Spatial Variation of Plant-Available Phosphorus in ... - PubAg - USDA

Spatial Variation of Plant-Available Phosphorus in Pastures with Contrasting Management Thomas J. Sauer* and David W. Meek ABSTRACT

history of poultry litter application (Sharpley et al., 1993; Kingery et al., 1994). Elevated soil P concentrations result in increased P concentration in runoff from littertreated soils (Sharpley, 1995; Pote et al., 1996; Sauer et al., 2000a). Threshold soil P levels or P indices are now being implemented as part of comprehensive nutrient management plans for animal feeding operations. Although other soil properties are important to pasture management (e.g., organic matter, other nutrients, and pH), water quality concerns have made plant-available P concentration a key factor in determining whether and how much poultry litter may be applied in the future. Phosphorus distribution in soils receiving poultry litter can be affected by variation in P concentration in the litter and the uniformity of litter application. Total P content in poultry litter varies from 10 to 20 g kg⫺1 and is affected by the number of flocks between cleanouts, P content and form in diet, and type of bedding material (Stephenson et al., 1990; Barnett, 1994). Ndegwa et al. (1991) measured P distribution within different litter particle-size classes and found the distribution of P to be uniform among the size fractions. Wilhoit et al. (1993) found uneven litter distribution after application with a typical horizontal spinner, centrifugal-type spreader as most smaller sized litter particles fell within 3.7 m of the spreader. The coefficient of uniformity for litter distribution was only 50% at the manufacturerrecommended swath width of 12.2 m but improved to 11% at a swath width of 8.5 m. Phosphorus content in the litter-size fractions was not evaluated in this study, however, it is clear that uniform distribution of poultry litter would be necessary to assure uniform P distribution on the soil surface. Redistribution of nutrients by grazing animals adds another level of complexity to nutrient management in pastures receiving poultry litter. Nutrients tend to accumulate in soil where animals lounge, especially near shade, water sources, and mineral supplement feeders (West et al., 1989; Mathews et al., 1994; Schomberg et al., 2000). From 60 to 99% of the nutrients ingested by grazing animals are returned to the soil in feces and urine (Barrow, 1987; Haynes and Williams, 1993). Nutrients excreted primarily in feces, including P, are likely to be concentrated in shallow soil layers beneath the dung patch and at varying distances up to five times the area of the dung patch (Peterson et al., 1956; MacDiarmid and Watkin, 1972; During and Weeda, 1973). Grazing management impacts soil nutrient distributions as stock densities and duration of grazing periods affect distribution of excreta. Rotational grazing systems, typified by high stock densities over short grazing intervals,

Land application of animal manure, at rates based on soil nutrient content or crop requirements, optimizes nutrient recycling and minimizes offsite environmental impacts. The objective of this research was to characterize the spatial variation of plant-available P and other soil properties (C, N, and pH) in two pastures having contrasting grazing and poultry litter management. One site (Cellar Ridge) was a lightly grazed 6-ha tall fescue (Festuca arundinacea Schreb.) pasture with limited poultry litter application and the other (Haxton) was a 9.5-ha tall fescue pasture with annual poultry litter application and intensive rotational grazing for 10 yr. Soil cores (0–0.15 m) were collected on a 30-m grid at both sites and analyzed for plant-available P (Mehlich-3 extract), total C and N (combustion method), and pH (1:1 water/0.01 M CaCl2). Cellar Ridge had significantly less Mehlich-3 extractable P (32 vs. 341 mg kg⫺1), more acid pH (5.25 vs. 5.73), and significantly greater C (23.3 vs. 16.3 g C kg⫺1) and N (1.76 vs. 1.54 g N kg⫺1). Spatial dependence over approximately 1 to 3 lag distances with a consistent orientation (across ridge) was observed for all parameters at Cellar Ridge. No spatial dependence was observed for Mehlich-3 P, C, N, or pH at the Haxton site (all parameters exhibiting nugget effect). Ten years of poultry litter application likely eliminated spatial structure for these properties. Further research is needed to determine whether additional costs associated with grid sampling and variable rate litter application can be justified.

R

esearch on the spatial variation of soil properties has enabled the development of land management strategies that incorporate site-specific soil information. Variable-rate fertilizer application based on measured patterns of plant-available nutrients has become a widely employed practice aimed at increasing crop yield and nutrient-use efficiency (Wollenhaupt et al., 1994; Stein et al., 1997; Timlin et al., 1998). Costs of implementing variable rate technology, including those associated with soil sampling and analysis, can be significant and are justified only by increased profitability or improved environmental stewardship (Sawyer, 1994; Schepers et al., 2000; Schmidt et al., 2002). The Ozark Highlands of the USA have large areas of intensive and expanding poultry production. Poultry litter (manure with bedding material) is applied to permanent pastures of the region and often serves as the sole nutrient source for forage growth. The litter is routinely applied at a rate to meet forage N requirements. As most plants require several times more N than other macronutrients, excessive P is applied. Phosphorus accumulation in soils is common in areas with an extended

T.J. Sauer, USDA-ARS, Poultry Production and Product Safety Research Unit, Univ. of Arkansas, Fayetteville, AR 72701; currently at: USDA-ARS, National Soil Tilth Lab., 2150 Pammel Drive, Ames, IA 50011-4420; and David W. Meek, USDA-ARS, National Soil Tilth Lab., 2150 Pammel Drive, Ames, IA 50011-4420. Received 26 July 2002. *Corresponding author ([email protected]). Published in Soil Sci. Soc. Am. J. 67:826–836 (2003).

Abbreviations: STP, soil test P.

