PRECISION AGRICULTURE Soil-Sampling Alternatives and Variable-Rate Liming for a Soybean–Corn Rotation Agustin A. Bianchini and Antonio P. Mallarino* ABSTRACT graphical coordinates of each sample or cells to create Precision agriculture technologies can be used to manage soil pH. maps. The sampling intensity required for effective use This study compared soil-sampling schemes for pH and evaluated of VR technology (VRT) is not clearly defined and may variable-rate (VR) liming for soybean [Glycine max (L.) Merr.] and be different for different soil tests, fields, and geographic corn (Zea mays L.). Global positioning systems and yield monitors regions. Research (Wollenhaupt et al., 1994; Franzen were used in two fields. Soils were Typic Hapludolls, Aquic Haplu- and Peck, 1995) showed that grid soil sampling at densi- dolls, and Typic Endoaquolls. Treatments were a control and a fixed- ties of 0.4 and 0.1 ha further increase accuracy of soil rate (FR) or VR liming based on a 0.2-ha soil-sampling scheme. Soil test mapping. However, research (Wollenhaupt et al., pH (15-cm depth) ranged from 5.4 to 8.4, and most subsoils were 1994; Mallarino, 1996) has also shown that soil test vari- calcareous. Treatments were applied before soybean to long and repli- cated strips. Grain yield and soil pH were measured during 3 yr in ability within cells of that size can be very high and that one field and 2 yr in the other. Initial pH; canopy photos; and soil use of systematically aligned sampling points may lead survey, elevation, and electrical-conductivity maps were used to simu- to large error when soil test patterns tend to be cyclic. late sampling schemes based on larger cells or zones. Liming increased Zone sampling has recently been suggested to reduce (P 0.05) corn yield in one site-year (230 kg ha 1), and liming meth- number of samples and sampling costs while main- ods did not differ. The VR method applied less lime (56–61%) and taining acceptable information about nutrient variation reduced pH variability in one field. The lack of response was explained within fields. Sampling by zone assumes that sampling by high subsoil pH and high small-scale variation of topsoil pH. Sam- areas can be identified on the basis of zones with differ- pling schemes based on 0.7-ha cells or zones identified smaller acid and ent soil or crop characteristics across a field and that alkaline areas than schemes based on small cells. Results suggest that patterns are likely to remain temporally stable (Franzen yield response from lime is not likely when calcareous subsoils are present and topsoil pH is as low as 5.5. A VR liming method would et al., 2000). Criteria used to delineate management apply less lime than a FR method in soils similar to those in this study. zones vary. Topography and soil and crop canopy im- ages can be used to identify management zones because they tend to reflect different soil properties, are nonin- vasive, and may be of low cost (Franzen et al., 1998; T he value of liming acid soils to increase soil pH to values optimal for crops is well known. McLean and Brown (1984) presented in a detailed review the Schepers et al., 2000). Soil electrical conductivity (EC), which can be estimated using noninvasive electromag- beneficial effects of lime in corn and soybean for the netic induction methods, has been useful to estimate U.S. Corn Belt. Soil testing is a useful diagnostic tool, topsoil depth (to a claypan or other root growth–limiting but high variability is often observed for pH and other layer) and physical and chemical soil properties and to nutrients in farmer’s fields (Cahn et al., 1994; Cambar- explain yield variability (Doolittle et al., 1994; Kitchen della et al., 1994; Mallarino, 1996; Pierce and Warncke, et al., 1999; Myers et al., 2000). Yield maps can be 2000). Variability patterns of these nutrients sometimes used to define different soil productivity areas, which are related to soil map units but not when fertilization together with other layers of information, can be used and liming have increased soil test values and created as a basis for VR fertilization (Stafford et al., 1999). new patterns of variability (Franzen and Peck, 1995; Colvin et al. (1997) showed, however, that stable within- Mallarino, 1996). Fields where fertilizers have been field yield patterns over time are observed in some fields banded or where high rates of nutrients and manure but not in others. were used show large small-scale nutrient variability Benefits from VR liming may include larger yield (Peck and Melsted, 1973; Mallarino, 1996). Different increases in acidic areas and lime savings in high-pH sampling schemes can be used to collect soil samples areas, but crop response should offset likely higher costs from fields. Grid soil sampling began to be used exten- of soil sampling and application (Pierce and Warncke, sively in the early 1990s in the Corn Belt, and it refers 2000). Bongiovanni and Lowenberg-DeBoer (2000) to a process whereby a field is divided into many smaller simulated corn and soybean yields using soil pH re- cells for sampling purposes. Soil sampling for pH, P, sponse functions from small-plot data and predicted and K often is based on square 1- to 2-ha grids (Sawyer, larger annual returns with site-specific pH management. 1994). The results of analyses are combined with geo- Soil test data from a field sampled by Borgelt et al. (1994) suggested that 3.4 to 4.5 Mg ha 1 lime was needed Dep. of Agron., Iowa State Univ., Ames, IA 50011. Iowa Agric. and Home Econ. Exp. Stn. Journal Paper no. J-19556. Project 4062. This Abbreviations: CCE, calcium carbonate equivalent; EC, electrical project was supported in part by the Iowa Soybean Promotion Board. conductivity; DGPS, differential global positioning system; FR, fixed Received 8 Oct. 2001. *Corresponding author (email@example.com). rate; NNA, nearest-neighbor analysis; RCBD, randomized complete block design; SD, standard deviation; VR, variable rate; VRT, vari- Published in Agron. J. 94:1355–1366 (2002). able-rate technology. 1355 1356 AGRONOMY JOURNAL, VOL. 94, NOVEMBER–DECEMBER 2002 and that a uniform rate would have resulted in overlim- native soil-sampling schemes for describing soil pH vari- ing of 9 to 12% of the field and underliming of 37 to ability over a field and (ii) assess the impacts of FR and 41% of the field. Mulla et al. (2000) estimated lime VR lime application methods on soil pH and grain yield requirements of a 12-ha field by collecting soil samples of a corn–soybean rotation using production agricul- from cells of various sizes (9 by 9, 18 by 18, or 100 by ture equipment. 100 m) and by simulating a sampling scheme based on near-infrared reflectance images of bare soil and MATERIALS AND METHODS soybean canopy. Areas needing lime were 1.3 ha for the 9- by 9-m scheme, 3.4 ha for the 18- by 18-m scheme, Grain Yield and Soil pH Response Study none for the 100- by 100-m scheme, and 0.6 ha for the A field response study was conducted from 1998 to 2000 targeted sampling scheme. Heiniger and Meijer (2000) in one field (Field 1) and from 1999 to 2000 in another field used soil samples collected on 1-ha square grids from (Field 2) using a strip-trial methodology, DGPS, VRT, and four eastern U.S. states to estimate amounts of lime grain yield monitors. The fields were located in central Iowa required for uniform or VR application. Based on simu- (Boone County), had soils of the Clarion–Nicollet–Webster lated corn yield response and soil pH data, they con- soil association, and were managed with a 2-yr corn–soybean cluded that use of VR lime application would have rotation. Areas of approximately 15 ha in Field 1 and 18 ha in Field 2 that were located at least 50 m from field borders resulted in an average profit increase of $4.03 ha 1 com- were selected for the experiments. Besides the Clarion, Ni- pared with the uniform application. Pierce and Warncke collet, and Webster soil series, the Field 1 had the series (2000) applied five lime treatments for corn and soybean Canisteo (Typic Hapludoll), Crippin (Aquic Hapludoll), to small field plots (4.5 by 30.5 m) located according to Harps (Typic Calciaquoll), and Storden (Typic Eutrodepts). interpolated surfaces from soil samples collected from Field 2 had the series Canisteo, Clarion, Harps, Nicollet, and 30.5-, 61-, and 91.5-m cells. They reported that grid two variants (silty-clay loam and mucky-silt loam) of Okoboji soil sampling did not accurately predict soil pH or lime (Cumulic Vertic, Endoaquoll). Treatments were a control, a requirements for corn or soybean. FR lime application method, and a VR application method Yield monitor maps, differential global positioning based on a surfaced map from an intensive grid soil-sampling system (DGPS) receivers in combines, and a strip-trial scheme and were applied once before the first soybean crop. There were four replications (blocks) in Field 1 and three in methodology can be used to evaluate the effects of VRT Field 2. The individual treatment strip size was 18 by 624 m or other site-specific management practices (Oyarzabal in Field 1 and 24 by 900 m in Field 2. The strip width was et al., 1996; Colvin et al., 1997; Mallarino and Wittry, decided on the basis of the spreading width of the commercial 1997; Mallarino et al., 2001). Treatments are applied to lime spreaders used and the width of the farm combine head- narrow (usually the width is a multiple of the equipment ers. Permanent plastic pipes were buried at each corner of width used to apply the treatments) and long strips the experimental areas, and the geographical coordinates were (generally the length of the field), and crops are har- recorded with a hand-held DGPS receiver. In Field 1, soil vested with combines equipped with yield monitors and samples were collected in November 1997; the lime was ap- DGPS receivers. However, the flow meter data of the plied on 23 Apr. 1998; and soybean was planted on 10 May. In Field 2, soil samples were collected in October 1998; the yield monitor cannot be expected to resolve detailed lime was applied on 16 Dec. 1998; and soybean was planted yield variation over spatial intervals of less than approxi- on 28 May 1999. Soil samples (12 cores, 15-cm depth) were mately 20 to 25 m (Lark et al., 1997). Much of the collected from areas approximately 80 m2 in size randomly research on VR liming discussed previously focused on located (using computer software) within 0.2-ha cells (a grid- describing soil pH variation using various sampling point sampling procedure). Samples were dried in an oven at strategies and simulated responses to lime. Moreover, 35 C, ground to pass a 2-mm screen, and analyzed in dupli- when lime was applied, treatments did not compare cates. Soil pH was measured using a 1:1 (w/v) soil/water ratio yield response to FR and VR application using equip- and a 10-min shaking time (Watson and Brown, 1998). Lime ment used by farmers. requirement was calculated for each sampling point using the SMP (Shoemaker–McLean–Pratt) buffer method as described The objectives of this study were to (i) compare alter- by Watson and Brown (1998). The same lime source was used in both fields, had a 91% Table 1. Field areas and lime rates for the fixed-rate and variable- calcium carbonate equivalent (CCE) neutralizing value, and rate liming methods. was predominantly calcitic (230 g kg 1 Ca and 25 g kg 1 Mg). Field Treatment pH class Field area Lime rate† All the material passed through a 4.75-mm screen, 93% through a 2.36-mm screen, and 34% though a 0.25-mm screen. ha Mg ha 1 CCE‡ The lime was spread with commercial broadcast spreaders 1 Fixed All 14.6 5.77 Variable All 14.6 2.54 (spinners) equipped with DGPS receivers and controllers. The 5.7 1.3 6.73 equipment was calibrated by the commercial applicator fol- 5.7–6.2 5.7 4.56 lowing manufacturer’s recommendations. For the VR method, 6.2 7.6 0 lime was applied only when soil pH was 6.3. The lime rates 2 Fixed All 18.0 4.62 were calculated to raise pH to 6.5 and ranged from 0 to 8.2 Variable All 18.0 1.80 5.7 3.6 4.69 Mg ha 1 CCE in both fields (average rates applied for various 5.7–6.2 5.6 2.30 pH ranges are shown in Table 1). Lime application surfaced 6.2 8.8 0 maps were prepared from point-sampling data using the in- † Rates for the variable-rate method are weighed averages (lime was not verse distance method with a distance-weighing exponent applied with the variable-rate method when soil pH was 6.2). value of 2 (Wollenhaupt et al., 1994). The fixed lime rates ‡ CCE, calcium carbonate equivalent. were 5.77 Mg ha 1 CCE in Field 1 and 4.62 Mg ha 1 CCE in BIANCHINI & MALLARINO: SOIL SAMPLING FOR pH AND VARIABLE-RATE LIMING 1357 Field 2 and were applied uniformly along all strips of the FR treatment evaluation. One or two 9-m-wide combine passes method within each field. Based on the collaborators inputs were used from each soybean strip, and two to four 6-m-wide (farmer and local cooperative), the FR used in Field 1 was combine passes were used from each corn strip. The yield based on the average lime requirement of areas with pH monitor data were carefully analyzed for common errors such 5.8. In Field 2, the FR was defined as the average lime require- as incorrect geographic coordinates due to partial loss of good ment of areas with pH 6.3. Iowa State University current differential correction, the effects of waterways, and incorrect recommendations for corn and soybean are to lime soils with settings in the time lag for the grain path through the combine. pH 6.3 (15-cm depth) and raise it to pH 6.5 (Voss et al., Affected data were corrected (such as grain path lags) or 1999), except for a few soil associations with high-pH subsoil deleted (for example, yield points near waterways and when where lime is recommended only below pH 6.0 (but a target the combine stopped within the trial area). Grain yield and pH of 6.5 is still used). The lime was incorporated to a 12- to soil pH data were exported from ArcView to appropriate files 15-cm depth by chisel plowing and disking. Uniform rates of N, P, and K were applied by the farmers following Iowa State for statistical analysis with SAS (SAS Inst., 1996). University recommendations based on soil testing (for P and Grain yield responses to the treatments were analyzed using K) and corn yield potential (for N). three statistical procedures. One procedure assumed a ran- To evaluate the impact of the lime treatments on soil pH domized complete block design (RCBD) for which the yield over time, soil samples were collected from all strips immedi- input data were yield means of each strip (the experimental ately before the lime application and after each crop harvest units). In a second procedure, the spatial correlation of yield using a more intensive sampling scheme than the one used to was accounted for in the RCBD-ANOVA by nearest-neighbor define the VRs. In Field 1, soil samples were collected in April analysis (NNA). The NNA was used to calculate values of a (spring) 1998 immediately before liming and in late October covariate that was included in the RCBD-ANOVA for each or November (fall) 1998, 1999, and 2000. In Field 2, samples field following a procedure used before (Hinz, 1987; Hinz and were taken in November (fall) 1998 immediately before liming Lagus, 1991; Mallarino et al., 1998; Mallarino et al., 2001). and again