Agronomy Journal Grow Your Career With ASA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published in Agron J 99:822-832 (2007)
DOI: 10.2134/agronj2006.0172
© 2007 American Society of Agronomy
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Bermudez, M.
Right arrow Articles by Mallarino, A. P.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Bermudez, M.
Right arrow Articles by Mallarino, A. P.
Agricola
Right arrow Articles by Bermudez, M.
Right arrow Articles by Mallarino, A. P.
Related Collections
Right arrow Phosphorus
Right arrow Nutrient Management
Right arrow Soil Fertility and Productivity

Site-Specific Analysis & Management

Impacts of Variable-Rate Phosphorus Fertilization Based on Dense Grid Soil Sampling on Soil-Test Phosphorus and Grain Yield of Corn and Soybean

Manuel Bermudez and Antonio P. Mallarino*

Dep. of Agronomy, Iowa State Univ., Ames, IA 50011

* Corresponding author (apmallar{at}iastate.edu)

Received for publication June 9, 2006.

   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Most agricultural fields have high soil-test phosphorus (STP) variability. Variable-rate (VR) technology facilitates application of different P rates over a field and could improve nutrient application and crop yield. Replicated strip trials (6–12 ha) were established at six Iowa fields and were evaluated during 4 yr to compare VR and fixed-rate (FR) P fertilization for corn (Zea mays L.)–soybean (Glycine max L. Merr.) rotations. Fields had median Mehlich-3 STP ≤ 20 mg P kg–1, although minimum and maximum values within each field were 4 to 18 and 22 to 62 mg P kg–1, respectively. Treatments replicated at least three times were a control, VR based on STP from grid sampling (0.06- to 0.08-ha grids), and FR based on median STP. Treatments were applied with commercial spreaders and grain was harvested with combines equipped with yield monitors and global positioning systems (GPS). Phosphorus increased yield in 13 site-years and application methods differed in 1 site-year, when FR increased yield further. On average, VR applied 12.4% less P and reduced STP variability in five fields compared with FR. Semivariograms and SD showed that fertilization, especially VR, often reduced yield variability and seldom increased it. High STP variability at a small scale and P recommendations to maximize yield and buildup STP in low-testing soils might explain a lack of yield differences between application methods. Although VR did not increase yield compared with FR, it managed P better and showed potential for reducing excess P loss from fields through reduced P application to high-testing field areas.

Abbreviations: GPS, global positioning system • FR, fixed rate • GIS, geographical information systems • RCBD, randomized complete-block design • STP, soil-test phosphorus • VR, variable rate


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
PRECISION AGRICULTURE TECHNOLOGIES such as yield monitors, GPS equipment, and geographical information systems (GIS) are widely used in the U.S. Midwest for mapping soil test values and grain yield. Most agricultural fields usually include several soil map units that may have different nutrient supplying capabilities. Spatial variability of soil properties can range from regional to subcentimeter scales. Studies of STP spatial variability have shown large within-field variability even in fields apparently uniform in other soil properties (Peck and Melsted, 1973; Franzen and Peck, 1995; Mallarino and Wittry, 2004). Applying a FR of fertilizer to an entire field may be inefficient because it may overfertilize some areas and underfertilize others. Furthermore, a fixed application may increase nutrient loss to surface and ground water by adding nutrients to high-testing areas. Variable-rate technology allows for changes in fertilizer rates on-the-go and better control of the amount of fertilizer applied to specific field areas.

Some have estimated the value of VR from crop response trials based on fixed nutrient rates applied to long and narrow strips across a field (Carr et al., 1991; Rehm and Lamb, 2000). Others have directly compared FR and VR fertilization for crops such as barley (Hordeum vulgare L., sorghum [Sorghum bicolor (L.) Moench], and wheat (Wibawa et al., 1993; Mulla et al., 1992; Yang et al., 2001) using a variety criteria for soil sampling and deciding about fertilization rates. These studies showed inconsistent yield differences between VR and FR although often, but not always, less fertilizer was applied with VR. Studies with corn and soybean are scarce. Anderson and Bullock (1998) compared VR and FR of P-K mixtures for corn and soybean based on a 1-ha grid soil sampling and found no yield response to fertilization in any field, although some fields had low-testing areas according to local interpretations. Lowenberg-DeBoer and Aghib (1999) studied fertilization with P-K mixtures using FR, VR based on soil sampling of 1.2-ha cells, and VR based on sampling by soil series for corn, soybean, and wheat over 12 site-years in six Midwestern farms. Soil-test results from the 1.2-ha soil sampling showed that P and K varied from low to high in all fields but the field average always was at optimum levels for the crops. The VR method sometimes increased yield over FR but seldom increased net returns when fertilization and soil sampling costs were included. The amount of fertilizer applied with VR compared with FR usually was slightly greater for K and slightly smaller for P. A risk analysis indicated that fertilization based on soil series resulted in the lowest economic risk.

