Masters Theses

Date of Award

12-1997

Degree Type

Thesis

Degree Name

Master of Science

Major

Environmental Engineering

Major Professor

Bruce A. Tschantz

Committee Members

James Smoot, William Miller

Abstract

The Tennessee Valley Authority (TVA) operates an integrated multi-use reservoir system. Accurate and reliable forecasts of expected precipitation are a critical segment of the decision-making process in the scheduling of daily reservoir operations and for long-term decision-making. TVA receives daily quantitative precipitation forecasts (QPF) for the current day plus the nine following days from a weather forecast vendor. Application of TVA historical precipitation data in conjunction with the vendor forecast could increase the effectiveness of the precipitation forecasts in the long-term (6-10 day period in this context) portion of the forecasts.

Available historical precipitation data consists of 107 years of monthly average precipitation values for the two major divisions of the Tennessee River basin. TVA maintains a data base of this information and also published monthly and annual precipitation reports from 1936 through early 1980. Twenty-seven months of daily forecast and observed values from October 1994 through December 1996 were similarly obtained from the TVA data base. These values consist of forecast (and observed values) for a particular forecast date for each of 5 QPF sub-basins within the Tennessee River basin.

In the first phase of the analysis, the historical data was subjected to statistical smoothing to identify possible long-term trends. Factors such as rain gage spatial distribution, climatic change, topography, etc., which could influence recorded amounts, were also considered. No long-term historical trends were indicated. The second phase of the analysis consisted of a comparison of observed and vendor forecast precipitation values. and various historical forecast precipitation values derived from statistical analysis. Each of the 5 QPF sub-basins were examined for 3 forecast periods. Cumulative precipitation volumes for the forecast periods were compared. Analysis of residuals included an examination for trends and a comparison of the residual values.

Results indicate that the forecast expected rainfall quantities for the short-term forecast cannot be significantly enhanced using the historical data. However, a significant improvement in the forecast of expected rainfall quantities could be realized in the long term forecasts. The vendor tends to under-forecast precipitation for this period by almost 50 percent. A forecast based on historical data could thus be used to replace or supplement a vendor forecast for this period. Moreover, an apparent variation exists in vendor forecasts for the Southwest QPF sub-basin as compared to comparable forecasts for the 4 other QPF sub-basins. Supplementary analysis of the Southwest QPF sub-basin to determine the cause of the variation is recommended. Additional study is likewise recommended to identify any seasonal trends in the forecast volumes or residuals. Further analysis could also be performed as more precipitation data becomes available.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS