Raheel Osman,Yan Zhu,Weixing Cao,Zhifeng Ding,Meng Wang,Leilei Liu,Liang Tang,Bing Liu*

National Engineering and Technology Center for Information Agriculture/Engineering Research Center of Smart Agriculture,Ministry of Education/Key Laboratory for Crop

System Analysis and Decision Making,Ministry of Agriculture and Rural Affairs/Jiangsu Key Laboratory for Information Agriculture/Jiangsu Collaborative Innovation Center for Modern Crop Production,Nanjing Agricultural University,Nanjing 210095,Jiangsu,China

ABSTRACT Extreme high-temperature stress (HTS) associated with climate change poses potential threats to wheat grain yield and quality.Wheat grain protein concentration (GPC) is a determinant of wheat quality for human nutrition and is often neglected in attempts to assess climate change impacts on wheat production.Crop models are useful tools for quantification of temperature impacts on grain yield and quality.Current crop models either cannot simulate or can simulate only partially the effects of HTS on crop N dynamics and grain N accumulation.There is a paucity of observational data on crop N and grain quality collected under systematic HTS scenarios to develop algorithms for model improvement as well as evaluate crop models.Two-year phytotron experiments were conducted with two wheat cultivars under HTS at anthesis,grain filling,and both stages.HTS significantly reduced total aboveground N and increased the rate of grain N accumulation,while total aboveground N and the rate of grain N accumulation were more sensitive to HTS at anthesis than at grain filling.The observed relationships between total aboveground N,rate of grain N accumulation,and HTS were quantified and incorporated into the WheatGrow model.The new HTS routines improved simulation of the dynamics of total aboveground N,grain N accumulation,and GPC by the model.The improved model provided better estimates of total aboveground N,grain N accumulation,and GPC under HTS (the normalized root mean square error was reduced by 40%,85%,and 80%,respectively) than the original WheatGrow model.The improvements in the model enhance its applicability to the assessment of climate change effects on wheat grain quality by reducing the uncertainties of simulating N dynamics and grain quality under HTS.

Keywords:Heat stress Total aboveground N Grain N accumulation Grain protein concentration Model improvement WheatGrow model

1.Introduction

Wheat is the third most important cereal crop in the world by production,providing 20% of daily protein and calories for 4.5 billion people [1].In the past few decades,wheat yield and quality have maintained pace with global consumer demand through both breeding efforts and management practices[2,3].However,ensuring wheat grain quality,especially under extreme hightemperature stress (HTS) is challenging,as it is crucial to human nutrition,commodity value,and end-use functional properties[4–6].Protein governs the end-use quality of grain products,as it is considered the principal component of wheat grain quality[5,7].High-temperature stress (HTS) events have been projected[8,9] as a risk in the near future for deteriorating grain quality,in particular grain protein concentration (GPC).The effects of HTS on wheat have been previously quantified in terms of duration,level,and growth stage at which stress occurs [6,10,11].Specifically,HTS influences wheat quality by altering grain starch content and GPC [5,6,9].However,increase in GPC under HTS despite reduced grain protein accumulation indicates that the reduction in grain mass is greater than that of protein [4,12,13].

Simulation analysis is considered to be a powerful tool for examining and interpreting the impacts of environmental and management factors on grain quality [14].In simulations of N dynamics,crop N uptake is generally simulated using a daily minimum of N demand and potential N uptake by the plant [4,15].Generally,critical N concentration or optimum N concentration is applied to calculate daily N demand (in CERES-Wheat [16],APSIM-Wheat [17],CROSYST [18],DAISY [19],EPIC [20],FASSET[21],STICS[22],HERMES[23],and MONICA[24]).In other models,crop N is divided into three compartments:structural N concentration,green leaf area N concentration,and storage N concentration(luxury consumption) [14,25].Potential N uptake is divided into two phases:pre-and post-anthesis N uptake.Pre-anthesis N uptake is a function of leaf area index (LAI) and aboveground biomass,whereas post-anthesis N uptake is a function of postanthesis thermal time or grain yield [25,26].Pan et al.[25] simulated plant N dynamics in wheat and accounted for genetic and environmental factors such as N and water supply,but did not consider the impact of temperature on post-anthesis N uptake.Likewise,effects of temperature stress on N dynamics,in particular on N uptake and demand,are ignored in most current crop models.