826

SAUER & MEEK: SPATIAL VARIATION OF PLANT-AVAILABLE PHOSPHORUS

have been found to enhance the uniformity of excreta return (Peterson and Gerrish, 1996). Geostatistical methods are now routinely employed to describe spatial variation of soil properties. One purpose of these techniques is to predict values of a variable at nonsampled locations using ordinary kriging. Related statistical issues include understanding variability structure and quantification of the uncertainty in these estimates. The methods, of course, are based on some underlying assumptions such as stationarity. Analyses that are modified to better meet these assumptions may result in better predictions and understanding of the matters at hand. For example, Cressie’s (1993) analysis of coal ash data revealed a large-scale trend present in the east-west direction. When stationarity was assumed in this example, the resulting east-west empirical directional semivariogram showed spatial dependence. Analyzing residuals from the set with the trend removed, however, resulted in an empirical directional semivariogram that showed no east-west spatial dependence (i.e., a pure nugget model). Thus, spatial models can consist of a large-scale trend component, one or more intermediate scale components, and a small-scale component, which is generally the component to be modeled with a semivariogram. In practice, often a two-component model (trend ⫹ spatial error) will suffice. Spatial prediction for such modeling approaches can be done with median polish kriging and universal kriging. Variable rate poultry litter application is being considered as a practice to improve P management in intensive poultry production areas of the mid-south and southeastern USA. To enable improved nutrient management in grazing systems of these regions, better information is needed on the spatial variation of soil properties, especially plant-available P. The objective of this research was to characterize the spatial variation of plantavailable P, soil C, N, and pH within two grazed paddocks in northwestern Arkansas having contrasting management intensities. MATERIALS AND METHODS Field Sites Two field sites with contrasting management intensities were selected for grid soil sampling (Fig. 1). The low intensity site was a 6-ha paddock located on the University of Arkansas’ Department of Animal Science research substation in western Washington County, Arkansas (36⬚ 7⬘ N lat., 94⬚ 21⬘W long.). This site, referred to as Cellar Ridge, had a history of lowintensity, set stocking grazing management of endophyte-infected tall fescue with essentially no external nutrient additions. A set stocking grazing system was used on Cellar Ridge as cattle were allowed access to this paddock and several other paddocks on the substation for the entire grazing season (March to November). The paddock has an irregular shape and is located across a north-south ridge that slopes gradually to the south and more steeply to the east and west. Slopes range from 1 to 15% and the paddock is bounded by hardwood forest on all sides. Two soils are mapped on Cellar Ridge, Nixa (loamy-skeletal, siliceous, active, mesic Glossic Fragiudults) and Clarksville cherty silt loams (loamy-skeletal, siliceous, semiactive, mesic Typic Paleudults) (Harper et al., 1969). Additional details on properties of soils found at this

827

Fig. 1. Soil maps of Cellar Ridge (low intensity, Nixa and Clarksville soils) and Haxton (high intensity, Captina and Tonti soils) field sites. Dashed lines indicate boundaries of paddocks sampled on 30-m grids. Square on Haxton map indicates location of the 30 by 30-m cell around grid point (5,3) that was sampled at the 3-m spacing.

site can be found in Sauer et al. (1998, 2000a). The Cellar Ridge site, with it’s small area, forested margins, and relatively steep slopes, typifies a paddock that might begin receiving poultry litter if a new poultry production facility were to be constructed in the vicinity. The intensively managed site, referred to as the Haxton site, was located in eastern Benton County, Arkansas (36⬚ 17⬘ 30″ N lat., 94⬚ 14⬘ W long.). This 9.5-ha paddock was under intensive forage production (endophyte-infected tall fescue) and rotational grazing management. Poultry litter has been applied annually in the fall since 1989, before 1989 this site also had very limited outside nutrient additions. The paddock is essentially rectangular in shape, fenced on all sides, and has slopes ranging from 0 to 3%. Rotational grazing management at the Haxton site involved grazing a large herd of cattle on the paddock for several days to consume the available forage. The cattle would then be rotated to another paddock and returned to the sampled paddock several times during the growing season at intervals dictated by forage growth patterns. A cutting of hay was also harvested from the paddock in the spring each year. Over 90% of the paddock area is mapped as Captina silt loam (fine-silty, siliceous, active, mesic Typic Fragiudults) with the remaining area on the east boundary mapped as Tonti cherty silt loam (fine-loamy, mixed, active,

828

SOIL SCI. SOC. AM. J., VOL. 67, MAY–JUNE 2003

mesic Typic Fragiudults) (Phillips and Harper, 1977). The Haxton site, with its flat slopes and relatively deep and lessstony soils typifies a productive improved pasture site of the Ozark Highlands.

Sampling and Analytical Procedures Uniform grid soil sampling at a 30-m spacing was completed at the Cellar Ridge and Haxton sites on 19 Jan. and 8 Feb. 1999. The 30-m grid spacing was selected to provide sufficient samples for the application of spatial statistical analyses on a paddock scale. One 0- to 0.15-m core was collected at each grid point using a 0.025-m diam. split-tube soil corer. At the Haxton site, a second grid (10 by 10) on a 3-m spacing was sampled around one of the 30-m grid points chosen at random (Row 5, Column 3 with grid reference point 0,0 in the southwest corner of the paddock). Soil cores were taken on this grid using the same procedures as for the 30-m grid. The 3-m grid sampling was included to determine the spatial variation of soil properties within one grid cell of the 30-m grid. Total number of sample points was 66, 111, and 100 for the Cellar Ridge, Haxton 30-m, and Haxton 3-m grids. The Haxton paddock was also divided into four quadrants within which eight cores were taken in a random pattern and composited for analysis to simulate a traditional soil sampling pattern. All soil samples were air dried and sieved to remove coarse fragments. Plant-available P was estimated from measurement of soil test P (STP) determined using the Mehlich-3 extraction method (Mehlich, 1984), which is the standard procedure in Arkansas. Concentrations of P in the Mehlich-3 extracts were quantified colorimetrically on a Skalar SanPlus1 analyzer (Skalar Inc., Norcross, GA). Total C and N were determined by grinding subsamples to a fine powder on a roller mill (Bailey Mfg., Inc., Norwalk, IA) followed by analysis on a Fison NA 1500 Elemental Analyzer (ThermoQuest Corp., Austin, TX). Soil pH was determined on 1:1 (w/v) pastes of soil with deionized water and 0.01 M CaCl2.