in November 1999 and 2000. One composite sample The input data were means of yield monitor points recorded (12 cores, 15-cm depth) was collected from an area approxi- for areas delineated by the width of the combine header (9 m mately 80 m2 in size located at the center of each of 144 cells for soybean and 6 m for corn) and the length of the soil- in Field 1 and 180 cells in Field 2. The width of each cell sampling cell along the crop rows (52 m in Field 1 and 45 m coincided with the strip width (18 m in Field 1 and 24 m in in Field 2). The individual data recorded by the yield monitors Field 2), and the length along the strips was 52 m in Field 1 were not directly used because of the known lack of accuracy and 45 m in Field 2. Thus, the area represented by each sample of yield monitors over distances shorter than 20 to 25 m (Lark approximately corresponded to 0.1-ha (0.09 ha in Field 1 and et al., 1997). The first step in the calculation was to obtain 0.11 ha in Field 2). yield residuals by removing treatment and block effects with Two additional sets of soil samples were collected from each field. In fall 1998, composite subsoil samples (three 5-cm- a RCBD-ANOVA. Afterwards, covariate values were calcu- diam. cores) were collected from selected sampling points lated by subtracting each yield residual from the mean value corresponding to the soil series present. Fourteen areas were of its four neighbors (one from each north, south, east, and sampled in Field 1 and 23 in Field 2. Each core was collected west directions). The third procedure assessed treatment ef- to a 91-cm depth and was divided into six 15-cm sections. Soil fects separately for parts of the fields with different pH follow- was analyzed for pH, and samples with pH 7.5 were analyzed ing procedures used by Oyarzabal et al. (1996) and Mallarino for CaCO3 and MgCO3 (Dreimanis, 1962) to calculate CCE. et al. (1998, 2001). In the second set, soil samples (15-cm depth) were collected The yield and pH data input were means for areas defined from transects laid out along strips that received the FR and by 0.1-ha soil-sampling cells. The initial pH values were used VR treatments in two replications of each field and were to classify each cell into five pH classes ( 5.70, 5.70–6.29, analyzed for soil pH. The transects (four in each field) were 6.3–7.2, or 7.2). There were at least 13 cells in a pH class, laid out where pH data from the cell soil-sampling scheme and the maximum number was 61. The F test from a one-way suggested high pH variability along the strips. Composite soil ANOVA was used to estimate the consistency of lime effects samples (eight cores) were collected from 4.5-m2 areas spaced for each pH class. The numerator mean square (between 6 m along 142 m in Field 1 and 135 m in Field 2. groups) represented variation introduced by the treatments, Grain yield was measured and recorded using a combine and the denominator mean square (within groups) repre- equipped with an impact flow-rate yield monitor (Ag Leader sented the average variation within treatments for cells with Technol., Ames, IA) and a real-time DGPS receiver. Differen- a similar pH classification. Tables with grain yield data for tial corrections were obtained through the U.S. Coast Guard each pH class do not show results for the VR method for field AM signal. The monitors recorded yield data with a 9-s interval areas with soil pH 6.2 because this method was not a distinct in 1998 and 1999 and a 1-s interval in 2000. The monitor treatment for these areas. Data for the treatment labeled con- was calibrated outside the experimental areas of the fields trol for the two high-pH classes are means of the control and by weighing all grain harvested along several (at least four) combine passes over the entire length of the fields. Grain VR lime treatments, and statistical tests correspond to an moisture was determined on the go by a sensor located in the orthogonal contrast with the FR method. combine auger, and grain yield was corrected to 155 g kg 1 The effect of the lime treatments on soil pH from each H2O for corn and 130 g kg 1 H2O for soybean. Each combine sampling date was evaluated using two procedures. One proce- pass was identified with a unique number that was recorded dure assessed treatment effects on pH by an ANOVA that with the georeferenced yield data. The raw yield data were assumed a conventional RCBD and for which input data were exported into ArcView (Environ. Syst. Res. Inst., Redlands, pH means for each strip. The second assessed treatment effects CA). Yield data were unaffected by field borders because the on pH for areas of the field with pH within each pH class experimental areas were located at least 50 m from any border. defined for the yield analyses and was the same type of AN- Yield data from combine harvest passes that may have in- OVA used to assess treatment effects on grain yield for areas cluded crop rows from two treatment strips were not used for with different soil pH. 1358 AGRONOMY JOURNAL, VOL. 94, NOVEMBER–DECEMBER 2002 Soil pH Assessment with Various Information collected and characterized for the individual Soil-Sampling Schemes zoning approaches was used to identify an integrated manage- ment-zone approach for both fields. The maximum number Simulations of soil-sampling schemes of various intensities of zones that could have been defined was very large (784 for were conducted for the two fields based on the soil samples Field 1, for example, from seven yield zones, seven soil-series collected immediately before liming. This methodology was zones, four elevation zones, and four EC zones), and many previously developed and used by others (Franzen and Peck, would be very small and would have irregular shapes. Using 1995; Mulla et al., 2000; Pierce and Warncke, 2000). Six simu- our knowledge for these fields (including remote sensing, pro- lated schemes were sampling of 0.3-ha grid cell, 0.3-ha grid duction system, and equipment requirements), we identified point, 0.7-ha grid cell, 0.7-ha grid point, soil series, and man- nine integrated zones in Field 1 and six in Field 2. This ap- agement zone. A vector map with associated information for proach for identifying management zones integrates farmers’ each sampling scheme was created using ArcView. The pH preferences into the zone identification process (Fleming et data of the 0.3-ha grid cell were calculated by averaging the al., 2000). As an example, Fig. 1 shows yield, elevation, EC, point data for three contiguous cells across each row of cells. soil series, and management zones for one field. The pH for The pH data for the 0.3-ha grid-point scheme corresponded each management zone is the mean of corresponding sampling to the single sampling point at the center cell of the same points of the 0.1-ha cells. three cells. The pH data of the 0.7-ha grid cell were calculated Soil pH data for the six schemes were compared by study by averaging the point data for eight contiguous cells in Field 1 of several descriptive statistics and GIS maps. Field areas (four cells across strips and two along strips) and six contiguous represented by each pH class were calculated for each scheme cells in Field 2 (three cells across strips and two along strips). to determine how the schemes would have estimated the size The pH data of the 0.7-ha grid point were identified by ran- of the area that should receive lime. The two lower pH classes domly selecting one sampling point from the cells used to were merged into one class for pH 6.3 to represent in one calculate mean pH for the 0.7-ha cells. The pH data for the class the area that could have been limed. Average soil pH soil-series