Published online 1 January 2007
Published in Agron J 99:158-165 (2007)
DOI: 10.2134/agronj2006.0090
© 2007 American Society of Agronomy
677 S. Segoe Rd., Madison, WI 53711 USA
Modeling
Estimating Grain and Straw Nitrogen Concentration in Grain Crops Based on Aboveground Nitrogen Concentration and Harvest Index
Armen R. Kemaniana,*,
Claudio O. Stöckleb and
David R. Hugginsc
a Biological Systems Engineering Dep., Washington State Univ., Pullman, WA 99164-6120 (current address: Texas Agricultural Experiment Station, Blackland Research and Extension Center, Temple, TX 76502)
b Biological Systems Engineering Dep., Washington State Univ., Pullman, WA 99164-6120
c USDA-ARS, Washington State Univ., Pullman, WA 99164-6421
* Corresponding author (armen{at}brc.tamus.edu)
Received for publication March 24, 2006.
 |
ABSTRACT
|
|---|
Simulating grain (Ng) and straw (Ns) nitrogen (N) concentration is of paramount importance in cropping systems simulation models. In this paper we present a simple model to partition N between grain and straw at harvest for barley (Hordeum vulgare L.), wheat (Triticum aestivum L.), maize (Zea mays L.), and sorghum (Sorghum bicolor Moench). The principle of the model is to partition the aboveground N at physiologic maturity based on the relative availability of biomass and N to the grain. The inputs for the model are the harvest index (HI), representing the relative availability of biomass to the grain, and the aboveground N concentration (Nt) at harvest, representing the availability of N. The model has five parameters, of which four (the maximum and minimum achievable grain and straw N concentrations) are readily available; the parameter C requires calibration. The model was calibrated and tested for these four species without differentiating genotypes within species. The testing included diverse experiments in wheat; comparing observed and estimated Ng the relative RMSE ranged from 3 to 10% (five experiments) and was 31% in one experiment in which the estimated Ng exceeded consistently the observed values. For barley, maize, and sorghum, the data availability for testing was limited, but the model performed well (relative RMSE values of 7, 7, and 18%, respectively). Therefore, the model proposed seems to be robust. It remains to be determined if the parameters and the method are useful to discriminate genotypic differences in Ng within a species and if the method can be applied to legume crops.
Abbreviations: HI, harvest index • Ng, grain nitrogen concentration • NHI, N harvest index • Ns, straw nitrogen concentration • Nt, N concentration in aboveground biomass at physiologic maturity
|
INTRODUCTION
|
|---|
SIMULATING grain (Ng) and straw (Ns) nitrogen (N) concentration is of paramount importance in cropping systems simulation models. Ng is a major quality determinant of cereal and legume crops. For crop simulation models to be useful in helping producers make informed decision regarding N management, they must provide accurate estimates of Ng. In addition, accurate estimates of the N removed with the grain are needed to keep accurate N balances in short- and long-term simulations.
The basic approach to simulate Ng in process-oriented crop models is to allocate dry matter and N to the grain during grain filling depending on the balance between the grain demand and the supply of these two resources. The degree to which the demand is satisfied by the supply depends on environmental and crop conditions affecting photosynthesis and on the N status of the crop. The approach used by Ritchie et al. (1985) in wheat, which was modified by Asseng et al. (2002), assumes that the daily demand of dry matter and N for each grain is independent. The demand is determined by the maximum daily grain growth and N deposition rates, which are empiric functions of temperature. The optimum temperature for N deposition in the grain is higher than that for dry matter, and therefore the simulated Ng tends to increase as temperature increases. The supply of dry matter depends on current photosynthesis and pre-stored reserves, and the supply of N depends on the N concentration of roots, leaves, and stems, which can be depleted until they reach a minimum allowable N concentration. Larmure and Munier-Jolain (2004) proposed a conceptually similar approach to model Ng in peas. This model is not linked to a comprehensive cropping system simulation model and requires considerable input of physiologic parameters to run (number of grains and individual grain growth rate at each reproductive node, rate of progression of the beginning and end of grain filling along the nodes in the stem, and genotype-dependent maximum grain and N deposition rate).
Jamieson and Semenov (2000) followed a slightly different approach. They assumed that the minimum Ng is 15 g N kg–1 and that the N harvest index (NHI) increases linearly during grain filling as a function of thermal time, so that the NHI at physiologic maturity is 0.8. An allowance is made for NHI to be greater than 0.8 in the event that the demand of N by the grain is met by postanthesis N uptake. The practical effect is that Ng is basically determined by the supply of total dry matter during grain filling: the lower the supply of dry matter, the higher Ng. None of these models was built as a generic model for grain crops.
