Hits:

Journal:Agricultural and Forest Meteorology

Key Words:reference evapotranspiration, Penman-Monteith equation, Ångström-Prescott formula, global solar radiation, machine learning, support vector machine

Abstract:Accurate estimation of reference evapotranspiration (ET0) is of great importance for regional water resources planning and irrigation scheduling. The FAO56 recommended Penman-Monteith (P-M) model is widely adopted as the standard method for ET0 estimation, but its application is usually restricted by limited meteorological data worldwide, especially global solar radiation (Rs). This study provided two possible solutions to deal with the missing Rs data in ET0 estimation in China mainland. In the first solution, Rs data were estimated with the Ångström-Prescott (A-P) formula and daily sunshine hours. The values of two A-P formula fundamental coefficients a and b were obtained through three ways: (1) estimated based on limited Rs measurements at 80 solar radiation measurement stations (or site-calibrated); (2) recommended by the FAO-56 manual (or FAO-recommended); and (3) estimated based on the altitude and latitude of each weather station through the support vector machine algorithm (or SVM-estimated). The second solution used the SVM algorithm and available weather variables without Rs to directly estimate the ET0 values. The results showed that the FAO-recommended coefficients a and b were separately overestimated and underestimated in China mainland, which generated the largest simulation errors of Rs (average determination coefficient (R2) = 0.855; average root mean square error (RMSE) = 3.105 MJ m-2 d-1). Rs estimations with the SVM-estimated coefficients obtained similar accuracy as the site-calibrated coefficients. However, the transfer errors from Rs estimations to ET0 estimations were reduced by using the P-M model for all of the three kinds of coefficients. Compared with the Rs-based models, the estimation accuracy of the SVM-ET0 model yielded the highest accuracy during the train stage (R2 = 0.973; RMSE = 0.293 mm d-1) but largest errors during the test stage (R2 = 0.935; RMSE = 0.480 mm d-1). Thus, the SVM-estimated a and b coefficients were recommended for the estimation of Rs, which were further used in daily ET0 estimations with the P-M model in China when only Rs data are missing. Considering the complexity in the programming, the SVM-ET0 model is not recommended for daily ET0 estimation.

First Author:Chen Shang

Indexed by:Journal paper

Volume:316

Page Number:108864

Translation or Not:no

Date of Publication:2022-02-16

Included Journals:SCI

Date of Publication:2022-02-16