scheme were the mean pH of all of the 0.1-ha sam- and standard deviation (SD) were determined for the soil- pling points included within each soil series. map, large-grid, and integrated management-zone schemes Management zones were identified using five different ap- and also for the yield, elevation, and EC zones used to deter- proaches. Four approaches used individual attributes (yield, mine the integrated management-zone scheme. Also, F tests soil series, elevation, and EC maps), and one approach inte- based on a two-way ANOVA using SAS PROC GLM (SAS grated this information into a management-zone scheme. For Inst., 1996) were used to compare pH variability between and the yield zones, yield monitor maps from growing seasons within zones for the zone-sampling schemes (soil-map, yield, before treatment application (three maps for Field 1 and two elevation, EC, and integrated management-zone schemes). maps for Field 2) were used to create one yield-zone map for For each scheme, the numerator of the F ratio was the mean each field following a two-step procedure. First, four to five square arising from differences between average pH across areas with different yield levels were delineated using Arc- zones, and the denominator was a pooled mean square for View in maps from each crop using equal intervals. Second, pH variation within zones. A statistically significant difference these maps were used (through visual observations) to create would suggest that the sampling scheme was effective in identi- one map for each field that described seven yield zones in fying field areas with contrasting soil pH. The size of the F Field 1 and six zones in Field 2. Some field areas had consis- value can be interpreted as an index of the effectiveness of tently higher or lower yield over time compared with other each sampling scheme to reduce within-zone pH variability areas and were identified as separate zones. At least one zone and increase pH differences between zones. in each field corresponded to areas containing large temporal yield variability. For the soil-map zones, soil series (seven in Field 1 and six in Field 2) were obtained from digitized RESULTS AND DISCUSSION (1:12 000 scale) soil survey maps (Andrews and Dideriksen, 1981). Elevation models and EC maps for the elevation and EC Lime Use and Soil pH zones were obtained after harvesting the 1998 crops by driving The VR method applied 56% less lime than the FR a vehicle equipped with a high-precision DGPS receiver (4000 in Field 1 and 61% less in Field 2 (Table 1). The average Total Station with a real-time kinematic system, Trimble, Sun- variable lime rates used in areas with soil pH 5.7 were nyvale, CA) and an electromagnetic induction sensor (EM-38, markedly higher than the FR in Field 1 (17%) but were Geonics Limited, Mississauga, ON, Canada). Elevation and EC data for 320 observations (points) per hectare were im- only slightly higher in Field 2 (1%). In areas with soil ported into ArcView to create surface maps. The elevation pH values between 5.7 and 6.2, the VR method applied range was approximately 8 m in both fields. The EC values significantly less lime than the FR method in both fields ranged from approximately 8 to 70 mS m 1 in each field. Both (18% less in Field 1 and 50% less in Field 2). Other elevation and EC values were mapped into four equidistant research has shown that VR application reduces lime classes. An aerial digital color image (1-m resolution) of the application rates (Heiniger and Meijer, 2000). The dif- soybean canopy was taken from each field in late June of one ferences in lime use may vary, however, depending on year. Each image was imported into ArcView, and although the distribution of pH values and lime needs. In our zones based on color differences were not delineated, visual fields, although soil pH ranged from 5.5 to 8.2 in Field observations of contrasting color differences were used to help 1 and from 5.4 to 8.4 in Field 2, large areas had pH create the integrated management-zone maps. The photos 6.2 and were not limed with the VR method. Also, there showed small areas ( 10% of the experimental areas) with chlorotic soybean canopy. There was no attempt to identify is no widely accepted criterion to decide the fixed lime the reason for the chlorosis. In this soil association, soybean rate to use in field experiments or when producers chlorosis at early stages usually is associated with excess mois- choose to apply a uniform rate over a field. ture, Fe deficiency induced by high soil pH, or severe infesta- Table 2 shows mean soil pH data for each treatment tion with soybean cyst nematode (Heterodera glycines). and sampling date. Data from the first sampling date BIANCHINI & MALLARINO: SOIL SAMPLING FOR pH AND VARIABLE-RATE LIMING 1359 Fig. 1. Examples (using Field 2) of grain-yield zones, elevation zones, electrical-conductivity zones, soil-series zones, and integrated manage- ment zones. (immediately before applying the lime treatments) 0.05). Liming did not increase soil pH significantly in showed that pre-existing pH of areas that would receive Field 1, but there were increasing trends for both appli- the three treatments did not differ significantly (P cation methods in all sampling dates. Lime increased Table 2. Descriptive statistics of soil pH for each treatment across the two fields sampled. Descriptive statistics‡ Lime effect§ Field Sampling date† Treatment Mean Max Min Range SD Lime F-V pH P F 1 Initial (spring 1998) No lime 6.78 8.18 5.55 2.63 0.96 NS NS Fixed 6.81 8.18 5.68 2.50 0.95 Variable 6.78 8.20 5.50 2.70 0.95 Fall 1998 No lime 6.62 8.10 5.30 2.80 1.10 NS NS Fixed 6.80 8.25 5.63 2.62 0.97 Variable 6.88 8.50 5.75 2.75 0.92 Fall 1999 No lime 6.54 8.05 5.35 2.70 1.03 NS NS Fixed 6.78 8.00 5.60 2.40 0.92 Variable 6.79 8.13 5.65 2.48 0.91 Fall 2000 No lime 6.65 8.18 5.40 2.78 1.06 NS NS Fixed 6.95 8.10 5.85 2.25 0.89 Variable 6.99 8.15 5.75 2.40 0.83 2 Initial (fall 1998) No lime 6.62 8.10 5.40 2.70 0.96 NS NS Fixed 6.52 8.05 5.28 2.77 0.88 Variable 6.61 8.35 5.25 3.10 0.96 Fall 1999 No lime 6.65 8.10 5.53 2.58 0.88 0.04 0.05 Fixed 6.73 8.10 5.60 2.50 0.72 Variable 6.90 8.05 5.45 2.60 0.79 Fall 2000 No lime 6.69 8.23 5.73 2.50 0.91 NS 0.06 Fixed 6.66 8.18 5.45 2.73 0.82 Variable 6.89 8.18 5.40 2.78 0.87 † The initial soil sampling was done immediately before applying the lime (spring 1998 in Field 1 and fall 1998 in Field 2). Any other sampling was done in fall (October or November) of each year. ‡ Max, maximum soil test value; Min, minimum soil test value; SD, standard deviation. § Lime, orthogonal comparison of the control vs. the mean of the two limed treatments; F-V, orthogonal comparison of the fixed-rate and variable-rate lime treatments. ¶ NS, not significant at P 0.1. 