Improving Fertilizer Recommendations for Corn - The
Nebraska Soil Fertility Project (NSFP)
A. Dobermann1, J. Blumenthal1,2, R. Ferguson1,3, C. Shapiro1,4, D. Tarkalson1,5, C. Wortmann1,6, D. Walters1
1Dept. of Agronomy and
Horticulture, University of Nebraska-Lincoln
2Panhandle Research and Extension
Center, Scottsbluff
3South-Central Research and
Extension Center, Clay Center
4Northeast Research and Extension
Center, Concord
5West-Central Research and
Extension Center, North Platte
6Southeast Research and Extension
Center, Lincoln
The fertilizer recommendations presently used by the University of Nebraska (UN-L) have not been thoroughly documented. Written documentation is not available because the process involves statistical analysis of research data mixed with the judgment of individuals or by a committee (Hergert et al., 1997). The primary database for the UN-L fertilizer recommendations for corn is about 25 (N) or 40 years (P, K) old. LB 284 passed by the legislature in 1949 established the Outstate Testing Program within the Department of Agronomy for fertility and crop variety research. A public soil testing service was initiated in 1949, reaching a maximum of 21,000 samples analyzed annually in the mid 1950s. Commercial soil testing operations began in the state in the mid to late 1950s. Since then, the number of commercial soil samples analyzed by the university laboratory (Soil and Plant Analytical Laboratory SPAL) has declined to about 6000 samples per year, whereas private laboratories currently process about 140,000 soil samples collected in Nebraska. In the early 1960s annual meetings between the university agronomy staff and commercial laboratories were held to share soil fertility research information. However, after a few years these collaborative efforts slowed down and the different laboratories operating in the state developed their own fertilizer recommendations. Collaboration between the soil fertility faculty at UN-L and the private soil testing laboratories became strained due to different opinions about fertilizer recommendation philosophies. Due to the publication of research comparing various fertilizer programs, relationships were further strained (Olson et al. 1982, Olson et al. 1987)
Under the Outstate Testing Program, calibration and correlation research was conducted in Nebraska until 1964 and formed the basis for N, P, K, Zn and other essential plant nutrient guidelines for Nebraska. Limited availability of competitive funds for research on fertilizer recommendations has hampered regular updates of fertilizer recommendations. The only exception is the present N recommendation (third revision, 1995). Research for this was conducted from 1976 to 1982 and made possible by the so-called energy funds given to Nebraska because of overcharging by energy companies in the late 1970s. Fertilizer check-off programs have not been used in Nebraska to support soil testing and plant nutrient management research and education needed for regular updates of fertilizer recommendations.
On May 22, 2001, the Nebraska Legislature approved LB 329, which allocated $300,000 of excess funds accumulated in the Fertilizers and Soil Conditioners Administrative Fund to the University of Nebraska for conducting research on more precise nutrient management in irrigated corn systems. The project described below will use these funds for verifying and possibly updating the fertilizer recommendations for corn.
Currently used soil testing and fertilizer management recommendations are the result of many years of research and field verification conducted in Nebraska and other Midwestern states. However, many recommendations and Best Management Practices (BMP) in use today were developed decades ago. It is reasonable to question whether these recommendations are applicable with precision agriculture technology, particularly at yield levels that exceed those achieved during the original calibration research. We need to validate and update our knowledge for high yielding production systems and for changing production technologies. The sections below summarize some of the key issues to be addressed.
The University of Nebraskas algorithm for estimating N fertilizer recommendations in corn predicts the amount of N needed for achieving a certain yield goal as a function of soil organic matter (SOM), soil nitrate content in spring, and N credits from previous crop, manure and irrigation (Hergert et al., 1995; Shapiro et al., 2001).
