Local sensitivity analysis of the AquaCrop model outputs for wheat under Semi-Arid water stress condition

Abstract

The FAO AquaCrop model has been extensively reported to simulate wheat growth and productivity in response to environmental conditions in many parts of the world. However, the calibration of the model could be tedious due to its large number of input parameters. The complexity in the model evaluation could be simplified by conducting a prior sensitivity analysis (SA), which information on it is hard to come-by in the North-eastern Nigeria. The SA of the model’s output variables to its input parameters was conducted using the local sensitivity analysis (LSA) technique. An early maturing REYNA-28 wheat variety was used under water deficit conditions in the semi-arid North-eastern Nigeria. The analysis revealed that the simulation of grain yield was highly influenced by days-to-flowering (DtF), normalized water productivity (WP*), reference harvest index (HIo), crop coefficient when the canopy is complete but prior to senescence (KcTrx) and maximum effective rooting depth (Zx) with sensitivity coefficients (SCs) of 1.23, 1.05, 0.83, 0.75 and 0.61, respectively. Biomass yield was highly sensitive to days-to-emergence (DtE), WP*, KcTrx, number of plants per hectare (den), soil surface covered by individual seedlings at 90 % emergence (ccs) and initial canopy cover (cco). The sensitivity of canopy cover was more to its related parameters such as DtE, maximum canopy cover (CCx), days-to-maximum canopy cover (DtCCx), canopy growth coefficient (CGC), ccs, cco, den and days-to-start of senescence (DtSS). Stress parameters were found to be either insensitive or with negligible sensitivity except lower soil water depletion threshold for canopy expansion (Pexlw). The analysis also revealed that the model outputs were insensitive to half of the model’s input parameters. These parameters could be fixed within their ranges in order to simplify the model and ease its calibration. The influential/sensitive parameters on the other hand require higher consideration during data collection, fine-tuning and calibration. This work can be validated using different SA techniques and wheat variety and under different environmental condition.