Joe Vistad is a Graduate Student (MEngr) in the Civil Engineering Department at the University of North Dakota. He received his Bachelor of Science degree in Civil Engineering at UND in May of 2008. Currently his research is in regional flood frequency analysis with the goal of improving upon the standard methods of flood frequency calculation used for projects in the United States. Joseph.Vistad@und.edu Phone: 701-521-0552

Regional Flood Frequency Analysis in the Missouri River Basin Based on L-moments and GLS Regression

Fellow: Joseph D. Vistad

Advisor: Dr. Howe Lim, Assistant Professor, Department of Civil Engineering, University of North Dakota

Matching Support: University of North Dakota

Degree Progress: MEngr in Civil Engineering with a Water Resource Concentration expected in December 2009.

Description of the project:

The Interagency Advisory Committee on Water Data of the United States Geological Survey published Bulletin 17B “Guidelines for Determining Flood Flow Frequency,” which sets out the methods to be used for finding extreme event flows for all streams in the United States. The intent of Bulletin 17B was to give a standardized method of flood frequency calculation in the United States. The Bulletin has served its purpose extremely well since its final revision in 1982, however; the time has come for updates using today’s technology and modern statistical methods. Two such modern methods will be used in this project.

• GLS regression aims to improve the correlation of basin characteristics to predictive flood frequency values in ungauged streams.
• L-Moment based Index flood methods will be used to evaluate the flood quantiles for various regions. The advantage of regionalization is that it gives a larger data set to be studied rather then data for a single site.

This study will performed according to the following procedure:

• Annual maximum flow data is collected from more then 2000 sites across the Missouri River Basin. Data is compiled from past and present gauging stations which are operated by the United States Geological Survey.
• The data will be analyzed and grouped into homogeneous regions based upon statistical analysis, mainly of the L-moment variants. The clustered regions will then undergo L-moment statistical calculation and fit to the best distribution and determine the mean peak flow as well as various event estimates.
• The mean flow will then be compiled along with the physical characteristics of the basins using GLS regression analysis to give a predictive equation that can be used to find the mean peak flow at any site in the region, gauged or ungauged.
• A flood index plot will also be created to give a comprehensive indicator of frequency prediction across the basin. This growth curve can then be used in conjunction with the calculated regression equation to find the magnitude of a flood of any desired frequency at any point across the basin.

Project Objective: 

This study will provide a comprehensive set of L-moment index flood growth curves and mean flow regression equations for all homogeneous regions throughout the entire Missouri River Basin.

The project will also provide a good alternative and comparison to the current Bulletin 17B procedures. It is believed that with the flexibility of distributions, the estimates provided by the proposed study will be more robust and can be used with greater confidence.

This study can then serve a basis for future research and can be viewed by others to determine if changes in the policy governing flood estimations in the US are required.

Significance of Research:

The state of North Dakota, as well as other states across the nation, has invested a great deal into flood protection and recovery over recent years. The purpose of this study is to develop improvements to the currently used Bulletin 17B methods as inaccurate flood estimates can have a tremendous economic impact due to too frequent flooding of structures which are under designed. There can also be savings in infrastructure development if structures are currently being over designed.

Advisor: Howe Lim Assistant Professor of Civil Engineering University North Dakota howelim@mail.und.edu