1360 AGRONOMY JOURNAL, VOL. 94, NOVEMBER–DECEMBER 2002 Table 3. Soil pH for different sampling dates, treatments, and pH classes for two fields. Soil pH by sampling date and pH class Spring 1998† Fall 1998‡ Fall 1999 Fall 2000 Field pH class Treatment pH P F§ pH P F pH P F pH P F 1 5.7 No lime 5.63 NS 5.45 0.01¶ 5.49 0.01¶ 5.55 0.01¶ Fixed 5.69 5.65 5.69 5.95 Variable 5.61 6.00 5.93 6.23 5.7–6.29 No lime 5.99 NS 5.70 0.01¶ 5.70 0.01 5.78 0.01¶ Fixed 5.97 5.95 5.97 6.13 Variable 5.96 6.18 5.98 6.29 6.3–7.2 No lime 6.76 NS 6.50 NS 6.54 NS 6.59 0.09 Fixed 6.57 6.86 6.65 6.95 7.2 No lime 7.84 NS 7.76 NS 7.68 NS 7.81 NS Fixed 7.87 7.82 7.79 7.92 2 5.7 No lime 5.55 NS 5.78 0.01 5.93 NS Fixed 5.51 6.01 5.82 Variable 5.51 6.12 5.99 5.7–6.29 No lime 6.00 NS 6.15 0.01 6.10 0.02 Fixed 6.03 6.45 6.33 Variable 6.03 6.34 6.34 6.3–7.2 No lime 6.61 NS 6.75 NS 6.69 NS Fixed 6.58 6.71 6.70 7.2 No lime 7.85 NS 7.79 NS 7.89 NS Fixed 7.81 7.60 7.66 † Initial soil sampling immediately before liming in Field 1. ‡ Second soil sampling (first after liming) in Field 1 and initial soil sampling (immediately before liming) in Field 2. § Probability of orthogonal comparisons between the control and the mean of the two application methods for the two lower pH classes and between the fixed-rate method and the mean of the control and variable-rate method for the two higher pH classes (no lime was applied with the variable-rate method when pH was 6.2). ¶ An orthogonal comparison between the fixed-rate and variable-rate methods was significant at P 0.1. soil pH in the 1999 sampling date of Field 2, and the VR no effect in Field 2 (Table 4). The means for the RCBD method increased soil pH more than the FR method. analysis correspond to observed yields, and means for Results for the 2000 sampling date of Field 2 are difficult the RCBD-NNA are least square means that were ad- to explain because only the VR method seemed to have justed for the spatial correlation of yield. The RCBD increased soil pH. The lime main effect was not signifi- and RCBD-NNA means for each treatment were almost cant (P 0.12), and the comparison between applica- exactly the same, a result that was observed in other tion methods was significant at P 0.06. Strips that studies (Mallarino et al., 1998). However, adjusting for received the VR treatment had less soil pH variability spatial correlation with NNA reduced the standard error (SD) than the control or FR treatments in the fall 1998 of treatment means. The 1999 corn showed a positive and 2000 sampling dates of Field 1. In Field 2, the FR small response to lime (230 kg ha 1), which was statisti- treatment had the lowest SD in the two sampling dates cally significant with both methods of analysis. The after the lime application. However, it should be noted RCBD-NNA method of analysis suggested a very small that the initial SD for plots that would later receive the negative soybean response in 1998 (P 0.07). This FR treatment was lower than for the other treatments. negative response was mainly due to a lower yield of A reduction in variability from either FR or VR liming the VR method and is difficult to explain. can be explained by a larger pH increase in acid areas Table 5 shows yield means by pH class, treatment, than in high-pH areas. and year for both fields. In Field 1, liming had no influ- Table 3 shows mean soil pH data for each treatment ence on yield in any pH class. This result makes the and each of four pH classes. Results for the first soil- small field-average positive response to lime of the 1999 sampling date (before liming) indicated no significant corn crop in Field 1 difficult to explain although there differences between treatments. As expected, the lime was a small nonsignificant responsive trend for the 5.7 treatments usually increased soil pH (P 0.05) in the to 6.29 pH class. The corn crop of Field 2 showed lower two more acidic pH classes (except for one acidic pH yield (P 0.01) for the FR method compared with the class in the 2000 sampling date of Field 2). The VR control in the 6.3 to 7.2 pH class. This result would be method increased pH more than the FR method for soil possible if excess lime had detrimental effects on yield within the most acidic class of Field 1 but not in Field through a reduction in availability of other nutrients 2. A larger pH increase in areas with the most acid soil (McLean and Brown, 1984). However, we did not detect with the VR method is reasonable because more lime treatment differences for soils with pH 7.2. was applied with this method than with the FR method. There could be several reasons for the small or nonex- The FR method did not affect soil pH in the neutral or istent crop response to lime. One likely reason is the high-pH classes. presence of high-pH (calcareous) subsoils in both fields. Eighty-nine percent of the deep sampling points had Grain Yield Response to Lime soil pH 7.4 and were calcareous ( 2% CCE) at some The lime treatments had little effect on mean corn depth (0–91 cm). Thirty-eight percent were calcareous and soybean yields along the strips of Field 1 and had at all depths, and 51% were calcareous at a 30- to 9-cm BIANCHINI & MALLARINO: SOIL SAMPLING FOR pH AND VARIABLE-RATE LIMING 1361 Table 4. Effect of lime application on corn and soybean yields evaluated by two methods of analysis. Method of statistical analysis† RCBD RCBD-NNA Field Crop Year Treatment Yield SE‡ Lime§ F-V¶ Yield SE Lime F-V 1 1 kg ha P F kg ha P F 1 Soybean 1998 No lime 3 980 71 NS# NS 3 988 23 0.07 NS Fixed 3 975 3 954 Variable 3 887 3 896 Corn 1999 No lime 11 118 101 0.1 NS 11 122 58 0.02 NS Fixed 11 337 11 324 Variable 11 376 11 386 Soybean 2000 No lime 3 154 23 NS NS 3 148 5 NS NS Fixed 3 156 3 160 Variable 3 143 3 145 2 Soybean 1999 No lime 3 306 32 NS NS 3 301 12 NS NS Fixed 3 304 3 305 Variable 3 277 3 281 Corn 2000 No lime 9 184 27 NS NS 9 178 10 NS NS Fixed 9 138 9 143 Variable 9 181 9 181 † RCBD, randomized complete block design (observed means and statistics); RCBD-NNA, least square means and statistics from RCBD analysis combined with nearest-neighbor analysis (NNA). ‡ SE, average standard error of the least square means. § Lime, significance of the orthogonal comparison between the control and the mean of the two application methods. ¶ F-V, significance of the orthogonal comparison between the two application methods. # NS, not significant at P 0.1. depth. Several sampling points had acid soil in the 0- distances of about 50 m in most transects. In some sec- to 15-cm layer but had calcareous subsoil below that tions, soil pH varied about 2 pH units over a 12-m depth. It is likely that a potential detrimental impact of distance although sometimes changes were more grad- acid pH of surface soil on crop yield was offset by high- ual. There was a good agreement between the transect pH subsoil. Current Iowa State University lime recom- data and the cell data even with such a high small-scale mendations for corn and soybean (Voss et al., 1999) variability in Field 1, which suggests that for this portion consider a soil pH 6.0 (15-cm depth) sufficient for these of the field, the cell data accurately represented the pH crops when subsoils are calcareous although advise lim- of the small area sampled. However, there was more ing to pH 6.5 if lime is required. Our data suggest that discrepancy between the transect and cell data in two the critical pH level should be lower. transects of Field 2. This result may be explained by Another possible reason for a lack of response to VR high soil pH variability along multiple directions, which liming was very high small-scale variability of soil pH. coincides with results of previous research for P and K Figure 2 shows soil pH data for the intensive sampling (Mallarino, 1996). The high small-scale pH variation conducted along eight transects and the corresponding suggests that the pH class assignment for VR liming 0.1-ha cell data. Soil pH varied from 5.4 up to 8.0 over based on a 0.2-ha grid sampling may have not been Table 5. Soybean and corn yield by pH class and treatment for two fields. Field Year Crop pH class No lime Fixed Variable Statistics† 1 kg ha P F 1 1998 Soybean 5.7 4 513 4 477 4 603 NS‡ 5.7–6.29 4 228 4 287 4 196 NS 6.3–7.2 4 089 4 031 NS 7.2 3 497 3 537 NS 1999 Corn 5.7 11 713 11 716 11 583 NS 5.7–6.29 11 508 11 674 11 637 NS 6.3–7.2 11 556 11 587 NS 7.2 10 816 10 923 NS 2000 Soybean 5.7 3 192 3 226 3 163 NS 5.7–6.29 3 134 3 166 3 180 NS 6.3–7.2 3 226 3 230 NS 7.2 3 110 3 110 NS 2 1999 Soybean 5.7 3 690 3 539 3 609 NS 5.7–6.29 3 485 3 424 3 400 NS 6.3–7.2 3 327 3 389 NS 7.2 2 940 2 979 NS 2000 Corn 5.7 8 891 8 913 8 685 NS 5.7–6.29 9 009 8 833 8 949 0.01 6.3–7.2 9 196 8 934 NS 7.2 9 234 9 315 NS † Statistical significance of orthogonal comparisons between the control and the mean of two application methods for the two lower pH classes and between the fixed-rate method and the mean of the control and variable-rate method for the two higher pH classes (no lime was applied with the variable-rate method when pH was 6.2). ‡ NS, not significant at P 0.1. An orthogonal comparison between the two application methods never was significant at P 0.1. 