The variability in a lake's stage depends upon many factors, including surface-water flows, meteorological conditions, and hydrogeologic characteristics near the lake. An understanding of the factors controlling lake-stage variability for a population of lakes may be helpful to water managers who set regulatory levels for lakes. The goal of this study is to determine whether lake-stage variability can be predicted using multiple linear regression and readily available lake and basin characteristics defined for each lake.
Regressions were evaluated for a recent 10-year period (1996-2005) and for a historical 10-year period (1954-63). Ground-water pumping is considered to have affected stage at many of the 98 lakes included in the recent period analysis, and not to have affected stage at the 20 lakes included in the historical period analysis. For the recent period, regression models had coefficients of determination (R2) values ranging from 0.60 to 0.74, and up to five explanatory variables. Standard errors ranged from 21 to 37 percent of the average stage variability. Net leakage was the most important explanatory variable in regressions describing the full range and low range in stage variability for the recent period. The most important explanatory variable in the model predicting the high range in stage variability was the height over median lake stage at which surface-water outflow would occur. Other explanatory variables in final regression models for the recent period included the range in annual rainfall for the period and several variables related to local and regional hydrogeology: (1) ground-water pumping within 1 mile of each lake, (2) the amount of ground-water inflow (by category), (3) the head gradient between the lake and the Upper Floridan aquifer, and (4) the thickness of the intermediate confining unit. Many of the variables in final regression models are related to hydrogeologic characteristics, underscoring the importance of ground-water exchange in controlling the stage of karst lakes in Florida. Regression equations were used to predict lake-stage variability for the recent period for 12 additional lakes, and the median difference between predicted and observed values ranged from 11 to 23 percent.
Coefficients of determination for the historical period were considerably lower (maximum R2 of 0.28) than for the recent period. Reasons for these low R2 values are probably related to the small number of lakes (20) with stage data for an equivalent time period that were unaffected by ground-water pumping, the similarity of many of the lake types (large surface-water drainage lakes), and the greater uncertainty in defining historical basin characteristics. The lack of lake-stage data unaffected by ground-water pumping and the poor regression results obtained for that group of lakes limit the ability to predict natural lake-stage variability using this method in west-central Florida.
Additional publication details
USGS Numbered Series
Regression Analysis of Stage Variability for West-Central Florida Lakes