On the determination of crop water requirements, climate data are essential but are often limited due to the farm field's lack of weather station. For this reason, it is necessary to consider plant water requirements with various potential evapotranspiration (ETp) models with various weather input parameters, including Artificial Neural Network (ANN) models. The objectives of this paper were 1) to develop ANN models to estimate ETp, 2) to compare various ETp models (empirical models) including ANN models with the FAO standard models, 3) to analyze crop water requirements by the models, and 4) to determine the recommended input parameters for estimating ETp. The analysis was performed based on the measurement of weather parameters data in the two rice planting seasons, i.e., April - August 2017 and January - May 2018. There are 8 ETp models (empirical models) and 3 ANN models with a combination of input parameters. The results of this study indicated that the ANN-2 model with solar radiation input parameters was the best ANN model with R2 values 0.91-0.92 and RMSE 0.284 mm and 0.287 mm for the 2017 and 2018 planting seasons. ETp Turc model, one of the empirical ETp models with parameters input of air temperature and solar radiation, was the best model with the highest R2 and the lowest RMSE. Therefore, these two models were the best models with total ETp values closed to the ETp FAO standard. In addition, the parameters of air temperature and solar radiation are recommended parameters to be measured in the determination of crop water requirements using the ETp Turc model. But if there is only one parameter that can be measured, it is recommended to measure solar radiation with ANN-2 models to determine potential evapotranspiration.
Copyrights © 2020