1362 AGRONOMY JOURNAL, VOL. 94, NOVEMBER–DECEMBER 2002 entirely correct and could partly explain a lack of re- Webster soil association), even the very intensive grid sponse to the VR lime method in apparently most acidic soil sampling used may not represent soil pH variability areas. For example, pH data from a Field 2 transect well and may not produce a reasonable interpolated (VR treatment, Replication 2) suggests that lime is re- map. Furthermore, even if soil samples were collected quired, but the cell pH data from the same area suggests with the extremely high intensity used in the transects, that no lime is required (soil pH is 6.3). These observa- current VRT equipment used by cooperatives or distrib- tions suggest that in these soils (Clarion–Nicollet– utors cannot manage such a small-scale variation. Fig. 2. Intensive soil-sampling data from transects compared to the cell data for the fixed-rate and variable-rate application methods in Fields 1 and 2. BIANCHINI & MALLARINO: SOIL SAMPLING FOR pH AND VARIABLE-RATE LIMING 1363 The influence of factors other than pH on yield could range and SD within a field were smaller for the soil-map also explain small and inconsistent response to lime and and management-zone schemes than for more intensive especially the small negative response trend of soybean sampling schemes. This suggests that these schemes in 1998. There was a negative linear relationship (P were effective in separating areas with contrasting pH. 0.05) between soybean yield and soil pH of unlimed The smaller pH range for most grid-cell and zone-sam- areas in 1998 and 1999 (data not shown), which ex- pling schemes suggests, however, that large sampling plained 45% of the yield variability in 1998 (Field 1) units pool areas with large pH variation. The soil-map and 54% in 1999 (Field 2). Thus, an apparent negative scheme had the lowest pH range and was the least effec- effect of the FR liming on soybean yield for high-pH tive in separating areas with distinctly different pH in areas could be explained by low yield in high-pH areas. Field 1. The soil-map and management-zone schemes Correlations between corn yield and soil pH were nega- were less effective in Field 2. The size of field areas that tive in 1999 and explained 46% of yield variability but would be classified into four pH classes by each sampling were positive in 2000 and explained 36% of yield vari- strategy varied markedly. The two most acid pH classes ability. These relationships likely are explained by dif- were merged in one class because this pH range repre- ferences in soil moisture. The low-lying and high-pH sents the area with greatest potential for yield increase. soils of this soil association (such as the series Canisteo, In Field 1, the less intensive sampling schemes resulted Harps, Okoboji, and Webster) are prone to excessive in a smaller area that would be limed compared with moisture in seasons with above-average rainfall. The more intensive schemes. However, this was not always 50-yr average rainfall recorded in a weather station lo- the case in Field 2, probably because one large manage- cated 10 km from the fields (Perry, IA) for the March– ment zone (10.1 ha) with a mean pH of 6.03 significantly September period is 640 mm (U.S. Dep. of Commerce, increased the area that would be limed. The least inten- 1951–2000). The 1998 and 1999 rainfall for the same sive sampling schemes also resulted in smaller high-pH period was 816 and 827 mm, respectively, but was 408 mm areas, especially in Field 1. in 2000. In wet years, like 1998 and 1999, excessive mois- Table 7 shows within-zone mean soil pH and statistics ture may limit yield in the low areas, but in dry years used to compare the pH variability within zones for the (like 2000), the same areas may have an advantage com- soil-map, yield, elevation, EC, and integrated manage- pared with the rest of the field, especially with corn. ment-zone schemes. In Field 1, the soil-map and inte- Kaspar et al. (2000) worked on similar soils and found grated management-zone schemes had the largest pH a negative correlation between corn yield and elevation difference across zones, and the SD for the zones was when rainfall was less than normal during the growing intermediate compared with other schemes. The yield- season but a positive correlation when rainfall was zone scheme had the smallest pH range across zones, greater than normal. Moreover, Jaynes and Colvin (1997) found that the yield spatial pattern and structure and it was the only sampling scheme that would have vary over time for this soil association mainly due to resulted in no lime requirement (soil pH was 6.2 for changing rainfall patterns. all zones). Although SD was low in one yield zone (0.13 in Zone 5), much higher SD values for other yield zones suggest that this scheme was less effective than other Soil pH Assessment with Various schemes in reducing within-zone variability. The eleva- Soil-Sampling Schemes tion-zone scheme had the lowest range of SD and an The mean pH values for sampling units of various intermediate pH range across the units, which confirms sampling schemes ranged from 6.6 to 6.9 for Field 1 and other research results (Franzen et al., 1998; Luchiari et from 6.6 to 7.0 for Field 2 (Table 6). However, the pH al., 2000) suggesting that elevation may be a good source Table 6. Area for each sampling unit, number of sampling units, and descriptive statistics of soil pH for different soil-sampling schemes. Soil pH Total area by pH class Number Field Sampling scheme Area† of units‡ Mean Range SD§ 6.3 6.3–7.2 7.2 ha ha 1 Small grid point 0.1 144 6.81 2.70 0.94 6.9 1.4 6.1 Medium grid point 0.3 48 6.78 2.63 0.95 6.9 1.5 6.0 Medium grid cell 0.3 48 6.81 2.38 0.86 6.3 1.8 6.3 Large grid point 0.7 18 6.93 2.48 0.94 5.6 1.8 7.0 Large grid cell 0.7 18 6.81 2.08 0.75 4.1 4.8 5.5 Soil-map zones 0.5–4.6 7 6.64 1.75 0.73 5.4 5.1 3.9 Management zones 0.5–2.1 9 6.68 2.03 0.60 4.5 7.3 2.6 2 Small grid point 0.1 180 6.60 3.10 0.93 9.2 3.1 5.7 Medium grid point 0.3 60 6.62 2.70 0.95 9.9 2.1 6.0 Medium grid cell 0.3 60 6.60 2.65 0.87 9.3 2.7 6.0 Large grid point 0.7 30 6.68 2.70 0.95 10.2 1.1 6.7 Large grid cell 0.7 30 6.57 2.32 0.72 6.7 6.7 4.6 Soil-map zones 0.9–6.7 6 6.95 1.86 0.80 9.7 4.2 4.1 Management zones 1.0–10.1 6 6.91 1.82 0.81 11.9 0.0 6.1 †Size of each sampling unit. The two numbers indicate the smallest and largest sampling zones. ‡Number of sampling units for each soil-sampling scheme. §SD, standard deviation. 