Forecasting Dry Season Streamflow on the Peace River At Arcadia, Florida, USA

• Published in: Journal of the American Water Resources Association 2006-08, Vol.42 (4), p.851-862
• Main Authors: Coley, D.M, Waylen, P.R
• Format: Article
• Language: English
• Subjects: climatic factors; drought; dry season; Earth sciences; Earth, ocean, space; El Nino Southern Oscillation (ENSO); Exact sciences and technology; Florida; Freshwater; Hydrology. Hydrogeology; mathematical models; model validation; prediction; rivers; rivers/streams; spring; statistical models; statistics; stream flow; Water resources; water supply; winter
• ISSN: 1093-474X; 1752-1688
• DOI: 10.1111/j.1752-1688.2006.tb04499.x

The Peace River at Arcadia, Florida, is a municipal water supply supplement for southwestern Florida. Consequently, probabilities of encountering low flows during the dry season are of critical importance. Since the association between Pacific Ocean sea surface temperatures (SSTs) and seasonal streamflow variability in the southeastern United States is well documented, it is reasonable to generate forecasts based on this information. Here, employing historic records of minimum, mean, and maximum flows during winter (JFM) and spring (AMJ), upper and lower terciles define "above normal," "normal," and "below normal" levels of each variable. A probability distribution model describes the likelihood of these seasonal variables conditioned upon Pacific SSTs from the previous summer (JAS). Model calibration is based upon 40 (of 50) years of record employing stratified random sampling to ensure equal representation from each decade. The model is validated against the remaining 10 samples and the process repeated 100 times. Each conditional probability distribution yields varying probabilities of observing flow variables within defined categories. Generally, a warm (cold) Pacific is associated with higher (lower) flows. To test model skill, the forecast is constrained to be the most probable category in each calibration year, with significance tested by chi-square frequency tables. For all variables, the tables indicate high levels of association between forecast and observed terciles and forecast skill, particularly during winter. During spring the pattern is less clear, possibly due to the variable starting date of the summer rainy season. This simple technique suggests that Pacific SSTs provide a good forecast of low flows.