Vida Atashi

Date of Award

January 2023

Document Type


Degree Name

Doctor of Philosophy (PhD)


Civil Engineering

First Advisor

Yeo H. Lim

Second Advisor

Taufique H. Mahmood


In recent decades, flooding has become a major issue in many areas of the Upper Midwest. Many rivers and streams in the region had considerable increases in mean annual peak flows during this period, which was driven by a combination of natural factors including discharge synchrony with the spring thaw, ice jams, glacial lake plain, and a decrease in gradient downstream. The Red River of the North is a prominent river in the United States and Canada's Upper Midwest. It flows from its headwaters in Minnesota and North Dakota to Lake Winnipeg in Manitoba. The river is well-known for its spring floods, which can cause havoc on communities along its banks. There is an increasing need to improve the characterization and identification of precursors in the Red River basin that affect the hydrological conditions that cause spring snowmelt floods and improve predictions to reduce Red River flood damage. This dissertation has developed different research that concerns the dynamics of floods in the Red River basin by integrating hydrological, hydraulic, and machine-learning models. The primary objectives were to improve flood prediction accuracy by deriving the parameters of the Muskingum Routing method using discharge measurements obtained by an Autonomous Surface Vehicle, to predict scour potential of the river through HEC-RAS modeling, and to provide an estimate of the flood progression downstream based on the flow characteristics. The study also compared the effectiveness of Seasonal Autoregressive Integrated Moving Average (SARIMA), Random Forest (RF), and Long Short-Term Memory (LSTM) algorithms for flood prediction. Additionally, the research investigated the surface water area variation and response to wet and dry seasons across the entire Red River basin, which can inform the development of effective flood mitigation strategies. The results of this study contributed to a better understanding of flood control strategies in the Red River Basin and helped to inform policy decisions related to flood mitigation in the region. Ultimately, this research aimed to understand the complex dynamics of the RRB and derive hydrological and hydraulic models that could help to improve flood prediction. The first research developed a linear and nonlinear Muskingum model with lateral inflows for flood routing in the Red River Basin using Salp Swarm Algorithm (SSA). The distributed Muskingum model is introduced to improve the accuracy and efficiency of the calculations. The study focuses on developing a linear and nonlinear Muskingum model for the Grand Forks and Drayton USGS stations deriving the parameters of the Muskingum Routing method using discharge measurements based on spatial variable exponent parameters. The suggested approach minimizes the Sum of Square Errors (SSE) between observed and routed outflows. The results show for an icy river like Red River, the Muskingum method proposed is a convenient way to predict outflow hydrographs caused by snowmelt. The second study improved flood inundation mapping accuracy in flood-prone rivers, such as the Red River of the North, by using simulation tools in HEC-RAS for flood modeling and determining Manning's n coefficient. An Autonomous Surface Vehicle (ASV) was used to collect bathymetry and discharge data, including a flood event with a 16.5-year return period in 2022. The results showed that Manning's n-coefficient of 0.07 and 0.15 for the channel and overbanks, respectively, agreed well with the observed and simulated water level values under steady flow conditions. The study also demonstrated the efficiency of using ASVs for flood mapping and examined the scour potential and any local scour development in the streambed near the bridge piers. The third study of this dissertation used hourly level records from three USGS stations to evaluate water level predictions using three methods: SARIMA, RF, and LSTM. The LSTM method outperformed the other methods, demonstrating high precision for flood water level prediction. The results showed that the LSTM method was a reliable choice for predicting flood water levels up to one week in advance. This study contributes to the development of data-driven forecasting systems that provide cost-effective solutions and improved performance in simulating the complex physical processes of floods using mathematical expressions. This last study focused on the spatiotemporal dynamics of surface water area in the Red River Basin (RRB) by using a high-resolution global surface water dataset to investigate the changes in surface water extent from 1990 to 2019. The results showed that there were four distinct phases of variation in surface water: wetting (1990-2001), dry (2002-2005), recent wetting (2006-2013), and recent drying (2014-2019). The transition from bare land to permanent and seasonal water area was observed during the wetting phase, while the other phases experienced relatively little fluctuation. Overall, this study contributes to a better understanding of the spatiotemporal variation of surface water area in the RRB and provides insights into the impact of recent wetting and drying periods on the lakes and wetlands of the RRB.