Nrate (lb/A) = 35 + (1.2 EY) (8 NO3-N) (0.14 EY x OM) other N credits
EY = expected yield, 105% of 5-year average (bu/A)
NO3-N = root zone soil residual nitrate-N in 2-4 ft depth (ppm)
OM = soil organic matter (%)
The N algorithm has been validated to generally estimate N needs well in numerous on-farm demonstration studies in Nebraska during the past 15 years, but there are also situations where in over- or under-estimates N need (Ferguson et al., 1991). In the Mid-Nebraska Demonstration Project (1992-1997), the N algorithm under-recommended N 14% (24 of 170 sites) and over-recommended N 52% of the time. In regional studies, soil-test-based N algorithms were shown to have worked well for yield levels up to about 14 Mg ha-1 (220 bu acre-1), but tended to overpredict N rates in years with low response to fertilizer due to unfavorable climate or inadequate soil NO3 testing at sites with recent manure history or where alfalfa was the previous crop (Bundy et al., 1999). Uncertainties include:
a) Current farm yield averages exceed the average maximum yields achieved in the original N response studies. The present N algorithm for corn is based on 81 site-years of N rate experiments conducted on irrigated and rainfed land from 1976 to 1982. Nitrogen rates ranged from 0-280 lb/A (irrigated) or 0-180 lb/A (dryland) in 40 or 20 lb/A increments. However, the majority of sites were located in eastern Nebraska, particularly in the Northeast and the data set included 30 dryland sites with lower yields and different N response than that at irrigated sites. Maximum yields ranged from 55 to 196 bu/A, but averaged 153 bu/A for irrigated and 102 bu/A for dryland corn or 134 bu/A for all data sets together. This compares to average dryland corn yields of 111 bu/A and irrigated corn yields of 157 bu/A achieved in Nebraska during the 1996 to 1999 period. In other words, current farm yield averages exceed the average maximum yields achieved in the original N response studies and many irrigated corn farmers routinely harvest more than 200 bu/A corn, for which we have no adequate calibration.
b) Spatial and temporal variation of indigenous N supply. Nitrogen rates to achieve the maximum yield ranged from 0 to 220 lb/A, but averaged 79 lb/A. This compares to a statewide average N use on corn of 145 lb N/A in recent years (USDA, 2001). Moreover, at 28 sites there was no yield response to N and those sites mostly included dryland (21 sites). Nitrogen rates for maximum yield ranged from only 23 lb/A for dryland corn or 90 lb/A for irrigated corn on fine-textured soils to as high as 153 lb/A for irrigated corn on sandy soils. Check yields ranged from 35 to 168 bu/A (average of 96 bu/A for dryland, 102 bu/A for irrigated sandy soils, 125 bu/A for irrigated fine soils), indicating large spatial and temporal variation in the indigenous N supply. Understanding climatic effects and previous crops on indigenous N supply (Peterson et al., 1990; Yamoah et al., 2000) would be a key for fine-tuning of N recommendations according to major agroecological zones (AEZ) or major differences in cropping practices.
c) Predicting maximum yield potential based on natural resources. The N algorithm assumes a constant internal crop N requirement (bu yield per lb N taken up by the plant) and is very sensitive to specifying the expected yield (yield goal) in advance. The recommendation is that EY should be about 105% of the past 5-yr yield average. Knowing the true climatic yield potential (theoretically achievable yield only limited by solar radiation and temperature) and its variation across the state and among years would allow (i) adjusting yield goals according to AEZ, (ii) identifying upper limits of attainable yield for irrigated corn production in each AEZ, (iii) assess yield probabilities for dryland corn, and (iv) adjust crop internal N requirements depending on the yield potential. Season-specific yield goals should not exceed 70-80% of the yield potential because the internal crop nutrient requirements increase as yields approach the yield potential (Dobermann, 2001). This is also the yield level at which financial returns are greatest under most market conditions.