In a previous Iowa study, Wittry and Mallarino (2004) assessed corn and soybean responses to FR and VR P fertilization in six fields (12 site-years). The soil sampling density in which they based VR varied across fields from 0.2- to 1.7-ha cells. The sparser density was less dense than the 1.0-ha grid sampling method commonly used in the Midwest. The fertilization methods did not influence crop responses to P in any field but VR applied 12 to 41% less P and reduced STP variability compared with the FR method. The authors explained the lack of response to VR by P fertilizer recommendations that encourage STP buildup in low-testing soils combined with a high STP variability at small scale that could not be correctly measured with the soil sampling density used.

Further research is needed to assess VR technology effects on yield and yield variability in corn–soybean rotations. The objectives of this study were to evaluate the grain yield responses of corn and soybean to P fertilization using FR and VR application methods, differences between these methods for field areas with different soil series or STP values, and effects of these methods on STP and yield variability. The study complements a previous Iowa study (Wittry and Mallarino, 2004) by using different fields, a denser soil sampling method, a data analysis method that accounted for spatial correlation of yield, and by studying FR and VR effects on yield variability.


   MATERIALS AND METHODS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Replicated strip-trials were conducted during 4 yr on six Iowa farmer's fields managed with corn–soybean rotations. The dominant soil series in the experimental areas (6–12 ha) were typical soils of Iowa and neighboring states (Table 1). The glacial-till derived Clarion (fine-loamy, mixed, superactive, mesic Typic Hapludoll), Webster (fine-loamy, mixed, superactive, mesic Typic Endoaquoll), and Canisteo (fine-loamy, mixed, superactive, calcareous, mesic Typic Endoaquoll) series predominated in Fields 1 to 4. These soils also are common in south-central Minnesota. The loess derived Marshall (fine-silty, mixed, superactive, mesic Typic Hapludoll) series predominated in Fields 5 and 6. This soil also is found in western Nebraska and northwestern Missouri. All fields had histories of fixed P fertilization. Management practices were those used by each farmer and, therefore, hybrids and cultivars, seeding rates, planting dates, and other practices varied among fields. In Fields 1 to 4, corn residues were chisel plowed after harvest in October or November (fall), and were disked before planting in April or early May (spring). Fields 5 and 6 were managed with no-till.


View this table:
[in this window]
[in a new window]

  		Table 1. Predominant soil series in the experimental areas and mean initial soil-test P (0–15 cm) for six fields.

 
Treatments were a control without P fertilization and P fertilization using FR or VR application methods. A randomized complete-block design (RCBD) was used in all fields, with three replications in Fields 1 and 2 and four in Fields 3 through 6. The strip (plot) width was 18.3 m for Fields 1 to 4 and 21.3 m for Fields 5 and 6, while strip length varied from 310 to 505 m across fields. Fig. 1 shows a schematic representation of the experimental design and soil sampling areas. In Fields 5 and 6, two blocks were separated from the others because they followed contoured terraces. Measurements were made with a measuring tape, permanent plastic pipes were buried at each trial corner, and coordinates were recorded with a hand-held GPS receiver. All experiments were evaluated for 4 yr. Experiments in Fields 1 and 2 were established in 1998, and crop sequences were soybean–corn–soybean–corn in Field 1 and corn–soybean–corn–soybean in Field 2. Experiments in Fields 3, 4, 5, and 6 were established in 1999, and crop sequences were corn–soybean–corn–soybean in Field 3, soybean–corn–soybean–corn in Fields 4 and 5, and corn–soybean–soybean–corn in Field 6. A single code for field, crop, and year includes a field number (1–6), suffixes a and b to indicate the first and second crop after P application, and a suffix 2 as needed to identify the two crops after the second P application.


