Evaluation of AquaCrop model in predicting rice grain yield and biomass under water stress in different years

Introduction

Rice is one of the most important cereals in providing food security and the main food of billions of people in the Asian continent and many other parts of the world. Drought is the most important factor limiting rice production in paddy fields, which affects all stages of rice growth and development. Researches has shown that avoiding the conventional flood irrigation methods in rice cultivation and using alternative methods such as intermittent irrigation have a great effect on increasing water productivity and reducing its consumption. But in determining the most suitable alternative method of irrigation in each region, it should be noted that the effects of water shortage change with changes in the intensity, duration and time of its application. Today many Crop Growth Models (CGM's) have been designed to avoid the huge costs of conducting field research, speed up finding suitable solutions, and help to better understand and solve problems related to water movement in the soil and plant growth. Studies show that CGM's are able to consider the effect of different stresses such as water stress on dry matter production and grain yield during the growth period. One of the CGM’s is the AquaCrop model developed by FAO. The presence of factors such as moderate to severe water stress causes a significant decrease in the simulation accuracy by model, which is mentioned as one of the defects of the model. Most of studies with AquaCrop model have been limited to data collected in short periods. Therefore, the aim of this research is to evaluate the efficiency and accuracy of the AquaCrop model in simulating rice grain yield and biomass under multiple water stresses and during different years.



Materials and Methods

In order to evaluate the accuracy of the model in predicting the grain yield and biomass of rice, data collected from several research projects carried out in different years, the model was first calibrated and validated. The Hashemi variety used in all these projects, which is the most common variety cultivated in Guilan province. All agricultural operations of planting, growing and harvesting were carried out according to regional customs and the amounts of chemical fertilizers, herbicides and pesticides based on the recommendations of experts in agriculture and herbal medicine of the rice research institute. For the present study, a total of 45 irrigation treatments were selected from the previous projects carried out in the lands of the Country Rice Research Institute, Lahijan and Soumesara, of which 31 treatments were used for the calibration section and 14 treatments were used for model validation. The model was implemented for each irrigation treatment separately and the grain yield and biomass values obtained from the simulation were analyzed with the measured values based on the statistical indicators used in this research.



Results and Discussion

Based on the results of calibration of the model, the range of observed grain yield was 2100 to 4870 with an average of 3765 kg.ha-1. This is while the corresponding values simulated in the calibration conditions by the model were equal to 1749 to 4704 with an average of 3748 kg.ha-1. The accuracy of the model is low at the low limit of performance, but very accurate at the high and average performance limits. This phenomenon can be attributed to the estimation error of the model in the presence of environmental stresses such as water and fertilizer stress, which has also been mentioned in the studies of other researchers. Also, the values of RMSE and NRMSE for yield simulation were equal to 309.65 kg.ha-1 and 8.22%, respectively, which indicates very good accuracy in model calibration. Also, RMSE and NRMSE values for biomass simulation in calibration conditions were equal to 596.31 kg.ha-1 and 6.41%, respectively. The values of RMSE and NRMSE for the simulation of performance in the validation conditions were equal to 168.42 kg.ha-1 and 10.30%, which indicates good accuracy in model validation. Also, the values of RMSE and NRMSE for the simulation of biomass were equal to 554.71 kg.ha-1 and 12.90%, which shows the good accuracy of the model in validation. Examining the results of the model in different water stresses showed that with the increase of water stress from permanent waterlogging to high stress with the addition of irrigation cycles, the amount of model error in yield simulation increases.



Conclusion

In general, the AquaCrop model has good accuracy in simulating the grain yield and biomass of Hashemi variety rice, but the more severe the amount of water stress, the accuracy of the model decreases and its error increases. This problem is attributed to the structure of the model and the mathematical equations used in it, as well as the measured data. However, the AquaCrop model has many advantages such as the need for less data, the ability to be used for a wide range of crops and the user-friendliness of the model and it is recommended to use the AquaCrop model in different irrigation managements, especially in conditions without severe water stress, where the model has very good accuracy. But it is recommended due to the advantages of the AquaCrop model, such as the need for less data, the ability to be used for a wide range of crops and the user-friendliness of the model, its use in different irrigation managements, especially in conditions without extreme water stress, where the model has very good accuracy.