Preliminaries to watershed instrumentation system design
Watershed instrumentation system design consists of identifying the locations and types of stream gauges, rain gauges, and other instruments needed for collection of input data for hydrological models, such that the models are useful in achieving objectives. The design is based on assessing the impact of various alternatives on the uncertainty in the model output. This assessment can occur only if certain preliminaries are adequately established: model selection, spatial resolution, temporal resolution, and the uncertainty analysis method. In this dissertation, the Soil and Water Assessment Tool (SWAT) model is selected as most appropriate for the site and objectives of concern, and applied to the Little Washita River Experimental Watershed (LWREW) in Oklahoma, and subwatershed I of the Little River Experimental Watershed (LREWswI) in Georgia. For spatial resolution, SWAT subbasin delineation is to most nearly match that of the field, and the number of hydrologic response units (HRUs) aggregated to 102 for LREWswI provides nearly identical results to those of an unwieldy 10,700 HRUs, the finest resolution possible. For temporal resolution, a novel event-adaptive time series may prevent the underestimation of event flow (as shown by the 11% increase and better matching of peaks for LREWswI), and provide better estimates of parameters, including the baseflow recession constant for LWREW. For uncertainty analysis applied to a synthetic LREWswI, a novel likelihood function estimates water budget components with an average error of 1.2%, while a conventional likelihood function leads to an average error of 13.3%. Also, the use of storm volume uncertainty parameters are to be included in the likelihood function if both the parameters and the flow quantities are to be estimated unbiasedly. These preliminaries help establish the foundation for watershed instrumentation system design.