Figure 1
View larger version (37K):
[in this window]
[in a new window]

  		Fig. 1. Representation of the experimental design showing two of three to four replications (C, control; FR, fixed-rate P fertilization; VR, variable-rate P fertilization). The strip width varied from 18.3 to 21.3 m across fields and its length varied from 310 to 505 m.

 
Table 2 shows the P fertilizer rates used for each application method, which were based on Iowa State University Soil and Plant Analysis Laboratory recommendations for 2-yr corn–soybean rotations applied once before the first crop (either corn or soybean). Therefore, the treatments were applied twice to the same strips (before the first crop and before the third crop). The laboratory recommendations were 70 kg P ha–1 for the Very Low interpretation class, 54 kg P ha–1 for Low, 34 kg P ha–1 for Optimum, and no fertilization for High or Very High, and have changed only slightly since 2002 (Sawyer et al., 2002). The rates applied with FR deviated from recommended rates in Field 5a (34 instead of 54 kg P ha–1) and Fields 4a2 and 5a2 (24 kg P ha–1 instead of zero) because the farmers wanted to apply the same rate planned for the rest of each field.


View this table:
[in this window]
[in a new window]

  		Table 2. Target fixed P rates (FR) and variable P rates (VR) for six fields.

 
Composite soil samples (15-cm depth) used to determine P application rates were collected before applying treatments (twice) using a systematic, grid-point sampling method (Wollenhaupt et al., 1994) that matched the experimental layout. This is the sampling depth recommended in Iowa for P and K in fields managed with no-till or chisel-plow tillage (Sawyer et al., 2002). The grid lines separation across strips coincided with the strip width (18.3 or 21.3 m) and the separation along strips was 45 m in Fields 1–4 and 30 m in Fields 5 and 6 (Fig. 1). Therefore, the number of grid cells along each strip was eight in Fields 1 and 2, 11 in Fields 3 and 4, and 10 in Fields 5 and 6. Soil cores for each sample (10–12 cores) were collected from 100-m2 areas at the center of each cell and represented 0.08-ha cells in Fields 1 to 4 and 0.06-ha cells in Fields 5 and 6. Soil samples were analyzed by the Mehlich-3 P test with a colorimetric determination of extracted P (Frank et al., 1998). Table 3 shows descriptive statistics for STP and distribution of values within Iowa interpretation classes for the initial soil samples. Rates for the FR method were defined from median STP of samples collected across all strips before the first treatment application, and only from all FR strips before applying this treatment for a second time after harvesting the second crop. Rates for the VR method were defined from STP of samples collected across all strips before the first treatment application, and only from VR strips before applying this treatment for a second time. Interpolated P application maps for each field suitable for controllers of commercial VR applicators were prepared in ArcView GIS (Environmental Systems Research Inst., Redlands, CA) by using an inverse-distance method with a distance-weighing exponent of two (Wollenhaupt et al., 1994).


View this table:
[in this window]
[in a new window]

  		Table 3. Descriptive statistics for initial soil-test values and soil-test P distribution according to Iowa State University soil-test interpretation classes for six fields.

 
Granulated mono-ammonium phosphate fertilizer was broadcast with commercial VR spreaders (double spinner delivery systems in Fields 3 and 4, and air-powered delivery systems in Fields 1, 2, 5, and 6) equipped with GPS receivers and controllers. Fertilizer was applied in the fall after harvest (October or November) except for the first treatment application in Field 6 where was applied in spring (March) before planting. Fertilizers were incorporated into the soil by disking, except for Fields 5 and 6 that were managed with no-till. These are the usual management practices in the area and, furthermore, Iowa research has shown no difference between broadcast and P fertilizer placement methods for corn or soybean managed with no-till and chisel-plow tillage (Bordoli and Mallarino, 1998; Borges and Mallarino, 2000). Corrective N rates to offset the small N content of the phosphate fertilizer were used only for corn in Fields 5 and 6, and an additional rate of 120 kg N ha–1 was applied across all treatments. No corrective N rate was applied at other fields, but the highest N rate recommended in Iowa for corn after soybean (168 kg N ha–1) was applied across all treatments. A K fertilizer rate equivalent to the 2-yr average K removal in corn and soybean grain (120 kg K ha–1) was applied across all treatments before the first and third crops. Mean initial soil pH ranged from 5.6 to 6.7 across fields and mean organic matter ranged from 36 to 60 g kg–1 (not shown).

