Short-Run Water Demand Forecast Accuracy for the Tampa Bay Area

Short-term water demand forecasts are often employed for operational planning purposes. The success of operational decision-making depends in part on the quality of those forecasts. This study examines the accuracy of one-week-ahead demand projections produced between June 2006 and February 2014 by Tampa Bay Water, a large wholesale water supplier located in western Florida. Tampa Bay Water uses lagged historical data on weather conditions and previously recorded water consumption to predict demand in each of eight zones within the service area. For this study, random walk forecasts were developed to serve as benchmarks for the evaluation of the water supplier's original ex-ante forecasts. Both descriptive prediction-error summary statistics and formal statistical tests were used to gauge predictive accuracy. The results generally indicate that the one-week-ahead forecasts produced by Tampa Bay Water compare favorably against the benchmark projections.