Numerous approaches have been developed for estimating hourly reference evapotranspiration ET0, most of which require numerous meteorological data. In many areas, the necessary data are lacking and new techniques are required. The objectives of this study are: (1) to develop artificial neural networks for estimating hourly reference evapotranspiration from limited weather data; (2) to evaluate the reliability of obtained artificial neural networks (ANNs) and Food and Agricultural Organization—56 Penman Monteith (FAO-56 PM) equation compared to the lysimeter measurements; (3) to test the performance of the FAO-56 PM equation for hourly daytime periods using rc=70 s m−1 (PM70) and using a lower rc=50 s m−1 (PM50); and (4) to evaluate the reliability of obtained ANNs compared to the FAO-56 PM equation using an hourly dataset from a variety of locations. The accuracy of two reduced-set artificial neural networks (ANNTR and ANNTHR) and two FAO-56 Penman-Monteith equations with different canopy resistance values (PM50 and PM70) was assessed using hourly lysimeter data from Davis, California. The ANNTR required only two parameters (temperature and radiation) as inputs. Temperature, humidity and (RnG) term were used as inputs in the ANNTHR. The ANNTR and PM50 were best at estimating hourly grass ET0. The ANNTR approach was additionally tested using hourly FAO-56 PM ET0 data from California Irrigation Management Information System (CIMIS) dataset. The overall results recommended Radial Basis Function (RBF) network for estimating hourly ET0 from limited weather data. Also, the results support the introduction of new value for canopy resistance (rc=50 s m−1) in the hourly FAO-56 PM equation.

This content is only available as a PDF.