Early studies modeling grain quality treated grain N as a key parameter defining nutritional and functional properties [4].To simulate grain N accumulation,a simple harvest index method was used by STICS [22] and SIRIUS [27],and the latter model was modified by Jamieson and Semenov [14] to use a more complex source–sink approach.Other models,such as CERES-wheat [16],AFRCWHEAT2 [28],APSIM-Wheat,and Nwheat [29],apply independent functions for simulation of grain dry matter and grain N accumulation [14].However,the uncertainty in simulating GPC is usually high[30].For example,the average root mean squared relative error of 20 wheat models in simulating the GPC in four contrasting locations was 34% [31].Grain N accumulation usually increases linearly with mean temperature in these models [32].The majority of crop models are restricted to simulating kernel size and grain N concentration.However,SiriusQuality [26] accounts for structural proteins (albumin-globulin and amphiphilic proteins) and storage proteins (glutenin and gliadin) under normal temperature [4].

Most studies have not considered the effects of extreme HTS on N dynamics and grain protein deposition,as revealed by recent model evaluations [33–35].In these model evaluation studies,some crop models show relatively good agreement with observations under normal temperatures.However,underestimation of extreme-temperature effects on grain quality parameters has been noted[36].Accordingly,Sánchez et al.[37]emphasized the need to define response functions for extreme temperature stress,whereas Chenu et al.[30]suggested the need to model the influence of HTS on protein aggregation.Numerous authors [11,38,39] have proposed functions for quantifying HTS effects on phenology,photosynthesis,respiration,and leaf area index to simulate biomass and grain yield under HTS in wheat crop models.But little attention has been paid to improving models for simulating grain quality under HTS.

There is a scarcity of observational data on crop N and grain quality collected under systematic HTS conditions,suitable for developing algorithms for model improvement and evaluation of crop models.In this context,Craufurd et al.[40] emphasized the need to conduct experiments in crop science in order to acquire reliable data for model evaluation and improvement,particularly under extreme temperatures.

Using comprehensive data collected from environmentcontrolled phytotron experiments under HTS,this study aimed to determine and quantify the impacts of extreme HTS at critical growth stages on N uptake,grain N accumulation,and GPC in wheat.A further objective was to use this knowledge to improve predictions of grain protein dynamics under extreme HTS in the WheatGrow model [25].

2.Materials and methods

2.1.Data sources

Two-year phytotron experiments were performed.Environment-controlled phytotrons (FYS-10,Yuheng Instrument Co.Ltd.,Nanjing,China) were made of transparent glass with dimensions of 3.4 m × 3.2 m × 2.8 m (length × width × height).Relative humidity and air temperature were controlled with bubbling systems and air conditioners.Fans were installed at the front and rear of each phytotron to maintain a CO2concentration similar to that of the environment.HTS was achieved by precise control of the diurnal fluctuations of temperature and humidity inside the phytotrons,similar to those in the environment (Fig.S1).Temperature and relative humidity were recorded every 1 min during the treatment periods using EM50 data loggers(Decagon Devices,Inc.,Pullman,WA,USA).A halogen lamp was used to provide sufficient radiation for wheat growth.At noon,the light intensity inside the phytotron was approximately 1380 μmol m-1s-1during sunshine and 240 μmol m-1s-1under cloud cover.Pots were placed and rotated every day to random positions to minimize positional effects.The plants were removed from the phytotrons after the extreme HTS treatments and kept in ambient conditions until harvest.

In pot experiments,two common wheat cultivars (Yangmai 16 and Xumai 30)were used.Pots with dimensions of 30 cm×25 cm(diameter×height)were filled with sieved yellow–brown soil.The soil had available nitrogen,phosphorus,and potassium contents of respectively 150 mg kg-1,58 mg kg-1,and 96 mg kg-1with organic matter of 25 mg kg-1.Ten plants per pot were maintained and placed under normal optimum conditions before and after HTS treatments.Fertilizer;18.3 g N m-2,10.2 g P2O5m-2,and 18.3 g K2O m-2was applied before sowing and another 18.3 g N m-2during the jointing stage of wheat.All other practices,including irrigation and pesticide application,followed local wheat farming standards to ensure that there was no water or nitrogen stress during the experiments.The pots used for HTS treatments were surrounded by extra pots to limit side effects.Once the wheat plants reached target growth stages,pots were transferred into the phytotrons for HTS treatments.

Anthesis and grain filling were selected as treatment stages for HTS treatments,as these two stages are most susceptible to HTS worldwide[13,41,42].Based on previous studies[10,43],the average temperature range of 19–22°C is the optimum for anthesis and grain filling.However,30 °C is considered as the temperature threshold for winter wheat cultivars for the post-anthesis period[10,36].Accordingly,the control temperature of 27 °C/17 °C,and the maximum temperature of 39 °C/29 °C were selected for HTS.Four temperature treatments were applied:27°C/17°C(Tmax/Tmin,T1),31 °C/21 °C (Tmax/Tmin,T2),35 °C/25 °C (Tmax/Tmin,T3),and 39 °C/29 °C (Tmax/Tmin,T4).According to previous studies[5,6,11],short-term HTS during anthesis and grain filling is detrimental to wheat yield and quality.In winter wheat in China,individual HTS events usually span fewer than 6 consecutive days based on our previous study [43] on the occurrence of postheading heat stress across the major wheat producing regions.Accordingly,3 and 6 days were selected as the durations of HTS treatment.The HTS treatments included individual stress treatments at anthesis (Zadoks 61) stage,grain filling (10 days after anthesis) stage,and combined stress treatments at both anthesis and grain filling (Table S1).The pots were grouped randomly into three blocks before HTS treatments.When wheat plants reached target treatment stages,nine pots from each block were picked randomly and transferred into each phytotron.Four phytotrons were used for HTS treatments for four temperature levels.A total of 27 pots were subjected to each treatment combination(cultivar × temperature level × duration × growth stage).