The objective of this paper is to present a simple model of N partitioning between grain and straw at harvest. The inputs for the model are the harvest index (HI) and the aboveground biomass N concentration at physiologic maturity (Nt). This information is readily produced by cropping systems simulation models like CropSyst (Stöckle et al., 2003) and EPIC (Williams, 1995), which calculate Nt directly (i.e., independently of Ng and Ns). The model requires minimum calibration to accommodate differences between genotypes or species.
|
MODEL DESCRIPTION
|
|---|
The basic assumptions of the model are (i) there is a minimum Ng (Ngn) and Ns (Nsn) that must be satisfied for growth to take place; (ii) there is a maximum Ng (Ngx) and Ns (Nsx) that cannot be exceeded; (iii) the grain (Ngd) and straw (Nsd) N demands above the minimum concentrations are given by Ngd = Ngx – Ngn and Nsd = Nsx – Nsn, respectively; (iv) at harvest, all N above the minimum concentration (Na) is considered available for allocation to grain or straw; (v) the proportion of Na allocated to the grain depends on the grain N demand Ngd and the total aboveground N demand (Ngd + Nsd). These assumptions have been compiled using the functional equations shown below.
The actual Ng depends on how much of Na is allocated to the grain and on HI as follows:
 | [1] |
where Na is the N available for allocation expressed as a concentration quantity:
 | [2] |
where Nt is the aboveground biomass N concentration at physiologic maturity, and Pg is a grain partitioning factor computed as
 | [3] |
Multiplying Na from Eq. [2] by the aboveground biomass gives the N mass in excess of that required to satisfy the minimum concentration of grain and straw and is therefore available for allocation to grain or straw. The term within brackets in the first line of Eq. [3] represents fractionally what would be the partition of Na to the grain if Ngx and Nsx are met; under such conditions R = 1 as explained below. Similarly, multiplying Ng from Eq. [1] by the grain yield gives the grain N mass. The power R is computed as follows:
 | [4] |
The term within brackets represents the fraction of the N needed to reach the maximum concentration in the aboveground biomass that is satisfied by Na and can be interpreted as the degree of "saturation" on N of the aboveground biomass. If the aboveground N biomass satisfies only Ngn and Nsn, then R = 0 because Na = 0; if it is sufficient to satisfy Ngx and Nsx, then R = 1. The power C is a dimensionless empiric factor that allows adjusting Pg for cultivar or species effects: the higher the value of C, the higher the priority of the grain as a sink for Na. The grain partitioning factor Pg is therefore the partitioning of Na to grain if grain and straw reach their maximum N concentration, adjusted through R by the actual availability of N.
The parameters Ngx, Ngn, Nsx, and Nsn are considered constants that depend on the species or cultivar. Therefore, to compute Ng based on Eq. [1], the only inputs required are Nt and HI. Once Ng has been determined, Ns can be calculated from:
 | [5] |
|
MATERIALS AND METHODS
|
|---|
Data from numerous sources for wheat, barley, maize, and sorghum were collected and used to calibrate and test the model. The specific information collected was HI, Nt, Ng, and Ns. The criteria for selecting data were that besides having available HI, Nt, Ng, and Ns, the data showed a reasonable range of variation in HI, Ng, or both. Data sets with the widest range of variation in one of these variables were favored for calibration. The parameters Ngx and Ngn were not calibrated but were derived from an analysis of several data sets showing the apparent biological boundaries of these parameters for each species. For wheat and barley, the parameter C was calibrated using a data set from the Cook Agronomy Farm (46°47' N, 117°5' W, elevation 773–815 m) located 8 km north east of Pullman, WA, in the years 1999 and 2001 (spring wheat) and 2000 (spring barley) (Huggins, unpublished data). For maize, the parameter C was calibrated using a limited data set given in Huggins et al. (2001) and Derby et al. (2005). For sorghum, the parameter C was calibrated using a limited data set given in Kamoshita et al. (1998a). For testing purposes, we used several data sets collected for our own team or retrieved from the literature. The optimization was performed by setting an algorithm seeking the least square difference between observed and predicted Ng by changing the parameter C.
