Modeling Channel Degradation at the Watershed Scale: A Comparison of GWLF, SWAT, and CONCEPTS

TR Number
Journal Title
Journal ISSN
Volume Title
Virginia Tech

In 2005 an assessment of existing Total Maximum Daily Load studies by the U.S. Environmental Protection Agency showed sediment as the fourth leading cause of water quality impairment. A source assessment is important in developing a successful TMDL. Past research efforts have focused on controlling erosion sources in agricultural and urban land areas. New research suggests major contributions to overall sediment loads may be due to stream channel degradation. Monitoring and modeling techniques to assess the contribution of channel sediment to overall sediment load are needed to determine the reductions necessary to meet water quality standards. This research focused on testing the ability of watershed and reach-scale models to predict stream channel degradation. Model predictions were compared to estimates developed from a system of erosion pins and scour chains.

A 500-m experimental reach in Blacksburg, VA, USA, was selected as the focus of channel degradation monitoring and modeling efforts. A series of over 250 erosion pins and seven scour chains were installed systematically throughout the experimental reach. A monthly monitoring program measured channel degradation for the period from July 2005 - June 2006. Point data were interpolated across individual bank segments to produce an estimate of soil erosion volume. Measured soil bulk densities were then used to calculate the estimated mass loading to Stroubles Creek from channel degradation.

Two watershed models and one reach-scale model were developed to predict sediment loading to the stream channel from channel degradation. The Generalized Watershed Loading Function (GWLF) was selected to represent watershed models with limited channel degradation process detail; the Soil and Water Assessment Tool (SWAT) represented the level of channel degradation detail seen in the majority of watershed models; and the CONservation Channel Evolution and Pollutant Transport System (CONCEPTS) reach-scale model was used to evaluate the effectiveness of a detailed process model. Monthly model predictions were compared to retreat rates measured using the erosion pin network.

Sediment loading to the stream from bank retreat was estimated as 41 tonnes/yr, based on erosion pin measurements. GWLF, SWAT, and CONCEPTS predicted stream channel sediment contributions of 8 tonnes/yr, 1500 tonnes/yr and 4 tonnes/yr, respectively. Theil-Sen non-parametric simple linear regression was used to test agreement between monthly model predictions and erosion pin estimates. No significant agreement was found between any model predictions and measured retreat, using a conservative a-value of 0.2. GWLF model predictions underpredicted measured channel degradation, but most closely approximated observed data. This result is likely due to similarities in climate and watershed characteristics for the Stroubles Creek watershed and the Pennsylvania watershed used in the empirical model development. SWAT predicted retreat rates exceeded measured values by two orders of magnitude. This result is explained by the inability of SWAT to predict daily flow and sediment discharge. Highly sensitive channel degradation parameters and the lack of calibration data also contributed to SWAT simulation error. CONCEPTS simulation predicted monthly retreat rates slightly less than GWLF. The lack of agreement between CONCEPTS simulation and observed data was mainly the result of limited input data availability. SWAT daily discharge predictions were used as CONCEPTS input data and likely contributed to poor model agreement. Poor estimation of sensitive sediment input parameters may have also contributed to underpredictions by CONCEPTS. Results showed the potential of screening-level watershed models in channel degradation prediction and the importance of flow and sediment time series discharge data in detailed process-based simulation. The limited flexibility of the GWLF channel degradation algorithm makes it unsuitable for evaluating the effects of stream restoration. SWAT and CONCEPTS should only be used for evaluation if appropriate input data are available.

Future research will focus on the development of a long-term flow and sediment monitoring data set. Few long-term data sets of this nature exist, making channel degradation modeling difficult. Development of long-term data will allow more accurate modeling and better assessment of channel restoration impacts on channel degradation. Further modeling with GWLF in geographic regions outside the Eastern United States is also needed to determine the scope of applicability of the GWLF channel degradation empirical relationship. Additional research should also focus on the significance of subaerial processes for watersheds of various sizes and on the development of algorithms to simulate these processes.

SWAT, CONCEPTS, GWLF, channel erosion