Data Analysis Classical Statistics Classical descriptive statistics were used to determine measures of central tendency and dispersion for the Cellar Ridge, Haxton 30-m, Haxton 3-m, and Haxton random composite data sets and to assess overall differences among data sets. Two-sample t tests at P ⫽ 0.05 were used to identify statistically significant differences between means of the Cellar Ridge vs. Haxton 30-m and Haxton 30-m vs. Haxton 3-m data sets. Statistical significance of differences between the Haxton 30-m vs. Haxton random composites and Haxton 3-m vs. grid point 5,3 were evaluated using the 95% confidence intervals for the grid data in each comparison (Steel and Torrie, 1980). Spatial Statistics Geostatistical techniques were employed to delineate the spatial structure of soil properties within each of the Cellar Ridge and Haxton data sets. Large-scale trends in many of the measured variables were detected via multiple graphical and analytical tests. Simple first- and second-order response surfaces in the coordinates generally did not adequately model any of the trends. Thus, median polish residuals were estimated in these cases (Cressie, 1986). Otherwise, the raw data 1 Names of proprietary products are necessary to report factually on available data; however, the USDA neither guarantees nor warrants the standard of the products, and the use of the product by USDA implies no approval to the exclusion of others that may also be suitable.

were used to develop the empirical semivariogram. Both isotropic and directional regular and robust semivariograms (Hawkins and Cressie, 1984) were developed and examined for each variable in each location; denoted ␥(h) and ␥r(h). For all variogram estimation, lag tolerance was restricted to less than half the shortest distance between sample locations. Two sets of directional variograms were estimated for each variable. The first set was along azimuth angles of 0, 45, 90, and 135⬚ with an azimuth angle tolerance of 22.5⬚ and a maximum bandwidth of 45⬚. The second set was along azimuth angles of 0, 22.5, 45, 67.5, 90, and 112.5⬚ with an azimuth angle tolerance of 11.25⬚ and a maximum bandwidth of 22.5⬚. Semivariogram models were considered when the apparent sill/␥(1) ⬎ 1.5 or when ␥(hmax)/␥(1) ⬎ 1.5 (i.e., from a model fit, the sill to nugget variance ratio exceeded 1.5). Following the practical rule of Journel and Huijbregts (1978), only points in the spatial series for distances up to the half-length of the domain were considered. When modeled, at least two theoretical forms were considered for each selected empirical semivariogram. To avoid the bias of ordinary least squares estimates and the complexities of generalized least squares estimates, the variogram models were estimated using Cressie’s (1993, p. 96–7) pragmatic compromise weight that offers a balance between the efficiency of general least squares and the simplicity of ordinary least squares. For an example of this procedure, see Gotway (1991). Considered models with the associated scales of fluctuation (␪), are reported in Meek (2002). Hence, the scale of fluctuation was estimated only when significant spatial dependence was present, the semivariogram model was bounded, and the nugget model was inappropriate. The scale of fluctuation is reported for two reasons: (i) asymptotic models do not have a range but can have an arbitrary practical range stipulated while ␪ is consistently defined for all bounded models (Meek, 2001) and, (ii) independent averaging intervals or areas can be readily defined from ␪ values (Vanmarcke, 1983). Maps of variables were created with ordinary kriging using weighted values for a minimum of nine nearest-neighbor data points in the interpolations. Analyses were completed using the SAS (SAS, 1996) and S⫹ Spatial Stats (Kaluzny et al., 1998)

RESULTS Classical Statistical Analyses Standard summary statistics for the Cellar Ridge and Haxton 30-m data sets indicate statistically significant differences between the two sites for each parameter (Tables 1 and 2). Values from sampling sites near the north and south borders at the Haxton site were consistently outliers when compared with those from the interior grid. Proximity to fences and the influence of windblown dust from the crushed limestone on the unpaved road (P adsorption by deposited CaCO3) may have influenced soil P content in these areas. Removal of values for the north and south grid rows produced a data set for the spatial analyses, now with 96 sample locations, that met stationarity standards and had a semivariogram that was reasonably modeled as a pure nugget. To assure an unbiased comparison of data sets by both classical and spatial statistics, all values reported for 30-m grid at the Haxton site exclude the extreme north and south grid rows. Cellar Ridge had significantly less Mehlich-3 P and more acid pH (in both H2O and CaCl2) than at the Haxton site. These findings for the Haxton site are simi-

829


Table 1. Comparison of soil property values for different sites and sampling patterns. No. of samples

Data set

Mehlich-3 P

Total C

kg⫺1

mg g i) Haxton 30-m means vs. Cellar Ridge 30-m means: Haxton 30-m 96 341a† 16.3b Cellar Ridge 66 32b 23.3a ii) Haxton 30-m means vs. Haxton 3-m means: 30-m 96 341a 16.3a 3-m 100 307b 15.4b iii) Haxton 30-m means vs. quadrant (random composite) means: 30-m 96 341 16.3a Random composites 4 327 14.9b iv) Haxton 30-m grid point at center of 3-m grid (5,3) vs. Haxton 3-m means: Grid point 5,3 1 208b 13.2b 3-m 100 307a 15.4a

Total N

C/N

pH in H20

pH in CaCl2

1.54b 1.76a

10.6b 13.5a

5.73a 5.25b

5.29a 4.68b

1.54a 1.46b

10.6 10.6

5.73a 5.39b

5.29a 5.02b

1.54a 1.40b

10.6 10.6

5.73a 5.41b

5.29a 5.10b

1.30b 1.46a

10.2b 10.6a

5.50a 5.39b

5.12a 5.02b

kg⫺1

† For i and ii, values in a column followed by different letters are significantly different at P ⫽ 0.05 as determined by two-sample t tests. Significant differences for iii and iv, also indicated by different letters, are based on 95% confidence intervals for the 30-m and 3-m data, respectively.

