Neural Network based Predictors for Evaporation Estimation at Jabalpur in Central India
Free water evaporation is an imperative parameter for estimation of crop water requirement, and irrigation scheduling. This study aims to evaluate different techniques to estimate evaporation with weather parameters inputs. Multilayer Perception (MLR), Radial Basis Function (RBF) based neural network, traditional statistical Linear Regression (LR) approach and conventional empirical methods of Linacre and Christianson were used to estimate the evaporation at Jabalpur station situated under Kymore Plateau and Satpura Hills Agro-climatic Zone of Madhya Pradesh in the Central India. The weather parameters considered for estimation of evaporation are temperature, humidity, sunshine hours and wind speed. Results indicate that MLP and RBF based models with input of all selected weather parameters is able to estimate evaporation much precisely than LR and empirical approaches. It was found that higher accuracy may be obtained with multiple weather data input and low accuracy with only temperature input. It was observed that with temperature used as input the performance accuracy reduces in estimating evaporation with the selected models. However, neural network approach seems to produce better results as compared to statistical and empirical approach. The neural network based model RBF found more efficient in estimation of evaporation as compared to MLP. This study suggests that evaporation can be estimated by RBF model of a station, where there is no standard instrument available for its observation.
Empirical methods, Linear regression, Machine learning, RMSE, Weather parameters
Full Text: PDF (downloaded 405 times)
- There are currently no refbacks.