摘要：Artificial neural network (AN) was used to assess the reference evapotranspiration under missing or limited climatic parameters as input variables. The climatic data from year 1991-2014 i.e. 24 years was used for study. The results indicated that temperature based ANN architecture 2-2-1 and 3-4-1 found suitable for estimation of reference evapotranspiration under humid conditions. For mass based ANN model 4-4-1 and 5-4-1 architectures found appropriate for forecasting of evapotranspiration. The ANN architecture 6-2-1 gives good outcome than other architectures when all climatic variables were considered in the input layers. The study found that the different ANN architectures may be used under limited or missing data conditions. Temperature based, mass based and combination based models can be used for estimation of evapotranspiration by selecting the ideal nodes in hidden layer.