Depending on the choice of parameters and on the values of HI and Nt, the computed Ng can exceed the allowable maximum (Ngx) or fell below the allowable minimum (Ngn) in extreme cases, when dealing with very high or very low Nt or HI. Similarly, Pg can exceed unity in the computations. Therefore, if in the computation Ng > Ngx, then Ng is set to Ngx; if Ng < Ngn, then Ng is set to Ngn. Similarly, if Pg > 1, then Pg is set to 1.
|
RESULTS
|
|---|
Calibration
We analyzed information on Ng and Ns to define objectively Ngx, Ngn, Nsx, and Nsn for these crops. Selected results are shown in Table 1. For wheat and barley, Ngx seems to be between 35 and 40 g kg–1 and Ngn between 11 and 12 g kg–1. For comparison, Ng of soybean is typically 60 g kg–1 (e.g., Huggins et al., 2001). It is likely that there is genotypic variation in these parameters; however, the information reviewed prevents drawing definite conclusions in that regard. For straw, Nsn and Nsx are in the order of 2 and 14 g kg–1. Larmure and Munier-Jolain (2004) discussed the possibility that crops well nourished with N can have higher Ngn than crops with low N status. We explored the impact of changing Ngn and other parameters of the model in the sensitivity analysis presented in the Discussion section.
View this table:
[in this window]
[in a new window]
|
Table 1. Selected information on grain and straw maximum and minimum nitrogen concentration at harvest (Ngx, Ngn, Nsx, Nsn, respectively) of wheat, barley, maize, and sorghum.
|
|
Maize and sorghum have generally lower Ng than wheat or barley. We found difficulties in finding relatively high Ng or Ns in experiments with these crops. For maize, the Ng of hybrids typically grown by producers rarely exceeds 15 g kg–1 in field conditions (Table 1). Uribelarrea et al. (2004) presented useful information on the biological aptitude of maize to produce grains with high or low Ng by using hybrids generated from the Illinois Protein strains, obtained under several cycles of selection for low and high Ng. They showed Ng ranging from 7 to 29 g kg–1 (Table 1). The minimum values are in accord with those presented by Bodley (2004) and Derby et al. (2005). Wyss et al. (1991) presented a surprisingly high value of Ng of 47 g kg–1 for the Illinois Protein strain line selected for high protein, a major difference compared with values from hybrids. Kamoshita et al. (1998a) presented data for sorghum showing that Ng can reach values comparable to those of wheat or barley (29 g kg–1), albeit in crops with extreme postanthesis stress and ample N supply. This value is similar to that reported for maize (Uribelarrea et al., 2004). We assumed that Ngn reported for maize applies for sorghum as well. Maximum and minimum Ns values for sorghum straw are similar to those reported for barley and wheat (ca. 14 and 2 g kg–1) (Table 1), and we assume that they also apply to maize. Our choices for Ngx, Ngn, Nsx, and Nsn for these four crops are presented in Table 2.
View this table:
[in this window]
[in a new window]
|
Table 2. Grain and straw maximum and minimum nitrogen concentration at harvest (Ngx, Ngn, Nsx, Nsn, respectively) and the optimized value for the parameter C of wheat, barley, maize, and sorghum used to estimate grain and straw nitrogen concentration at harvest.
|
|
We used a set of experiments for each crop to estimate the parameter C. The results of the calibration are shown in Fig. 1
. For spring wheat, we used information collected in Pullman, WA, in which the source of variation was N fertilization rates and within-field spatial variation. The agreement between predicted and observed values was reasonably good (C = 0.72, r2 = 0.92, n = 336, RMSE = 0.8 g kg–1). In the case of barley, the information was also collected in Pullman, WA, and, similar to wheat, the calibration yielded very good results (C = 0.19, r2 = 0.86, n = 139, RMSE = 0.9 g kg–1). In both cases, there was a tendency for the results obtained with the calibrated model to overestimate the lower Ng and to underestimate the higher Ng values, as reflected by the slopes between predicted versus observed (ca. 0.9) reported in Fig. 1. For maize and sorghum, we do not have the abundance of data we have for wheat and barley. Therefore, we combined information from Derby et al. (2005) and Huggins et al. (2001) to calibrate the parameter C for maize and used one experiment reported in Kamoshita et al. (1998a) to calibrate the parameter for sorghum. The range of Ng in the case of maize was fairly narrow. Nevertheless, for maize and sorghum, the results of the calibration were satisfactory (Fig. 1). The values of C obtained in the calibration are summarized in Table 2.
