Validation of daily surface water depth model of the Greater Everglades based on real-time stage monitoring and aerial ground elevation survey

Abstract

As part of the Everglades Depth Estimation Network (EDEN) project, this paper describes validation of raster-based daily surface water depth models of the Greater Everglades in Florida developed using real-time stage data and elevation data obtained from a survey with an aerial height finder. Daily median stage data obtained at over 200 locations were interpolated using the multiquadric radial basis function. Surface water depth was obtained by subtracting a digital elevation model from the interpolated stage raster. The model was validated with 751 independent field measurements of surface water depth between 1999 and 2004. Correlations between prediction error and both density of the monitoring gages and distance from a major linear geographic feature, such as a canal, were weak, suggesting that the error does not depend on these features. South Florida has distinct dry and wet seasons and the study area is dominated by sawgrass and wet prairie. Seasonality and ground vegetation type significantly affect prediction error. Correlation between observed and predicted water depth was high for all combination of season and vegetation type (0.83–0.96). Model validation using an equivalence test provided evidence of equivalence between predicted and observed water depths in dry season prairie-dominated and wet season sawgrass-dominated areas with the strict test and in dry season sawgrass-dominated areas with the liberal test, but not in wet season prairie-dominated areas. Equivalence between observed and predicted water depth for both dry season sawgrass- and wet season prairie-dominated areas were confirmed with the strict test after further model calibrations using linear regressions.