North-Central Section - 47th Annual Meeting (2-3 May 2013)

Paper No. 4
Presentation Time: 4:30 PM

ESTABLISHMENT OF UNIVERSAL REGRESSION MODELS FOR PREDICTION OF STREAM MORPHOLOGY BASED ON RELIEF, CLIMATE & WATERSHED VARIABLES


JHA, Rajan, Environmental & Water Resource Engineering, Virginia Tech, 800 newport terrace, Blacksburg, VA 24060, rajan@vt.edu

Are stream properties decoupled from watershed characteristics? Or else do watershed characteristics dictate the channel morphology? If they dictate then can we predict the values of stream properties (Bankfull discharge, Width, Depth, Channel slope, Sinuosity and Meander wavelength) based on the value of its watershed variables (namely: Rainfall/Runoff intensity, Relief, Drainage area , Valley slope , Sediment supply , Watershed elevation , Forest cover, Urban cover, Grass Cover , type of vegetation, Bank material, Soil type, Tectonic events)? The answer to these questions can be very critical in establishing universal relationships that could help us predict the values of hydraulic geometry for any stream across the globe. This research is exactly based on finding answers to these questions and quantitatively figuring out the degree of dependency of the watershed inputs to the stream variables. In “Part I” we do a qualitative study of the watershed characteristics and reason how perturbations in any one of the characteristic can lead to change in one or all of the stream properties. We try figuring the threshold values of change of each stream property (width, depth, channel gradient and others) and thereby determining how a stream accomplishes its objective of maintaining a “Quasi-Equilibrium state.”In “Part II” we do an empirical study of formulating dimensionless regression equations in order to predict bankfull hydraulic geometry. The results from Part II can be very helpful in deciding which dimensionless watershed variable has the most dominant affect on each dimensionless channel property. Based on a large data set of 600 data points, a cumulative universal regression model is developed. Later the data set is segregated into state/region wise and equations are developed separately for streams of 20 different states mainly belonging to USA, Canada, UK & New Zealand. The regression results clearly indicate “annual average rainfall with distribution, drainage area and mean basin elevation” as the most important and significant parameters which when combined integrate the effects of other watershed variables namely: urban cover, forest cover, sediment supply, etc.