2.2.Plant sampling

Samples were taken from uniform wheat tillers that simultaneously reached the desired treatment growth stage (anthesis and grain filling).Sampling was performed on the day of treatment and at every fifth day after HTS treatment.The aboveground parts of sampled plants were divided into leaves,stems and sheaths,and spikes.All the samples were oven-dried at 105 °C for 30 min and then at 80 °C to constant mass for measurements of dry weight and N concentration.The semi-micro-Kjeldahl method [44] was used for determination of the total N concentration of leaves,stems and sheaths,and spikes.GPC was calculated by multiplying grain N percentage by weight by a conversion factor of 5.7 [45].

The total aboveground N was calculated as the sum of N accumulation in leaves,stem and sheaths,and spikes.The rate of grain N accumulation was determined by dividing grain N content(mg N kernal-1) by the number of days from anthesis to physiological maturity,as shown in Eq.(1).Relative total aboveground N and rate of grain N accumulation were calculated as the ratio of the absolute values under different treatments to the corresponding values from the control treatment (T1 treatment) for the same treatment stage and cultivar (Eq.(2)) [36].

2.3.The WheatGrow model

The WheatGrow model [11] is a process-based wheat growth simulation model with six sub-modules:(i) apical and phenological development,(ii) photosynthesis and biomass accumulation,(iii) biomass partitioning and organ development,(iv) grain yield and quality formation,(v) soil water balance,and (vi) nitrogen dynamics.Model evaluation studies [13,46] have been conducted for different wheat cultivars across the major wheat-producing region of China.Recent model improvements in WheatGrow have included the incorporation of HTS effects on wheat development,leaf senescence,biomass growth,biomass partitioning,and yield formation [11,38].However,simulations of the effects of HTS on plant N dynamics and grain quality have not been studied adequately.The grain quality formation sub-model,including simulation of the plant N uptake and grain N accumulation in individual wheat kernels,was fully described by Pan et al.[25].The effects on GPC of environment and management factors,including different sowing dates,water regimes,and N fertilization rates have been quantified and investigated in field experiments.However,the WheatGrow model produced unsatisfactory results under HTS for GPC and grain protein yield in our previous model-testing study[36].Accordingly,in the present study,we developed new process-based algorithms to simulate HTS influence on N dynamics and rate of grain N accumulation and then incorporated the algorithms into the routines proposed by Pan et al.[25],to improve the simulations of aboveground N,grain N accumulation,and GPC in the WheatGrow model under HTS conditions.

In the original WheatGrow model,the rate of single-kernel N accumulation depends mostly on mean temperature and the availability of N during grain filling [25].The N required for grain protein accumulation is derived from post-anthesis N uptake and N remobilization from the vegetative organs.

2.3.1.Total N uptake by plant

In the original WheatGrow model,the total N uptake of the plant is the sum of pre and post-anthesis N uptake:

Pre-and post-anthesis N uptake have been shown[25,47]to be different.Pre-anthesis N uptake is described [25] as the sum of N associated with leaf area expansion (Nleaf) and N associated with non-leaf biomass(Nnon-leaf),depending on the LAI and aboveground biomass (more details in Supplementary file),Eqs.S1–S3.

After anthesis,N uptake is assumed to be a function of grain weight (GW,g m-2) (Eq.(4)).

where PNRpost-anthesisis the potential N-uptake rate,with a value of 0.6 N m-2day-1[25,45].PLFD (days) is the physiological filling duration (grain-filling duration under optimal temperature conditions),which varies with cultivar.f(Wi) andf(Npi) are the effects of water and plant N status on plant N uptake and are calculated by separate modules in WheatGrow [48,49].ANA is the effect of the amount of N at the time of anthesis on post-anthesis N uptake and is calculated by Eq.(5).

where Nuptakethresholdis the threshold for post-anthesis N uptake,with a value of 16.23 g m-2[25].Higher pre-anthesis N uptake than the Nuptakethresholdwould inhibit post-anthesis plant N uptake,according to Pan et al.[25].

