Abstract:
Precise estimates of evaporation rates from lakes are needed for various research and management purposes. Although a number of models exist for estimation of lake evaporation, most of them, being site specific, need to be calibrated locally or it is required to develop a region specific model. In recent years, ANNs are successfully being used to model various hydrological processes such as rainfall-runoff, stream flow forecasting, reservoir operation etc. However, there are only a handful of reported studies in the area of evaporation and no studies have been reported on tropical lakes. The present study has been taken up for a tropical lake; Lake Pichhola in Udaipur, with the objective to develop an Artificial Neural Network (ANN) Model to estimate lake evaporation and to evaluate its performance. Daily data of five years (2002-2006) have been used in the study. Based on the results of various experiments with different numbers of neurons, an ANN model with 3 neurons and 145 epochs has been selected as the best ANN architecture for the model. The performance of the model has been tested vis-a-vis a Multiple Linear Regression Model. The results of ANN are superior than the MLR model. Based on the overall performance, the ANN model is found to be reasonably suitable to estimate the evaporation from the lake.
