Bathymetry is the studies of underwater depth of a lake or ocean floors. Implementing bathymetric survey is expensive, time consuming and laborious process because it involved in the measurements for depth at point of the large area; it can be reduced using remote sensing techniques. The depth may be determined from the satellite images by establishing the correlation between lake depth and spectral reflectance of the image. Artificial Neural Network (ANN) techniques were used in this context has an advantage of the depth estimation possibility, without refining the depth causing scattering from environmental factors (e.g. bottom material and vegetation). In this research work, investigations have been done on the usage of the empirical models for depth determination and also on the suitability of bands. Least squares regression analysis has failed to produce desired result. ANN based Cascade Forward (CF) back propagation model was tested for single and multi band. Its result showed that using multiband (2&3) has produced comparatively better results than the classical regression analysis. The validation is performed by the various standard statistical tests such as Chi-square, t test and Mean Absolute Percentage error. Linear regression plot of measured depth and actual depth were also plotted.