| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a USDA-ARS Wheat, Sorghum, and Forage Research Unit, 344 Keim Hall, Univ. Nebraska, Lincoln, NE 68583 USA
b USDA-ARS, National Soil Tilth Lab, 2150 Pammel Dr., Ames, IA 50011 USA
c USDA-ARS, 215 Johnson Hall, Washington State Univ., Pullman, WA 99164-6421 USA
d Dep. Soil, Water, and Climate, Univ. of Minnesota, 1991 Buford Circle, St. Paul, MN 55108 USA
e USDA-ARS, Bushland, TX 79012 USA
jbrejda{at}unlserve.unl.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and methodsNOTESResults and discussionREFERENCES |
|---|
Abbreviations: MBC, microbial biomass carbon • MEP, Mehlich extractable phosphorus • MLRA, Major Land Resource Areas • MWD, mean-weight diameter • NRCS, Natural Resources Conservation Service • NRI, National Resource Inventory • PMN, potentially mineralizable nitrogen • PSU, primary sampling units • TOC, total organic carbon • WSA, water stable aggregates
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and methodsNOTESResults and discussionREFERENCES |
|---|
However, soils can be degraded by mechanisms other than soil erosion. Degradation can result from loss of organic C, compaction, salinization, acidification, alkalinization, excessive nutrient enrichment or depletion, chemical or heavy metal contamination, or reduced diversity and activity of soil organisms. Thus, an assessment of soil condition must go beyond measuring soil erosion and consider soil properties that are affected by other forms of soil degradation.
Currently, there is interest in collecting soil samples at NRI sample points to improve NRCS' ability to monitor changes in soil and natural resource condition. The soil samples could be analyzed for appropriate physical, chemical, or biological properties that serve as indicators of resource condition. However, the diversity of soils across the USA could hinder the success of a national scale assessment to monitor changes in soil properties using the NRI. Such an assessment may be more feasible at a regional scale if each region contains similar soil and land-use patterns so that precise and accurate estimates of soil properties can be made. A regional-scale assessment may be further improved if sampling was confined to key benchmark soils within a region.
Major Land Resource Areas are geographic units of several thousand hectares in extent that contain similar patterns of soils, climate, water resources, and land uses (USDA-SCS, 1981). They are important in agricultural planning at the state, regional, and national levels (USDA-SCS, 1981). Thus, the MLRA is an appropriate regional scale unit for monitoring soil condition and developing conservation policy and practices to address soil degradation problems.
The statistical analysis of data from a regional-scale assessment requires assumptions concerning the probability distribution and variability of the population to develop appropriate estimators of population parameters (Upchurch and Edmonds, 1991). Most parametric univariate and multivariate statistical analyses require the data to be normally distributed and have equal variances (Sokal and Rohlf, 1981; Johnson and Wichern, 1992; Sharma, 1996). Violation of these assumptions can reduce the power of the statistical tests and may lead to incorrect conclusions (Sharma, 1996). Further, the quality of inferences made using these methods depends on how closely the true parent population resembles a normal distribution (Johnson and Wichern, 1992).
At plot and field scales, the probability distributions exhibited by many soil properties are often non-normally distributed (Edmonds and Lentner, 1987; Edmonds et al., 1988; Parkin et al., 1988; Parkin and Robinson, 1992, 1994, Starr et al., 1992; 1995). In Iowa, Cambardella et al. (1994) reported that in one 36-ha field, 25 of 27 soil properties were non-normally distributed, and in another 96-ha field, 4 of 14 soil properties were non-normally distributed. In Missouri, 11 of 12 soil properties from 12 map unit transects were non-normally distributed (Young et al., 1998) and 55 of 60 soil properties in a 40-ha field pasture were non-normally distributed (Young et al., 1999). However, the distribution and variability of soil properties is scale dependent (Beckett and Webster, 1971; Starr et al., 1992; Parkin, 1993; Seyfried and Wilcox, 1995). As scale increases, smaller scale sources of variability such as landform, landscape position, and land-use may become subsumed into regional scale variability (Beckett and Webster, 1971; Hillel, 1991; Seyfried and Wilcox, 1995). Thus, the probability distribution of soil properties observed at plot and field scales may not apply at the regional scale. Information on the probability distribution and variability of soil properties at the regional scale could improve NRCS' ability to assess changes in soil properties using the NRI, following changes in land management and soil conservation practices.
