Influence of Sample Size on Chemical and Physical Soil Measurements

Understanding the distribution of soil properties with respect to sample size is essential for designing efficient soil sampling strategies. This study was conducted to evaluate the influence of the size of soil samples on the distribution of estimated soil physical (bulk density and water content), chemical (pH and orthophosphate P), and biological (NO3-N and denitrification rates) properties. The experiment was conducted at Beltsville, MD, on a Beltsville silt loam soil (fine-loamy, mixed, mesic Typic Fragiudult). Soil samples were collected from the surface layer (0-16 cm) using five different sized soil coring tubes, 1.7 to 5.4 cm in diameter, and one 20 by 30 cm block. Thirty-six samples of each sample size were collected twice before and twice after planting corn (Zea mays L.). Frequency distributions of denitrification rates, NOa-N, and orthophosphate P values were approximately lognormal, while the values for bulk density, water content, and pH were approximately Gaussian. All soil parameters except bulk density exhibited spatially dependent results. The effect of sample size varied with the soil property. The most consistent size effect was that smaller diameter samples gave smaller means, greater skewness, and higher variances compared with the large-block sample statistics. S are commonly sampled with the purpose of measuring biological, chemical, or physical properties of the soil. Soil properties naturally vary in space and time due to soil formation processes, climate, crop growth, and tillage practices. Variation may also be due to under sampling, sample handling, and analyses. Many questions related to soil sampling have been studied and have been summarily presented (Cline, 1944; Peterson and Calvin, 1986; Sabbe and Marx, 1987). The size (area and volume) of the soil sample may be especially important for soil parameters that can undergo rapid transformations, and are dependent on local conditions (denitrification, nitrification, and plant growth). Classical statistics indicates that sample variance should decrease as sample support size increases. Where sample results are spatially interdependent, Zhang et al. (1990) have shown that sample variance and sample support size can be related to the variogram function. Yet the relationship of the physical size of the sample to various statistical parameters (mean, variance, and frequency distribution) and how these may vary with tune or under different cultural practices is not adequately understood. Although there have been several studies on the number of observations required to estimate the mean and variance of soil parameters (Bole and Pittman, 1976; Cameron et al., 1971; Keogh and Maples, 1967), only a few have related various measured soil parameters to sample size. Hassan et al. (1983) studied Cl leaching by collecting 18 soil cores of 2.1and 7.9-cm-diam. each. They observed higher recoveries, and greater variation, J.L. Starr and J.J. Meisinger, USDA-ARS, Natural Resources Inst., Environmental Chemistry Lab., Beltsville, MD 20705; and T.B. Parkin, USDA-ARS, National Soil Tilth Lab., 2150 Pammel Dr., Ames, LA 50011. Received 10 Feb. 94. ""Corresponding author ([email protected]). Published in Soil Sci. Soc. Am. J. 59:713-719 (1995). among data sets from the smaller diameter samples. In a similar study, Rice and Bowman (1988) found spatial variability to have a greater effect than sample size. Baker et al. (1989) collected 35 samples for each of four different sample sizes (1.9-20.3-cm-diam. cores) and observed a small sample size effect with samples from conventional-till sites but not from no-till sites. They concluded that a 5.1-cm-diam. soil core was best for predicting residual soil NC-3-N. Parkin et al. (1987) collected 36 samples for each of six sample sizes. They found that sample diameters >4.15 cm provided the most reliable estimates of denitrification rates and estimated that 10 to 15 kg of soil was necessary to obtain a representative soil mass for estimating natural denitrification rates. In the same experiment, Starr et al. (1992) reported that the samples of diameters <2.15 cm often did not allow accurate estimation of the spatial variation of soil NO3. The purpose of this study was to evaluate the influence of sample size on the sample mean and variance of: bulk density (pt,), water content, pH, denitrification rate, and NO3-N and orthophosphate P concentrations. MATERIALS AND METHODS Field Site and Sampling Samples were collected from a Beltsville silt loam soil site located on the University of Maryland Plant Research Farm, Beltsville, MD. The soil at this site had no visual indications of heterogeneity, such as cracks, color variations, worm holes, buried crop residues, etc. Previous history at the site was 12 yr of continuous corn. Tillage treatments were no-tillage culture (NT) except fall disking, and conventional plow tillage (PT). Fertilizer N was spread by hand on both tillage treatments in April at the rate of 168 kg N ha~' of NHtNOs. The PT area was plowed at the end of May, and both tillage treatments were planted to corn in 76-cm rows in early June. Two tillage treatments and four sampling times were included to provide a range of field conditions for the same soil series and location. Any surface residues, especially in NT, were brushed aside before soil sampling and were not included in the sample. Four sampling times were studied (Table 1) to increase the opportunity for biological factors (mineralization, nitrification, and plant growth) to influence soil parameters. Therefore, the tillage treatments and times of sample collection were used to broaden the scope of this research. Soil samples were collected from 36 experimental blocks (20 by 30 cm) at each of four dates (designated as experiments), using close(Exp. 1 and 2) or wide-block (Exp. 3 and 4) spacing patterns (Table 1). In the first two sampling times (before planting corn), the 36 blocks were placed direcdy adjacent to each other in a six by six grid for a total sampling area of 2.16 m as shown in Fig. 1. In the two summer experiments, the 36 blocks were placed between corn rows on 76-cm centers in a six by six grid for Exp. 3 and a three by 12 grid for Exp. 4. This sampling design allowed random sampling within the blocks, as described below, and the capacity to assess spatial dependency. Abbreviations: NT, no-tillage; PT, conventional plow tillage; BW, BoxWhisker; CV, coefficient of variation.