Journal of Statistics: Advances in Theory and ApplicationsAuthor's: MOHAMMAD ALI GHORBANI, VIJAY P. SINGH, RASOUL DANESHFARAZ and MAHSA HASANPOUR KASHANI
Pages: [15] - [36]
Received Date: April 5, 2012
Submitted by:
Evaporation estimation is an important task in irrigation water demand estimation, water resources management, and determination of water budgets. In recent years, artificial intelligence models have been applied more successfully to model evaporation and other hydrological processes. In this study, daily evaporation, as a dependent variable, was modelled using genetic programming (GP), which is one of the artificial intelligence models and an evolutionary data-driven modelling approach. We used data of four years (1987-1990) of five independent climatic variables including air temperature, solar radiation, wind speed, pressure, and humidity for two weather stations in California, USA. The GP models were trained and validated by using various combinations of the independent climatic variables. The performance of the selected GP model was compared with two artificial intelligence models, i.e., artificial neural network (ANN) and neuro-fuzzy (NF), and a traditional linear model by Stephen and Stewart (SS). The goodness of fit for the models was evaluated by using five performance criteria, namely, coefficient of determination, mean square error, mean absolute relative error, coefficient of efficiency, and index of agreement. The results obtained provided evidence that the GP model is capable of accurately modelling of evaporation and is a viable alternative to other artificial intelligence and traditional models.
artificial intelligence model, evaporation, genetic programming, modelling, pan evaporation.
