Illinois Fertilizer
Conference Proceedings
January 24-26, 2005
| Home | 2005 Index | Search |
|
|
| Home | 2005 Index | Search |
E.D. Nafziger, R.G. Hoeft, Eric Adee, R.E. Dunker, S.A. Ebelhar, and L.E. Paul
1
Recent research on corn has tended to show variability in N response. Brown et al. (1993) reported that economically optimal N rates among 77 sites in Illinois ranged from zero to more than 200 lb N per acre. Results from other studies show similar variability in time and space. Even with such variability, results over environments have been combined and used to develop an N fertilizer rate guideline in Illinois based on anticipated corn yield (Hoeft and Peck, 2002). This guideline suggests providing 1.2 lb of N (or a different factor based on the relative prices of corn and N) for each bushel of expected yield for corn following corn, with credits given when corn follows a legume or when manure has been applied to the field. While we know that yields, and thus actual N requirement, cannot be predicted with accuracy, the use of this guideline has proven to be satisfactory in most years and on most fields.
Data from a long–term previous crop x N rate study at Monmouth, Illinois revealed that the economically optimal N rate was 143 lb N/acre for corn following corn, with a yield at the optimal N rate of 146 bu/acre, and so an N requirement of slightly less than 1 lb N/bu (Bullock and Bullock, 1994). For corn following soybean, the optimal N rate was 99 lb N/acre and the yield at that N rate was 174 bu/acre. This is substantially less than the amount of N that would be recommended: 174 bu/acre times 1.2 minus 40 lb N credit for soybean is 169 lb N/acre.
The present study was designed to assess the response to N rate of corn following corn and corn following soybean, over a number of years and locations in Illinois, in order to find predictive relationships to help improve the correspondence between N rate and actual crop use of N.
Rotations to support this study were established in 1998, and data collection on N rate response has been ongoing since 1999. The study is being conducted at the following sites and soil types (with expected corn yield), on the six University of Illinois Crop Sciences Research and Education Centers: DeKalb – Flanagan sil (175); Monmouth – Sable sicl (180); Urbana – Drummer sicl (170); Perry – Clarksdale sil (140); Brownstown – Cisne sil (115); and Dixon Springs – Belknap sil (bottomland – 140) and Grantsburg sil (upland – 120). The study at the Dixon Springs upland site began one year later than at the other sites.
A split–plot design was used, with previous crop—corn or soybean—as main plots, and N rates—0, 45, 90, 135, 180, and 225 lb N/acre—on corn split within main plots. Corn followed corn on the same set of plots each year, with each N rate assigned to the same subplot. Soybean was planted into the third main plot each year, in preparation for corn with N rates the following year. Subplot sizes ranged from 10 x 30 to 20 x 50 ft.
Harvest for yield was done on the center two rows of each subplot. Yield data were analyzed using nonlinear regression (PROC NLIN) with the quadratic model. Where the Q–P model did not fit the data well—when yields declined at the higher N rates and/or when the model did not meet convergence criteria—the data were fit to a quadratic model. Economically optimal N rates were calculated from the quadratic function in each case using a cost:price ratio ($ per lb of N:$ per bushel of corn) of 0.10.
This research has now generated yield data of the N rate response of corn following corn and corn following soybean for six years at seven locations, for a total of 42 site–years for each rotation. Nearly all of the data are fit reasonably well by a quadratic + plateau function, though in 2000 and 2002 at Brownstown, the yield of corn following soybean was seriously damaged by wet soil conditions early in the season (the corn following soybean is in the same set of main plots in alternating years, hence in both of these years). While these were not the only two trials that showed no response to N rate, yields were very low (less than 30 bushels per acre) on these plots, and were lower than in the corn following corn plots. For these reasons we elected to delete the data from these two site–years from the analysis. In 2004, when corn following soybean was back in the same plots as in 2000 and 2002, yields and N responses were both relatively high at Brownstown. Since Perry does not cluster as described below, data from this location will be discussed in a later report.
One major reason for undertaking this work was to see if N rate responses and ensuing recommendations might differ among different locations. To test this in its broadest sense in Illinois, we clustered data from the 18 site–years representing Northern Illinois – six years at DeKalb, Monmouth, and Urbana – and from the 16 site–years in Southern Illinois – six years at Brownstown, Dixon Springs upland, and Dixon Springs bottomland, less the two site–years described above at Brownstown.
In both Northern and Southern Illinois, there appears to be some correlation between corn yield (OY) and the optimum N rate (ONR) needed to produce that yield (Figure 1). These correlations are not very strong, but in both regions, several "outlying" points tend to weaken the relationship. As indicated by the equations on Figure 1, the optimum N rate (lb per acre) is 83 + 0.56 x OY in Northern Illinois, and 36 + 0.92 x OY in Southern Illinois, suggesting that "base" amounts need to be higher in Northern Illinois, but that the change in response to (expected) yield might be more gradual than in Southern Illinois. The R2 values are not very high, explaining only 27 and 35 percent of the total yield variability in Northern and Southern Illinois, respectively, but they are statistically significant, and they provide some indication that recommending N rates based on yield goal (at least to the extent that yield goals predict actual yields) for corn following corn might have some validity. Unless we can learn more about what causes some of the outlying points in such data, though, it will be difficult to have a great deal of confidence in this approach. The average ONR and OY values for corn following corn are 178.5 lb N and 170.6 bu, giving a ratio of 1.05 lb N/bu for Northern Illinois, and 149.5 and 123.1 in Southern Illinois, for a ratio of 1.21 lb N/bu.
