Scientific Investigations Report 2009–5010
Prepared in cooperation with
St. Johns River Water Management District
Select an option:
Purpose and Scope
Methods of Investigations
Data Collection and Analyses
Relations between Water Use and Meteorological Parameters/Drought Indices
Single-Variant Regression Analyses
Influence of System Memory
Multivariant Regression Analyses
Water-use data collected between 1992 and 2006 at eight municipal water-supply utilities in east-central and northeast Florida were analyzed to identify seasonal trends in use and to quantify monthly variations. Regression analyses were applied to identify significant correlations between water use and selected meteorological parameters and drought indices. Selected parameters and indices include precipitation (P), air temperature (T), potential evapotranspiration (PET), available water (P-PET), monthly changes in these parameters (ΔP, ΔT, ΔPET, Δ(P-PET), the Palmer Drought Severity Index (PDSI), and the Standardized Precipitation Index (SPI). Selected utilities include the City of Daytona Beach (Daytona), the City of Eustis (Eustis), Gainesville Regional Utilities (GRU), Jacksonville Electric Authority (JEA), Orange County Utilities (OCU), Orlando Utilities Commission (OUC), Seminole County Utilities (SCU), and the City of St. Augustine (St. Augustine). Water-use rates at these utilities in 2006 ranged from about 3.2 million gallons per day at Eustis to about 131 million gallons per day at JEA.
Total water-use rates increased at all utilities throughout the 15-year period of record, ranging from about 4 percent at Daytona to greater than 200 percent at OCU and SCU. Metered rates, however, decreased at six of the eight utilities, ranging from about 2 percent at OCU and OUC to about 17 percent at Eustis. Decreases in metered rates occurred because the number of metered connections increased at a greater rate than did total water use, suggesting that factors other than just population growth may play important roles in water-use dynamics. Given the absence of a concurrent trend in precipitation, these decreases can likely be attributed to changes in non-climatic factors such as water-use type, usage of reclaimed water, water-use restrictions, demographics, and so forth. When averaged for the eight utilities, metered water-use rates depict a clear seasonal pattern in which rates were lowest in the winter and greatest in the late spring. Averaged water-use rates ranged from about 9 percent below the 15-year daily mean in January to about 11 percent above the daily mean in May.
Water-use rates were found to be statistically correlated to meteorological parameters and drought indices, and to be influenced by system memory. Metered rates (in gallons per day per active metered connection) were consistently found to be influenced by P, T, PET, and P-PET and changes in these parameters that occurred in prior months. In the single-variant analyses, best correlations were obtained by fitting polynomial functions to plots of metered rates versus moving-averaged values of selected parameters (R2 values greater than 0.50 at three of eight sites). Overall, metered water-use rates were best correlated with the 3- to 4-month moving average of ΔT or ΔPET (R2 values up to 0.66), whereas the full suite of meteorological parameters was best correlated with metered rates at Daytona and least correlated with rates at St. Augustine. Similarly, metered rates were substantially better correlated with moving-averaged values of precipitation (significant at all eight sites) than with single (current) monthly values (significant at only three sites). Total and metered water-use rates were positively correlated with T, PET, ΔP, ΔT, and ΔPET, and negatively correlated with P, P-PET, Δ(P-PET), PDSI, and SPI. The drought indices were better correlated with total water-use rates than with metered rates, whereas metered rates were better correlated with meteorological parameters.
Multivariant analyses produced fits of the data that explained a greater degree of the variance in metered rates than did the single-variant analyses. Adjusted R2 values for the “best” models ranged from 0.79 at JEA to 0.29 at St. Augustine and exceeded 0.60 at five of eight sites. The amount of available water (P-PET) was the single parameter most common to the best models (six of eight sites), whereas ΔT or ΔPET were included at five of eight sites. The moving average of at least one parameter was present in seven of the eight best models, indicating the influence of water-use memory to changing climatic conditions. Monthly P and ΔP were better correlated with metered water-use rates in the multivariant regression analyses (significant at six of eight sites) than in the single-variant analyses (significant at only three sites). This contrast can be attributed to the fact that the multivariant analyses better isolate the effects of precipitation on water use by factoring out the offsetting effects of other parameters, such as temperature, which are directly (and not inversely) related to water use.
The best model equations determined from the multivariant analyses were used to predict metered water-use rates for 2007, a relatively dry year. The average error between predicted and measured results, an indication of model bias, ranged from -53 gpdm (gallons per day per active metered connection) at OCU (overpredicted rates) to 42 gpdm at GRU (underpredicted rates). These biased results are likely due, in part, to factors not considered in the analyses such as increased usage of reclaimed water and changes in the ratio of commercial-to-residential users in 2007. Based collectively on the error and regression statistics developed for the predicted results, the best predictions were made at Eustis while the poorest predictions were made at GRU.
Murray, L.C, Jr., 2009, Relations between Municipal Water Use and Selected Meteorological Parameters and Drought Indices, East-Central and Northeast Florida: U.S. Geological Survey Scientific Investigations Report 2009–5010, 31 p.
U.S. Geological Survey
Florida Integrated Science Center
12703 Research Parkway
Orlando, FL 32826
L. Chi Murray firstname.lastname@example.org
|Accessibility FOIA Privacy Policies and Notices||U.S. Department of the Interior, U.S. Geological Survey
Persistent URL: http://pubs.usgs.gov/sir/2009/5010
Page Contact Information: USGS Publishing Network
Page Last Modified: Thursday, 10-Jan-2013 19:27:42 EST