View larger version (29K):
[in this window]
[in a new window]
|
Fig. 1. Calibration of the parameter C for wheat, barley, maize, and sorghum. For wheat and barley, the data are from Pullman, WA; for maize, data are from 1 yr from Derby et al. (2005) and from Huggins et al. (2001); for sorghum, data are from Kamoshita et al. (1998a). Wheat average harvest index (HI) and aboveground nitrogen concentration (Nt) were 0.42 (range 0.30–0.56) and 11.5 (range 8.1–16.7 g kg–1); barley average HI and Nt were 0.43 (range 0.35–0.53) and 9.6 (range 6–15.3 g kg–1); maize average HI and Nt were 0.55 (range 0.36–0.68) and 8.3 (range 5.1–10.8 g kg–1); sorghum average HI and Nt were 0.37 (range 0.18–0.48) and 10.4 (range 6.3–15.8 g kg–1). RMSE and MAD are RMSE and mean absolute difference between observed and predicted Ng.
|
|
Model Testing
Several data sets independent from those used in the calibration were used for model testing. Figure 2
shows the testing results for six different experiments with wheat. An overall evaluation indicates an excellent performance of the model across a range of localities, N fertilization rates, water availability, rotations, and cultivars. The Ng data reported by Fischer (1993) and Fischer et al. (1993) for spring wheat, corresponding to several N rates and application timing, were satisfactorily estimated by the model, except for one point that was overestimated. This point corresponded to the maximum N application rate of the experiment (240 kg N ha–1). The observed Nt and HI were 13 g kg–1 and 0.35, respectively, for which the model predicts Ng of 27.5 g kg–1 (NHI = 0.74), whereas the observed value was 23.5 g kg–1 (NHI = 0.63). McDonald (1992) reported the average Ng for three spring wheat cultivars at four different sites and with different N fertilization rates. The agreement between estimated and observed Ng was excellent (Fig. 2); the model captured the effect of the environment and the effect of the fertilization rate in each site. Halvorson et al. (2004) presented data for winter wheat for 9 yr with five N fertilization rates. Although the parameter C was calibrated for spring wheat, we tested the model for their winter wheat data as well. The overall agreement between estimated and observed Ng was good, with a tendency of the model to overestimate Ng at the higher end. Within each year, the model represented correctly the increase in Ng with increasing N application rate. Except for one year, the estimated Ng was within 10% of the observed value.
Wuest and Cassman (1992) and Huggins (1991) presented experiments in which the timing of N application was varied to favor N uptake during grain filling. The Wuest and Cassman (1992) experiments were conducted in irrigated wheat with the N applied pre-planting and at anthesis. The Huggins (1991) experiment was conducted in a Mediterranean climate where precipitation after anthesis is scarce. Therefore, N was applied at planting and in the fall of the previous year to allow N to penetrate deep in the profile with the infiltrating water during winter and early spring. Results of estimated versus observed Ng for both experiments are shown in Fig. 2. In the experiment of Wuest and Cassman (1992), the model overestimated Ng but correctly represented the increasing Ng at increasing N fertilization rates. Similarly, timing and rate of N fertilization affected Ng in the Huggins (1991) experiment, and the model correctly represented the effect of both variables on Ng (Fig. 2). Adding all or a fraction of the N in fall, as opposed to adding all the N in spring, caused increases in Nt and Ng at harvest of 15 and 10%, respectively, averaged over all N fertilization rates. A second experiment reported by Huggins (1991) involved tillage (no-till versus conventional tillage), preceding crop (Austrian winter peas or winter wheat), and N fertilization rates (range 0–200 kg N ha–1). The model correctly represented the increase in Ng with increasing fertilization rate (Fig. 2).
Diseases affect yield and the deposition of N in the grain. Dimmock and Gooding (2002) reviewed the effect of diseases on Ng and concluded that rusts (Puccinia spp.) and powdery mildew (Erysiphe graminis) infections decrease Ng and increase Ns, but Septoria spp. infections tend to increase Ng, with exceptions. We can speculate that Ng data obtained from plots affected by rusts or powdery mildew will be overestimated by the model. Olesen et al. (2000) presented 2 yr of data for winter wheat grown in Denmark. Treatments included irrigation and N fertilization timing. We compared the Ng reported by these authors with that estimated with our model and found a gross overestimation of Ng (Fig. 2). The absolute Ng values reported were relatively low (average 18.6 and 17.5 g kg–1 for 1996 and 1997, respectively), but the average Ns values were high (average 9.9 and 7.5 g kg–1 for 1996 and 1997, respectively). The authors indicated that a serious infestation of mildew was present in 1996 and that an infestation of Septoria was present in 1997. Therefore, we surmise that the overestimation by the model is due to the effect of mildew, which limits more the N yield than the total yield and thus decreases Ng. However, the argument is weakened when one considers that the effects of Septoria are ambiguous (Dimmock and Gooding, 2002). The application of fungicide in that experiment, which decreased the magnitude of the infections but failed to eliminate them, caused an increase in Ng in both years, consistent with the idea that diseases may explain a portion of the departure of the predicted Ng with respect to the observed. The model seemed to overestimate Ng in all of the irrigated experiments (Fig. 2; one case in Fischer et al., 1993; Wuest and Cassman, 1992; Olesen et al., 2000).