The objective of this study was to evaluate the hypothesis that the probability distribution of 17 physical, chemical, and biological soil properties are normally distributed at the regional scale. If this hypothesis is rejected, we evaluated the hypothesis that these soil properties are log-normally distributed. The log-normal distribution was chosen as an alternative to the normal distribution because several studies at the plot and field scales indicate that many soil properties are log-normally distributed (Parkin et al., 1988; Parkin and Robinson, 1994). Because these results may be used in designing future large-scale monitoring programs of soil resources, we compared the magnitude of change that may be detected in a future sampling assuming either a normal or log-normal distribution.
|
Materials and methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and methodsNOTESResults and discussionREFERENCES |
|---|
|
Major Land Resource Area 67 comprises the Central High Plains and covers 74 410 km2 in eastern Colorado, southeastern Wyoming, and western Nebraska (Fig. 1). Elevation ranges from 1100 to 1800 m, increasing gradually from east to west. Average annual precipitation ranges from 325 to 425 mm with maximum precipitation in late spring and early autumn. Average annual temperature ranges from 7 to 10°C. About 25% of the land is farmed to wheat and other small grains, and 60% is native rangeland of mid- and short-grass species. The remainder of the area is irrigated and used to grow corn, alfalfa (Medicago sativa L.), sugar beets (Beta vulgaris L.), and vegetables (USDA-SCS, 1981).
Major Land Resource Area 77 comprises the Southern High Plains and covers 126 780 km2 in eastern New Mexico, northwestern Texas, and the panhandle of Oklahoma (Fig. 1). Elevation ranges from 800 to 2000 m, increasing gradually from southeast to northwest. Average annual precipitation ranges from 375 to 550 mm, but fluctuates widely from year to year. Average annual temperature ranges from 13 to 17°C. About 33% of the land is dryland farmed to winter wheat, grain sorghum [Sorghum bicolor (L.) Moench], and cotton (Gossypium hirsutum L.). About 40% is native rangeland. The remainder of the area is irrigated and used to grow corn, soybean, alfalfa, and vegetables (USDA, SCS, 1981).
Sampling Design
A subset of established NRI sample points was sampled within each MLRA. The design of the NRI is a stratified two-stage area sample (Nusser and Goebel, 1997; Nusser et al., 1998). The 36 sections (1 section = 259 ha or 1 mi2) within each township are placed into three groups of 12 sections each. The groups are called strata and each stratum is 3.22 by 9.66 km (2 by 6 mi) in size. The first stratum is composed of Sections 1 through 12, the second stratum of Sections 13 through 24, and the third stratum of Sections 25 through 36. The purpose of stratification is to ensure that the sample points are well distributed over each county and MLRA.
In the first stage of sampling, two primary sampling units (PSU) were randomly selected within each stratum. Each PSU represents a 64.8-ha (quarter-section or 160 acres) area, 0.8 km (0.5 mi) on each side. In the second stage of sampling, two sample points were selected within each PSU according to a restricted randomization procedure. Detailed description on sample point selection within a PSU is found in Goebel and Baker (1982). Because the sample points used in this study are a subset of established NRI sample points and have been monitored during previous inventories, aerial photographs and geographic coordinates were available for each point. Field crews used the aerial photos and geographic coordinates in combination with global positioning system technology to locate each sample point in the field.