Grain yield was measured using combines equipped with yield monitors and real-time GPS receivers. The yield monitors used were impact flow-rate sensors Ag Leader 2000 (Ag Leader Technology, Ames, IA), Green Star (John Deere Inc., Moline, IL), or Micro-Trak (Micro-Trak Systems, Inc., Eagle Lake, MN). Differential corrections were obtained through the U.S. Coast Guard AM signal. The spatial accuracy was checked in several field locations with a hand-held GPS receiver. Yield data were unaffected by borders because at least 40 m at each strip end were harvested but not used. While harvesting, each combine trip (a 4.5-m swath) was identified with a unique number that was recorded with the georeferenced yield data. A sensor located in the combine augers determined grain moisture, and yield was corrected to 155 g kg–1 H2O for corn and 130 g kg–1 H2O for soybean. The yield data were analyzed for common yield monitor problems from grass waterways or combine stops, and affected data were deleted using ArcView.

Treatment effects on yield were analyzed for strip (plots) averages and for areas within fields. For strip averages, yield responses were analyzed using the mean yield for the strips as input data. For Fields 1 to 4 (where strips were contiguous), we used an ANOVA procedure for RCBD that accounted for spatial correlation of yield with nearest neighbor analysis using SAS (SAS Institute, Release 6.11, Cary, NC). Hinz and Lagus (1991) and Stroup et al. (1994) discussed the basis for this procedure and potential applications. More recently, the procedure was adapted to a strip-trial methodology and was described in detail by Mallarino et al. (1998, 2000) and Bermudez and Mallarino (2002). Yield input data were means of yield monitor points (8–12) recorded at 1-s intervals for areas delineated by the width of the combine head and 15 to 31 m (depending on the field) along the crop rows. Individual yield monitor records were not directly used because of insufficient accuracy at shorter distances (Lark et al., 1997). Yield residuals after removing treatment and block effects with a conventional RCBD ANOVA were used to calculate covariate values for a covariance analysis, which were calculated by subtracting each yield residual from the mean value of its four yield residual neighbors. A conventional ANOVA based on strip yield means was used for Fields 5 and 6 because strips followed a terraced field and two blocks were not contiguous to others. For all fields, the treatment sums of squares were partitioned into comparisons of the control vs. the mean of fertilized treatments and between the two application methods.

For analysis of treatment effects on yield for areas within fields, yield responses were assessed by two procedures. Procedure 1 analyzed treatment effects on yield for field areas testing within Iowa STP interpretation classes by a procedure developed by Oyarzabal et al. (1996), and recently used by Bermudez and Mallarino (2002) and Wittry and Mallarino (2004). Yield input data were means for the grid cells defined by the width of each treatment strip and the separation distance of the soil sampling grid lines. The STP input data of analyses for the first two crops were values from soil samples collected from the entire experimental area before the first P application and for the last two crops were values from samples collected from the control strips. Three yield means (one for each treatment) corresponded to one initial STP value. To assess the consistency of treatment effects for field areas testing within different STP classes for each crop and field, we used ANOVA for a RCBD for each STP class in which sources of variation were replications (blocks) and P treatments. Procedure 2 used similar data management and ANOVA test treatment effects for different soil series. Yield data for areas encompassed by each treatment, replication, and soil series from digitized soil-survey maps at a 1:12 000 scale (Iowa Cooperative Soil Survey, 2002) were averaged using ArcView. Values were not used for these two procedures when there were less than three yield cells for any STP class or soil series and when a class or soil was not represented in at least two replications.

The yield variation for each treatment was assessed by SD and geostatistical parameters using yield averages for the small cells described above. One unidirectional semivariogram was calculated along strips of each treatment using S-Plus version 6.0 and Spatial Statistics Supplement (Insightful Corp., 2001, Seattle, WA). Journel and Huijbregts (1978) discuss theoretical and practical geostatistical concepts used in this study. Semivariance values were calculated for minimum and maximum lag distances of 15 m and 60% of the strip length (120 to 320 m depending on the field), respectively. A spherical model that estimates nugget, sill, and range parameters was the best-fitting model in most instances and is the only one presented. The nugget represents random variability and spatially structured variability at a scale smaller than the minimum distance between samples. The range is the maximum distance two measurements are correlated. As the distance between samples increases up to the range, the semivariance increases from the nugget value toward a maximum called the sill. The difference between these two values represents spatially structured variability. Nugget and sill semivariances are the most relevant for this study and are the ones presented and discussed.