2.3.2.Grain N accumulation

In the WheatGrow model,grain N accumulation is calculated with the grain N accumulation rate on a single-kernel basis.The N concentration of the grain is multiplied by 5.7 to calculate GPC.The initial grain N content was taken as 0.105 mg N kernel-1[25,45].The individual rate of grain N accumulation (GNRi,mg N kernel-1day-1) is calculated using Eq.(6).

where GNRmis the maximum rate of grain N accumulation and varies with cultivar.f(Wi),f(Npi),andf(Ti) are stress factors for quantifying water,plant N,and average temperature effects on GNRi,and are calculated using separate modules in WheatGrow.Fig.1 shows the relationship between the daily mean temperature and the rate of grain N accumulation.f(Ti) is calculated as

Fig.1.Relationship between daily mean temperature and rate of grain N accumulation in the original WheatGrow model.

The optimal temperature (Topt) was set at 24.2 °C,according to Pan et al.[25].Tmean,idenotes the daily mean temperature andTmin,ithe daily minimum temperature.

GNAiin Eq.(6)is the N available to the grain during dayi(mg N kernel-1day-1),and was determined by the N uptake during the post-anthesis period and N remobilization from the vegetative plant parts,including the leaf,stem,and chaff.More details of the calculations of GNAiare found in Pan et al.[25].

2.4.Model calibration and evaluation

The simulations from the improved and the original Wheat-Grow version were compared to evaluate the performance of the newly developed algorithms to simulate extreme-temperature effects on plant N and grain protein dynamics.For this purpose,independently measured datasets from different growing seasons of the phytotron experiments were used for model calibration and validation simulations.The experimental datasets from the 2015–2016 growing season were used to calibrate the models,and the datasets from 2016 to 2017 were used to validate the models to achieve a reasonable model evaluation(Table S1).Calibration of genetic parameters was performed for the original and improved models to best match the simulated wheat growth and quality variables to the observed data.Table 1 shows the genetic parameters used in the simulations for the two cultivars.

Table 1 Genotypic parameters used in model simulations under high-temperature stress for two wheat cultivars.

Values for LAI,leaf weight,stem weight,total aboveground biomass,grain weight,and kernel number used during the simulations of N dynamics in the HTS experiments were simulated with other sub-models of WheatGrow.Model improvement for simulating HTS effects on phenology,LAI,biomass,grain weight,and kernel number was performed as previously [11,38] described.

2.5.Data analysis

Four statistical indices were used to determine and compare model performances.They were (1) normalized root mean square error (NRMSE),(2) mean bias error (MBE),(3) index of agreement(d),and (4) mean square error(MSE).The NRMSE provides a measure of the relative difference between the observed and simulated values.Model simulations having NRMSE <10% were identified as excellent,those having NRMSE of 10–30%as satisfactory,and those having NRMSE >30% as poor.MBE is the mean prediction error,representing the systematic error of a simulation model to undershoot or overshoot the forecast.The MBE indicates the degree to which the observed values are under-or oversimulated by the model.The statisticdshows the ratio of the mean square error and the potential error.The range ofdis 0–1,with advalue closer to 1 representing better simulation results.MSE was used to produce a single value for the goodness of fit of the regression line.MSE with smaller values implies better performance of the simulation model.These statistical indices follow R?tter et al.[50]and Liu et al.[11].The four indices were calculated by Eqs.(8)–(11).

whereXobs,i,Xsim,iare the observed and simulated values for theith dataset andis the mean of the observed values.

3.Results

3.1.Model improvement:simulating extreme high-temperature stress effects on N uptake and grain N accumulation

The original WheatGrow model did not incorporate the effect of extreme HTS on N uptake and rate of grain N accumulation.We accordingly quantified the relationship between N uptake,rate of grain N accumulation,and extreme HTS,to simulate extreme HTS effects.

3.1.1.Total N uptake by plant

In the original WheatGrow model,no extreme-temperature effects were considered for the simulation of plant N uptake.However,a significant negative relationship between relative aboveground N accumulation and HTS at anthesis and grain filling was noted in Fig.2.For this reason,algorithms for HTS effects on N uptake (f(HTS_Nuptakei)) were introduced into the Wheat-Grow model to simulate the impact of HTS on aboveground N(Eq.(12)).

For HTS treatments at anthesis and grain filling stages,f(HTS_Nuptakei)was quantified by Eq.(13),as the best nonlinear curve fit between N uptake and heat degree days (HDD) (Fig.2).HDD was used to quantify the HTS,as it considered both the duration and intensity of HTS events.

whereSHTS_Nuptakeis the slope of the curve and reflects the sensitivity of N uptake to HTS.No differences in the response of N uptake to HTS at anthesis and grain filling were observed,and accordingly the same equation was used for HTS at these two stages.Genetic differences inSHTS_Nuptakebetween the two cultivars could be observed(Fig.2).Accordingly,SHTS_Nuptakewas set as 0.51 and 0.57 for Yangmai 16 and Xumai 30,respectively,under post-anthesis HTS.The exact values of this genetic parameter for the two cultivars were chosen by curve fitting based on the observed values in Fig.2.