In MLRAs 9 and 105, 200 points were initially selected for sampling in each MLRA. However, some points were inaccessible, or fell on homesteads, urban areas, road pavement, or rock outcrops, and were not sampled. As a result, only 149 points were sampled in MLRA 9, and 186 points in MLRA 105. Points were selected at random, irrespective of soil series or land-use patterns.
In MLRAs 67 and 77, 100 points were initially selected for sampling in each MLRA, with the restriction that sampling was confined to a single soil series. This was to evaluate the assumption that sampling a single soil series would result in less variability, improving the ability to detect change with different management practices. In MLRA 67, sampling was confined to the Ascalon soil, and in MLRA 77 sampling was confined to the Amarillo soil. These soils were chosen because they are benchmark soils for each area and have a wide geographic distribution. Both soils formed in alluvium. The NRCS database indicated that the 100 points selected in each MLRA were located on the designated soil series. However, to ensure the quality of the data, a soil pit was dug and the series at each sample point was determined in the field by an NRCS soil scientist who was part of the sampling crew. If the sample point was located on an inclusion or for other reasons did not fit the series description, the point was not sampled. As a result, only 64 points were sampled in MLRA 67 and 47 points in MLRA 77.
Soil Sampling and Analysis
At each sample point duplicate 1000 cm3 soil samples were collected. If the soil had been recently cultivated the soil samples were taken from the 0- to 10-cm depth. If the soil had not been cultivated, the samples were taken from the 0- to 2.5- and 2.5- to 10-cm depths. However, for this analysis all data were analyzed for the 0- to 10-cm depth by taking a weighted average of samples from the 0- to 2.5- and 2.5- to 10-cm depths. Sampling was limited to the surface 10 cm because the soil surface is the most sensitive to change as a result of changes in land-use or conservation practices. One of the two soil samples was used for biological analysis and was placed in a cooler with ice packs for transport to the lab. The second sample was used for physical and chemical analysis and was sent to the lab without refrigeration.
Samples collected for biological analysis were analyzed for microbial biomass carbon (MBC), potentially mineralizable nitrogen (PMN), and basal respiration. Microbial biomass C was determined by fumigation and direct extraction with 0.5 M K2SO4 on duplicate 50-g samples of 4-mm, sieved, field moist soil (Tate et al., 1988). Organic C in the fumigated and nonfumigated extracts was measured using a Dohrmann DC-180 carbon analyzer (Rosemount Analytical Services, Santa Clara, CA) calibrated with potassium phthalate standards. Microbial biomass C was calculated using the correction factor (k = 0.33) of Sparling and West (1988). Basal respiration and PMN were measured on the <2-mm fraction using the method of Drinkwater et al. (1996) with the following modifications. Forty grams of soil were used in the analysis instead of 10 g, and the samples were incubated at 25°C instead of 30°C.
A 100-g subsample of the soil collected for physical and chemical analysis was analyzed for water stable aggregates (WSA). For samples from MLRAs 9, 67, and 77, the soil was sieved through an 8-mm screen to remove coarse fragments and WSA were determined using 4-, 2-, 1-, 0.5-, and 0.25-mm screen sizes. Samples from MLRA 105 were sieved through an 4-mm screen to remove coarse fragments and WSA were determined using 2, 1, 0.5, and 0.25 mm screen sizes. Aggregate weights from each size class were used to calculate a mean-weight diameter (MWD) as described by Kemper and Rosenau (1986).
A duplicate soil sample was sieved through a 2-mm screen and analyzed for sand, silt, and clay content (pipette method), pH (1:1 soil/water), total organic C (TOC) by dry combustion measured with a Leco SC-444 analyzer (Leco Corp., St. Joseph, MI), total N by dry combustion measured with a Leco FP-438 analyzer, cation exchange capacity (CEC) at pH 7 by ammonium acetate extraction measured with a Kjeltec Auto 1035 Analyzer (Tecator, Perstorp Analytical Inc., Florence, MA)1 , and exchangeable calcium (Ca), magnesium (Mg), potassium (K), and sodium (Na) at pH 7, by ammonium acetate extraction measured with a Perkins-Elmer AA 5000 (Perkins-Elmer Corp., Norwalk, CT), and extractable acidity by BaCl2-triethanolamine solution buffered at pH 8.2 and back-titrated with HCl. Standard soil survey laboratory methods (USDA-NRCS, 1996) were used in these analyses. The samples were also analyzed for Mehlich extractable phosphorus (MEP) (Mehlich, 1984) measured with an inductively-coupled plasma emission.