d) Understanding and predicting legume credits. Uncertainty about appropriate legume credits, particularly soybean credit. The N algorithm uses a 45 lb N/A credit for soybean as the previous crop, but other studies suggest that a higher credit of about 65 lb/A or 1 lb/A for each bu/A of previous soybean yield may be more appropriate (Peterson and Varvel, 1989; Varvel, 2000).
e) The N rate calculated cannot account for the temporal variation in crop N demand. The algorithm aims at predicting the optimal N rate for average climatic conditions and some general suggestions for time and method of N application are provided (Shapiro et al., 2001). Studies with corn in Nebraska indicate the importance of early season N management (Varvel et al., 1997), depth distribution of residual soil nitrate (Walters and Goesch, 1999),(Binder et al., 2000) as well as delayed N sidedressings (Bigeriego et al., 1979) for increasing N use efficiency, but the currently recommended a-priori N algorithm does not provide sufficient management criteria for N timing or in-season adjustments. Most methods proposed for in-season N management are corrective methods that employ diagnostic tools such as a chlorophyll meter (Peterson et al., 1993; Varvel et al., 1997; Shapiro, 1999), remote sensing (Blackmer et al., 1996), or on-the-go sensors (Lammel et al., 2001) to determine the need for an N topdressing.
Although technology development is proceeding rapidly, these techniques are not yet widely used in Nebraska. At present, this approach relies on empirical comparison with an over- or under-fertilized reference strip to assess whether an additional yield response to N is likely to occur. Moreover, corrective approaches require careful N management at all key growth stages to avoid that N deficiency occurs at critical growth stages. If N deficiency occurs during early vegetative growth of maize, correcting it with late-season N applications is unlikely to fully compensate for the yield loss associated with yield components formed during early growth (Binder et al., 2000). However, if the diagnostic tools used would allow establishing quantitative relationships between reflectance and biomass (Bouman et al., 1992) and between reflectance and nitrogen status (Blackmer et al., 1996), future improvements in interpretation can be made by applying concepts such as critical N dilution curves for a certain yield target (Greenwood et al., 1990) or by using sensed plant N status information as a forcing function in crop simulation models (predictive-corrective N management).
f) Prices are not build into the UN-L N recommendation. What is the best response model to apply for determination of optimum N rate (Sander et al., 1994)? We assume that the N algorithm was based on multi-year studies and should therefore represent an average economic N response at typical grain : N price ratios. Although optimal fertilizer rates tend to be not very sensitive to normal price fluctuations, yield has far more influence on profitability than prices of fertilizer or rates applied. In years with large disturbance of price rations (e.g., in spring 2001) and for high-yielding systems the ability to account for prices of grain and N would improve fertilizer recommendations. Prices may not have been the direct subject of the research below but risk is also an important factor (Helmers et al., 2001).
UN-L recommendations for P and K are based on the sufficiency concept, whereas many private laboratories use buildup-maintenance recommendations or recommendation that include a yield goal. At present, the UN-L recommendations suggest not to apply P above 15 ppm Bray-P and not to apply K above 124 ppm exchangeable K. These recommendations were mainly based on the outstate testing fertilizer experiments conducted in Nebraska in the 1950s and 1960s, and perhaps also inspired by the work done by Bray and colleagues in Illinois (Bray, 1944; Bray, 1945).
The original raw data are not available to us, but treatment means were published in annual Outstate Testing Circulars. The research base for establishing calibrations for the Bray-1 and Olsen-P soil tests was published by R.A. Olson et al. in 1952 (Olson et al., 1954). Supplemental research conducted later refined them in terms of categories used and rate differences according to P placement. Similarly, K recommendations published in 1964 had three levels (<75 ppm 25-50 lb K2O/A; 75-150 ppm usually no K recommended, but check strips should be tried in the field; >150 ppm no K recommended), which were later further divided into a total of 5 categories and two application methods (Hergert et al., 1995).