Treatment effects on STP were evaluated by analyzing soil samples collected after harvesting the third crop of the study using procedures described before. These samples reflected the effects of P application for the two rotation cycles (before the first and third crops). Differences in mean STP for strips receiving no P or P with FR or VR application methods were assessed by a conventional ANOVA for RCBD while SD was used to describe the STP variability for each treatment.


   RESULTS AND DISCUSSION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Yield Response to Phosphorus and Fertilizer Application Methods
Phosphorus fertilization (mean of both application methods) increased (P ≤ 0.05) grain yield as represented by treatment strip averages in 13 site-years (eight corn crops and five soybean crops). Fertilization increased yield of all crops in Fields 1 and 2, the two corn crops in Field 3, the last crop (corn) in Field 4, the last two crops in Field 6 (Table 4), and did not increase yield at other site-years. The fertilizer application method affected (P ≤ 0.05) grain yield only in Field 2b, where P applied using FR increased soybean yield more than P applied using VR. The difference between FR and VR was statistically significant at Field 5a2 but only because the yield of the control was intermediate and actually there was no yield response to P by any method (comparisons of each method and the control were not significant even at P ≤ 0.1).


View this table:
[in this window]
[in a new window]

  		Table 4. Corn and soybean grain yield response to P applied with two fertilization methods.

 
Soil-test results from samples collected before the first treatment application (Table 3) showed that STP encompassed at least four Iowa interpretation classes in all fields. The proportion of grid cells testing Very Low or Low ranged from 8% in Field 5 to 92% in Field 6. Large to moderate yield response to P should be expected when STP is Very Low or Low (Mallarino, 1997; Sawyer et al., 2002; Dodd and Mallarino, 2005). Consistent yield responses to P in Fields 1 and 2 can be explained by large (75 and 67%, respectively) low-testing areas (≤16 mg P kg–1). In Field 3, 34% of the area tested Low in STP but only 5% tested Very Low. A consistent response from corn but not from soybean agrees with results reported by Dodd and Mallarino (2005), who showed slightly lower STP requirements for soybean than for corn. A response only of the last crop in Field 4 and no response of any crop in Field 5 can be explained by small initial low-testing areas (25 and 8%, respectively). A lack of significant response of the first crop in Field 6 (Field 6a) and a responsive trend for the second crop (Field 6b) significant at only P ≤ 0.10 cannot be explained with certainty. This field was managed with no-till and was the only field in which the P was applied in spring. Although Iowa research (Bordoli and Mallarino, 1998; Borges and Mallarino, 2000) showed no P placement differences for no-till corn and soybean for P applied in the fall, perhaps a broadcast application near planting time was not efficient for the first crop in this field. This possible reason cannot explain the weak responsive trend of the second crop, although such a response could have become statistically significant if we had been able to account for spatial correlation of yield as for other fields. Previous work (Mallarino et al., 1998, 2000; Bermudez and Mallarino, 2002) showed that small differences often become significant by using spatial analysis in conjunction with ANOVA.

The average amount of P fertilizer applied varied considerably between application methods but often was less for VR than for FR (Table 2). The VR method applied less P than FR in 9 site-years (3–29 kg P ha–1 less) and more in 3 site-years (6–11 kg P ha–1 more), which on average across all site-years resulted in VR applying 9.8 kg P ha–1 or 29.4% less than FR. Although differences are explained by differences in STP between fields, attention is needed when interpreting results for Fields 4a2, 5a, and 5a2 for reasons explained in the methods section. In Field 5a, we applied an FR rate (34 kg P ha–1) lower than the rate called for by median STP (54 kg P ha–1). In Fields 4a2 and 5a2, we applied an FR of 24 kg P ha–1 when no P fertilizer was needed according to high median STP of the FR strips. Consideration of the FR rates that should have been applied for Fields 4a2, 5a, and 5a2 indicates that VR would have applied less P than FR in four fields (Fields 1, 2, 3, and 6), the same P as FR in Field 4, more P than FR in Field 5 (14.5 kg P ha–1 more), and on average would have applied 4.2 kg P ha–1 or 12.4% less than FR. Based on the lack of response to P in Fields 4a2, 5a, and 5a2, we believe the latter estimates more appropriately represent differences in rates between application methods for this study.