lar to those observed for other sites having an extended history of poultry litter application (Sharpley et al., 1993; Kingery et al., 1994). Annual poultry litter application provides a significant P input and, with pH normally ⬎ 6.0, raises the pH of acid soils commonly found in the Ozarks. Sharpley et al. (1993) and Kingery et al. (1994) also found significant increases in C and N in pastures with long-term poultry litter application. The Haxton site, however, had significantly less C, N, and a lower C/N ratio than Cellar Ridge. Variation about the means, as expressed by the coefficient of variation (CV ⫽ 100 ␦/␮), was on average 13% greater at Cellar Ridge (range 0.4% greater for H2O pH to 46% greater for Mehlich-3 P). There were significant differences between the Haxton 3- and 30-m sampling grids for C, N, H2O pH, and CaCl2 pH with means for the 30-m grid significantly greater. Coefficients of variation for the 30-m grid were also greater for each parameter by an average of 5%

(range 1% greater for north to 13% greater for Mehlich3 P). Mehlich-3 P for the Haxton 30-m grid was not significantly different from the random composite samples. For the other parameters, the Haxton 30-m grid values were greater than the four random composite samples for all parameters except C/N ratio. However, the differences between means were small, with the largest difference for total C with the random composite 10% less than the 30-m mean. Comparison of values for parameters from the 3-m grid and the single point of the 30-m grid at its center (5,3) allows assessment of the uncertainty in soil properties within the 30-m cell represented by that single data point. Mehlich-3 P at grid point (5,3) was significantly less than the mean from the 100, 3-m grid samples. Soil C, N, and C/N ratio were also significantly less for grid point 5,3 but pH was significantly greater than the means from the 3-m grid. Absolute differences between values

Table 2. Summary statistics for the three raw and detrended data sets. Variable Name

Raw data Units

␮†

Low intensity (Cellar Ridge) 30-m grid (n ⫽ 66) Mehlich-3 P mg kg⫺1 32 23.3 Total C g kg⫺1 1 ⫺ 1.76 Total N g kg C/N 13.5 5.25 pH in H2O 4.68 pH in CaCl2 High intensity (Haxton) 30-m grid (n ⫽ 96) 341 Mehlich-3 P mg kg⫺1 16.3 Total C g kg⫺1 1.54 Total N g kg⫺1 C/N 10.6 5.73 pH in H2O 5.29 pH in CaCl2 High intensity (Haxton) 3-m grid (n ⫽ 100) 307 Mehlich-3 P mg kg⫺1 15.4 Total C g kg⫺1 1.46 Total N g kg⫺1 C/N 10.6 5.39 pH in H2O 5.02 pH in CaCl2

Detrended data

Final model

Md

␦

Range

␦mp

␪

Direction

24 22.7 1.70 13.2 5.24 4.68

27 6.7 0.56 1.27 0.33 0.38

5–117 12.0–56.6 0.70–4.1 11.6–17.3 4.48–6.22 3.91–5.74

24 5.6 0.44 1.05 0.28 0.32

44 57 106 ∞‡ 52 57

E-W E-W E-W E-W E-W E-W

332 16.0 1.50 10.6 5.76 5.30

97 3.0 0.26 0.66 0.27 0.26

184–656 7.4–28.5 0.80–2.5 8.6–12.1 4.87–6.48 4.51–5.84

88 2.8 0.24 0.61 0.25 0.26

0§ 0 0 0 0 0

NA¶ NA NA NA NA NA

302 15.1 1.40 10.6 5.38 5.04

71 2.3 0.24 0.54 0.19 0.19

160–524 10.5–21.9 1.0–2.2 9.5–11.73 4.95–5.77 4.56–5.36

62 2.0 0.22 0.48 0.16 0.15

0 0 0 0 0 0

NA NA NA NA NA NA

† ␮ ⫽ mean; Md ⫽ median; ␦ ⫽ standard deviation; Range ⫽ minimum and maximum; ␦mp ⫽ standard deviation for median polished residual set; ␪ ⫽ scale of fluctuation (distance in m). ‡ ∞ ⫽ unbounded. § 0 ⫽ nugget effect. ¶ NA ⫽ not applicable.

830


from grid point 5,3 and the 3-m means varied from ⬍2% for CaCl2 pH to ⬎30% for Mehlich-3 P.

Spatial Analyses Significant spatial correlations were detected for all parameters at Cellar Ridge although the scale for C/N ratio was unbounded (Table 2). The orientation of the spatial dependence for all parameters with bounded structure was east-west and occurred over approximately 1 to 3 lag distances. Areas of elevated Mehlich3 P occurred at two locations, one on Nixa soil near the center of the paddock and one on Clarksville soil on the east margin of the paddock (Fig. 2a). High concentrations of soil P in these areas could not be attributed to any site factors other than the location on Nixa soil

was immediately adjacent to a field road often traversed by cattle. Spatial distribution of all other parameters were similar to Mehlich-3 P in that there was an area of elevated levels located on Nixa soil near the center of the paddock (data not shown). Less C and N and more acid pH were consistently observed along the paddock margins, especially on the NW and SE. In contrast to Cellar Ridge, no parameter exhibited any spatial dependence for either sampling scale at the Haxton site as a nugget effect was detected in each semivariogram (Table 2). The nugget effect represents unexplained or random variance, which is often attributed to measurement error or variability at a scale smaller than the sampling scale (Trangmar et al., 1985). Large-scale trend in Mehlich-3 P was present along the

Fig. 2. (a) Contour map of Mehlich-3 extractable soil P (STP) at the Cellar Ridge (low intensity) site and (b) contour map of the relative standard error (RSE) for the kriged estimates based on a STP of the sill ⫽ 27.58 mg kg⫺1.


north-south direction of the 30-m grid with an approximate plateau-shaped profile. Mehlich-3 P was found at greater concentrations over a large area near the center of the Haxton paddock (Fig. 3). There was a small area of P accumulation (⬎450 mg kg⫺1) near shade trees adjacent to the west border of the paddock and along the east border, but no evidence of P accumulation near the water supply that was also located on the east border. Soil pH was approximately 0.8 units greater near the unpaved road on the south paddock boundary as compared to the N boundary (data not shown). Again, this observation is attributed to CaCO3 blown onto the paddock from the unpaved road. Soil C, N, and C/N ratio varied over relatively narrow ranges with no distinct patterns related to any paddock features. A northwest-southeast oriented ridge of high Meh-

Fig. 2. Continued.