From 1973 to 1984, these recommendations were compared with buildup-and-maintenance strategies used by many private laboratories operating in Nebraska in a 12-year study at four sites (Mead, Concord, Clay Center, North Platte) and some accompanying studies at other locations. These experiments suggested that (i) the sufficiency concept was more profitable than buildup-maintenance recommendations because yields were similar and less P and K were applied, (ii) the soil test categories used were sufficient to describe the response to P, and (iii) no response to K was found. (Olson et al., 1982; McCallister et al., 1987). These and other studies concluded that deep, temperate region Mollisols common in Nebraska appear to supply significant amounts of K and often also P from native mineral reserves for an indefinite period so that efforts should concentrate on refining sampling, methods for sample analysis and interpretation, and methods for fertilizer placement (Olson et al., 1987). Nutrient balance calculations suggest, however, that, except where manure is applied, many producers in Nebraska have been running negative P and K input-output balances during the past 20 years, increasing the risk potential for marginal deficiencies at increasing yield levels. Uncertainties include:
a)
Shortage of calibration data for specific production
systems. The validity of the existing soil test categories for different
crops, different tillage systems, and high yield levels is unknown. Correlation
and calibration research has not kept pace with changes in cropping systems and
technologies (Hergert et al., 1997). The present P and K
recommendations are mainly based on the fertilizer trials conducted during the
1950s and 1960s, with highest yields up to about 170 bu/A. In many studies,
yields were well below 100 bu/A and the data included more dryland than
irrigated corn experiments. More recent research conducted in Iowa gives
support to the present recommendations, but also illustrates that arbitrary
choice of percentage sufficiency levels is a questionable practice because
direct economic analysis cannot be applied to relative yields (Mallarino and Blackmer,
1992). The Following give support to current P recommendations:
In 21 no-till trials conducted in Iowa, phosphorus fertilizer increased no-till
soybean yield (yields ranged from 1.8-4.3 Mg/ha) when soil test P was less than
9 ppm for 0 to 15 cm depth (Borges and Mallarino, 2000). In 26 no-till trials conducted in Iowa, corn yield was
increased in 8 of 11 cases with application of P in starter fertilizer where
soil test P was less than 12 ppm, but in no cases where soil test P was more
than 12 ppm (Bordoli and Mallarino, 1998). Corn yields ranged from 5.4 to 12.9 Mg/ha. The response was similar for 14 kg P ha-1
as for higher rates.
b) Need to include additional soil physical and chemical properties in recommendations. No consideration of different soil types and variation in subsoil P and K in present recommendations. Studies conducted from 1951 to 1956 at 139 locations of the Outstate Testing Program showed that differences in soil fertility status (pH, P, K) of Nebraska soils were closely associated with soil series (Olson et al., 1958). Although it was recognized that subsoil nutrient supply affects crop nutrient uptake under some conditions, this information did not produce P and K recommendations that included subsoil values. Recommendations were not made because there was close correlation of subsoil P and K status with that of topsoil. (Olson et al., 1958). Thirty to 40 years of farming and fertilizer additions have likely changed the topsoil to subsoil ratios of nutrients and also introduced more variability in P and K. The original correlation of the unfarmed soil probably no longer exists and may be management dependent. Similarly, although differentiations of recommended rates were made for band placement of P or K vs. broadcast application, these were rate recommendations and not differentiated by soil properties.
c) Insufficient differentiation by tillage. There has not been a coordinated effort for corn to conduct P or K correlation and calibration research for no-till or ridge-till systems in Nebraska. Research conducted included an experiment on ridge-till and P at North Platte (G. Hergert) and a significant body of research on P placement (Sleight et al., 1984; Raun et al., 1987; Eghball and Sander, 1989a; Eghball and Sander, 1989b; Eghball et al., 1990), but this has not led to differentiated fertilizer recommendations.