We tested the model for spring barley using data collected by Huggins (unpublished) at the Cook Agronomy Farm and data presented by Bulman and Smith (1993b) for three cultivars. We tested the model for winter barley using data from Delogu et al. (1998). The testing shows good agreement for the Pullman data (Fig. 3
). For the Bulman and Smith data, the model correctly predicted an increase in Ng as the N fertilization rate increased but increasingly overestimated Ng as the fertilization rate increased. For the control with no N applied, the model predicted Ng correctly. We do not have an explanation for the overestimation, but it is plausible that the parameters used were inappropriate for the condition of their experiment. It is worth noting that they reported the average for three cultivars, not the data by cultivar. The averaging could be masking genotypic differences not considered in the model parameters. The Ng data for winter barley of Delogu et al. (1998) were very well estimated by the model (Fig. 3).
View larger version (17K):
[in this window]
[in a new window]
|
Fig. 3. Testing of the model for barley, maize, and sorghum. The data for spring barley are from Pullman, WA (Huggins, unpublished) and from Bulman and Smith (1993b) in an experiment in Canada; their data is the average of three cultivars. Data from Delogu et al. (1998) are for winter barley growing at three nitrogen (N) fertilization rates (0, 80, and 140 kg N ha–1); each point is the average of 2 yr. Data for maize are from Bodley (2004) and Derby et al. (2005) for maize grown at different N fertilization rates in Pullman, WA and Oakes, ND, respectively. Data from Mehdi et al. (1999) correspond to 2 yr and different tillage practices. The data for sorghum from Kamoshita et al. (1998b) are from three hybrids grown at 0 and 240 kg N ha–1, and data from Traore and Maranville (1999) are for different genotypes adapted to the experimental area (Nebraska) or adapted to tropical growing conditions.
|
|
For maize, the testing was performed using the data presented by Bodley (2004), Derby et al. (2005) (data from a different year than that used in the testing), and Mehdi et al. (1999) (Fig. 3). The model underestimated Ng from Bodley's (2004) data but represented well the tendency of Ng to increase with increasing fertilization rate. Similarly, the model slightly underestimated the values given by Derby et al. (2005); however, the predicted values were within 10% of the observed Ng, except for one case that departed 13% from the observed. The Ng data presented by Mehdi et al. (1999) were very well estimated by the model. The two clusters of data belong to different years. The estimated Ng was within 5% of the observed Ng. In general, Ng of maize is between 10 and 15 g kg–1, a relatively narrow range compared with that of wheat or barley (Fig. 1
–3).
For sorghum, the testing was performed using data from Kamoshita et al. (1998b) for three hybrids grown at 0 and 240 kg N ha–1 and from Traore and Maranville (1999) for different genotypes adapted to the experimental area (Nebraska) or adapted to tropical growing conditions (Fig. 3). For both data sets, the model estimated the observed Ng reasonably well. However, for two points from Traore and Maranville (1999), the model overestimated Ng by 14 and 25%. In one case (14%), the overestimation corresponds to a line adapted to the experimental area growing conditions, and the reasons for the overestimations are not clear. The case in which the overestimation was the greatest (25%) corresponds to a genotype adapted to tropical conditions. In the experiment, the reported HI for that genotype was 0.07, an extremely low value for sorghum. The model seems to have difficulties handling extreme conditions. No data regarding the environmental and agronomic conditions of the plots were provided, and events such as frost could have affected grain filling in this tropical genotype.
|
DISCUSSION
|
|---|
The method proposed to estimate Ng is simple and requires minimal inputs. The principle of the model is similar to that used in mechanistic models: It is based on the relative availability of carbon or total biomass and N. The HI represents the "availability" of biomass for the grain, and Nt represents the availability of N. The allocation of N to the grain is made at harvest, not on a day-by-day basis, as is done in mechanistic models with daily time-step. It can be argued that daily (or even hourly) information generated during the simulation is not efficiently used when the final decision on how much N is allocated in the grain is made at harvest.