Yield Responses for Field Areas Testing Within Different Soil-Test Interpretation Classes
Analysis of grain yield response for field areas testing within different STP classes showed that yield response to P was observed only when STP was Optimum or less (≤20 mg P kg–1). The corn responses to P (P ≤ 0.05) in these field areas (Table 5) were observed for all fields in which strip-average responses were detected. We cannot reasonably explain an odd yield decrease due to P fertilization with both application methods in areas of Field 5b testing Optimum in STP. Probably this was a random effect because the P rates applied should not decrease yield. The application methods differed for corn only for the Very Low class of Field 2a2, where VR increased yield more than FR. This result seems reasonable because this field had large areas testing Very Low and VR applied more P than FR for this class. However, the analysis of strip averages for this field and crop (Table 4) showed no difference between methods. For example, the application methods did not differ for areas testing Low in this field but yield from strip averages tended to be higher with FR. These differences can be explained by nonsignificant responsive trends in other field areas. We could not analyze responses for high-testing areas of this field due to insufficient replication.


View this table:
[in this window]
[in a new window]

  		Table 5. Corn yield response to P for field areas testing in different soil-test interpretation classes.

 
Analysis of soybean responses for field areas with different STP (Table 6) showed significant responses (P ≤ 0.05) for areas testing Optimum or less in all fields where strip-average responses were observed except for Field 2b, where there was no significant yield response in areas testing Low or High. Yield responses were also observed where STP was Very Low in Fields 3b and 6b, in which no response was observed for strip averages. A discrepancy between results for strip averages and field areas could be expected because of the relative impact of treatment differences for yields representing different proportions of the field on average yield response and statistical tests. The application methods differed for soybean only in areas of Field 1a testing Optimum, where VR increased yield more than FR. This was probably a spurious and random difference because VR applied less P than FR in areas testing Optimum of this field and the methods did not differ in low-testing areas.


View this table:
[in this window]
[in a new window]

  		Table 6. Soybean yield response to P for field areas testing in different soil-test interpretation classes.

 
Yield Responses for Field Areas with Different Soil Series
Because of imposed replication requirements, analysis of yield response to P for field areas with different soil series was done for two dominant series for Field 1 and 2 and three dominant series for Fields 3 and 4. Therefore, results for Fields 5 and 6 are not shown. Trials at Fields 1, 2, 3, and 4 were conducted on soils of the same association, although the proportion of the different soils in the experimental areas differed. Data for corn (Table 7) and soybean (Table 8) showed that responses sometimes differed among soil series within these fields. The Clarion soil was more responsive to P than the other soils in 5 site-years (Field 1b2 and 3a for corn and Field 1a, 2b, and 2b2 for soybean) but responses were larger for other soils in Fields 3a2 and 4b2 or there were no differences. Lower STP for the Clarion soil in Fields 1 and 2 (Table 1) could explain a larger yield response for this soil in these fields. However, STP was within the same interpretation classes or was higher for the Clarion soil in Fields 3 and 4.


View this table:
[in this window]
[in a new window]

  		Table 7. Corn grain yield response to P for field areas with different soil series.

 

View this table:
[in this window]
[in a new window]

  		Table 8. Soybean grain yield response to P for field areas with different soil series.

 
Wittry and Mallarino (2004), working on other fields within this soil association, also found a higher frequency of response to P for the Clarion soil. All soils of this association formed on loam glacial till, but the Clarion series occupies higher and steeper landscape positions and is better drained than the Canisteo, Nicollet, or Webster series. Speculation about reasons for this different response is risky because the soils differ in many other properties. The response to P application method seldom differed across soils. The FR method produced slightly higher yield than VR in the Nicollet soil of Field 3a2 but lower yield in the Clarion soil of Field 1a. We could not explain these differences based on STP distribution for the soils or P rates applied, and most likely were random results.