d) Conflicting response to K, even on low-testing soils. It is generally believed that illite clays are a major source of K supply in Nebraska soils, including coarse-textured soils (Fawzi and Drew, 1966), which often lead to no response to K fertilizer applications. Studies on potassium did not find any significant yield response to K on sandy soils in northeast Nebraska testing low in exchangeable K (Rehm et al., 1981; Rehm et al., 1983; Rehm and Sorensen, 1985). More recent studies at a similar site provided conflicting results. Yield response to K occurred in one hybrid in 2000, but preliminary 2001 data suggest little effect of the K rate on corn yields (Dobermann and Shapiro, unpublished). However, variability in soil test K was large and marginal P deficiency may have masked the K rate effects. More statistical analysis will be done to clarify this. In these recent studies, however, K affected stalk strength characteristics in certain hybrids, which is an important consideration for reducing harvest losses (Dobermann, 2001). On Mollisols at irrigated and non-irrigated sites testing high in K, there was often a yield depression at the highest K rate (McCallister et al., 1987), but reasons for this are not well understood. In contrast, however, research on very high yields (>250 bu/A) conducted at Lincoln from 1999 to 2001 suggests much larger crop K requirements per unit yield as yields approach the yield potential (Dobermann, 2001).
e) Uncertainty about profit maximization, i.e., does the sufficiency concept optimize yield at the right level? Are critical levels derived from relative yield response curves truly independent from absolute yield levels and therefore yield goals? The economics of fertilizer use are not built into current UN-L or industry corn recommendations in an explicit way, so profit optimization in making a recommendation or assessing long-term P and K management strategies are not possible. For P and K, the sufficiency approach appeared more profitable in the long-term lab comparison studies (Olson et al., 1982), but it includes some risk (reliance on soil testing only). The sufficiency approach has not been evaluated for its long-term implications. The fertilizer recommendation approaches used by private laboratories in the lab comparison studies were not documented. They may have overestimated the true crop nutrient needs for the yield achieved, which obviously made them less profitable (see below).
f) Crop nutrient requirements. Management concepts that include a yield goal require an assumption about the amount of total plant nutrient uptake per unit yield. At present, these are typically single coefficients that were derived from earlier field experiments and they assume linearity between crop yield and nutrient accumulation. However, such constants tend to overestimate the simulated optimal nutrient requirements (Dobermann, 2001). There is also a tendency that data from research experiments conducted on soils with high background levels of P and K may overestimate true crop P and K requirements for a situation of optimal balanced nutrition. Recent analysis of published data suggests that earlier estimates may have been too high and that those numbers also depend on the yield level in relation to the yield potential. For example, in earlier publications the Potash and Phosphate Institute (PPI) cited P removal with grain as 0.44 lb P2O5 per bu yield, a number which was derived from maximum-yield studies. However, review of a large number of published experimental data produced an average of 0.31 lb P2O5/bu for all studies or 0.38 lb P2O5/bu for yields >200 bu/A (Scott Murrell, PPI, unpublished data). For a yield goal of 200 bu/A, these three different estimates would translate into a fertilizer P range of 62 to 88 lb P2O5/A to replenish grain removal. Obviously, which number to choose is not a trivial issue because this greatly affects the recommended P rate in a replenishment of crop removal approach and therefore the economic performance over time.
g) Prediction of soil test changes due to crop removal and fertilizer/organic additions. Buildup-and-maintenance strategies for P and K management require the definition of empirical functions that describe the increase or decrease in soil test P or K as a function of the amount of fertilizer P or K applied and crop removal, as well as the definition of a target soil test level. There is no generally accepted, reproducible method for establishing such functions and it is unclear whether most of the empirical choices currently in use properly account for the economics of fertilizer use.