Fig.2.The relationship between relative N uptake during the grain-filling period and heat degree days (HDD) for HTS treatments at anthesis and grain filling in two growing seasons.The relative N uptake was calculated as the ratio of the absolute values under the different treatments to the corresponding values from the control treatment (T1 treatment) for the same treatment,stage,and cultivar.

HDDirefers to the cumulative heat degree days (HDi) beyond the threshold temperature (Th) on theith day after treatment and was calculated by Eqs.(14) and (15).

where HDtis the hourly heat degree days andTt(°C)is the air temperature at thetth hour of a day,as recorded by EM50 data loggers installed in the phytotrons.The temperature fluctuations in phytotrons were precisely controlled to simulate the diurnal change patterns observed under field conditions (Fig.S1).The threshold temperature (Th) for HTS was set at 30 °C [5,51].

3.1.2.Grain N accumulation

According to previous studies [9,29],HTS increases the rate of grain N accumulation during the grain filling period,a finding corresponding with the findings of this study based on the HTS experiments (Fig.3).Because the original WheatGrow model did not account for the effects of extreme-temperature events on the rate of grain N accumulation,f(HTS_GrainNi)was integrated into the WheatGrow model to quantify these effects (Eq.(16)).

For the effect of HTS at anthesis and grain filling on the rate of grain N accumulation,f(HTS_GrainNi)was quantified as

whereSHTS_GrainNis the genetic parameter indicating the sensitivity of the rate of grain N accumulation to HTS.SHTS_GrainNfor the two cultivars was evaluated by nonlinear fitting as shown in Fig.3.There were genotypic differences between cultivars for HTS at anthesis,whereas no genotypic difference was noted for HTS at grain filling.Accordingly,0.61 and 0.46 were set as genetic parameters for Yangmai 16 and Xumai 30,respectively,at anthesis,while 0.46 was assigned as a genotypic parameter for both cultivars at grain filling.The exact values of this genetic parameter for the two cultivars were estimated using curve fitting based on the observed values in Fig.3.

Fig.3.The relationship between the relative rate of grain N accumulation during the grain-filling period and the heat degree days (HDD) for HTS treatments at (a)anthesis and (b) grain filling in two growing seasons.The relative rate of grain N accumulation was calculated as the ratio of the absolute values under the different treatments to the corresponding values from the control treatment (T1 treatment) for the same treatment,stage,and cultivar.

The data collected from the first year (2015–2016) experiment were used to calibrate the two winter wheat cultivars in the WheatGrow model.The parameter values for Yangmai 16 and Xumai 30 are given in Table 1 for aboveground N and grain N accumulation.The original WheatGrow introduced the maximum rate of grain N accumulation,and PLFD to quantify the genetic differences in grain-filling duration and rate of grain N accumulation between different cultivars.We adjusted the values of these two parameters by trial and error with the observed datasets in model calibration.To better explain these genetic differences as well as to expedite future application of the improved model,we accordingly introduced two HTS sensitivity parameters in Table 1 to quantify the genetic heat-stress sensitivity in the improved WheatGrow model for aboveground N and rate of grain N accumulation.The exact values of these genetic parameters were determined using curve fitting based on the values in Figs.2 and 3,when the algorithms for model improvement were developed.Lastly,model evaluation for both original and improved WheatGrow models was executed after model calibration by comparing simulated aboveground N,grain N accumulation,and GPC with the measured values from the growing season of 2016–2017.

3.2.Simulation of total aboveground N under HTS

Aboveground N decreases with increasing temperature levels and durations,as shown in Fig.4 and S2.The observed values of total aboveground N show that there were clear differences among different temperature levels and these differences became more apparent with increase in duration of HTS.Figs.5 and 6 and S3 and S4 show the comparison of observed and simulated total aboveground N dynamics for D1(3 days)and D2(6 days)durations for Yangmai 16 and Xumai 30 under HTS for the season of 2016–2017.HTS treatments (T2,T3,and T4) at anthesis and grain filling and combined HTS at both anthesis and grain filling reduced total aboveground N.This reduction increased in severity with rising temperature levels during HTS.Among the growth stages,total aboveground N showed more sensitivity to HTS at anthesis than at the grain-filling stage.HTS caused more reduction in aboveground N for D2 than for D1.The original WheatGrow model tended to reproduce some reduction in total aboveground N under HTS but still undervalued heat stress effects,especially in D2(Figs.6 and S4).However,after new HTS function (Eq.(13)) was incorporated,there was considerable improvement in the simulation results of total aboveground N under normal and HTS conditions (Fig.7).Table 2 shows the statistical indices,which reveal a decrease in NRMSE of total aboveground N simulation for Yangmai 16 and Xumai 30 under HTS treatments from respectively 7.43%to 3.78%and 5.14%to 3.52%for the improved,relative to the original,WheatGrow model.The MBE declined from 8.26 kg ha-1to 2.43 kg ha-1and 4.70 kg ha-1to -2.11 kg ha-1,whereasdincreased from 0.92 to 0.98 and 0.95 to 0.98 for Yangmai 16 and Xumai 30,respectively,under HTS.The decreases in MSE for Yangmai 16 and Xumai 30 were respectively 74% and 53% for the improved version of the WheatGrow model.Finally,the new routines of total aboveground N and HTS functions greatly improved simulations of N uptake under HTS treatments.

