Identifying Flood Periods in Fargo-Moorhead Area

Faculty Mentor: Dr. Trung Le

Ice-melting is the major reason for floods in the Fargo-Moorhead area.  Flooding impacts many aspects of aquatic environment including infrastructures, hydraulic structures and the life of aquatic habitat.  This problem is generally complex due to the interaction of floodwaters and the period of ice melting.  We propose the development of a new machine learning framework to identify flooding period.  The proposed framework will enable predition of flooding in cold regions.  The Red River of the North Basin (Miller and Frink, 1984) is an ideal location to study floods due to its topographic features and economic importance.  It drains 105412 sq. km at the international boundary.  The basin is remarkably flat with a small difference in elevation between the upper part (Wahpeton-North Dakota), which is at 287 meters above sea level, and the lower part (Winnipeg-Canada), which is at 218m.  When the flooding conditions occur, the entire valley can become the floodplain.  The basin floods regularly with the major floods having occurred in 1950, 1966, 1979, 1996, 1997 and 2009 (USACE, 2018).  There are thirteen USGS gaging stations and the basin is covered by LIDar data provided by North Dakota State Water Commission.  The proposed framework will be applied to study floods in the Fargo/Moorhead area using the USGS measurement data.  We envision the use of machine learning algorithms for predicting flow discharge during flooding.  Our framework will provide a powerful simulation-based toolbox for planning, designing, and monitoring riverine infrastructure in a wide range of river networks.  Our approach can be applied to: 1) identify critical areas for flood protection and evaluate alternatives among mitigation strategies, 2) assess the aquatic environment and propose strategies for stream restoration, and 3) estimate hydrodynamic impacts of in-stream structures as well as their associated risks.  The proposed framework will enable the transformation approach in hydraulic engineering practice, which largely relies today on legacy models.

Teacher and/or Community College Faculty Component: The teachers will work closely with the graduate student to carry out the development of machine learning model to predict flooding for the downtown area of Fargo-Moorhead region.  The topographic data from LiDAR will be used to generate three-dimensional model with the details of the buildings and their types.  The teachers will explore the virtual environement in which floodwaters interact with the buildings.  Using 3-D printing, a model of downtown Fargo-Moorhead can be replicated as a physical model for educational purpose.  Students can tinker with this physical model to understand the concepts of flow velocity, flowrate, water surface slope, forces on the buildings and bridges, scour and sediment transport.