Quinoa yield modeling revealed that conservative parameters of the AquaCrop model are not conservative: Evidences of planting methods and irrigation managements

Abstract

Water scarcity is the major significant environmental challenge affecting agricultural productivity, particularly in the semi-arid regions of the world. To cope with this challenge that threatens crop production, adapting proper field management is necessary to stream into sustainable crop production. In this route, using crop growth models is a strong and reliable approach to identifying the best management practices. Nevertheless, before use, crop models must be tuned and calibrated for the specific conditions of the location and field management. This study evaluates the accuracy of the water-driven AquaCrop model in simulating quinoa growth, yield, and soil water content (SWC) under varying planting methods and irrigation levels. The model was run under two conditions: globally recommended default conservative parameters and fine-tuned calibrated parameters. Field experiments were conducted in two growing seasons (2017 and 2018) considering different irrigation levels (I1:100 % of crop water requirement (WR), I2: 75 %WR, and I3: 50 %WR) and planting methods (P1: Basin, P2: on-ridge and P3: in-furrow planting method) in a semi-arid warm area. The results demonstrated AquaCrop's ability to simulate soil water content with good accuracy as normalized roots mean square error (NRMSE) and Willmott index of agreement (d) values were 12.2 % and 0.71 in the calibration step and 13.1 % and 0.75 in the validation step, respectively. AquaCrop could simulate the quinoa yield and biomass with reasonable accuracy at both validation and calibration steps with low NRMSE (9.4–14.1 %) and d values (0.61–0.67). The corresponding values for the validation step were 8–17.6 % and 0.93–0.94. Variations in deficit irrigation treatments introduced additional variability, particularly affecting SWC simulations. In addition, AquaCrop demonstrated reasonable accuracy across both calibration and validation steps, with d values ranging from 0.7 to 0.97 and NRMSE values between 7 % and 25 % for in-season biomass, crop evapotranspiration (ET_c), water productivity (WP_c), and canopy cover. Canopy cover was underestimated by 12.2 %, especially in the 75 % and 50 % WR treatments, however, RMSE, NRMSE, and d, pooled over all treatments, were 9.18 %, 13.58 %, and 0.90, respectively, in the validation step. Despite slight overestimations in grain yield and biomass in-furrow planting methods, the model provided reliable output, underscoring the impact of planting techniques on water use efficiency. Furthermore, we evaluated the default AquaCrop conservative parameters for simulating quinoa yield and biomass, revealing significant inaccuracies. Biomass was underestimated by 10.7 %, while grain yield was overestimated by 60.8 %, largely due to differences in quinoa cultivars, growth periods, and assumptions about water stress tolerance. Adjusting crop parameters to reflect moderate water stress tolerance improved model accuracy, emphasizing the need for region-specific calibration. This clearly implies that the conservative parameters recommended by the model developers are not globally valid.