The meaning of HI and Nt in the model is illustrated in Fig. 4
, where Ng is shown as a function of Nt and HI. For a given HI, Ng increases as Nt increases. For a given Nt, Ng decreases as HI increases, reflecting the dilution effect of increasing HI on Ng. The parameter C effect is also illustrated in Fig. 4 using the parameters calibrated for wheat and maize. For both crops, we fixed HI to 0.45 and graphed the change in Ng as a function of Nt. Wheat, which has a C constant greater than that of maize, tends to favor the grain as N sink rather than the straw. In maize, the priority given to the grain is moderated compared with that of wheat. The reasons for such difference in the physiology of these two crops are not clear. Elucidating the reasons could help in developing cultivars for high or low Ng. The model clearly shows that increasing HI while keeping Nt unchanged leads to a decrease in Ng. This is not desirable in crops like hard red spring wheat, for which the objective is to achieve Ng above approximately 20 g kg–1, but it is a logical way of keeping low Ng in malting barley, where Ng above approximately 20 g kg–1 is detrimental to the malt quality.
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 4. Modeled variation of the grain nitrogen (N) concentration in response to aboveground biomass N concentration at harvest and to the harvest index using the parameters fitted for wheat and maize.
|
|
Table 3 presents a sensitivity analysis of the parameters based on the calibration for wheat. All the parameters were increased or decreased by 20%, and the relative change in Ng was tabulated for several combinations of HI and Nt. The parameter that affected the Ng estimations the most was Ngx. One reason for that is that it is numerically the parameter with the maximum absolute value. In all cases, changing the parameters by 20% produced changes in Ng of less than 20%. In the worst case, changing Ngx by 20% changed Ng by 13% (Table 3). The parameter C showed relatively low sensitivity, with changes in Ng of less than 4% in response to changes in C of 20%. If cultivars or species vary in the values of the parameters, detecting differences in just one parameter could be challenging. Genetic differences in Ng have been suggested in wheat (Sofield et al., 1977). Perhaps the Illinois Protein strain lines of maize represent the most striking case of genotypic differences in Ng and variables related to the N and carbon metabolism within a species (Dudley and Lambert, 2004). It seems clear that the model correctly discriminates physiologic differences among species, as illustrated by the differences in the parameters among the four crops. It remains to be proven if the parameters of the model are able to capture differences among genotypes within a species.
View this table:
[in this window]
[in a new window]
|
Table 3. Sensitivity analysis of the model parameters. The parameters calibrated for wheat and shown in Table 2 were changed by plus or minus 20%, and the relative change in grain nitrogen concentration (Ng) with respect to original calibration is reported for three harvest index (HI) and aboveground nitrogen concentration (Nt) values. Ngx, Ngn, Nsx, Nsn are the maximum and minimum, grain and straw nitrogen concentration, respectively; C is an empirical parameter of the model.
|
|
As presented, the model does not consider differences on how the final HI and Nt are achieved. For example, the effect of N uptake timing, if any, is not represented in the model. It can be proposed that two crops with identical HI and Nt, but with one acquiring all the N preanthesis and the other acquiring a sizable fraction of the N postanthesis, would differ in the final Ng. Data by Huggins (1991) and Wuest and Cassman (1992) strongly suggest that all the effect is contained in Nt and that timing per se does not affect Ng unless Nt is affected. The data analyzed for wheat also suggest that for irrigated crops the model tends to overestimate Ng (the model was validated for dryland spring wheat). A plausible explanation is that crops without water stress rely mostly on current photosynthesis for grain filling instead of reserves remobilization (Gallagher et al., 1975). The remobilization of reserves to the grain includes N compounds, whose remobilization is limited if grain filling is performed mostly with current photosynthesis. If that is the case, the model could accommodate this by making the parameter C a function of a water stress index during grain filling: the lower the water stress, the lower the value of C.
