Pan Evaporation Simulation Based on Daily Meteorological Data Using Soft Computing Techniques and Multiple Linear Regression

Abstract

Evaporation, a major component of hydrologic cycle, is an important parameter to many applications in water resource management, irrigation scheduling, and environmental studies. In this study, two soft computing techniques: (a) Artificial Neural Network (ANN), (b) Co-active Neuro-Fuzzy Inference System (CANFIS); and Multiple Linear Regression (MLR) were used to simulate daily pan evaporation (Ep) at Pantnagar, located at the foothills of Himalayas in the Uttarakhand state of India. Daily meteorological data such as maximum and minimum air temperature, relative humidity in the morning (7 AM) and afternoon (2 PM), wind speed, sun shine hours and pan evaporation form January 1, 2001 to December 31, 2004 were used for developing the ANN, CANFIS and MLR models. A comparison based on statistical indices such as root mean squared error (RMSE), coefficient of efficiency (CE) and correlation coefficient (r) was made among the estimated magnitudes of Ep by the ANN, CANFIS and the MLR models. The architecture of ANN and CANFIS were managed by NeuroSolutions 5.0 software produced by NeuroDimension, Inc., Florida. The architecture of ANN was designed with hyperbolic tangent activation function and Delta-Bar-Delta learning algorithm and similarly the architecture of CANFIS was designed with Gaussian membership function, Takagi-Sugeno-Kang fuzzy model, hyperbolic tangent activation function and Delta-Bar-Delta learning algorithm. The results indicated that the performance of ANN model with 6-9-1 architecture in general was superior to the CANFIS and MLR models; however, the performance of CANFIS models was better than MLR models. The ANN model with all input variables and single hidden layer was found to be the best in simulating Ep at Pantnagar.