Statistical Analysis
The distributions of the data were tested for normality using the D'Agostino-Pearson K2 test. This test is an omnibus 
2 test for detecting deviation from normality caused by either skewness or kurtosis, and has good power over a broad range of non-normal distributions (D'Agostino et al., 1990). Skewness was tested using the third sample moment test (
b1) and kurtosis was tested using the fourth sample moment test (b2). The mean, median, standard deviation, and coefficient of variation for each soil attribute was calculated using PROC UNIVARIATE in SAS, and the K2, b1 and b2 tests were calculated using a SAS macro provided by D'Agostino et al. (1990). The distribution was evaluated at the 
= 0.05 probability level. All soil properties that were not normally distributed were subjected to a natural log (loge) transformation and retested for normality using the procedure of D'Agostino et al. (1990).
One application of regional soil studies using the NRI is to detect changes in soil properties with a second sampling following changes in land management or soil conservation practices. To estimate the power to detect change in the soil properties at a regional scale, a least significant difference (LSD) value was calculated at the = 0.05 probability level. The LSD value was also reported as a percentage of the MLRA mean for comparisons between MLRAs. This value represents the minimum difference that may be detected by resampling the same sample points, assuming the variance of the soil properties remains the same during both sample periods.
|
Results and discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and methodsNOTESResults and discussionREFERENCES |
|---|
|
b1 > 0) (Fig. 2a)
. Right-tailed skewness resulted from the presence of a few exceptionally large values for each soil property, lengthening the right tail (Fig. 2a). In addition, most soil properties cannot have negative values, constraining the left tail at zero. Soil properties having particularly pronounced right-tailed skewness (b1 > 3) were MEP in MLRA 9, exchangeable K in MLRA 105, exchangeable Na in MLRAs 9 and 105, PMN in MLRA 77, and basal respiration in MLRA 105 (Table 1).
|
Right-tailed skewness resulted in the mean being larger than the median (Table 1), and the large values in the right tail of the distribution tended to inflate the variance. In addition, the standard deviations were proportional to the mean when compared across MLRAs, which is a common property of log-normally distributed data (Steel and Torrie, 1980; Sokal and Rohlf, 1981; Gomez and Gomez, 1984).
Success in achieving a normal distribution using a loge transformation varied among soil properties. Concentrations of MEP were normally distributed in all four MLRAs and TOC, total N, and basal respiration were normally distributed in three of four MLRAs after a loge transformation (Table 2) (Fig. 2b). In contrast, soil aggregate MWD, exchangeable K and Na, and PMN remained non-normally distributed in three of four MLRAs after a loge transformation (Table 2). Young et al. (1999) reported that stratifying sample points by landform improved the proportion of normally distributed soil variables. However, many soil properties remained non-normally distributed, indicating stratification was only partially effective in explaining distributional patterns of soil properties in a landscape setting.
|
b1, and b2 statistics. However, with CEC in MLRA 77, exchangeable acidity in MLRA 9, percent silt in MLRAs 9 and 105, and percent clay in MLRAs 9 and 67 the loge transformation increased the degree of non-normality (Tables 1 and 2). The loge transformation also reduced variability of the data by as much as twofold, as assessed by comparing the CVs from before and after transformation (Tables 1 and 2). Other benefits of the loge transformation were that median values were very close to the mean, and that standard deviations tended to vary independent of the mean (Table 2), two important conditions of normality.