831

lich-3 P was located near the center of the 3-m grid with another smaller area of high concentration along the east margin (Fig. 4). The range of soil P values within this 900 m2 area was 364 mg kg⫺1 or 77% of the range observed for the entire paddock. Soil C and N also had areas of greater concentrations near the center of the grid. Soil pH showed a general gradient from less acid in the west to more acid in the east with no indication of influence from windblown dust from the unpaved road (data not shown).

DISCUSSION Spatial variation of soil properties can be the product of soil forming factors (parent material, climate, biological activity, topography, and time) acting over a contin-

832


Fig. 3. Contour map of Mehlich-3 extractable P (STP) for the 30-m sampling grid at the Haxton (high intensity) site. Grid point 5,3, about which the 3-m sampling grid was centered, was located at easting 90 and northing 150.

uum of space and time or of imposed soil management practices, such as tillage, cropping system, and fertilizer or manure application (Trangmar et al., 1985). Morphological factors were expected to have a greater influence at Cellar Ridge because of greater topography, less intensive grazing, and lack of nutrient inputs. The Nixa and Clarksville soils of Cellar Ridge have similar properties in their surface layers, the primary difference between the soils being that the Nixa soil has a cherty fragipan at 0.4 to 0.8 m (Harper et al., 1969). Topography is likely the key soil-forming factor affecting spatial variation of soil properties at Cellar Ridge. The relatively long, steep slopes are the product of spatially dependent weathering and erosional processes. Runoff routing continues to affect soil properties via soil and nutrient transport from higher to lower elevations. All parameters exhibited spatial structure at Cellar Ridge and did so across the ridge (east-west orientation), indi-

cating that position on the slope had a significant impact on the spatial distribution of surface soil properties. The influence of morphological factors at the Haxton site was expected to be secondary to management factors as the basic soil forming factors are assumed to be acting uniformly across the paddock. The majority of the paddock is mapped as Captina soil (fine-silty, siliceous, active, mesic Typic Fragiudult), which is very similar to the Tonti soil that has slightly greater slopes and increased coarse fragment content (Phillips and Harper, 1977). In the years immediately before the change in management intensity (1989), the Haxton site also had very little outside nutrient additions. Because of the recent intensification of litter application and rotational grazing at this site, management factors were expected to have a greater influence on the observed spatial patterns of surface soil properties. Poultry litter was applied annually at the Haxton site


833

Fig. 4. Contour map of Mehlich-3 extractable P (STP) for the 3-m sampling grid at the Haxton (high intensity) site. Grid point 5,3 of the 30-m sampling grid was located in the center of the map.

using the same application techniques. A spreader truck drove once around the border of the paddock and then in parallel north-south passes until the paddock area was completely covered. Given the known difficulty in distributing poultry litter uniformly (Wilhoit et al., 1993), multiple years with the same direction of spreader travel might be expected to lead to some linear patterns of soil properties, especially for relatively immobile nutrients like P. Instead, lower CV’s and a lack of spatial structure for all soil properties were observed at the Haxton site. Small variation in the spreader paths from year-to-year and redistribution of nutrients via cattle urine and feces may have obscured any spatial structure induced by the path of spreader travel in any one application. No data are available on the spatial variation of STP at the Haxton site before intensification of its management. However, as soil texture would not change with poultry litter application or grazing management, parti-

cle-size analyses were completed on Haxton samples to determine whether some spatial structure in soil properties existed before management intensification. As particle-size analysis was not included in the original experimental design, sufficient sample for analysis by the hydrometer method (Gee and Bauder, 1986) was available for only 70 and 72 of the 3- and 30-m grid samples. Measured sand, silt, and clay contents were very consistent with means ⫾ SD of 13.1 ⫾ 1.3, 69.1 ⫾ 1.8, and 17.8 ⫾ 1.9% for the 3-m grid and 13.4 ⫾ 1.7, 68.5 ⫾ 2.2, and 18.1 ⫾ 2.3% for the 30-m grid. All particle-size class data for the 3-m grid can reasonably be considered to be stationary although a slight trend mainly along the north-south direction was observed. Directional variograms reveal some spatial dependence along the northing, which can be modeled. All the variogram models suggest a minimum of 12- to 15-m averaging length along the northing (less than four lag increments). For the 30-m grid, the clay and sand data met sta-