Unlike corn following corn, corn following soybean showed no relationship between OY and ONR, in either Northern or Southern Illinois (Figure 2). In Northern Illinois, yields averaged 194 bu/acre and the average optimum N rate was 135 lb/acre, giving a ratio of only 0.9 lb N per bushel of yield after adding the 40–lb N credit. In southern Illinois, the average optimum yield was 136 bu/acre and the N rate required to produced the OY averaged 135 lb/acre, which is the same average ONR as in Northern Illinois. The N:yield ratio for Southern Illinois was 1.3 lb N per bushel after adding the soybean N credit to the average ONR value.
These data can be used to make a rather strong case that recommending N rates based on expected yield does little to match N rates with N fertilizer requirements. The suggested use of 1.2 lb N per bushel of expected yield appears to underestimate actual N needs for corn following soybean in lower–yielding soils of Southern Illinois, and to overestimate N needs for high–yielding soils of Northern Illinois. It might be possible to simply adjust the ratio depending on the level of expected yield. For example, we could use 1.3 lb N per bushel of expected yield or so for fields expected to yield less than 150 bu/acre, and 1.0 or 0.9 for fields expected to yield more than 150 bu/acre. Even this measure, while it would help to reduce over–application in high–yielding fields, would do little to adjust N rates to expected yield among fields in such broad categories. There is no simple way to improve recommendations for a field when there is no relationship between the actual yield in a given year and the N rate required to reach that yield.
One possible approach to this dilemma is to examine a collection of N rate optima such as those we are reporting here, and to make probability statements based on these data about how much N will be needed for a given field in a given year. We can begin by examining the distribution of optima, including cumulative probabilities that a certain N rate will satisfy N requirements for corn in a randomly–chosen field. Figure 3 shows such distributions for both Northern and Southern Illinois. The data from Northern Illinois show a little bit more tendency to cluster toward the center than do the data from Southern Illinois, but the optima are spread over a fairly wide range of values in both cases, and given that optima are not related to yield (Figure 2), it is clear that optimum N rates are (and are going to be) very difficult to predict with any accuracy.
What can we do if yields show no relationship with optimum N rate, and there seems to be little possibility that anything we can observe or do will predict ONR at the beginning of the season? One approach is to use the data to project what N rate is needed to meet "sufficiency" (N rate equal to or greater than the optimum N rate) a certain percentage of the time. To illustrate this approach, the lines from Figure 3 are re–plotted on Figure 4. The fact that they are close to one another means that this approach will suggest similar N rates in both Northern and Southern Illinois. From the shape of the lines, we can see that it takes about 200 lb N/acre in Southern Illinois to avoid deficiency 80 percent of the time, while in Northern Illinois it takes about 180 lb N/acre to attain this.
While the approach illustrated in Figure 4 is useful, it still does not provide information on the costs (in lost yield or excessive N) of choosing an N rate that ends up being too low or too high for a given field. One way to do this is to calculate "return to N" (RTN) using response curves generated through research. This is done by calculating the additional yield produced by using a given rate of N (that is, yield at that N rate minus yield without N), multiplying the added yield times the price of corn, then subtracting the cost of N, which is the N rate times the price of N. For each response curve, RTN reaches a maximum at the optimum N rate, and decreases when N rates are higher or lower than the optimum. Figure 5 shows the average RTN at different N rates for corn following soybean in Northern and Southern Illinois. To do this, we calculated RTN for each N response curve, and then averaged these values over all response curves.
Once we have a sufficient number of response curves on which to base RTN curves, we think that it is appropriate to set N rate recommendations as the rates that produce the higher RTN. This could be a single N rate found by locating the highest point in the RTN curve. The highest RTN value of $93.50 per acre occurs at 146 lb N/acre for Northern Illinois. For Southern Illinois, the highest RTN is $76.04 at 149 lb N/acre. These findings would support using 145–150 lb N per acre for corn following soybean in Illinois. We also see that RTN values decline slowly at N rates higher or lower than at the maxima (Figure 5). Ranges of N rates over which the RTN is within $1.00 per acre of the maximum are about 125–172 lb N/acre for Southern Illinois and 126–165 lb N/acre for Northern Illinois. This might support an N rate recommendation of 125 to 170 lb N/acre for corn following soybean in Illinois. We will need to include additional N response data to confirm such rates, but until and unless we find a way to better predict N responses for particular fields or parts of fields, such an approach may constitute our best guess at N rates that will provide highest returns while minimizing over–application.
Figure 1. Relationship between optimum N rates and yields for corn following corn.
Figure 2. Relationship between optimum N rates and yields for corn following soybean.
1 E.D. Nafziger and R.G. Hoeft are Professors, E. Adee is Principal Research Specialist, and R.E. Dunker, S.A. Ebelhar, and L.E. Paul are Agronomists, Dep. of Crop Sciences, Univ. of Illinois, Urbana, IL.
Brown, H.M, R.G. Hoeft, and E.D. Nafziger. 1993. Evaluation of three N recommendation systems for corn yield and residual soil nitrate. Ill. Fert. Conf. Proc., R.G. Hoeft (ed.). pp. 43–49.
Bullock, D.S. and D.G. Bullock. 1994. Calculation of optimal nitrogen fertilizer rates. Agron. J. 86:921–923.
Hoeft, R.G. and T.R. Peck. 2002. Soil testing and fertility. In Illinois Agronomy Handbook, 23rd Edition. College of Agricultural, Consumer, and Environmental Sciences, Dept. Of Crop Sciences, UI Extension, University of Illinois.
