Simulating Hydrologic Effects of Raised Roads within a Low-Relief Watershed

Raised roads and their appurtenances (i.e., ditches and culverts) can considerably modify the natural hydrologic processes (e.g., flow direction and runoff spatial distribution) of low-relief watersheds. However, these effects are not well understood and are poorly documented in literature. Also, there is a serious lack of effective approach that can be used to describe these effects in common simulation models, including the Soil and Water Assessment Tool (SWAT). As a result, these effects have been rarely taken into account by the existing model applications. The results from those applications are likely to be misleading for watershed management decisions. The objectives of this study were to (1) develop an approach that enables the incorporation of the aforementioned effects into hydrologic models (e.g., SWAT), and (2) evaluate whether and how a model’s simulation performance can be improved by taking these effects into account. The evaluation was conducted for the low-relief Wild Rice River watershed located in northwestern Minnesota by comparing the SWAT-NoRoad model, which is based on a “conventional flow direction map” derived using the deterministic eight-neighbor (D8) algorithm, with the SWAT-Road model, which is based on a “corrected flow direction map” generated by using the approach developed in this study. The corrected map provided more realistic flow directions for the regions adjacent to the raised roads than the conventional map. Compared with the SWAT-NoRoad model, the SWAT-Road model had a compatible performance for a calibration period but was more robust and did a better job in reproducing the observed streamflows for a validation period at two U.S. Geological Survey gauging stations, as indicated by the larger values for Nash-Sutcliffe ($E_j^2>0.60$) and determination ($R^2>0.68$) coefficients, as well as from the watershed perspective view, as indicated by the larger values for performance virtue (${\mathrm{PV}}_k>0.64$). In addition, the spatial patterns of water yield predicted by the SWAT-Road model were more accurate because they are more similar to those of the topographic wet index values than the spatial patterns predicted by the SWAT-NoRoad model. Furthermore, if the raised roads were neglected, the mean daily water yields for the study area would be erroneously predicted by up to 0.6 mm. A reasonable generalization of this study is that hydrologic effects of raised roads are important for accurately simulating watersheds with a low topographic relief and thus need to be taken into account.