Table 2 Statistical indices for validation of the original and improved WheatGrow models in simulating total aboveground N and grain protein concentration (GPC) in environmentcontrolled phytotron experiments under different high temperature levels.

3.3.Simulation of grain protein concentration under HTS

Fig.4.Effects of different high temperature levels (T1–T4) and durations (D1–D2) on total aboveground N of Yangmai 16 during the 2016–2017 growing season.Different letters indicate significant differences according to Tukey’s HSD test (P ≤0.05).Vertical bars represent the standard deviation of the mean.

Fig.5.Comparison of simulated and observed aboveground N dynamics of Yangmai 16 with the original(a–c)and the improved(d–f)WheatGrow model under three days of heat stress treatments in the 2016–2017 growing season.DOY,day of year.

Fig.6.Comparison of simulated and observed aboveground N dynamics of Yangmai 16 with the original(a–c)and the improved(d–f) WheatGrow model under six days of heat stress treatments in the 2016–2017 growing season.DOY,day of year.

Fig.7.Comparison of simulated and observed total aboveground N under heat stress treatments with the original and the improved WheatGrow model for calibration(a and b) and validation (c and d).

Fig.8.Comparison of simulated and observed grain protein concentration dynamics of Yangmai 16 with the original(a–c)and the improved(d–f)WheatGrow model under three-day heat stress treatments in the 2016–2017 growing season.DAA,day after anthesis.

Fig.9.Comparison of simulated and observed grain protein concentration dynamics of Yangmai 16 with the original(a–c)and the improved(d–f)WheatGrow model under six-day heat stress treatments in 2016–2017 growing season.DAA,day after anthesis.

Grain protein concentration (GPC) increased with HTS levels and durations.Figs.8 and 9 and S10–S11 depict the observed and simulated dynamics of GPC under HTS for Yangmai 16 and Xumai 30.Among different stages,a large increase in GPC occurred under combined stresses at anthesis and grain filling followed by HTS at anthesis.After incorporating the heat-stress effects on rate of grain N accumulation(Eq.(17)),the simulations of grain N accumulation under HTS were significantly improved (more details in Supplementary file).The original WheatGrow model underestimated the increase of GPC under HTS,whereas the improved model simulated the dynamics of GPC satisfactorily under HTS(Fig.10).Moreover,the improved WheatGrow model showed reasonable responses for the impacts of HTS at different growth stages,temperatures,and durations,as shown in Figs.8 and 9 and S10 and S11.The NRMSE of GPC for Yangmai 16 and Xumai 30 based on the improved WheatGrow model was increased by 81% and 79%,respectively,compared with simulations with the original WheatGrow model.Likewise,from the original Wheat-Grow model to the improved WheatGrow model,the MBE decreased from -2.53% to -0.22% and -1.90% to 0.01%,while thedvalue increased from 0.62 to 0.98 and 0.63 to 0.98 for Yangmai 16 and Xumai 30,respectively.Finally,the improved WheatGrow model findings revealed a decrease in MSE by 96% for both cultivars(Table 2).In summary,the improved WheatGrow model simulated GPC dynamics well under normal (T1) and HTS conditions(T2,T3,and T4).

4.Discussion

Fig.10.Comparison of simulated and observed grain protein concentration under heat stress treatments with the original and the improved WheatGrow model for calibration (a and b) and validation (c and d).

