Towards a Systems Modelling Approach for a Large-Scale Canadian Prairie Watershed

Abstract

The hydrological processes of a large-scale prairie watershed pose a number of challenges for a modeller, and are difficult to parameterize in a hydrological model. Prairie land surface heterogeneity includes wetland hydrology along with dynamic surface storage capacity and cascading arrangements of depressional surface storages or ponds, which eventually lead to a fill-spill type runoff propagation. Fill-spill runoff is movement of excess rainfall or snowmelt from pond to pond, which is different from traditional runoff propagation methods. For this reason, a special parameterization of fill-spill type runoff propagation as well as land surface heterogeneity is required for model development for a prairie watershed. The probability distribution model (PDM) concept has been tested for the prairies in the past and appears to be suitable to parameterize the runoff propagation. Besides the challenging prairie topography, large watersheds may contain lakes and reservoirs that are used for various purposes like water supply, agriculture, and recreation. Water resources management practices at interconnected, controlled and uncontrolled lake systems also pose challenges to the modellers. The effect of lakes may be insignificant for a prairie watershed located in a headwater area, however, the effects of interconnected lake system are significant and require appropriate parameterization to model. The main objectives of this thesis are to investigate the hydrological connectivity in the context of runoff processes and to assess the effect of water resources management practices in the prairie region of Canada. The Qu’Appelle River basin (QRB), located in the Canadian Prairie region, contains a network of multiple controlled and uncontrolled lakes, and hence is a suitable large-scale watershed for this study. The prairie hydrological processes in association with the interactions of multiple lakes create a cascading hydrological system in the QRB and a hybrid modelling approach looks to be a promising methodology to develop a systems model for this watershed. Instead of using a single modelling tool to address this system, two modelling tools are used in this study. MESH (Modelisation Environmentale Communautaire (MEC)—Surface and Hydrology) is a widely used hydrological modelling tool in Canada and contains runoff generation algorithms suited for different types of land surface schemes. For the prairies, an existing runoff generation algorithm was developed using the PDM concept (known as PDMROF). The PDM concept assumes that runoff is a function of dynamic storage capacity, which is represented using a probability distribution function of surface storage capacities. In the direction of simulating prairie runoff processes, there is a scope to improve the parameterization of PDMROF by addressing its limitation of not being able to simulate interflow. An interflow component was added in the parameterization of PDMROF using an approximate solution of Richards’ equation that was found in another existing runoff generation algorithm within MESH named ‘WATROF’. The algorithm developed in this study is known as ‘LATFLOW’, which was compared with the existing runoff generation algorithms used in MESH across three different prairie and non-prairie watersheds with areas ranging from 600 to 2500 km2. Comparison results suggest that LATFLOW performed better than the PDMROF algorithm for simulating streamflow in prairie watersheds and performed also reasonably well in a non-prairie watershed. Owing to the better performance of LATFLOW, it was used in developing a hydrological model for the QRB, which has an area of ~50,000 km2. Due to the presence of multiple lakes within the QRB, a lake system model was developed using the system dynamics (SD) approach to simulate operations and interactions of the lake system of the QRB. This model was developed using measured outflow from the tributaries and lake water levels. The calibrated SD model was able to simulate streamflow and lake levels to a high degree of accuracy, indicating that the lake system of the QRB is suitable for the task under consideration. In order to combine both MESH and the SD model to simulate streamflow in the QRB, two approaches were implemented, which are named as ‘top-down’ and ‘bottom-up’ approaches. In the top-down approach, simulated outflow from MESH models for the headwater tributaries of the QRB were input to the lake SD model to estimate streamflow at the outlet of the QRB near Welby, which is located near the Saskatchewan and Manitoba border. In the bottom-up approach, naturalized streamflow was generated at the outlet of the QRB by using the lake SD model, and the MESH parameters for the QRB were estimated considering the naturalized hydrological system of the QRB. Results indicate that simulated streamflow is underestimated for some tributaries, but the timing of peak flows is captured. The hybridization of the hydrological model and lake SD model is considered as a viable option when handling watersheds such as QRB where interactions between lakes play an important role in association with complex prairie hydrological processes. Investigations were also carried out to identify the effect of the choice of the calibration period on the model performance, and the results indicate that the QRB exhibits different streamflow patterns at a decadal scale and efficient modelling requires further knowledge about incorporating these patterns within the modelling framework.