JOURNAL OF HYDROINFORMATICS, vol.16, no.1, pp.50-59, 2014 (SCI-Expanded)
In this study, water depth distribution (bathymetric map) in a eutrophic shallow lake was determined using a WorldView-2 multispectral satellite image. Lake Eymir in Ankara (Turkey) was the study site. In order to generate the bathymetric map of the lake, image and data processing, and modelling were applied. First, the bands that would be used in depth prediction models were determined through statistical and multicollinearity analyses. Then, data screening was performed based on the standard deviation of standardized residuals (SD_SR) of depth values determined through preliminary linear regression models. This analysis indicated the sampling points utilized in depth modelling. Finally, linear and non-linear regression models were developed to predict the depths in Lake Eymir based on remotely sensed data. The non-linear regression model performed slightly better compared to the linear one in predicting the depths in Lake Eymir. Coefficients of determination (R-2) up to 0.90 were achieved. In general, the bathymetric map was in agreement with observations except at re-suspension areas. Yet, regression models were successful in defining the shallow depths at shore, as well as at the inlet and outlet of the lake. Moreover, deeper locations were successfully identified.