Soil variability and collecting a representative sample

Soil testing is the foundation of crop nutrient management, the reliability of which is based on the collection of a representative field sample, appropriate soil test method, accurate laboratory analysis, and the nutrient recommendation. Agricultural soils are inherently spatially heterogeneous as a consequence of soil parent material, landscape topography, and past agricultural management practices (e.g., tillage, crop history, application of manure, and/or soil amendments). Reduced-tillage systems result in vertical stratification of immobile nutrients (P, K, and Zn) near the soil surface. The collection of a representative soil sample is often a trade-off between the practicality of sampling versus the ability to obtain a certain level of soil test accuracy and precision of the field being sampled.

Soil-sampling strategies fall into three categories based on the scale of the management unit: whole-field composite, management zone, and systematic or spatial grid sampling. Whole-field sampling entails a composite of multiple soil cores denoting the field, whereas a management zone strategy is based on the collection of composite samples from a field management unit (often 5–20 ac) based on a common factor (e.g., topography, soil type, yield history, etc.). For whole-field and management zone composite samples, soil test results only represent the mean concentration with no estimate of the variance (Peterson & Calvin, 1968). Spatial grid sampling is based on systematically subdividing a field into cells, typically of 1.1-, 2.5-, 4.4-, or 10-ac with collection of a soil sample from each grid cell and the subsequent generation of a spatial distribution map of soil test results. The 2.5-ac grid size is most common, but 1.0-ac grids are often used to gain a more detailed spatial structure of fields with high soil test variability. Whole-field composites are used with uniform nutrient or soil amendment applications. Zone and grid strategies are utilized in conjunction with variable-rate applications of soil amendments or nutrients over across a field.

With the advent of soil testing in the early 20th century, Cooperative Extension recommended a whole-field composite based on 15 to 20 subsamples collected in a “zig-zag” pattern across the field. Cameron et al. (1971) reported that for whole fields, 20 cores provided a mean estimate within 10% for phosphorus 70% of the time but were inadequate on highly variable fields. Swenson et el. (1984) reported that 20 cores provided a mean estimate within 15% for nitrate 80% of the time. Subsequent work by Franzen and Peck (1993) showed substantial field spatial variability for soil test P and K and suggested sampling intensities of one sample per acre would be required to characterize spatial variation in most fields in Illinois. However, spatial dependencies of soil test P are inconsistent across fields (Lawrence et al., 2020). Those with high soil test P values tend to have higher spatial variability (Mallarino, 1996; Clodfelter et al., 2005) associated with fields with a past applications of livestock and poultry manures.

Often high variability of immobile nutrients are associated with fertilizer bands from previous nutrient applications. Kitchen et. al. (1990) suggested in situations where the location of the band is known, samplers should avoid the band area completely or collect one in the band for every 20 when the band spacing has been 30 inches. When the location of the band is not known, Shapiro (1988) and Hooker (1976) have suggested collecting 20 to more than 100 cores for a field composite sample. Most agronomists, however, find this impractical.

Further complicating the collection of a representative soil sample is variation in soil properties with depth. It’s well documented that soil texture, CaCO₃, and soil organic matter (SOM) vary by depth. But increasingly, stratification of soil test P, K, and Zn near the soil surface has been noted with the transition from conventional tillage to reduced-tillage cropping systems across the U.S. the past 30 years. Reduced-tillage systems result in decreased incorporation of crop residue and reduced soil mixing, resulting in both surface stratification and greater spatial heterogeneity of immobile nutrients (Wolkowski, 2004). Overall, the collection of a representative soil sample for nutrient management depends on the nutrient management strategy, the tillage practice, and the costs associated with sampling and laboratory analysis.

Study Design

Over the past 15 years, a study was conducted to assess whole-field and grid-point soil sample test uncertainty. The objective was to evaluated whole-field and grid-point P, K, and pH variability. Whole-field soil test data, based on a 2.5-ac grid, was collected from 47 field sites across eight states representing four tillage systems: conventional (CT), minimum till (MT), ridge till (RT), and no-till (NT). A second study entailed the collection of 12 individual soil cores from 42 fields across the western and central U.S. The cores were 2 by 9 inches at a single grid point each within a 10-ft radius of the center point based on a randomized structural design. Each soil core was dried, pulverized, thoroughly mixed, and analyzed for pH, SMP buffer pH, Bray P1 phosphorus, ammonium acetate K, SOM, and DTPA-extractable zinc, in triplicate. Laboratory quality control procedures included standard reference soils, blanks and duplicates. At 11 field sites, at a single grid point, soils were sampled incrementally by depth to 12 inches to assess nutrient stratification.