Interest in sulfur
mainly emerged in the early 1960s, particularly on sandy soils in northern and
northeast Nebraska. Earlier soil test research found that Ca-phosphate worked
best as a S extractant for Nebraska soils (Fox et al., 1964), but the usefulness of this
soil test has been questioned because no relationships between test results and
the rate of S needed for optimum production has been developed. Soil texture and
organic matter appear to be reliable enough to predict probable response to S,
with little extra gain achieved by the Ca-phosphate extractant (Rehm, 2000). Research on sulfur in
Nebraska was mainly conducted by G. Rehm (Rehm,
1978; Rehm, 1984; Rehm, 1993). There is some controversy about the present S recommendations that
justifies more research, especially for no-till on fine-textured hillside
soils. Some producers seem to apply more sulfur that what UN-L recommends.
UN-L lime recommendations are not based on maximizing profit, but rather
increasing the pH to approximately neutral 6.5. This has the appearance of a
build and maintenance approach. The UN-L soil test results will generate a lime
recommendation at pH 6.2 or lower, while advising that lime use is likely to be
profitable if pH is less than 5.7. This was done because different crops have
different critical pH levels, liming is costly, and soil pH changes slowly.
However, field research has not shown consistent yield response unless surface
soil pH has been below 5.5. Liming recommendations for several important
Nebraska crops were given in Penas (1988) and were probably based on a
review of the literature and expert opinion. In other research, the lime requirement values of 74 acid sandy
soils from northern Nebraska were assessed by eight different laboratory test,
suggesting that the currently used soil tests work well as compared to incubating
soil with CaCO3 (Alabi et al., 1986). However, correlation studies of lime
requirement values for five methods with selected soil properties showed that
no single property was correlated with all methods. While we need additional liming research, we should perhaps
also consider revising our lime recommendations to maximize profit. Because
liming is a long-term investment, land ownership and tenancy issues influence
the liming decision. Other issues that are important to the discussion are the
effect of surface soil pH on herbicide efficacy and the effectiveness of liming
for no-till situations.
Research on starter fertilizer in corn started prior to 1950 and was expanded through field trials conducted from 1951 to 1960 (Langin et al., 1962). Results of these studies were conflicting because both yield increases and yield depressions with starter fertilizer were found. However, no or negative yield response was typically observed at soil test P levels >15 ppm, which was also confirmed in studies conducted in Minnesota (Rehm et al., 1988). Most research on starter fertilizers for row crop production in Nebraska has been for tilled conditions. Soybean yield was increased with band application of P on low P soils but there was no P effect on medium P soil (Sander et al., 1990); the average increase due to knifing P was 0.12 Mg ha-1 over 8 locations. Trials were conducted on farmers fields to compare their application of starter fertilizer to no starter (Penas, 1990). The rate, composition, and method of application varied by farm. The average corn response to starter fertilizer when available soil P was less than or equal to 10 ppm was 0.23 Mg ha-1 with yield increases in 60% of the trials; otherwise, the mean gain was 0.13 Mg ha-1 with significant increases in 17% of the trials. Corn yield was increased by an average of 0.62 Mg ha-1 when starter fertilizer was applied to sandy and sandy loam soils, while the mean response was 0.08 Mg ha-1 on fine texture soils. In all cases, the starter fertilizer applied on the sandy and sandy loam soils contained N, P and S. Six of these 52 on-farm trials were under no-till conditions; there was a response in only one case, and S was included in the starter fertilizer in that case.
UN-L
recommendations (Penas and Hergert,
1990)
are based on the assumption that starter fertilizer is a good method to get P
in the ground but it often does not add much benefit when soil tests are above
the critical level. Considering the low probability
of a yield response for tilled situations, research on starter fertilizer
should have less priority at this stage, although we recognize that there is
more evidence for yield response under no-till farming in states such as KS,
SD, IA, and MN. We recommend that this is a good area for fertilizer dealers
and others to conduct their own applied studies.
Both the sufficiency approach and buildup-maintenance strategies for P and K management rely on soil testing. Therefore, current fertilizer recommendations are very sensitive to the frequency and spatial density of sampling as well as the quality of sample analysis. Log-normal soil nutrient distributions may exist within as well as across fields, increas