Estimating selected parameters for the Kentucky Watershed Model from watershed characteristics

Abstract

Multiple linear equations to predict selected parameters for the Kentucky watershed model (KWM) are presented. The independent variables consist of easily determinable watershed characteristics. The relationship provides a means by which the KWM can be used to predict streamflows from ungaged drainage basins. Examples are given for five test watersheds. Results are variable. INTRODUCTION include watershed data from other states in the southern re-Competition among users for a relatively fixed supply of gion. water has focused attention on the need for effective means to plan for the optimum use of this vital resource. Parametric models, which lend themselves to computer application and attempt to simulate the hydrologic cycle, have become recognized as useful aids to such planning. One of the better known and more comprehensive parametric models is the Stanford watershed model [Crawford and Linsley, 1966]. Because of its popularity it has undergone numerous modifications [e.g., James, 1965; Claborne and Moore, 1970; $hanholtz et al., 1972; Ligon et al., 1969]. However, each modification is basically a soil moisture accounting procedure in which mathematical expressions are used to define relationships between elements of the hydrologic cycle and the interactions between its components. Numerous parameters are utilized in this process, some of which are derived from historical records, some from climatological data, others from physical watershed characteristics, and still other nonmeasurable entities from estimation by trial-and-adjustment or optimization procedures. The effectiveness of these routines depends on the availability of a streamflow record of sufficient length to calibrate the model. Typically, several different combinations of parameters are utilized before an acceptable match can be found between predicted and recorded streamflows. The need of prior calibration makes it difficult to apply existing watershed models to ungaged watersheds. To circumvent this problem, attempts have been made to correlate model parameters to measurable physical watershed characteristics. James [1972], reporting on work by Ross [1970], discusses linear regression relationships that were developed to relate plant available water capacity (AWC), permeability of the A soil horizon, and overland flow surface slope to selected parameters in the Kentucky watershed model (KWM). Jarboe and Haan [1974] used a multiregression approach to estimate parameters for use in a monthly water yield model. Ambaruch and Simmons [1973] also reported some success with a multiregression approach, using data from several Tennessee Valley Authority watersheds. The study reported herein was undertaken to extend the work of Ross [1970] and Ambaruch and Simmons [1973] to

Regionalization of SWAT Model Parameters for Use in Ungauged Watersheds

“…This method, however, ignores the heterogeneity among watersheds and has been found to give unsatisfactory results [10]. Several studies, including [6,7,[10][11][12][13][14] have used regression models that relate model parameters to watershed and climatic characteristics to determine regionalized parameter values. In summary, these studies have found regression to give suitable parameter estimates.…”

Predictions and Projections of Pine Productivity and Hydrology in Response to Climate Change Across the Southern United States

“…Researchers have long used United States Geologic Survey (USGS) streamflow data for hydrologic modeling, but traditionally, the emphasis was on model calibrations (James, 1972;Dawdy et al, 1972;Magette et al, 1976). Basin streamflow data are useful in broad-scale modeling, calibration, and validation because measurements integrate ecosystem water input, movement, and usage.…”

Selection of catchment descriptors for the physical similarity approach. Part II: Application

“…This problem instigated development of different methods for bypassing model calibration. One way is to identify appropriate mathematical relationships between the model parameters and some catchment descriptors (CDs) and then to apply these relationships in an ungauged catchment (Magette et al 1976;Jakeman et al 1992;Post et al 1998). Another option is to search for the nearest gauged catchment (in the geographical sense or in terms of similarity of CDs) and then to transfer a model parameter set from the most similar gauged catchments to an ungauged one (Parajka et al 2005;Oudin et al 2008).…”