Abstract
Abstract. Etrophication and flooding are perennial problems in agricultural watersheds of the northern Great Plains. A high proportion of annual runoff and nutrient transport occurs with snowmelt in this region. Extensive surface drainage modification, frozen soils, and frequent backwater or ice-damming impacts on flow measurement represent unique challenges to accurately modelling watershed-scale hydrological processes. A physically based, non-calibrated model created using the Cold Regions Hydrological Modelling platform (CRHM) was parameterized to simulate hydrological processes within a low slope, clay soil, and intensively surface drained agricultural watershed. These characteristics are common to most tributaries of the Red River of the north. Analysis of the observed water level records for the study watershed (La Salle River) indicates that ice cover and backwater issues at time of peak flow may impact the accuracy of both modelled and measured streamflows, highlighting the value of evaluating a non-calibrated model in this environment. Simulations best matched the streamflow record in years when peak and annual discharges were equal to or above the medians of 6.7 m3 s−1 and 1.25 × 107 m3, respectively, with an average Nash–Sutcliffe efficiency (NSE) of 0.76. Simulation of low-flow years (below the medians) was more challenging (average NSE < 0), with simulated discharge overestimated by 90 % on average. This result indicates the need for improved understanding of hydrological response in the watershed under drier conditions. Simulation during dry years was improved when infiltration was allowed prior to soil thaw, indicating the potential importance of preferential flow. Representation of in-channel dynamics and travel time under the flooded or ice-jam conditions should also receive attention in further model development efforts. Despite the complexities of the study watershed, simulations of flow for average to high-flow years and other components of the water balance were robust (snow water equivalency (SWE) and soil moisture). A sensitivity analysis of the flow routing model suggests a need for improved understanding of watershed functions under both dry and flooded conditions due to dynamic routing conditions, but overall CRHM is appropriate for simulation of hydrological processes in agricultural watersheds of the Red River. Falsifications of snow sublimation, snow transport, and infiltration to frozen soil processes in the validated base model indicate that these processes were very influential in stream discharge generation.
Highlights
The Red River Basin spans over 122 730 km2 and encompasses portions of Canada and the United States (North Dakota, South Dakota, and Minnesota), with almost three-quarters of its land used for agriculture (Benoy et al, 2016)
The interaction between agricultural management and the hydrology of cold regions remains a topic of ongoing research (Rahman et al, 2014; King et al, 2015) given the need for improved understanding of processes such as prevalence of preferential flowpaths, enhanced hydrological connectivity promoted by drainage, and impact of different cropping systems on runoff generation
Little research addressing specificities of agriculture in cold-region hydrology is available in the literature, this activity is quite relevant in northern latitude regions such as the northern Great Plains (North America; Desaulniers and Gritzner, 2006; Wishart, 2004; Sharp, 1952; Li et al, 2010), northwestern Europe (Scandinavia; Parry et al, 1988), and northern Asia (Wang et al, 2002; Blanke et al, 2007)
Summary
The Red River Basin spans over 122 730 km and encompasses portions of Canada (provinces of Manitoba and Saskatchewan) and the United States (North Dakota, South Dakota, and Minnesota), with almost three-quarters of its land used for agriculture (Benoy et al, 2016). Hydrological models have been utilized at varying spatial scales to model the hydrology of agricultural areas of cold-climate countries such as Finland (Grizzetti et al, 2003; Knisel and Turtola, 2000), Russia (Schierhorn et al, 2014a, b), and Canada (Yang et al, 2014, 2009). Crops represented only 30 % of the land use in a study in Finland using the Soil and Water Assessment Tool (SWAT), making direct connection to agricultural processes more challenging (Grizzetti et al, 2003), while an application of the GLEMS model in that country was done at plot scale (0.11 ha in area), providing detail but making scaling up a challenge (Knisel and Turtola, 2000). SWAT simulations in Russia have focussed on productivity aspects of wheat only rather than hydrological implications (Schierhorn et al, 2014a ,b) and SWAT exercises in Canada have focussed on finer-scale simulation of a specific management practices rather than evaluating process representation (i.e. 14.5 km watershed area; Yang et al, 2009), or were assessed on too coarse (i.e. monthly) a time step for physically based modelling of snowmelt (Yang et al, 2014)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