834


tionarity criteria, but the silt data showed a trend primarily in the northing direction with a low along the south and leveling over the N half of the paddock. Median polish residuals from the silt and raw data from the clay fraction were considered to be spatially independent (i.e., modeled with the nugget effect). The sand fraction, however, did show spatial dependence with a 34-m scale of fluctuation. While some directional differences were apparent for this fraction, for practical purposes an omnidirectional hole-effect model worked extremely well. The results of the spatial analyses of particle-size data for the 3- and 30-m grids indicate that some spatial structure exists for particle-size classes at these scales in the Haxton paddock. Failure to detect similar spatial patterns for any of the soil chemical properties including STP suggests that 10 yr of poultry litter application and rotational grazing may be responsible for the absence of spatial structure for these properties. Greater soil C at Cellar Ridge was unexpected given the consistently greater concentrations of C observed in pastures receiving poultry litter as reported by Sharpley et al. (1993) and Kingery et al. (1994). It is possible that the observed C concentration differences between the Haxton and Cellar Ridge paddocks do not represent effects from recent management practices but rather the legacy of earlier (i.e., decades previous) management or degradation. Another possibility is that the addition of poultry litter had the so-called “priming effect” whereby addition of excess N in applied fertilizer or manure leads to accelerated soil organic matter turnover (Kuzyakov et al., 2000). Further research including measurement of total microbial biomass in soil of the paddocks and C and N concentrations in undisturbed areas adjacent to the paddocks would be necessary to fully explain the differences in soil C and N observed in this study. Franzluebbers et al. (1999 and 2000) found significant increases in soil C and N with proximity to sources of water and shade and degree of fungal endophyte infection in tall fescue pastures in Georgia. Franzluebbers et al. (1999) suggest that presence of the toxic alkaloids produced by the endophyte reduces soil microbial activity, slowing decomposition of plant residues, and allowing more organic C and N to accumulate in the soil. The degree of endophyte infection was not quantified at either site in this study, so it is not known whether reduced decomposition rates because of varying levels of endophyte infection, as suggested by Franzluebbers et al. (1999), may have also contributed to differences in soil C content. Data from the random composite and 30-m sampling grid at the Haxton site provide insightful perspectives on the relative utility of different sampling scales/strategies. The random composite scheme (paddock divided in quarters, 8 cores collected randomly and composited within each quarter) represents a traditional approach to paddock soil sampling. Fewer cores are taken, exact sampling locations are not identified or recorded, and sample analytical costs are much less. Although some significant differences were observed between the two data sets, Mehlich-3 P concentration was not significantly different and mean values of most parameters

were not substantially different (from an agronomic perspective) than for the 30-m grid. To further test the utility of the grid sampling, the mean Mehlich-3 P concentration from five sets of 20 randomly selected samples from each grid were compared with the mean for the entire grid. In no instance were the means between random subsets significantly different from the mean of the complete grid (P ⬎ 0.05). Regardless of scale (3- or 30-m) or site (Haxton or Cellar Ridge), complete coverage with grid sampling did not produce a mean Mehlich-3 P concentration that was significantly different from a mean of 20 randomly selected subsamples from any grid. Daniels et al. (2001) reported similar results for measured spatial variation of STP in 12 pastures in western Arkansas and eastern Oklahoma. For six paddocks ranging in size from 1.1 to 4.2 ha, the probability of obtaining an STP estimate within the 95% confidence interval of the measured mean was ⬎0.5 if more than 15 subsamples were taken. Results from Daniels et al. (2001) and this study both indicate that, for small paddocks in the Ozark Highlands, analysis of ⬎15 samples per paddock is likely to provide a representative estimate of mean STP. When comparing means from the 3-m grid with values from the single 30-m grid core at its center, two of the six parameters were significantly greater than and four significantly less than the values for the single core. Mehlich-3 P for the single 30-m grid core was significantly less than the mean of the 3-m grid. Thus, for this cell of the 30-m grid within the Haxton paddock, whether one or 100 samples were collected and analyzed significantly affected the estimate of each parameter but in an inconsistent manner. One technique for addressing variation over small areas has been compositing several cores taken near each grid point. Data from the 3-m grid here shows that such a strategy would likely not have been effective in achieving an accurate estimate of STP for the entire 30-m grid cell. However, findings from the 3-m sample grid represent only one of 96, 30-m grid cells. It is unknown whether these observations are typical of the majority of 30-m grid cells within the paddock. Raun et al. (1998) and Solie et al. (1999) found significant spatial variation of Mehlich-3 P at even smaller spatial scales (⬍1 m) under bermudagrass (Cynodon dactylon L.) pasture in Oklahoma. While knowledge of spatial variation of STP at such small scales may be important for forage production, it may of little value with regard to variable rate technology since field-scale fertilizer or poultry litter application cannot be accomplished at such small scales with commercial application equipment. Thus, given the large variation in STP over such small distances in Ozark pastures, the design of soil sampling strategies should include consideration of the minimum scale at which nutrient application rates can be efficiently adjusted. If variable rate litter application were to be considered for Cellar Ridge, uncertainty in kriged estimates between data points (Fig. 2b) and the ability to apply poultry litter uniformly at varying rates would be important considerations. Given the direction (east-west) and scale (approximately 44 m) of STP spatial variation,


there is a high degree of uncertainty in kriged estimates between columns of data points. As commercial litter spreaders prefer to drive their vehicles in parallel passes along the long axis of paddocks, the direction of spreader travel would likely be north-south on Cellar Ridge. Assuming an optimum spreader pass-width of 8.5 m (Wilhoit et al., 1993), and 30 m between sampling points, approximately three of every five spreader passes would be within areas with relatively large uncertainty in STP estimates. The 30-m sampling interval used in this study is smaller than is commonly used by consultants or producers (Schepers et al., 2000). In addition, most commercial applicators use interpolation schemes much simpler than kriging (e.g., nearest neighbor or inverse distance weighting) that do not provide an estimate of uncertainty for the interpolated values. Thus, the relatively intensive sampling and additional analyses required to obtain estimates of uncertainty displayed here likely represent an approach that is costprohibitive under current economic conditions. Current emphasis of animal manure management is directed toward balancing the amount of applied P with the amount of P removed by the crop. Comprehensive nutrient management planning will consider STP concentrations and crop yield in determining manure application rates. The optimum Mehlich-3 P concentration for forage production in Arkansas is 50 to 100 mg kg⫺1 in the surface 0.15 m (Arkansas Cooperative Extension Service, 1998). After 10 yr of intensive management, the Haxton paddock had an average Mehlich-3 P concentration of 315 mg kg⫺1. All sampling points of the 30-m grid exceeded the threshold value of 150 mg kg⫺1 under consideration in Arkansas (Daniels et al., 2001). The recommended annual poultry litter application for pastures in the region is 4.5 to 6.7 Mg ha⫺1, which may supply from 45 to 130 kg of P ha⫺1. Maximum P removed in forage consumed by grazing animals is estimated at ⬍8 kg ha⫺1 (Sauer et al., 2000b). The rapid and substantial increases in STP in pasture soils receiving poultry litter can be attributed to the gross disparity between P inputs and removal.

