ESTIMATION OF EVAPOTRANSPIRATION OF CASSAVA OVER CROSS RIVER BASIN USING ARTIFICIAL NEURAL NETWORK MODEL

Abstract

This study dwells on the estimation of crop evapotranspiration (ET c ) of cassava over Cross River basin with the focus of estimating the crop evapotranspiration by the Artificial Neural Network (ANN) modelling strategies and comparing with 3 empirical models; Penman-Monteith (standard method), Blaney-Morin-Nigeria and Hargreaves-Samani. Relevant meteorological parameters were remotely sensed from Climatic Research Unit (CRU), University of East Anglia, Norwich, for 39 years (1979-2017) in order to estimate the crop evapotranspiration (ET c ), while Artificial Neural Network (ANN) was used for prediction of reference evapotranspiration over Cross River basin. Artificial Neural Network (ANN) tool-box is embedded in MATLAB R2017a, it is a high performance language in technical computing, which includes different toolbox for different field functions like the curve fitting box, bi-information toolbox, database box and others. The empirical models estimated crop evapotranspiration (ET c ) of cassava over Cross River basin as: Penman- Monteith (standard method) 2.7mm/day; Blaney-Morin-Nigeria 2.4mm/day and Hargreaves-Samani 3.0mm/day. The Artificial Neural Network (ANN) models performed better than the empirical models in terms of prediction as: BMNT3-3-1, R 2 of 0.9890 and RMSE of 0.000056mm/d; BMNL3-3-1, R 2 of 0.9883 and RMSE of 0.000177mm/d; and HAGT3-3-1, R 2 of 0.9038 and RMSE of 0.000754mm/d. ANN models estimated more precise and accurate values of crop evapotranspiration (ET c ) of cassava as: BMNT3-3-1 2.7mm/day; BMNL3-3-1 2.7mm/day and HAGT3-3-1 2.7mm/day. Further analysis revealed that, ANN models outperformed the existing empirical models. Therefore, the ANN models are efficient in water resources management and planning in Cross River basin.