Total aboveground N of wheat is either directly or indirectly affected by the processes of soil N supply and plant N demand.A decline in biomass,LAI,and grain yield [43] under HTS could reduce total aboveground N.An increase in HTS level and duration may also result in a high decrease in total aboveground N,in view of the negative relationship observed between total aboveground N and HDD(Fig.2).This result accords with previous findings.The HTS at anthesis,grain filling,and combined heat stress at anthesis and grain filling caused a reduction in grain weight (Fig.S12) and grain number [12,43],which ultimately led to reduced postanthesis N uptake owing to a reduction in sink size [26,52].There was more reduction of total aboveground N under HTS applied at both anthesis and grain filling than under HTS at individual stages,and this finding could be attributed to a greater reduction in kernel number,kernel size,and grain yield.The ability of the original WheatGrow model to simulate the effect of HTS on aboveground N uptake could be explained by the recent improvements by Liu et al.[38].Recent improvements in the WheatGrow model were made by incorporating reduction functions for phenology,growth,and grain yield to quantify the impact of HTS on these processes[11,38].However,other factors such as reduction in mineral N uptake by roots,disruption of enzymes involved in nutrient metabolism,xylem translocation,phloem cycling,root-to-shoot translocation of N,and depletion of labile C because of decrease in transport of shoot C to roots or increased root respiration may lead to a decrease in aboveground N under HTS[53].After the incorporation of a new temperature response function (Eq.(13)) to simulate the effect of HTS on total aboveground N,the improved WheatGrow model showed reasonable estimates for N uptake in model evaluation (Fig.7).

Another factor in the effect of HTS on grain quality is the decline in grain N accumulation.Previous studies[29,54]have shown that HTS restricts the rate of starch accumulation and accelerates grain N accumulation.Similarly,the present study also showed that the accumulation rate of grain N increases with temperature and duration(Fig.3).However,despite the increase in grain N accumulation rates,grain N accumulation at maturity may decrease owing to accelerated phenology and senescence[29].In a recent study,Nuttall et al.[8] applied an increment function for quantification of grain N accumulation under HTS.But according to other studies[29,54],the rate of grain N accumulation is the primary determinant of grain N accumulation.For this reason,a new increment function was integrated into the WheatGrow model for the rate of grain N accumulation under HTS,and improved WheatGrow model simulations of this response(more details in Supplementary file,Table S2).

An increase in GPC occurs under HTS,and our findings agreed with previous studies [6,13,29,55].Incorporation of the HTS temperature function improved the accuracy of the WheatGrow model for different cultivars and temperature levels(Table S3).A comparison of simulation estimates from the improved WheatGrow model with the GPC observed under heat stress by Asseng et al.[56]reveals that our improved model can satisfactorily simulate the effect of HTS on GPC.For instance,compared with the control,the absolute increase in observed GPC under heat shock treatment during grain filling with an HDD of 16.5°C day-1was 2.25%,while the absolute increase in GPC simulated by the improved Wheat-Grow model was 3.18% under HTS at a similar growth stage with an HDD of 17.6 °C day-1.

According to Nuttall et al.[4],crop models should account for genotypic parameters,particularly under extreme climate conditions,so that breeders can easily identify the genotypes of plants(by using the genotypic parameters) and improve grain yield and grain protein simultaneously.However,studies [57,58] that have improved the simulation of HTS effects have ignored the genetic variability of heat stress sensitivity.Although genetic variability and environmental variation in GPC have made experimentation and management more difficult [45],the inclusion of a genetic parameter for grain N accumulation and sensitivity of cultivars to HTS could increase the ability of the WheatGrow model to better capture the interactions between genotype,environment,and management.This is because there is an urgent need for improved wheat crop cultivars that fulfill the market demand for different GPCs [3].The parameters in the model developed in the present study(Eqs.(13)and(17))can be divided into two categories:constant and genetic parameters.Constant parameters are numbers that do not change for different cultivars.For example,in the WheatGrow model,potential N uptake rate is a constant with a value of 0.6 N m-2day-1and is assumed to be the same for all cultivars,following Pan et al.[25]and Asseng et al.[45].The values of constant parameters(0.042 and 0.083 in Eqs.(13)and(17),respectively)in this study were determined by curve fitting,and they are fixed for all other cultivars in future model applications.The genetic parameters for heat-stress sensitivity should vary for cultivars with different heat-stress sensitivities,and they were determined using the observed relationship for the two cultivars in the phytotron experiments (Figs.2 and 3).As a general approach to crop model application,genetic parameters for new cultivars should be calibrated with observed data in future applications of the improved model.Thus,the values of genetic parameters for new cultivars can be determined by calibration methods,such as trial and error,when observed grain quality data under HTS are available.By adjusting the values of these genetic parameters,we could determine the exact parameter values by minimizing the differences between simulated and observed GPC.However,when no observed GPC data under HTS are available for calibration,a possible value range of these parameters for different types of cultivars should be provided as default input or as a reference for deriving the values.This aim could be achieved by building a parameter library for the representative cultivars across the main wheatproducing region of China.This resource would greatly facilitate the application of our model to other cultivars.The values of genetic parameters for these representative cultivars might be determined from the literature and by experimentation [59].