SUMMARY AND CONCLUSIONS Spatial structure and orientation with slope direction was detected for all measured soil properties at the low management intensity Cellar Ridge site. Mehlich-3 P concentrations at Cellar Ridge were less than at the Haxton site and reflect the extended history of no litter application and low intensity grazing. By contrast, annual litter application and rotational grazing at the high intensity management Haxton site for 10 yr resulted in no spatial structure for any measured parameter. Concentrations of Mehlich-3 P at the Haxton site were much greater than the proposed threshold level, indicating that poultry litter inputs of P are significantly greater than P removal by grazing animals at this site. Failure to scale pasture P inputs to P removal in grazed or hayed forage quickly leads to plant-available P concentrations far in excess of agronomic requirements. It is unclear whether pastures with low manage-

835

ment intensity like Cellar Ridge, which displayed consistent spatial structure in measured soil properties including Mehlich-3 P, could benefit from variable rate litter application. Costs associated with implementing variable rate application would need to be justified and it is yet unproven whether site-specific management practices would efficiently eliminate areas deficient in plant-available P and/or prevent excessive accumulation of P in the surface soil layer. Further research is needed to determine whether variable rate litter application is a viable management practice with benefits for pasture production (forage yield and grazing animal performance) and/or environmental protection (reduced P runoff). ACKNOWLEDGMENTS The authors express appreciation to P. Doi, J. Martin, S. Compston, J. Murdoch, J. Many, A. Morrow, D. DenHaan, and B. Drake for assistance in completion of the field sampling and laboratory analyses.

REFERENCES Arkansas Cooperative Extension Service. 1998. Soil test recommendations guide. AGR 9. Arkansas Cooperative Ext. Service, Little Rock, AR. Barnett, G.M. 1994. Phosphorus forms in animal manure. Bioresour. Technol. 49:139–147. Barrow, N.J. 1987. Return of nutrients by animals. p. 181–186. In R. W. Snaydon (ed.) Managed grasslands. Elsevier, Oxford, England. Cressie, N.A.C. 1986. Kriging nonstationary data. J. Am. Stat. Assoc. 81:625–634. Cressie, N.A.C. 1993. Statistics for spatial data. John Wiley, New York. Daniels, M.D., P. Delaune, P.A. Moore, Jr., A. Mouromoustakos, S.L. Chapman, and J.M. Langston. 2001. Soil phosphorus variability in pastures: Implications for sampling and environmental management strategies. J. Environ. Qual. 30:2157–2165. During, C., and W.C. Weeda. 1973. Some effects of cattle dung on soil properties, pasture production, and nutrient uptake: I. Dung as a source of phosphorus. N.Z.J. Agric. Res. 16:423–430. Franzluebbers, A.J., N. Nazih, J.A. Stuedemann, J.J. Fuhrmann, H.H. Schomberg, and P.G. Hartel. 1999. Soil carbon and nitrogen pools under low- and high-endophyte-infected tall fescue. Soil Sci. Soc. Am. J. 63:1687–1694. Franzluebbers, A.J., J.A. Stuedemann, and H.H. Schomberg. 2000. Spatial distribution of soil carbon and nitrogen pools under grazed tall fescue. Soil Sci. Soc. Am. J. 64:635–639. Gee, G.W., and J.W. Bauder. 1986. Particle-size analysis. p. 383–411 In A. Klute (ed.) Methods of soil analysis. Part 1. 2nd ed. ASA, Madison, WI. Gotway, C.A. 1991. Fitting semivariogram models by weighted least squares. Comput. Geosci. 17:171–172. Harper, M.D., W.W. Phillips, and G.J. Haley. 1969. Soil Survey of Washington County, Arkansas. USDA, Soil Conservation Service, U.S. Gov. Print. Office, Washington, DC. Hawkins, D.M., and N.A.C. Cressie. 1984. Robust kriging—A proposal. J. Intl. Assoc. Math. Geol. 16:3–18. Haynes, R.J., and P.H. Williams. 1993. Nutrient cycling and soil fertility in the grazed pasture ecosystem. Adv. Agron. 49:119–199. Journel, A.G., and C.J. Huijbregts. 1978. Mining geostatistics. Academic Press, New York. Kaluzny, S.P., S.C. Vega, T.P. Cardoso, and A.A. Shelly. 1998. S⫹ spatial stats威: User’s manual for Windows威 and Unix威. SpringerVerlag, New York. Kingery, W.L., C.W. Wood, D.P. Delaney, J.C. Williams, and G.L. Mullins. 1994. Impact of long-term land application of broiler litter on environmentally related soil properties. J. Environ. Qual. 23: 139–147. Kuzyakov, Y., J.K. Friedel, and K. Stahr. 2000. Review of mechanisms