Summary Discussion of Differences between Phosphorus Application Methods
Three possible reasons could explain infrequent, small, and inconsistent differences observed between FR and VR fertilization methods, even for field areas testing low in P. One reason may be inadequate assessment of STP variability, even with the very dense sampling approach used in the study. An important reason we decided to conduct this study was because a previous study by Wittry and Mallarino (2004) used a less dense soil sampling method and they believed that could have explained a lack of differences between FR and VR in their study. However, our results also showed no difference between methods even though the grid-point sampling method we used (0.06–0.08 ha cells) was denser than in the previous study, and denser than any grid sampling method being used by farmers or consultants in the Corn Belt (1.0 to 1.8 ha cells). Another reason might be the use of 2-yr fertilizer recommendations for the corn–soybean rotation applied once before the first crop (because excess P likely is applied by both methods for the first crop), but this should not explain lack of differences for the second crop. The most likely reason for a lack of difference between application methods is use of P recommendations for low-testing soils designed to maximize yield and build-up STP over time to Optimum levels. This reason was suggested before by Anderson and Bullock (1998) and Wittry and Mallarino (2004). If the P application rates are higher than needed to maximize yield, any higher P application with VR than with FR would not result in higher yield unless the low-testing field areas are very large and extremely low in STP.

Effect of Variable-Rate Phosphorus Application on Yield Variability
Standard deviations showed that yield variability differed greatly across fields and crops. In corn, P application with VR or FR methods usually decreased or did not affect (P ≤ 0.05) yield variability (Table 9). Fertilization decreased variability in Fields 1b, 2a, 2a2, 3a, 4b2, and 6a, increased it in Field 3a2, and did not affect it clearly in other fields. Fertilization effects on soybean yield variability were less clear and consistent than in corn (Table 10). Fertilization decreased yield variability in Fields 3b and 4a, increased it in Fields 1a2, 2b, 3b2, and 5a, and did not affect it clearly in other fields. The VR method reduced corn yield variability compared with FR in Fields 2a, 2a2, and 5b and soybean yield variability in Fields 3b, 3b2, and 6b. However, VR increased corn yield variability in Field 5b2 and soybean variability in Fields 5a and 5a2.


View this table:
[in this window]
[in a new window]

  		Table 9. Effect of fixed rate and variable rate on corn yield variability and spatial structure.

 

View this table:
[in this window]
[in a new window]

  		Table 10. Effect of fixed rate and variable rate on soybean yield variability and spatial structure.

 
Study of semivariograms parameters sill (an estimate of total variability) and nugget (an estimate of the random variability), as well as their ratio (an index of the proportion of the random and spatially structured variability) also indicated differences between treatments for some crops and fields. In corn, both nugget and sill semivariances were consistently lower for fertilized soil than for the nonfertilized control except for Fields 3a2, 5b, and 6b2 (Table 9). These differences between fertilized and nonfertilized soils agreed with differences indicated by SD with few exceptions (in Fields 1b2, 5b, and 5b2). The spatial structure of the corn yield variability measured as the sill/nugget ratio was not affected by fertilization with three exceptions. In Fields 4b2 and 5b2, both fertilization methods reduced random variability proportionally more than spatially structured variability compared with the control, and in Field 6b2 the only clear difference was that FR significantly reduced random variability compared with control or VR treatments. We cannot explain this reduction in random variability satisfactorily because the yield maps for these fields (not shown) indicated very small variation for neighboring yield observations along fertilized strips where nugget semivariance was very small but we do not know the reason. In soybean fields (Table 10), P fertilization did not consistently affect sill and nugget semivariances. Fertilization increased both parameters in Fields 1a2, 2b, 5a, and 6a2, decreased them in Field 3b, and did not affect them in Field 1a. The VR method tended to reduce total variability in two-thirds of the fields compared with FR, a result that is in contrast with the less consistent results for SD.

The results indicated that P fertilization usually, although not always, decreased yield variability compared with nonfertilized strips. Bermudez and Mallarino (2004), using a similar methodology, found inconsistent effects of starter fertilizer containing P on corn yield variability in seven fields although it usually increased early growth variability. In our study, both SD and semivariances showed that fertilization by either application method often decreased corn yield variability and seldom increased it. Less consistent effects on soybean yield variability might be explained by less frequent and relatively smaller responses than for corn. We theorized that VR fertilization could reduce spatially structured variability by increasing yield in low-testing areas and not affecting yield in high-testing areas, but could increase random variability when yield was increased. The results showed that in the responsive sites, VR decreased yield variability compared with FR more frequently than increased it. This result appears to be in contradiction with a lack of yield difference between fertilization methods. This contradiction might be explained by small, variable, and statistically nonsignificant yield differences between application methods.