The ability to detect significant change in soil properties at the regional scale using non-transformed data varied depending on the property (Table 3) . With soil pH, a change of 3 to 4% of the regional mean may be detected with the sampling intensity used in this study. This suggests that it may be fairly easy to monitor soil acidification or alkalization problems at a regional scale. For monitoring changes in soil organic matter content, a change of 10 to 16% of the regional mean may be detected using nontransformed TOC content (Table 3), and 3 to 8% using loge transformed TOC content as the indicator (Table 4) . For monitoring excessive P enrichment or depletion, a change of 21 to 61% of the regional mean may be detected using nontransformed MEP concentration data (Table 3), and 4 to 24% of the regional mean using loge transformed MEP data (Table 4). Microbial biomass C and basal respiration have been proposed as biological indicators of chemical or heavy metal contamination (Anderson, 1994). Using nontransformed data a change of 15 to 29% of the regional mean may be detected for MBC, and 12 to 23% of the regional mean for basal respiration (Table 3). This difference may be reduced to 2 to 6% of the regional mean for MBC and 2 to 11% of the regional mean for basal respiration using loge transformed data (Table 4).
|
|
Our data indicate that many soil properties at the regional scale are log-normally distributed, in that normality was achieved after a log transformation (Parkin and Robinson, 1994). However, there were a few soil properties for which a loge transformation did not improve, or even worsened the degree of non-normality of the data. This suggests that with these soil properties other distributions may be present. These may include Poisson, negative bionomial, Weibull, gamma, and exponential distributions (Parkin and Robinson, 1994). Additional research is needed to characterize the distribution of the soil properties that were not normal or log-normally distributed so that appropriate statistical analyses may be applied to these data.
Our results also indicate that it is possible to monitor changes in surface soil properties at the regional scale using the NRI. However, the ability to detect significant change will vary depending on the soil property being evaluated. Soil pH may be used to monitor changes in soil acidification or alkalization with a high degree of precision. In addition, it may be possible to detect relatively small changes in soil C content (3–8% of the regional mean) using loge transformed TOC as the indicator. It may also be possible to detect significant changes in biological soil properties such as MBC and basal respiration at a regional scale using the NRI. However, these properties were measured at only a single point in time. Temporal variation in these properties should be evaluated to determine its impact on the magnitude of change that may be detected at the regional scale. Sampling soil properties as part of the NRI should improve NRCS' ability to monitor soil and natural resource condition on a regional scale.Natural Resources Conservation Service 1996; Soil Conservation Service 1981
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and methodsNOTESResults and discussionREFERENCES |
|---|
Received for publication January 4, 1999.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and methodsNOTESResults and discussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  K. Beven, D. Zhang, and A. Mermoud On the Value of Local Measurements for Prediction of Pesticide Transport at the Field Scale Vadose Zone J., March 8, 2006; 5(1): 222 - 233. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. S. Andrews, D. L. Karlen, and C. A. Cambardella The Soil Management Assessment Framework: A Quantitative Soil Quality Evaluation Method Soil Sci. Soc. Am. J., November 1, 2004; 68(6): 1945 - 1962. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C.-W. Chang, D. A. Laird, M. J. Mausbach, and C. R. Hurburgh Jr. Near-Infrared Reflectance Spectroscopy-Principal Components Regression Analyses of Soil Properties Soil Sci. Soc. Am. J., March 1, 2001; 65(2): 480 - 490. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. J. Brejda, D. L. Karlen, J. L. Smith, and D. L. Allan Identification of Regional Soil Quality Factors and Indicators: II. Northern Mississippi Loess Hills and Palouse Prairie Soil Sci. Soc. Am. J., November 1, 2000; 64(6): 2125 - 2135. [Abstract] [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Crop Science | |||
| Vadose Zone Journal | Journal of Plant Registrations | ||||
| Journal of Natural Resources and Life Sciences Education |
Journal of Environmental Quality |
||||
