Chinese Journal of Agrometeorology ›› 2016, Vol. 37 ›› Issue (04): 415-421.doi: 10.3969/j.issn.1000-6362.2016.04.005

Previous Articles     Next Articles

Comparison of Machine Learning Algorithms and Hargreaves Model for Reference Evapotranspiration Estimation in Sichuan Basin

FENG Yu, CUI Ning-bo, GONG Dao-zhi   

  1. 1.Institute of Environment and Sustainable Development in Agriculture, Chinese Academy of Agricultural Sciences/Key Laboratory of Dryland Agriculture of Ministry of Agriculture, Beijing 100081, China; 2.State Key Laboratory of Hydraulics and Mountain River Engineering & College of Water Resource and Hydropower, Sichuan University, Chengdu 610065; 3.Provincial Key Laboratory of Water-Saving Agriculture in Hill Areas of Southern China, Chengdu 610063
  • Received:2016-01-05 Online:2016-08-20 Published:2016-08-10

Abstract:

Reference evapotranspiration (ET0) is an essential component of agricultural water management, accurate estimation of ET0 is vital in irrigation scheduling. This study investigated the applicability of two machine learning algorithms, the generalized regression neural networks (GRNN) and wavelet neural networks (WNN), in modeling ET0 only with temperature data at Suining meteorological station, central Sichuan basin. The performances of GRNN and WNN models were compared with the empirical Hargreaves (HS1) and two calibrated Hargreaves (HS2, HS3) models. From the results, the root mean square error (RMSE), model efficiency (Ens) and coefficient of determination(R2) were 0.395mm×d-1, 0.924 and 0.902 for GRNN model, 0.401mm×d-1, 0.911 and 0.901 for WNN model, respectively. The performances of GRNN and WNN model were much better than HS1, HS2 and HS3 model. A further performances evaluation of GRNN and WNN model was conducted, which manifested the better applicability of GRNN and WNN models in western and eastern Sichuan basin.

Key words: Reference evapotranspiration, Temperature data, FAO-56 Penman-Monteith model, Machine learning algorithm, Hargreaves model