Results

Whole-field variability across six states indicate soil test P relative standard deviations (RSD) ranged from 36–69% for soils with mean concentrations 17 to 53 ppm (Table 1). Precision was consistently lower on NT fields than MT fields, likely associated with decreased soil incorporation of previous crop residues and/or fertilizer applications. Ammonium acetate extractable K RSD values ranged from 12 to 34% for soils with mean concentrations of 135–280 ppm K. Precision of pH ranged from 3.3 to 14.1% for soils ranging from 4.7 to 6.6 with the highest RSD noted for a NT field comprised of both acid and alkaline grid-point soil samples.

Grid-point multi-core composite mean soil test P concentrations decreased going from 2 to 12 cores with an associated improvement in precision across all tillage systems (Table 2). With only two soil cores composited, the RSD for the NT field was 51%, whereas the MT field was 38% and CT was 13%. Increasing soil core composites from 2 to 12 for the NT site improved precision from 51 to 44%. Improvements were noted for MT and CT fields but to a lesser extent. Results for soil K showed slightly higher levels of precision than soil test P but similar improvement with increases in the number of cores composited. The number of soil cores composited had no appreciable effect on soil pH precision.

Across 31 NT field sites, grid-point soil P precision based on a grid point 12-core composite ranged from 18–121% with a mean of 38%. The lowest grid-point precision was noted for a NT site with a mean soil test P concentration of 11 ppm but with individual cores ranging from 5–137 ppm, resulting in an RSD of 121%, heterogeneity likely associated with past band applications of P fertilizers. Soil test P precision of MT ranged 13–34% for nine fields with an average of 24% while that of CT ranged from 6–22% for seven fields with an average of 14%. Similar trends across tillage systems were noted for soil K with NT systems having the poorest precision.

Soil P and K stratification by depth was assessed at 10 sites across the western and central U.S. Results for three specific sites show soil test P concentrations were highly stratified in the top 0 to 1 inches for the NT and MT sites and roughly constant for depths below 6 inches, whereas CT site concentrations were constant (Table 3). Concentrations of soil test K decreased markedly from the 0- to 1-inch to the 2- to 3-inch depth and continued to slowly decline at the deepest depth across all NT and MT tillage systems. Soil test K concentrations had less pronounced stratification and were near constant for the 3- to 12-inch soil depths whereas soil test P continued to decrease.

Summary

Whole-field soil test P precision (RSD) across eight states averaged 42%, K 31%, and pH 3.1% across 47 fields evaluated. High variability of soil test P and K was associated with spatial variation in field topography, soil texture, and organic matter content. No-till sites had the highest field soil test P variation, followed by MT and CT sites. Based on a soil test P RSD of 42%, a randomly collected field composite sample of 20 cores would result in soil test P uncertainty of ± 18% of the mean 90% of the time, whereas a composite 40 cores would result in P uncertainty of ± 11% of the mean, 90% of the time. A 20-core composite of a NT field with low soil test P and prior band fertilizer P applications would have much greater uncertainty.

Grid-point soil sample precision indicates substantial improvement in precision, increasing from 2 to 12 soil core composites for soil test P and to a lesser extent for K. For NT sites, increasing the number of core composites substantially improved grid-point soil test P precision, especially for fields where fertilizer phosphorus had been previously banded. It was also noteworthy that NT grid-point precision, based on a 12-core composite, consistently exceeded 40% across sites when the mean soil test P concentrations were less than 15 ppm. Although optimum precision was obtained with 12 core composites per grid point, generally for CT sites, an eight-core grid-point composite with a soil test P RSD of 11–17% would result in a general soil test P uncertainty of ± 6 to 9% of the mean, 90% of the time.

Soil near-surface stratification of nutrients impact soil sample collection. Consistently, soil test P exhibited greater near-surface stratification than soil K, and high nutrient stratification contributed to poor grid-point precision. Care must be taken when sampling reduced-tillage systems to assure accurate depth control as shallow sampling may overestimate soil test P, whereas deep sampling may result in a bias low nutrient concentration.

These results underwrite the need to consider the field history, tillage, and the soil-sampling management approach (whole field, zone, or grid) when developing a soil-sampling strategy for nutrient management. In addition, they show how soil-sampling intensity impacts precision and soil test uncertainty. For reduced-till cropping systems and low-testing soils, consideration should be given to decreasing the spatial grid cell size and increasing the number of soil core composites to provide a more detailed map of spatial variability and improve both the soil test accuracy for effective nutrient management. Lastly, effective improvement of nutrient management should include a routine assessment of stratification of soil test P and K.