High tolerance to extreme temperatures has been suggested as one of the most critical crop adaptation strategies for future climate change by Gouache et al.[60] and Zheng et al.[61].With the information about HTS sensitivity,our improved model could assist in selecting cultivars according to their environment for high yield and targeted GPCs,especially under HTS conditions.Another critical question for the future,then,is how to manage the relationship between grain yield and grain N concentration,where the main objective is to increase the level and stability of yield and N concentration.For example,Asseng et al.[56] showed that a combination of delayed anthesis and higher grain-filling rate increases grain protein yield when N is not a limiting factor.Several studies[1,30,56] have focused on adaptation strategies for grain yield improvement.In this context,it is also critical to quantify the future effects of different adaptation strategies on grain quality,as adaptation strategies may have adverse effects on grain protein formation [56].Our improved model may help in selecting appropriate adaptation strategies to improve grain yield and quality at the same time under HTS scenarios.Validation with independent observed data has shown that the improved WheatGrow model can now simulate the response of two cultivars under various HTS levels and durations during the reproductive period.This capability is a critical prerequisite for the application of crop models to assess future climate change impacts on grain yield and quality under increasing heat-event scenarios.

In the present study,HDD was used to quantify HTS.However,the same HDD could be achieved with a very high temperature in a few days as with relatively low temperatures over a longer time,though the effects of chronic high temperature and heat shock could differ [62,63].For example,under the same heat load (the sum of°C above a base of 21/16°C),more reduction in grain weight was observed under acute heat shock than under chronic high temperature [63].However,because there have been no studies comparing the observed effects of acute heat shock and chronic heat stress on wheat GPC,we could not evaluate our model under chronic heat stress over a longer period.In this study,we focused specifically on the impacts of short-term heat stress.More detailed validation would be needed for applying the improved model to investigate the impacts of chronic high temperature on grain quality.

An increase in CO2concentration may lead to a decrease in GPC of up to 6.3%[64].The introduction of genotypes adapted to warmer temperatures could increase global wheat yield by 7%and protein yield by 2%,but the GPC will decrease by-1.1%,reflecting a relative change of-8.6%[56].For this reason,future model improvement must integrate the interaction of increased CO2concentration and elevated temperature.Given that we conducted the HTS experiments in environment-controlled phytotrons with only two cultivars,the results of the present study could reflect to some extent the effects of HTS on aboveground N,grain N accumulation,and GPC.Future studies are needed to validate the model improvements under diverse growing conditions and cultivar types,especially under field conditions,given that current response routines were based on observed data from pot experiments.Likewise,root growth in pot experiments could differ significantly from field conditions,and this difference could affect N dynamics under extreme temperature events.According to Talukder et al.[65],crops grown in controlled environments may respond differently to HTS than crops grown in the field,owing perhaps to greater soil volume accessible for root exploration,water,and nutrient uptake.In this regard,pot experiments are criticized for exposing root systems,which are sensitive to high temperatures,to unrealistic microclimate environments,unlike field conditions[66,67].For example,van Herwaarden et al.[66]reported that heat stress under controlled pot conditions could affect growth and senescence more than under field conditions.However,in that case,the duration of HTS was relatively short,whereas the daily variation of temperature in the phytotron mimicked HTS events under field conditions,possibly reducing the biases associated with pot experiments.To avoid the over-response of crops to HTS under controlled conditions and achieve reliable model predictions under field conditions,further calibration and testing for HTS under field conditions will help further improve and validate crop models.In particular,calibration of the parameters in the improved algorithms should be based on field observations.

5.Conclusions

Extreme high-temperature stress(HTS)at anthesis,grain filling,and both stages caused a decrease in aboveground N and accelerated the rate of grain N accumulation,which ultimately resulted in increased GPC.The observed effects of HTS on aboveground N and rate of grain N accumulation were quantified,and newly developed algorithms were integrated into the WheatGrow model.The better fit between observed and simulated values suggests that the improved WheatGrow model can reliably predict aboveground N uptake,grain protein accumulation,and GPC under HTS.The improved model provides better assessments of climate change for use under scenarios of increasing HTS.

CRediT authorship contribution statement

OsmanRaheelcurated data,performed analyses,and wrote the original draft.YanZhucurated data,performed analyses,and revised the manuscript.WeixingCaoconceptualized and revised the review.ZhifengDingcurated data and performed analyses.MengWangcurated data and performed analyses.LeileiLiuand LiangTangrevised the original draft.BingLiuconceptualized and wrote and edited the review.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was supported by the National Key Research and Development Program of China (2019YFA0607404),the Natural Science Foundation of Jiangsu Province (BK20180523),the National Science Fund for Distinguished Young Scholars(31725020),the National Natural Science Foundation of China(31801260,31872848,41961124008,and 32021004),and the China Scholarship Council.

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2020.10.001.