Date of Award

1-1-2012

Document Type

Campus Access Thesis

Department

Geography

First Advisor

Allan James

Abstract

Low-order headwater streams, a link between the upland landscape and larger streams within a watershed, affect watershed hydrology. Yet current widely available maps inadequately depict headwater streams, underestimating channel length or omitting channels entirely. Intermittent and perennial headwater streams each provide unique habitat. Until recently, it has not been possible to accurately delineate these streams or predict channel flow durations without fieldwork. Recent advances in remote sensing may allow more accurate headwater stream mapping.

Methods for mapping intermittent and perennial headwater streams with a GIS-based modeling approach coupled with field verification were adapted, tested, and evaluated. Four models were developed utilizing increasing numbers of predictor variables ranging from a univariate model based on drainage area to models including geomorphometric and physiographic variables. Current models are often based solely on a threshold of drainage area, or variables derived from a Digital Elevation Model (DEM) such as drainage area and slope. The North Carolina Division of Water Quality (NC DWQ) conducted an extensive headwater stream modeling study, and mapped intermittent and perennial channel origins in the field in the North Carolina Slate Belt ecoregion. The NC DWQ study developed a model using variables derived from a DEM. This thesis tested use of additional model variables generated from commonly available datasets

characterizing land cover for effectiveness in predicting the locations of intermittent and perennial channels. A secondary research objective was to explore model impacts from the use of a multi-direction flow routing algorithm and a different DEM.

Results show the NC DWQ model with multiple DEM-derived variables was substantially more accurate than a model based solely on drainage area, and the addition of new DEM-derived variables again increased this accuracy. The addition of physiographic variables produced the most accurate model by most metrics. Important from a conservation perspective, the model with physiographic variables had the highest true positive and lowest false negative values, erring on the side of over-predicting channel lengths. All models over-predicted total perennial channel length, although only the model with physiographic data over-predicted total intermittent channel length, while all other models under-predicted intermittent channel length.

Share

COinS