836


and quantification of priming effects. Soil Biol. Biochem. 32:1485– 1498. MacDiarmid, B.N., and B.R. Watkin. 1972. The cattle dung patch: 2. Effect of a dung patch on the chemical status of the soil, and ammonia nitrogen losses from the patch. J. Br. Grassl. Soc. 27: 43–48. Mathews, B.W., L.E. Sollenberger, V.D. Nair, and C.R. Staples. 1994. Impact of grazing management on soil nitrogen, phosphorus, potassium, and sulfur distribution. J. Environ. Qual. 23:1006–1013. Meek, D.W. 2001. A semiparametric method for estimating the scale of fluctuation. Comput. Geosci. 27:1245–1251. Meek, D.W. 2002. Another look at Clark’s adit silver series. p. 356–368 In. G. Milliken (ed.), Proc. 13th Appl. Stat. Agric. Conf., Stat. Dep., Kansas St. Univ., Manhattan, KS. Mehlich, A. 1984. Mehlich 3 soil test extractant: A modification of Mehlich 2 extractant. Commun. Soil Sci. Plant Anal. 15:1409–1416. Ndegwa, P.M., S.A. Thompson, and W.C. Merka. 1991. Fractionation of poultry litter for enhanced utilization. Trans. ASAE. 34:992–997. Peterson, P.R., and J.R. Gerrish. 1996. Grazing systems and spatial distribution of nutrients in pastures: Livestock management considerations. p. 203–212. In R.E. Joost and C. A. Roberts (ed.) Nutrient cycling in forage systems. Potash and Phosphate Institute and Foundation for Agronomic Research, Manhattan, KS. Peterson, R.G., W.W. Woodhouse, Jr., and H.L. Lucas. 1956. The distribution of excreta by freely grazing cattle and its effect on pasture fertility: II. Effect of returned excreta on the residual concentrations of some fertilizer elements. Agron. J. 48:444–449. Phillips, W.W., and M.D. Harper. 1977. Soil Survey of Benton County, Arkansas. USDA, Soil Conservation Service, U.S. Gov. Print. Office, Washington, DC. Pote, D.H., T.C. Daniel, A.N. Sharpley, P.A. Moore, Jr., D.R. Edwards, and D.J. Nichols. 1996. Relating extractable soil phosphorus to phosphorus losses in runoff. Soil Sci. Soc. Am. J. 60:855–859. Raun, W.R., J.B. Solie, G.V. Johnson, M.L. Stone, R.W. Whitney, H.L. Lees, H. Sembiring, and S.B. Philips. 1998. Microvariability in soil test, plant nutrient, and yield parameters in bermudagrass. Soil Sci. Soc. Am. J. 62:683–690. SAS Institute Inc. 1996. SAS/STAT technical report: Spatial prediction using the SAS system. SAS Inst. Inc., Cary, NC. Sauer, T.J., P.A. Moore, Jr., K.P. Coffey, and E.M. Rutledge. 1998. Characterizing the surface properties of soils at varying landscape positions in the Ozark Highlands. Soil Sci. 163:907–915. Sauer, T.J., T.C. Daniel, D.J. Nichols, C.P. West, P.A. Moore, Jr., and G.L. Wheeler. 2000a. Runoff water quality from poultry littertreated pasture and forest sites. J. Environ. Qual. 29:515–521. Sauer, T.J., K. VanDevender, C.V. Maxwell, M.B. Daniels, A.N. Sharpley, S.E. Watkins, J.L. Hatfield, C.P. West, and K.P. Coffey.

2000b. Phosphorus balance model to predict land base requirements for sustainable broiler feeding operations. p. 401–407 In J.A. Moore (ed.) Proc. 8th Int. Symp. Animal, Agricultural and Food Processing Wastes. Am. Soc. Agric. Eng., St. Joseph, MI. Sawyer, J.E. 1994. Concepts of variable rate technology with considerations for fertilizer application. J. Prod. Agric. 7:195–201. Schepers, J.S., M.R. Schlemmer, and R.B. Ferguson. 2000. Site-specific considerations for managing phosphorus. J. Environ. Qual. 29:125– 130. Schmidt, J.P., R.K. Taylor, and G.A. Milliken. 2002. Evaluating the potential for site-specific phosphorus applications without highdensity soil sampling. Soil Sci. Soc. Am. J. 66:276–283. Schomberg, H.H., J.A. Stuedemann, A.J. Franzluebbers, and S.R. Wilkinson. 2000. Spatial distribution of extractable phosphorus, potassium, and magnesium as influenced by fertilizer and tall fescue endophyte status. Agron. J. 92:981–986. Sharpley, A.N. 1995. Dependence of runoff phosphorus on extractable soil phosphorus. J. Environ. Qual. 24:920–926. Sharpley, A.N., S.J. Smith, and W.R. Bain. 1993. Nitrogen and phosphorus fate from long-term poultry litter applications to Oklahoma soils. Soil Sci. Soc. Am. J. 57:1131–1137. Solie, J.B., W.R. Raun, and M.L. Stone. 1999. Submeter spatial variability of selected soil and bermudagrass production variables. Soil Sci. Soc. Am. J. 63:1724–1733. Steel, R.G.D., and J.H. Torrie. 1980. Principles and Procedures of Statistics. 2nd ed. McGraw-Hill Book Co., New York. Stein, A., J. Brouwer, and J. Bouma. 1997. Methods for comparing spatial variability patterns of millet yield and soil data. Soil Sci. Soc. Am. J. 61:861–870. Stephenson, A.H., T.A. McCaskey, and B.G. Ruffin. 1990. A survey of broiler litter composition and potential value as a nutrient source. Biol. Wastes 34:1–9. Timlin, D.J., Y. Pachepsky, V.A. Snyder, and R.B. Bryant. 1998. Spatial and temporal variability of corn grain yield on a hillslope. Soil Sci. Soc. Am. J. 62:764–773. Trangmar, B.B., R.S. Yost, and G. Uehara. 1985. Application of geostatistics to spatial studies of soil properties. Adv. Agron. 38:45–94. Vanmarcke, E. 1983. Random fields: analysis and synthesis. MIT Press, Cambridge, MA. West, C.P., A.P. Mallarino, W.F. Wedin, and D.B. Marx. 1989. Spatial variation of soil chemical properties in grazed pastures. Soil Sci. Soc. Am. J. 53:784–789. Wilhoit, J.H., C.W. Wood, K.H. Yoo, and M.Y. Minkara. 1993. Evaluation of spreader distribution patterns for poultry litter. Appl. Eng. Agric. 9:359–363. Wollenhaupt, N.C., R.P. Wolkowski, and M.K. Clayton. 1994. Mapping soil test phosphorus and potassium for variable-rate fertilizer application. J. Prod. Agric. 7:441–448.