A disadvantage of this method is that the potential contribution of N remobilized from the roots is not represented in the model. We have tried this method only in nonlegume crops, but it would be relevant to calibrate the parameters or modify the method for a legume crop like soybean. Given that legumes are self-sufficient in N acquisition, we hypothesize that differences in Ng derive mostly from differences in HI. A major advantage of this model is the transparency with which Ng is determined. In a crop simulation model, it would be meaningless to expect, or even to obtain, a correct estimate of Ng, when the simulated Nt or HI depart from reality.
|
CONCLUSIONS
|
|---|
The method proposed here to partition N between grain and straw at harvest in grain crops seems to be robust. Four out of the five parameters in the model were obtained from field experiments, and one was calibrated based on observed values of HI, Nt, and Ng, which suggests that this model can be easily parameterized for other species or, if necessary, growing conditions. It remains to be determined if the parameters and the model are useful to discriminate genotypic differences in Ng within a species and if the model can be applied to legume crops.
|
ACKNOWLEDGMENTS
|
|---|
The authors acknowledge the generosity of Dr. Ardell D. Halvorson (USDA-ARS Fort Collins, CO) and Dr. Nathan E. Derby (Dep. of Soil Sci. of North Dakota State Univ.) for sharing the original data for winter wheat in Halvorson et al. (2004) and for maize in Derby et al. (2005), respectively. Dr. Richard T. Koening (Dep. Crop and Soil Sci. of Washington State Univ.) made valuable suggestions to the original manuscript. Funding for this research was provided by the Paul G. Allen Family Foundation through the Climate Friendly Farming Project of Washington State University's Center for Sustaining Agriculture & Natural Resources.
|
REFERENCES
|
|---|
- Asseng, S., A. Bar-Tal, J.W. Bowden, B.A. Keating, A. Van Herwaarden, J.A. Palta, N.I. Huth, and M.E. Probert. 2002. Simulation of grain protein content with APSIM-Nwheat. Eur. J. Agron. 16:25–42.[CrossRef]
- Birch, C.J., S. Fukai, and I.J. Broad. 1997. Estimation of responses of yield and grain protein concentration of malting barley to nitrogen fertiliser using plant nitrogen uptake. Aust. J. Agric. Res. 48:635–648.
- Bodley, M.K. 2004. Fertility requirements of no-till corn in a semi-arid environment. M.S. thesis. Washington State Univ., Pullman.
- Bonfila, D.J., A. Karnielib, M. Razb, I. Mufradia, S. Asidoa, H. Egozic, A. Hoffman, and Z. Schmilovitchc. 2004. Decision support system for improving wheat grain quality in the Mediterranean area of Israel. Field Crops Res. 89:153–163.
- Bulman, P., and D.L. Smith. 1993a. Grain protein response of spring barley to high rates and postanthesis application of fertilizer nitrogen. Agron. J. 85:1109–1113.[Abstract/Free Full Text]
- Bulman, P., and D.L. Smith. 1993b. Accumulation and redistribution of dry matter and nitrogen by spring barley. Agron. J. 85:1114–1121.[Abstract/Free Full Text]
- Delogu, G., L. Cativelli, N. Pecchioni, D. De Falcis, T. Maggiore, and A.M. Stanca. 1998. Uptake and agronomic efficiency of nitrogen in winter barley and winter wheat. Eur. J. Agron. 9:11–20.[CrossRef]
- Derby, N.E., D.D. Steele, J. Terpstra, R.E. Knighton, and F.X.M. Casey. 2005. Interactions of nitrogen, weather, soil, and irrigation on corn yield. Agron. J. 97:1342–1351.[Abstract/Free Full Text]
- Dimmock, J.P.R.E., and M.J. Gooding. 2002. The influence of foliar diseases, and their control by fungicides, on the protein concentration in wheat grain: A review. J. Agric. Sci. 138:349–366.
- Dudley, J.W., and R.J. Lambert. 2004. 100 generations of selection for oil and protein in corn. Plant Breed. Rev. 24:79–110.
- Fischer, R.A. 1993. Irrigated spring wheat and timing and amount of nitrogen fertilizer: II. Physiology of grain yield response. Field Crops Res. 33:57–80.
- Fischer, R.A., G.N. Howe, and Z. Ibraim. 1993. Irrigated spring wheat and timing and amount of nitrogen fertilizer: I. Grain yield and protein content. Field Crops Res. 33:37–56.
- Gallagher, J.N., P.V. Biscoe, and R.K. Scott. 1975. Barley and its environment. V. Stability of Grain Weight. J. Appl. Ecol. 12:319–336.[CrossRef]
- Halvorson, A.D., D.C. Nielsen, and C.A. Reule. 2004. Nitrogen fertilization and rotation effects on no-till dryland wheat production. Agron. J. 96:1196–1201.[Abstract/Free Full Text]
- Huggins, D.R. 1991. Redesigning no-tillage cropping systems: Alternatives for increasing productivity and nitrogen use efficiency. Ph.D. diss. Washington State Univ., Pullman.