1364 AGRONOMY JOURNAL, VOL. 94, NOVEMBER–DECEMBER 2002 Table 7. Soil pH means and standard deviation (SD) for soil-map, yield, elevation, electrical-conductivity (EC), and management- zone schemes. Soil-map zones Yield zones Elevation zones EC zones Management zones Field Zone pH SD Zone pH SD Zone pH SD Zone pH SD Zone pH SD 1 1 6.42 0.92 1 7.08 0.94 1 7.20 0.75 1 6.12 0.65 1 7.01 0.95 2 6.98 0.89 2 7.39 0.93 2 7.63 0.67 2 6.05 0.53 2 6.24 0.51 3 5.85 0.15 3 6.39 0.88 3 6.05 0.47 3 7.40 0.74 3 7.17 0.85 4 7.53 0.68 4 6.72 0.87 4 6.16 0.69 4 7.62 0.56 4 6.48 0.65 5 6.14 0.64 5 7.85 0.13 5 6.73 0.98 6 5.97 0.52 6 6.83 1.14 6 5.80 0.23 7 7.60 0.63 7 6.73 0.72 7 6.68 0.94 8 6.22 0.75 9 7.83 0.32 F tests† 16.7, P 0.001 4.0, P 0.001 54.4, P 0.011 55.4, P 0.001 13.3, P 0.001 2 1 5.99 0.60 1 5.96 0.56 1 7.80 0.26 1 6.04 0.43 1 7.85 0.14 2 6.65 0.87 2 7.81 0.19 2 6.87 0.93 2 6.17 0.73 2 7.29 0.76 3 6.14 0.50 3 7.51 0.64 3 5.89 0.45 3 7.35 0.79 3 6.03 0.68 4 7.49 0.27 4 7.36 0.71 4 6.27 0.23 4 7.80 0.19 4 6.27 0.17 5 7.60 - 5 6.26 0.17 5 6.28 0.23 6 7.85 0.30 6 5.99 0.63 6 7.75 0.36 F tests 48.8, P 0.001 52.3, P 0.001 57.9, P 0.001 56.5, P 0.001 55.5, P 0.001 † F value and probability for comparisons of between-zone and within-zone pH variability for each sampling scheme. of information to delineate management zones. All EC than schemes based on larger cells or zones but are zones also had low SD values, and the pH range across more expensive (especially in large fields) due to high zones was intermediate compared with other schemes. sampling and analysis costs. The effectiveness of tradi- Areas with high EC were associated with areas of high tional sampling by soil-map unit (based on commonly pH, and correlation coefficients were 0.67 for Field 1 used soil survey maps) still used by many producers and 0.70 for Field 2. These results should be interpreted could be improved by using other layers of information carefully, however, because areas with high EC and high to develop management-zone schemes. A management- pH tended to be at low elevations, were moderately zone scheme likely will represent small-scale soil test poorly drained, had higher egg counts of soybean cyst variability less accurately than intensive grid-sampling nematode, and yielded less seasons with higher precipi- schemes when the number of sampling units is compara- tation. The F tests showed a significant difference (P tively lower. However, the management-zone concept 0.01) for all schemes, which suggests that all schemes is flexible to accommodate different information layers identified field areas with contrasting soil pH. However, as well as different fields, sampling objectives, and eco- the sizes of the F values confirm previous interpretations nomic conditions. based on pH and SD in suggesting that, in this field, Future developments of on-the-go automated soil- the elevation and EC schemes were more effective in testing systems should markedly decrease the cost of reducing within-zone pH variability and increasing pH soil sampling and improve the accuracy of soil nutrient differences between zones. In contrast to results for maps (Sudduth et al., 1997). Birrell et al. (1999) and Field 1, pH values, SD, and F tests suggested that all Adamchuk et al. (1999) have developed real-time soil zone schemes provided similar information about pH nutrient analysis sensors to determine soil pH that variability. showed a reasonably good relationship (R2 0.83) with Our results coincide with other research in showing manually collected soil samples. Although these early that elevation and EC maps can be useful tools to delin- automated soil-sampling systems provide analysis of soil eate management zones although correlations among acidity with lower accuracy than standard laboratory elevation, EC, and soil properties, as well as the basis methods, they should improve the quality of the soil for their use to delineate management zones, vary among maps because much higher spatial resolution of soil fields and regions (Jaynes et al., 1995; Hartsock et al., sampling can be achieved (Adamchuk et al., 1999). 2000). Our work and that of the previous authors also However, in fields with very high small-scale pH vari- showed that soil series are not always properly identified ability, these advances should be accompanied by ad- by soil survey maps and that EC maps are useful tools vances in VRT equipment effectiveness to apply lime to complement soil survey maps. Furthermore, results accurately and precisely over very short distances. show that various options are available when farmers need to take soil samples to decide lime application in these soils. Specific recommendations that apply to all CONCLUSIONS fields and conditions are not possible because the effec- Fixed-rate or VR liming had no meaningful effect on tiveness of different sampling schemes and layers of crop yield even though a soil-sampling method more information vary greatly across regions, with the main intensive than any method being used in production objective of the soil sampling, and with economic con- agriculture showed that 15% of the areas had pH 5.4 siderations. Intensive grid-sampling schemes being used to 5.7 and 35% had pH 5.7 to 6.2. The VR method in the Corn Belt (which are based on 0.7- to 1.0-ha cells) usually increased pH of acidic areas more than the FR are more effective in describing small-scale variability method and applied 56% less lime in one field and 61% BIANCHINI & MALLARINO: SOIL SAMPLING FOR pH AND VARIABLE-RATE LIMING 1365 less in the other. The lack of yield response could be Franzen, D.W., A.D. Halvorson, and V.L. Hofman. 2000. Manage- ment zones for soil N and P levels in the Northern Great Plains explained by predominant calcareous subsoil and by [CD-ROM]. In P.C. Robert et al. (ed.) Proc. Int. Conf. on Precision very high small-scale pH variability although this possi- Agric., 5th, Bloomington, MN. 16–19 July 2000. ASA, CSSA, and bility could not be confirmed with the methods used. SSSA, Madison, WI. Irregular patterns with a variation of about 2 pH units Franzen, D.W., and T.R. Peck. 1995. Field soil sampling density for within 10 to 20 m were common. Although this variabil- variable rate fertilization. J. Prod. Agric. 8:568–574. Hartsock, N.J., T.G. Mueller, G.W. Thomas, R.I. Barnhisel, K.L. ity should not have affected crop response to FR liming, Wells, and S.A. Shearer. 2000. Soil electrical conductivity variability it may have reduced the effectiveness of VR liming. [CD-ROM]. In P.C. Robert et al. (ed.) Proc. Int. Conf. on Precision Results suggest that no lime is needed in this soil associa- Agric., 5th, Bloomington, MN. 16–19 July 2000. ASA, CSSA, and tion when the topsoil has pH 5.4 and subsoils are SSSA, Madison, WI. calcareous although other soils could respond differ- Heiniger, R.W., and A.J. Meijer. 2000. Why variable rate application of lime has increased grower profits and acceptance of precision ently. agriculture in the southeast [CD-ROM]. In P.C. Robert et al. (ed.) Soil pH information provided by a 0.1-ha point-grid Proc. Int. Conf. on Precision Agric., 5th, Bloomington, MN. 16–19 sampling, which is more intensive and costly than grid- July 2000. ASA, CSSA, and SSSA, Madison, WI. sampling schemes used in the Corn Belt, may not pro- Hinz, P.N. 1987. Nearest neighbor analysis in practice. Iowa State J. Res. 62:199–217. vide more useful pH data than less intensive zone-sam- Hinz, P.N., and J.P. Lagus. 1991. Evaluation of four covariate types pling schemes. This is due to extreme variation at a used for adjustment of spatial variability. p. 118–126. In Proc. Appl. scale much smaller than the distance between grid Stat. in Agric. Conf., Kansas State Univ., Manhattan, KS. 28–30 points. Zone-sampling schemes may not provide better Apr. 1991. Dep. of Stat., Kansas State Univ. Manhattan information about soil pH variability than intensive Jaynes, D.B., and T.S. Colvin. 1997. Spatiotemporal variability of corn and soybean yield. Agron. J. 89:30–37. grid-sampling schemes, but they offer more flexibility Jaynes, D.B., T.S. Colvin, and J. Ambuel. 