Date of Award

5-2017

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Legacy Department

Environmental Engineering

Committee Member

Lawrence C. Murdoch, Committee Chair

Committee Member

Mark A. Schlautman, Committee Co-Chair

Committee Member

John C. Hayes

Committee Member

Ronald W. Falta

Abstract

Sediment runoff from construction sites is a major cause of impairment of surface water bodies, and the restoration efforts cost billions of dollars annually in the USA. This may be because Best Management Practices (BMPs) designed to control water quantity and quality are improperly implemented, or the safety limits imposed by regulations are inadequate. To assess the collective effectiveness of BMPs, three small watersheds that underwent various degrees of urban development, and a fourth undeveloped reference watershed were monitored in South Carolina, USA. The primary objective of the study was to characterize changes in flow and sediment output with development, which included fully urbanized and construction-related land uses. The requirement to have accurate stream flowrates led to an additional study that evaluated flowmeters and flowrate estimation methods. Identification of a conceptual flaw in the Curve Number (CN) method, a popular rainfall-runoff model, led to additional studies that were aimed at overcoming its shortcomings. Paired watershed studies were performed with the objective of quantifying changes in streamflow and water quality due to development at the watershed-scale. A method based on the Revised Universal Soil Loss Equation was used for land use scale analysis, in which the contribution from each land use to sediment yield was quantified. Area-normalized stormflows and peak flows from developing watersheds were 2 to 9 times greater, and sediment yield (SY) and event mean concentrations were one to two orders of magnitude greater, than those from the reference watershed. Sediment contribution factor (10-5 t h MJ-1 mm-1), defined as SY per unit rainfall erosivity, for each land use with 95% confidence interval was: Forest = 4 ± 2, Pasture = 2 ± 2, Full Development = 18 ± 11, Active Development = 440 ± 120. These values can be used to predict potential increase in sediment yield due to a future development scenario. Construction activities were accompanied by various BMPs, and significant increases in flow and sediment occurred despite their use. Improvements to the implementation of BMPs and/or proper maintenance may be necessary to ensure that their protective goals are met. Stream flowrate is a fundamental quantity in any land-use change study as it is used to calculate stormflows, sediment output, and contaminant concentrations. Flowrate measurements made with a hand-held flowmeter, SonTek FlowTracker (FT), and a fixed flowmeter, ISCO 750 Area Velocity Module (AVM), revealed that the flowrate measured by the AVM was nearly twice as much as that measured by the FT. Tests in a flume showed that the instruments were functioning within the uncertainty specified by the manufacturer. They also showed that the AVM nearly averaged the velocity over the depth of the water column above it. So, the differences in flowrates likely occurred because the AVM excluded the low velocity regions near the bottom and the banks from its sampling volume, whereas these regions could be sampled with the FT. The flowrate estimate of FT was assumed to be accurate, and used to calibrate the following flowrate estimation methods using stage or velocity measurements of AVM as inputs: Rating Curve Method (RCM), Index-Velocity Methods (IVM-1 without stage, IVM-2 with stage), and Conveyance-Slope Method (CSM). The ranking of their overall performance was: CSM < IVM-1 < RCM < IVM-2. Except for one stream in the study area, measurement of stage alone was sufficient to estimate flowrates with reasonable accuracy. The Curve Numbers for watersheds in the study area were sought to estimate the increase in runoff potential due to development. The data showed that CN decreased with rainfall magnitude (P) and approached a constant at large P, whereas the conventional CN method assumes that CN is constant for a given set of watershed conditions. To resolve this discrepancy, a theoretical analysis involving the spatial distribution of initial abstraction (Ia) was derived. It shows that heterogeneity within the watershed causes all parameters in the CN method to vary with P, and become constant at large P. Based on this finding it was hypothesized that treating the parameters as functions of P can account for heterogeneity and improve the runoff predictions of the CN method. The performance of the modifications that treat Ia as a function of P, termed variable Ia models, was compared with that of the conventional CN models using runoff from a synthetic watershed with precisely defined heterogeneity. The hypothesis was proved to be true, and the variable Ia models provide a simple way to improve runoff predictions by accounting for watershed heterogeneity. To complement the inclusion of spatial variations (heterogeneity) by the variable Ia models and further improve the performance of the CN method, an approach to include temporal variations was sought. This was achieved by refining an existing method of including antecedent moisture (M) in the CN method. A suite of models that include variable Ia, M, or both was developed and evaluated using rainfall-runoff observations from nine watersheds from a range of hydrologic settings. Including M in the CN models significantly improved the accuracy of the runoff predictions, whereas including variable Ia alone resulted in modest improvements. The best performance, an increase in the Nash-Sutcliffe efficiency parameter by 0.4, was achieved when both modifications were included together. A single storage rainfall-runoff model (SSM) was developed based on the findings from the analysis of the CN method, which is dual storage model. The model formulation is justified by the observation that the filled portions of both storages in the CN method vary similarly with P. SSM was evaluated using observations from the same nine watersheds used to test the suite of modified CN models. SSM predicted the overall runoff, and the runoff from smaller events better than the conventional CN method, and it is conceptually simpler than the latter. The CN method is widely applied throughout the world by many hydrologists and watershed models. Incorporating the proposed modifications (variable Ia, M, or SSM) would significantly improve runoff predictions while only modestly increasing (or decreasing in case of SSM) the complexity of the method.

Share

COinS