- Huggins, D.R., G.W. Randall, and M.P. Russelle. 2001. Subsurface drain losses of water and nitrate following conversion of perennials to row crops. Agron. J. 93:477–486.[Abstract/Free Full Text]
- Jamieson, P.D., and M.A. Semenov. 2000. Modelling nitrogen uptake and redistribution in wheat. Field Crops Res. 68:21–29.[CrossRef]
- Kamoshita, A., S. Fukai, R.C. Muchow, and M. Cooper. 1998a. Genotypic variation for grain yield and grain nitrogen concentration among sorghum hybrids under different levels of nitrogen fertilizer and water supply. Aust. J. Agric. Res. 49:737–747.[CrossRef][ISI]
- Kamoshita, A., S. Fukai, R.C. Muchow, and M. Cooper. 1998b. Sorghum hybrid differences in grain yield and nitrogen concentration under low soil nitrogen availability: II. Hybrids with contrasting phenology. Aust. J. Agric. Res. 49:1277–1286.
- Larmure, A., and N.G. Munier-Jolain. 2004. A crop model component simulating N partitioning during seed filling in pea. Field Crops Res. 85:135–148.[CrossRef]
- McDonald, G.K. 1992. Effects of nitrogenous fertilizer on the growth, grain yield and grain protein concentration of wheat. Aust. J. Agric. Res. 43:949–967.[CrossRef]
- Mehdi, B.B., C.A. Madramootoo, and G.R. Mehuys. 1999. Yield and nitrogen content of corn under different tillage practices. Agron. J. 91:631–636.[Abstract/Free Full Text]
- Mi, G., L. Tang, F. Zhang, and J. Zhang. 2000. Is nitrogen uptake after anthesis in wheat regulated by sink size? Field Crops Res. 68:183–190.[CrossRef]
- Olesen, J.E., J.V. Mortensen, L.N. Jorgensen, and M.N. Andersen. 2000. Irrigation strategy, nitrogen application and fungicide control in winter wheat on a sandy soil: I. Yield, yield components and nitrogen uptake. J. Agric. Sci. 134:1–11.
- Rao, S.C., and T.H. Dao. 1992. Nitrogen placement and tillage effects on dry matter and nitrogen accumulation and redistribution in winter wheat. Agron. J. 84:1028–1032.[Abstract/Free Full Text]
- Ritchie, J.T., D.C. Godwin, and S. Otter-Nacke. 1985. CERES-Wheat. AGRISTARS Publ. YM-U3-04442-JSC-188892. Michigan State Univ., Ann Arbor.
- Sofield, I., I.F. Wardlaw, L.T. Evans, and S.Y. Zee. 1977. Nitrogen, phosporus and water contents during grain development and maturation in wheat. Aust. J. Plant Physiol. 4:799–810.
- Stöckle, C.O., M. Donatelli, and R. Nelson. 2003. CropSyst, a cropping system simulation model. Eur. J. Agron. 18:289–307.[CrossRef]
- Stoddard, F.L. 1999. Variation in grain mass, grain nitrogen, and starch B-granule content within wheat heads. Cereal Chem. 76:139–144.
- Traore, A., and J.W. Maranville. 1999. Nitrate reductase activity of diverse grain sorghum genotypes and its relationship to nitrogen use efficiency. Agron. J. 91:863–869.[Abstract/Free Full Text]
- Uribelarrea, M., F.E. Below, and S.P. Moose. 2004. Grain composition and productivity of maize hybrids derived from the Illinois protein strains in reponse to variable nitrogen supply. Crop Sci. 44:1593–1600.[Abstract/Free Full Text]
- Voltas, J., I. Romagosa, and J.L. Araus. 1997. Grain size and nitrogen accumulation in sink-reduced barley under Mediterranean conditions. Field Crops Res. 52:117–126.
- Williams, J.R. 1995. The EPIC Model. p. 909–1000. In V.P. Singh (ed.) Computer models of watershed hydrology. Water Resources Publ., Highlands Ranch, CO.
- Wuest, S.B., and K.G. Cassman. 1992. Fertilizer-nitrogen use efficiency of irrigated wheat: I. uptake efficiency of preplant vs. late-season application. Agron. J. 84:682–688.[Abstract/Free Full Text]
- Wyss, C.S., J.R. Czyzewicz, and F.E. Below. 1991. Source-sink control of grain composition in maize strains divergently selected for protein concentration. Crop Sci. 31:761–766.[Abstract/Free Full Text]
TOP
ABSTRACT
INTRODUCTION
