Development Of Crop Water Production Function For Winter Wheat Based On The CERES-Wheat Model

Crop-water production functions(CWPFs) is a function to describe the relationship between crop yield and the irrigation amount during the growth season, which has been widely used in irrigation scheduling and deficit irrigation studies. In commonly used CWPFs, the influence of water stress on crop yield was represented with a water sensitivity index. However, this index usually suffered from strong temporal and spatial variations. It changed significantly with the differences of climate, soil, crop cultivar etc. Generally, the CWPFs can only be obtained through specifically designed multi-stage drought experiments and statistical regression analyses. Unfortunately, those experiments were usually complex and both time and labor consuming, which impeded the further study and application of CWPFs. In this study, we investigated the possibility of development of CWPFs for winter wheat production based on the wheat simulation model of CERES-Wheat. The main idea was: to calibrate the CERES-Wheat model with experiment data that can be easily obtained in common field experiments; then to conduct the multi-stage drought experiments with the calibrated model; and finally to develop the required CWPFs based on the simulation outputs.Field experiments were conducted under a rainout shelter for winter wheat growing under water stresses at different growth stages in two seasons(Oct. 2012-Jun. 2013 and Oct. 2013-Jun. 2014). The whole growth season of wheat was divided into five growing stages(wintering, greening, jointing, heading, and grain filling). Water stress occurred at two continuous stages while irrigations were applied at other stages, which resulted in four different levels of water stress period(D1-D4). Two irrigation levels of 40 mm(I1) and 80 mm(I2) were applied. A total of eight treatments, with three replicates for each, followed a split-plot experiment design. An extra control treatment with irrigation at all five stages was arranged beside. To investigate the influences of water stress at different growth stages on the growth and yields of winter wheat, the dynamic changes of several eco-physical characteristics of wheat growth were measured and compared, including height, leaf area index, phenology, biomass, and yield. Secondly, based on two years of field experiments, the CERES-Wheat was calibrated and verified. A total of five different plans for model calibration and verification were designed and the DSSAT-GLUE, a program package for parameter estimation in DSSAT, was used to estimate the relevant genetic coefficients. Then the five plans were compared for the discrepancies between corresponding simulated and observed values of phenological dates, single grain weight, biomass, yield, and soil moisture so as to determine the accuracy of CERES-Wheat model to simulate the agro-ecological processes of winter wheat farming system in arid areas. Then was used to simulate wheat growth under different climatic scenarios of 56 years(1955.10-2012.6). The simulation results were used to obtain the corresponding water production function, which were compared with the water production function based on field experiment data. Some main conclusions have been drawn as follows:(1) The growth and development of wheat could be remarkably influenced by continuous water stresses occurred at vegetative stages. The height, LAI and biomass were the worst for all treatments, when water stress occurred at the stages of wintering and greening. However, the negative influences on wheat growth were not notable when water stress occurred after jointing stage. The average growth rate of height and LAI after jointing was about ten times as that before jointing. There were no notable differences of biomass between all of the treatments until the jointing stage. The biomass values of treatments with water stresses at wintering and greening stages were remarkably lower than other treatments. The irrigation later could not recover these serious biomass losses. Water stress could shorten the whole growth season of wheat, with a maximal 5-day advancing of maturation. At the same irrigation level, the heading and flowering stages could be delayed for 1-3 days for different levels of stress period. For the same irrigation level, relatively higher numbers of productive ears and seeds per ear could be obtained when water stress occurred at the heading and grain filling stages, but with a lower thousand-kernel weight. On the contrary, a relatively higher thousand-kernel weight could be achieved when irrigation was applied at the heading and grain filling stages, but with lower numbers of productive ears and seeds per ear. For irrigation levels of I1 and I2, yields were the lowest when water stress occurred at wintering and greening stages, which was only 42% of the control treatment. However, the treatments with the highest yield were different for different irrigation levels. For I1, it was the treatment with water stress at jointing and heading stages that had the highest yield, or about 63% of the control treatment. For I2, it was the treatment with water stress at greening and jointing stages, which had a yield of about 75% of the control treatment. There was a clear interaction between the intensity and occurring stage of water stress. In general, the greening and filling stages were the critical periods of water demand for winter wheat. Reasonable irrigation managements are needed at these two growth stages to guarantee a higher yield of winter wheat in arid region.(2) Two genetic coefficients P1V(days at optimum vernalizing temperature required to complete vernalization) and G3(Standard, non-stressed total dry weight of a single tiller at maturity) of CERES-Wheat model varied remarkably under different scenarios of water stress. The coefficients of variation were 19.07% and 16.34%, respectively. It suggests that the values of these two parameters were influenced heavily by genotype-environment interactions. The rest parameters were relatively independent of water stress scenarios since the coefficients of variation were all less than 10%. The DSSAT-GLUE package was proved to have good convergence since the estimated values of most genetic coefficients converged into narrow ranges. For output variables, the different plans of model calibration and verification show great discrepancy. Plan 1(model calibration with data from the CK treatments with sufficient irrigation and verification with data from the rest treatments in the two seasons) was proved to be the optimal one since its absolute relative error(ARE) and relative root mean squared error(RRMSE) for model calibration were the lowest, only 4.89% and 5.18%, respectively. When water stresses occurred during the heading and grain-filling stages, CERES-Wheat model was able to correctly simulate the dynamic changes in growth and development of wheat as well as the soil moisture content. However, when water stresses occurred during the wintering and greening stages, there were relatively large simulation errors. When water stress occurred earlier and severer, the simulation accuracy was lower. In addition, CERES-Wheat model cannot correctly simulate the phenological discrepancies caused by different water stress scenarios because current algorithm for phenology estimation was only based on temperature and photoperiod but neglecting the secondary effect of water stress. Thus an improvement on current phenology prediction algorithm of winter wheat is needed. The results of leave-one-out cross validation showed that the overall error was about 15-18% for CERES-Wheat model to simulate winter wheat growth and yield under different scenarios of water stresses designed in this study. In general, there were some limitations for CERES-Wheat model to simulate winter wheat growth in arid conditions. It is necessary to research into the mechanism and simulation method of winter growth responding to water stresses in early vegetative stage, if CERES-Wheat is expected to be applied more widely in the management and research of winter wheat production in arid and semi-arid areas in China.(3) For Jensen, Blank and Stewart models based on field experiment data in this study, the water sensitive indices were higher at both wintering and filling stages of winter wheat, and lower at jointing stage. On the contrary, the indices were high at wintering, jointing and filling stages for the Minhas model. Meanwhile, for Jensen, Blank, Stewart and Minhas models based on simulation results with CERES-Wheat model under different climatic conditions of 56 years(1955.10-2012.6), the water sensitive indices were higher at greening and filling stages and lowest at wintering stage.There are still some limitations in simulating the development growth and yield of winter wheat with CERES-Wheat under water stress conditions, in order to improve the simulation precision of CERES Wheat model in the arid and semi-arid region, the model must be improved accordingly. In addition, the water production function of winter wheat was completely under the condition of water controlled, and does not take into account the actual rainfed condition. In the later study, we will focus on the water production function under rainfed condition,and conduct validation for it, then will make the water production function under different situations procedural eventually, and embed it in DSSAT CERES-Wheat model.