1995. Yield mapping by to reduce the number of samples depending on particu- electromagnetic induction. p. 383–394. In P.C. Robert (ed.) Proc. lar field conditions, soil-sampling objectives, and eco- Int. Conf. on Site-Specific Manage. for Agric. Syst., 2nd, Blooming- nomic conditions. Although no sampling scheme will ton, MN. 27–30 Mar. 1994. ASA, CSSA, and SSSA, Madison, WI. alleviate the limitations of current VRT equipment to Kaspar, T.C., T.S. Colvin, D.B. Jaynes, D.L. Karlen, D.E. James, D.W. Meek, D. Pullido, and H. Butler. 2000. Estimating corn yield manage high small-scale variability, the results showed using six years of yield data and terrain attributes [CD-ROM]. In that VR liming is a better alternative to FR liming in P.C. Robert et al. (ed.) Proc. Int. Conf. on Precision Agric., 5th, these soils because it provides a reasonable way of Bloomington, MN. 16–19 July 2000. ASA, CSSA, and SSSA, Madi- avoiding lime application to large high-pH areas. son, WI. Kitchen, N.R., K.A. Sudduth, and S.T. Drummond. 1999. Soil electri- cal conductivity as a crop productivity measure for claypan soils. REFERENCES J. Prod. Agric. 12:607–617. Adamchuk, V.I., M.T. Morgan, and D.R. Ess. 1999. An automated Lark, R.M., J.V. Stafford, and H.C. Bolam. 1997. Limitations on the sampling system for measuring soil pH. Trans. ASAE 42:885–891. spatial resolution of yield mapping for combinable crops. J. Agric. Andrews, W.F., and R.O. Dideriksen. 1981. Soil survey of Boone Eng. Res. 66:183–193. County, Iowa. USDA, Washington, DC (available online at http:// Luchiari, A., J. Shanahan, D. Francis, M. Schlemmer, J. Schepers, M. icss.agron.iastate.edu). (Verified 25 July 2002.) Liebig, A. Schepers, and S. Payton. 2000. Strategies for establishing Birrell, S.J., J.W. Hummel, R.G. Hoeft, and T.R. Peck. 1999. Develop- management zones for site specific nutrient management [CD- ment and field testing of real-time soil nutrient sensors for precision ROM]. In P.C. Robert et al. (ed.) Proc. Int. Conf. on Precision fertilizer application. p. 65–70. In Proc. Illinois Fert. Conf., Peoria, Agric., 5th, Bloomington, MN. 16–19 July 2000. ASA, CSSA, and IL. 25–27 Jan. 1999. Univ. of Illinois, Urbana. SSSA, Madison, WI. Bongiovanni, R., and J. Lowenberg-DeBoer. 2000. Economics of vari- Mallarino, A.P. 1996. Spatial variability of phosphorus and potassium able rate lime in Indiana. Precis. Agric. 2:55–70. in no-tilled soils for two sampling scales. Soil Sci. Soc. Am. J. 60: Borgelt, S.C., S.W. Searcy, B.A. Stout, and D.J. Mulla. 1994. Spatially 1473–1481. variable liming rates: A method for determination. Trans. ASAE Mallarino, A.P., M.U. Haq, D. Wittry, and M. Bermudez. 2001. Varia- 37:1499–1507. tion in soybean response to early-season foliar fertilization among Cahn, M.D., J.W. Hummel, and B.H. Brouer. 1994. Spatial analysis and within fields. Agron. J. 93:1220–1226. of soil fertility for site-specific crop management. Soil Sci. Soc. Mallarino, A.P., and D.J. Wittry. 1997. Use of DGPS, yield monitors, Am. J. 58:1240–1248. soil testing, and variable rate technology to improve phosphorus Cambardella, C.A., T.B. Mooreman, J.M. Novak, T.B. Parkin, D.L. and potassium management. p. 267–275. In Proc. Integrated Crop Karlen, R.F. Turco, and A.E. Konopka. 1994. Field-scale variability Manage. Conf., Ames, IA. 17–18 Nov. 1997. Iowa State Univ. of soil properties in central Iowa soils. Soil Sci. Soc. Am. J. 58: Ext., Ames. 1501–1511. Mallarino, A.P., D.J. Wittry, D. Dousa, and P.N. Hinz. 1998. Variable- Colvin, T.S., D.B. Jaynes, D.L. Karlen, D.A. Laird, and J.R. Ambuel. rate phosphorus fertilization: On-farm research methods and evalu- 1997. Yield variability within a central Iowa field. Trans. ASAE ation for corn and soybean. p. 687–696. In P.C. Robert et al. (ed.) 40:883–889. Proc. Int. Conf. on Precision Agric., 4th, St. Paul, MN. 19–22 July Doolittle, J.A., K.A. Sudduth, N.R. Kitchen, and S.J. Indorante. 1994. 1998. ASA, CSSA, and SSSA, Madison, WI. Estimating depths to claypans using electromagnetic induction McLean, E.O., and J.R. Brown. 1984. Crop response to lime in the methods. J. Soil Water Conserv. 49:572–575. midwestern United States. p. 267–303. In F. Adams (ed.) Soil acidity Dreimanis, A. 1962. Quantitative gasometric determination of calcite and liming. Agron. Monogr. 12. 2nd ed. ASA, CSSA, and SSSA, and dolomite by using Chittick Apparatus. J. Sediment. Petrol. 32: Madison, WI. 520–529. Mulla, D.J., A.C. Sekely, and M. Beatty. 2000. Evaluation of remote Fleming, K.L., D.G. Westfall, D.W. Wiens, and M.C. Brodahl. 2000. sensing and targeted soil sampling for variable rate application of Evaluating farmer defined management zone maps for variable lime [CD-ROM]. In P.C. Robert et al. (ed.) Proc. Int. Conf. on rate fertilizer application. Precis. Agric. 2:201–215. Precision Agric., 5th, Bloomington, MN. 16–19 July 2000. ASA, Franzen, D.W., L.J. Cihacek, V.L. Hofman, and L.J. Swenson. 1998. CSSA, and SSSA, Madison, WI. Topography-based sampling compared to grid sampling in the Myers, D.B., N.R. Kitchen, K.A. Sudduth, and R.J. Miles. 2000. Esti- Northern Great Plains. J. Prod. Agric. 11:364–370. mation of a soil productivity index on claypan soils using soil electri- 1366 AGRONOMY JOURNAL, VOL. 94, NOVEMBER–DECEMBER 2002 cal conductivity [CD-ROM]. In P.C. Robert et al. (ed.) Proc. Int. to regionalize fields into potential management units. p. 225–237. Conf. on Precision Agric., 5th, Bloomington, MN. 16–19 July 2000. In P.C. Robert et al. (ed.) Proc. Int. Conf. on Precision Agric., 4th, ASA, CSSA, and SSSA, Madison, WI. St. Paul, MN. 19–22 July 1998. ASA, CSSA, and SSSA, Madi- Oyarzabal, E.S., A.P. Mallarino, and P.N. Hinz. 1996. Using precision son, WI. framing technologies for improving applied on-farm research. p. Sudduth, K.A., J.W. Hummel, and S.J. Birrell. 1997. Sensors for site- 379–387. In P.C. Robert et al. (ed.) Proc. Int. Conf. on Precision specific management. p. 183–210. In F.J. Pierce and E.J. Sadler Agric., 3rd, Bloomington, MN. 23–26 June 1996. ASA, CSSA, and (ed.) The state of site-specific management for agriculture. ASA, SSSA, Madison, WI. CSSA, and SSSA, Madison, WI. Peck, T.R., and S.W. Melsted. 1973. Field sampling for soil testing. United States Department of Commerce. 1951–2000. Climatological p. 67–75. In L.M. Walsh and J.D. Beacon (ed.) Soil testing and data. Natl. Oceanic and Atmos. Administration, Natl. Clim. Data plant analysis. SSSA Book Ser. 3. SSSA, Madison, WI. Cent., Asheville, NC (summaries available online at http://mesonet. agron.iastate.edu/climodat). (Verified 25 July 2002.) Pierce, F.J., and D.D. Warncke. 2000. Soil and crop response to vari- Voss, R.D., J.E. Sawyer, A.P. Mallarino, and R. Killorn. 1999. General able-rate liming for two Michigan fields. Soil Sci. Soc. Am. J. 64: guide for crop nutrient recommendations in Iowa. Publ. PM 1688 774–780. (Revised). Iowa State Univ. Ext., Ames. SAS Institute. 1996. SAS/STAT user’s guide. Release 6.11. SAS Inst., Watson, M.E., and J.R. Brown. 1998. pH and lime requirement. p. Cary, NC. 13–16. In J.R. Brown (ed.) Recommended chemical soil test proce- Sawyer, J.E. 1994. Concepts of variable-rate technology considera- dures for the North Central Region. Recommended North Central tions for fertilizer application. J. Prod. Agric. 7:195–201. Regional Res. Publ. 221 (Revised). SB 1001. Missouri Agric. Exp. Schepers, J.S., M.R. Schlemmer, and R.B. Ferguson. 2000. Site-specific Stn., Columbia. considerations for managing phosphorus. J. Environ. Qual. 29:125– Wollenhaupt, N.C., R.P. Wolkowski, and M.K. Clayton. 1994. Map- 130. ping soil test phosphorus and potassium for variable-rate fertilizer Stafford, J.V., R.M. Lark, and H.C. Bolam. 1999. Using yield maps application. J. Prod. Agric. 7:441–448.
Pages to are hidden for
"Soil-Sampling Alternatives and Variable-Rate Liming for a Soybean"Please download to view full document