Newly developed water productivity and harvest index models for maize in an arid region

Abstract

Simulating yield response to different irrigation scenarios is important for agricultural production, especially in the arid region where agriculture depends heavily on irrigation. To better predict yield under different irrigation scenarios, the variation of normalized water productivity (WP*) over the whole growing period of maize for seed production and the effect of different irrigation treatments on harvest index (HI) were investigated using field experiments from 2012 to 2015 in an arid region of northwest China. Two new non-linear dynamic WP* (WP*KR-L and WP*KR-S) models derived from the Logistic and Sigmoid equations, and four new HI (HIKR-J, HIKR-M, HIKR-B and HIKR-S) models developed on the basis of water deficit multiplicative or additive models at different growth stages were compared with the measurements and the WP* (WP*AC) and HI sub-model (HIAC) in the original AquaCrop model (Version 4.0). In addition, the WP*AC and HIAC models in the original AquaCrop model were replaced by the optimal WP* and HI models to build the AquaCrop-KR model. Then the yield simulated by the AquaCrop-KR model was compared with the measured yield and the yield simulated by the original AquaCrop model. The results show that both WP*KR-L and WP*KR-S models improved the simulation of final biomass, especially for the WP*KR-L model. The tested HI sub-models, namely HIKR-J, HIKR-M, HIKR-B and HIKR-S models had good performance to simulate HI under different irrigation scenarios, and the HIKR-M model was the best among all tested sub-models. When both WP*KR-L and HIKR-M sub-models were embedded into Aquacrop, the performance of the AquaCrop model was improved significantly to simulate yield, especially under severe water stress condition, with R2 increased from 0.496 to 0.653, NRMSE decreased from 26.2% to 16.1% and EF increased from 0.055 to 0.642.