Applying factor analysis and multiple linear regression analysis in developing water yield models for small Tennessee watersheds

The uncertainty that now exists in predicting water yield requires that large factors of safety be incorporated into the design of hydraulic structures. If a mathematical model could be developed from data that is readily available or easily measured, that would predict the water yield with greater accuracy, this might allow a reduction of the safety factors thereby lowering the costs of these projects. This study was designed to examine the feasibility of using factor analysis and multiple linear regression techniques in the develop-ment of mathematical models that would predict water yield from small watersheds in Tennessee on a seasonal and an annual basis. Twelve parameters were initially selected for study by use of factor analysis. Of these 12 parameters one was deleted by factor analysis. Multiple linear regression analyses were then performed using various combinations of data from watershed parameters and various time periods. From these analyses the following conclusions were drawn: 1. Factor analysis can be used to screen superfluous parameters and thereby reduce the number of parameters needed to char-acterize the hydrologic properties of watersheds. 2. Watersheds must be grouped using similar hydrologic char-acteristics and especially similar geologic characteristics. 3. Many of the prediction equations of this study indicate that as area increases, runoff decreases which is contrary to that which is generally reported. 4. Prediction equations can be derived from different parameters for the same watersheds, and these equations often produce satisfactory predictions as long as the data used for the prediction are near the mean values of the parameters used in deriving the equations. The best results are obtained using data collected over a long time period.