accuracy; however, their flow directions are highly dependent on topographic slope, making them sensitive to topographic input. In this study, a novel digital elevation model (DEM) smoothing method, named Ridge-TO-channel Directional TOpography (R2-D2) Smoothing, is developed to reduce this sensitivity. In the R2-D2 Smoothing framework, a given watershed is partitioned into subbasins and smoothing is applied along the lines connecting ridge to channels in each subbasin. A finite element kinematic wave model is used to evaluate the impact of the R2-D2 Smoothing on model sensitivity to topographic inputs. The study is conducted using multiple finite element meshes of the same watershed of equivalent resolution with slightly perturbed horizontal nodal positions, with elevations interpolated from underlying raw, resampled (a typical DEM smoothing technique) and R2-D2 DEMs. Performance is assessed using the coefficient of variation of the Nash-Sutcliffe efficiency (NSE), peak discharge rate, and total discharge volume relative to observations. The R2-D2 Smoothing improved NSE and peak discharge rate variability by about 21% and 74%, respectively, over DEM resampling, and improved total discharge volume variability by about 22% over the raw DEM, while the DEM resampling underperformed. Overall, the R2-D2 DEM produced the most consistent and accurate results across the set of meshes, helping to reduce the sensitivity of the model to topographic input. Finally, the R2-D2 Smoothing is developed as a stand-alone technique with broader application in rainfall-runoff modeling frameworks.

This publication has 17 references indexed in Scilit: