Illinois Fertilizer Conference Proceedings
January 27-29, 1992

Home 1992 Index Search

Elwood N-P-K Fertility Study

L.E. Paul & S.K. Sipp1



This study was initiated in 1963 to evaluate the yield response of corn and soybeans grown on a light colored soil to evaluate grain yield response to several combinations of levels of N, P, and.K. The experiment was conducted at the University of Illinois Northeastern Illinois Agronomy Research Center at Elwood, Illinois.


The objectives were to determine:

  1. The optimum rate of N, P, and K for corn.
  2. The interaction between N, P, and K for corn and soybeans.
  3. The relationship between soil test P and K and yield.
  4. The optimum rate of P and K for soybeans.

Materials and Methods

The experiment was conducted on a Blount Silt Loam (Fine, illitic, mesic Aeric Ochraqualf), a somewhat poorly drained, slowly permeable soil developed in shallow (0-20")loess over silty clay loam glacial till under native deciduous forest. This soil is most commonly found in level to gently sloping uplands and is fairly productive for agriculture. Plots were established in three blocks in each of two series with fertility treatment combinations randomly assigned. Corn and soybeans were grown in rotation, being alternated between the fields using conventional tillage and machine harvest. Fertilizer treatments were broadcast by hand in the spring each year prior to tillage. Nitrogen was applied prior to the corn crop as ammonium nitrate. Annually, phosphate was applied as 0-45-0 and potash as 0-0-60. Soil samples were collected prior to the first application of fertilizer treatments and starting in the fall of 1983 biennially.

Statistical Design

The study was designed as a randomized complete block factorial with three rates of N, P and K fertility with three replications. All combinations of fertilizer rates occur in each block and field. In 1967, N and K rates were increased because of low yield response to the original rates to these nutrients.

1963-66 treatments   1967-73 treatments
Nutrient and rate lb/acre   Nutrient and rate lb/acre
N P2O5 K2O   N P2O5 K2O
0 0 0   0 0 0
50 60 30   100 60 45
100 120 60   200 120 90

In 1973 a modified set of 27 treatment combinations using N rates of 0, 60, 120, 180, and 240 and P2O5 and K2O rates of 0, 30, 60, 90, and 1201bs/acre were established to provide a wider range of fertilizer rates; howerver, not all combinations of rates were possible. . A larger range of rates and greater number of rate increments were used to provide for more precise estimates of optimum fertilizer application rate. By using some of the modified treatment combinations in more than one data set, five different fertility experiments can be analyzed.

The five experiments were analyzed using sets of the modified combinations to investigate the following:

  1. Randomized complete block with 2 treatments to test if cessation of P fertilization following addition of P in excess of crop removal over time will affect grain yield.
  2. Randomized complete block, 1 factor factorial using five rates of N fertilizer with non-limiting P and K to determine grain yield response to N fertilizer applications.
  3. Randomized complete block 1 factor factorial using five rates of K fertilizer with non-limiting N and P to determine grain yield response to K fertilizer applications.
  4. Randomized complete block 2 factor factorial using two rates of N and 5 rates of fertilizer with non-limiting K to determine grain yield response to P fertilizer applications when N is limiting and not limiting to yield.
  5. Randomized complete block 2 factor factorial using five rates of P and 3 rates of K while N is not limiting to determine how grain yield response to rates of P fertilizer are affected by rates of K fertilizer.

In 1983, P and K fertilizer application was discontinued. Nitrogen applications and data collection continued to determine the residual effects of previous regular K and P fertilizer additions.

Results and Discussion

The data were initially tested using the analysis of variance procedure. For the first ten years, all data were included in a single 3 way factorial analysis. For the second ten years, the data were divided into groups and analyzed according to the design as discussed previously. The last five years (when P and K fertilizer rates were discontinued) were analyzed as if the rates were continuing to be applied to determine whether the effect of fertilizer rates would continue after cessation of applications.

The 10% confidence level was used in preparing the summaries in Tables 1-2. One might expect that a significant difference be indicated when there is no significant difference approximately 10% of the time.

Main Effects of Nutrients

Nitrogen application increased corn yield nearly every year. Soybean yield increases attributable to prior N fertilizer applications were significant about 20 percent of the time, about the same as expected from random variation.

In the analyses when phosphorus fertilizer rate was a factor grain yields of both corn and soybeans were significantly related with P fertilizer rate in most years. This relationship carried over into the years after the cessation of P fertilizer applications in the factorial studies. In the single treatment comparison, no significant differences in corn yield were produced between the treatment never receiving P fertilizer and the treatment which had received 120 lb P2O5/acre/yr for the previous ten years. Soybean yields were somewhat more responsive to the carryover P fertilizer producing significant differences in yield in 3 of 11 years.

In the first ten years of the study, potassium fertilizer produced a significant response in corn yield 7 years and in soybean yield only 2 years. After the experiment was modified, neither corn nor soybean yield responded significantly to K fertilizer at non-limiting N and P fertilizer levels. Significant response was observed in only two growing seasons for each crop. After the P and K fertilizer treatments were discontinued, grain yield response of both corn and soybeans continued approximately the same trends as had been established in the previous ten years.

The other combination of treatments which involved K as a main effect in the modified experiment indicate a grain yield response by both corn and soybeans to K fertilizer treatment during the time when fertilizer applications were applied; as well as after P and K fertilizer applications ceased.

Interaction Effect of Nutrients

Interaction effects indicate that a change in the level of one factor will affect the response to levels of a second factor. During the initial ten years of the study, corn grain yield responded to interactions of N and P fertilizer rates in 5 years. Significant interaction responses of corn and soybean yield to other combinations of nutrients occurred so infrequently as to be considered approaching random variation.

In the two analyses from data sets produced by the modified treatment combinations, some interaction effects were produced. In the N X P factorial, corn grain yield followed the pattern established in the initial 10 years of the study that about one-half of the years produced significant interaction effects during the period while annual applications of P fertilizer were used. After the P fertilizer applications ceased, no interactions were observed. Unexpectedly, soybean grain yield response to N-P rate interaction during the same time period closely parallels the corn response with about one half of the years indicating statistically significant differences.

The only significant response to P by K interactions occurred in the modified treatment phase of the study. In nearly one half of the years when P and K fertilizer were applied, the interaction effect of P by K produced a significant soybean yield response.

The data were divided into four groups to perform regression analyses to develop formulae for the grain yield response curves of the various nutrients. The grouping of the data enabled the comparison of yield responses under similar soil fertility programs. Rather than attempt to develop curves using a complex model including main and interaction effects of the nutrients, a regression analysis for each nutrient was run by year to determine response curves based on actual grain yield. Regression analyses of combined years within each group of data using relative yield (% of maximum yield for each year) were then performed. Analyses combining years in the group were run combining data using the following criteria:

  1. All years
  2. Years with typical curves regardless of significance of coefficients
  3. Years with typical curves with significant linear coefficient @ .25
  4. Years with typical curves with significant linear coefficient @ .10
  5. Years with typical curves with significant quadratic and linear coefficients @ .25
  6. Years with typical curves with significant quadratic coefficient @ .25 and significant linear coefficient @ .10
  7. Years with typical curves with significant quadratic and linear coefficients @ .10

Typical curves are convex curves produced by a negative quadratic coefficient and a positive linear coefficient which can be differentiated to calculate estimates of optimum fertility rate to produce maximum economic return. The values calculated from the statistical models from the combination of years which produced the highest probability of fit to the model are located in Table 3. In most cases, the optimum fertility rate estimate declines as one moves down through the combination criteria list.

As the model probability value decreases, actual data points fit the generated curves more closely. So, the data used to predict the P rates for corn and soybeans and the N rates for corn are well associated with the curves. In addition, the R square value which is the percentage of the variation that is attributable to the factors in the model must be considered in conjunction with the model probability. For example, the values generated by the N rate curves for corn have about 70% of the variation explained by N rate; the remaining variation is caused by factors not included in the model such as weather, tillage, other fertility rates, or any number of unknown factors. On the other hand, the R square values for the K rate curves indicate that one-third or less of the variation in the data is attributable to the K fertilizer rate. Consequently, the N values predicted by the study are more reliable estimates of the fertilizer rate required than are the predicted K rate estimates. The R square values associated with the P rates are intermediate compared to those generated by the N and K models, so the predicted rates must be considered accordingly.

The yield reduction values in Table 6 must be evaluated considering the model significance and R square values from Tables 3 & 4. In addition the variation in grain yield between years is large and is dependent on numerous factors in addition to soil fertility level.


1. What are the optimum rates of N, P, and K for corn?

The analyses of variance procedures indicate that statistically significant differences in crop yield exist between rates of the fertilizer elements. The regression analyses of N rate and corn yield over the several phases of the study (Table 3) indicate with a reasonable confidence that the N rate for optimum economic return is approximately 1401b/acre.

The P rate curves do not account for the variation as well as the N curves. The rates for optimum economic return are about 80 lb/acre when N is not limiting.

Although the K rate predictions for optimum yield and return appear reasonable compared to the Agronomy Handbook recommendations, prediction levels based on the data in this study would be questionable as most of the coefficients used to predict the optimum rates are not significant even at the .25 level.

2. Are there interactions between N, P, and/or K?

The analyses of variance indicate relatively low incidence of interaction between the fertilizer elements investigated by this experiment. The results indicate corn productivity is affected by the interaction of N by P in about half of the years. During the latter part of the study, the results suggest that both N by P and P by K interactions affect soybean productivity approaching half of the time.

The regression analyses suggest that there is an interaction between P and N rates for corn production since the predicted optimum P levels decrease at the 601b N/acre rate in the last fifteen years of the study. However, other than noting the existence of interaction effects, explanation is difficult because only two levels of N were included; one well in excess of optimum and the other well below optimum.

The interaction effects between P and K during the period when sufficient data to assess the interaction using regression analysis are available (the latter 15 years) are weak at best, probably in large part because of the low predictability of yield response to K.

3. Does the soil test accurately predict the response of corn to P and/or potash on this soil?

The data produced by this study cannot directly answer this question.

4. Does the response of soybeans follow corn with respect to P and K?

The analysis of variance performed indicates that corn and soybean yield response to P and K fertilizer is similar. Similarly, the regression analysis confirms that soybeans react to P and K fertilizer in a pattern very similar to corn; although, the optimum rates of each nutrient are slightly different for each crop.

5. Are the residual effects about the same as the year applied?

The analysis of variance suggests that effects of P and K rates on grain yield of both corn and soybeans continues for at least five years after application of the fertilizer is discontinued, although the predictability of the response curve decreases. However, this study cannot predict whether the yields produced during the period are the same as if rates had been applied since there is no data to compare yields with and without fertilizer applications.


Some notes of caution are in order. The interpretations of this data should be considered as guides and not rigid limits. In general, the application of these results should be limited to the soil type on which the experiment was performed; extension of these guidelines to similar soil types in the region should be approached cautiously and refined to suit the individual site. Regression values for optimum yield and economic return were determined with only years of normal response and therefore, the values may be too high for use over all years due to other factors, such as weather, disease, insects or other factors.

Nitrogen fertilizer rate is strongly associated with corn grain production. The optimum rates for both total yield and economic efficiency determined from this study appear to be in the range recommended for the productivity index for this soil and the yields produced over the history of the study. Soybean production is not affected in any predictable way by the N applied to corn in the prior year.

Corn yield is not as predictably affected by P fertilizer rate as by N fertilizer. Soybean yield response frequency is approximately equivalent to that found with corn. The optimum rates determined from this study are at about 1.5 times the maintenance level recommended by the Agronomy Handbook. Only about one-half of the variation in yield can be accounted for by P rate as compared to approximately three-fourths of the variation explained by N rate.

Corn yield response to K in this study was quite low. The optimum rates calculated from this study, while not unreasonable when compared to rates recommended in the Agronomy Handbook, are much lower in predictability than the estimates for either N or P.

Carryover effects from previous P and K fertilizer applications appears to last for at least five years. However, this study as conducted cannot definitely determine whether grain production from either corn or soybeans was affected by the cessation of P or K fertilizer application.

Although the study demonstrates the relationship of corn and soybean grain production to the P, and K soil tests, the predictability of the relationship is rather low. In addition, this study cannot define the soil test levels at which application of additional fertilizer becomes marginal.

Tables and Figures Referenced

Table 1: Crop response to fertilizer

Table 2: Crop yield response to fertilizer or soil fertility

Table 3: Optimum fertilizer rate for corn yield determined by regression

Table 4: Optimum fertilizer rate for soybean yield determined by regression

Table 5: Optimum P and K soil test values

Table 6: Percentage yield reduction from maximum when fertilizing for high P.I. yield based on regression curves

Table 7: Average Grain Yield

Footnotes and References

1Associate Agronomists, University of Il.

Back to 1992 Index

*Department of Crop Sciences
*University of Illinois Extension
*College of Agricultural, Consumer, and Environmental Sciences
*University of Illinois at Urbana-Champaign
*Illinois Fertilizer & Chemical Association
*Illinois Department of Agriculture
  Design by: Crop Sciences Computer and Web Support Group
Copyright © 2005 University of Illinois at Urbana-Champaign
Email site problems to the webmaster