Generalized Quadratic Synaptic Neural Networks for ETo Modeling

Abstract

This study aims at developing generalized quadratic synaptic neural (GQSN) based reference evapotranspiration (ETo) models corresponding to the Hargreaves (HG) method. The GQSN models were developed using pooled climate data from different locations under four agro-ecological regions (semi-arid, arid, sub-humid, and humid) in India. The inputs for the development of GQSN models include daily climate data of minimum and maximum air temperatures (Tmin and Tmax), extra terrestrial radiation (Ra) and altitude (alt) with different combinations, and the target consists of the FAO-56 Penman Monteith (FAO-56 PM) ETo. Comparisons of developed GQSN models with the generalized linear synaptic neural (GLSN) models were also made. Based on the comparisons, it is concluded that the GQSN and GLSN models performed better than the HG and calibrated HG (HG-C) methods. Comparison of GQSN and GLSN models, reveal that the GQSN models performed better than the GLSN models for all regions. Both GLSN and GQSN models with the inputs of Tmin, Tmax and Ra performed better compared to other combinations. Further, GLSN and GQSN models were applied to locations of model development and model testing to test the generalizing capability. The testing results suggest that the GQSN and GLSN models with the inputs of Tmin, Tmax and Ra have a good generalizing capability for all regions.