Effect of Variable-Rate Phosphorus Application on Soil-Test Phosphorus
Table 11 shows summarized STP results for soil samples collected after harvesting the third crop of the study. Soil-test P data from this sampling date are useful to study effects of P application methods on STP because they reflect the cumulative effects of two P applications (before the first and third crops). Fertilization increased STP (P ≤ 0.05) compared with the control treatment in all fields. Soil-test P for VR was lower than for FR in Fields 1, 3, and 4, although there was also a similar trend for Fields 4 and 5. This result agrees with usually smaller amounts of P fertilizer applied and lower yield variability for this application method in many fields. A more important result was that the STP variability was lower for VR than for FR with the only exception of Field 2, which had a much higher STP variability (at least twice) than all other fields (including the nonfertilized strips). A lower STP variability for VR is reasonable because this method was designed to apply higher P fertilizer rates to low-testing areas and no fertilizer to high-testing areas. Wittry and Mallarino (2004) observed a similar result. Therefore, although VR did not increase yield compared with FR, it did manage P application better because less unneeded P was applied to high-testing areas and it reduced within-field STP variability. Because research has shown linear or exponential increases in total or dissolved P loss from fields through surface runoff and subsurface drainage when STP increases (Klatt et al., 2003; Allen et al., 2006), our results strongly suggest that P application using VR can reduce P loss from fields compared with FR and could result in improved water quality.


View this table:
[in this window]
[in a new window]

  		Table 11. Soil-test P after harvesting the third crop of the study as affected by P fertilization with two application methods.

 

   CONCLUSIONS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
Phosphorus fertilization increased grain yield response as evaluated by strip averages in 13 of 24 site-years and the application methods differed only in 1 site-year, where FR increased yield more than VR. Yield responses for field areas testing within different STP interpretation classes showed responses in 16 site-years only when STP was ≤ 20 mg P kg–1. Fertilizer application methods differed only for areas testing Very Low of 1 site-year, where VR increased yield more than FR. Semivariograms and SD showed that VR usually, but not always, reduced within-field yield variability. A lack of difference between fertilization methods in-spite of using a very dense soil sampling approach might be explained by high STP variability at a small scale that current soil sampling methods and VR technology cannot manage. Also, the use of high fertilizer rates for low-testing soils to maximize yield and slowly build-up STP over time may have contributed to a lack of differences between FR and VR.

Although the VR fertilization method did not increase yield compared with FR, on average across all fields VR applied 12.4% less P fertilizer than FR, reduced within-field STP variability in most field, and reduced the P applied to high-testing field areas. Therefore, although VR did not increase yield compared with FR, it did better P management and showed potential for reducing excess P loss from fields to water resources through reduced P fertilization in high-testing field areas.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Project supported in part by the Iowa Soybean Association, the Leopold Center for Sustainable Agriculture, and the United Soybean Board.


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 




This article has been cited by other articles:


Home page
 Crop Sci.Home page
C. A. Jackson, J. M. Windes, P. Bregitzer, D. Obert, W. Price, and B. Brown
Phosphorus Fertility Effects on the Expression of the Low Phytic Acid Barley Phenotype
Crop Sci., August 7, 2009; 49(5): 1800 - 1806.
[Abstract] [Full Text] [PDF]

Home page
 Soil Sci.Home page
M. M. Barbazan, A. P. Mallarino, and J. E. Sawyer
Liquid Swine Manure Phosphorus Utilization for Corn and Soybean Production
Soil Sci. Soc. Am. J., March 1, 2009; 73(2): 654 - 662.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Bermudez, M.
Right arrow Articles by Mallarino, A. P.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Bermudez, M.
Right arrow Articles by Mallarino, A. P.
Agricola
Right arrow Articles by Bermudez, M.
Right arrow Articles by Mallarino, A. P.
Related Collections
Right arrow Phosphorus
Right arrow Nutrient Management
Right arrow Soil Fertility and Productivity

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Crop Science Vadose Zone Journal
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome