In Pursuit of Improving Peak Flow Prediction in the Canadian Prairies

Abstract

The prairies were subjected to multiple unprecedented floods over the past decade that caused major damages to agricultural and residential areas. Accurate prediction of the magnitude and timing of floods is important as it is an essential component of flood risk management programs. However, the accuracy of predicting floods and the associated flooding extents have not drawn much attention in the prairies due to difficulties in predicting prairie streamflow in general. Such difficulties are caused, mainly, by the limitations of the currently available modeling approaches in handling the pothole complexities – a dominant feature in prairie watersheds. This thesis focuses on improving the prediction of floods (peak flows), in particular, and the streamflow in general, along with the associated landscape pluvial and nival flooding extents that frequently occur in the complex pothole-dominated environment of the Canadian prairies. This aim is achieved through adapting/developing a set of models that are built and tested for the prairies to contribute to solving the flood prediction problem in the prairies. The first model is a new Hydrological model for the Prairie Region (HYPR), which is proposed as an engineering solution for the prediction of the flood peak in the prairies. HYPR is a modified version of the HBV model, developed by coupling the conceptual HBV model, for hydrological processes representation, and the Probability Distribution Model based RunOFf generation algorithm (PDMROF) for pothole representation. The second model is a novel Prairie Region Inundation MApping model (PRIMA), which is developed as a distributed hydrologic routing model for more accurate and comprehensive storage dynamics simulation and inundation mapping in the prairies. PRIMA uses a set of rules along with Manning’s equation (iteratively) to route the water over the landscape. The third model is the Modelisation Environmentale Communautaire (MEC)—Surface and Hydrology (MESH), which is modified by coupling it with PRIMA to improve the non-contributing area and potholes dynamic representation in complex land surface models for better prediction of peak flows and the associated flooding extents. In this model, called MESH-PRIMA, MESH handles the vertical fluxes calculations based on physically based equations and PRIMA routes the water over the landscape and accounts for the effect of potholes on changing the net runoff reaching the stream network. HYPR shows good simulation of the overall hydrograph and peak flows, on a daily resolution, as indicated by the Nash-Sutcliffe Efficiency (NSE) of 0.72 and NSE for flows over threshold (NSEOT) of 0.78, respectively, averaged over multiple prairie watersheds for the entire simulation period. Although HYPR’s process representation is simple, it shows acceptable simulation of internal hydrologic variables (e.g., accumulated snow on ground) when compared against field measurements. HYPR can be useful when data or computational resources are limited. As for PRIMA, it shows potential for simulating the inundation extents when compared against remote sensing observations of water extents with an accuracy of 85 % averaged over two prairie basins in Saskatchewan, Canada. PRIMA is three to eight times as computationally efficient as the recently developed Wetland DEM Ponding Model (WDPM). The MESH-PRIMA model shows an improved hydrograph and flood simulation on a daily resolution (NSE = 0.55 and NSEOT = 0.60, respectively) compared to the MESH model with its current prairie algorithm (NSE = 0.49 and NSEOT = 0.55, respectively) for the entire simulation period. More importantly, MESH-PRIMA can identify the spatial distribution of water over the landscape and quantify the spatial non-contributing area for different flood events. The proposed models in this thesis can be used for efficient pothole storage dynamics simulation, inundation mapping, streamflow, and peak flow prediction in the prairies. The models can be used for a wide spectrum of hydrologic or hydraulic purposes ranging from limited data, conceptual-lumped-operational mode (e.g., HYPR) to detailed data, physically based research mode (e.g., MESH-PRIMA). These models, especially MESH-PRIMA, improve our understanding of the complexities of the prairie hydrology and the impacts of land depressions on changing the watershed response. More importantly, the methods proposed in MESH-PIMA can be explicitly used in most land-surface schemes within earth system models, allowing for important application in climate change and numerical prediction systems that typically ignore this important prairie phenomenon.