HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (14) Citing Articles via Google Scholar Google Scholar Articles by Conant, R. T. Articles by Paustian, K. Search for Related Content PubMed PubMed Citation Articles by Conant, R. T. Articles by Paustian, K. GeoRef GeoRef Citation Agricola Articles by Conant, R. T. Articles by Paustian, K. Related Collections Soil Organic Matter Experiment Design Soil Analysis Spatial Distribution

TECHNICAL REPORTS
Plant and Environment Interactions

Spatial Variability of Soil Carbon in Forested and Cultivated Sites

Implications for Change Detection

^a Natural Resource Ecology Laboratory, Colorado State Univ., Fort Collins, CO 80523-1499
^b Environmental Resources Trust, 209 NW 58th Street, Seattle, WA 98107-2030

* Corresponding author (conant{at}nrel.colostate.edu)

Received for publication February 23, 2002.

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
The potential to sequester atmospheric carbon in agricultural and forest soils to offset greenhouse gas emissions has generated interest in measuring changes in soil carbon resulting from changes in land management. However, inherent spatial variability of soil carbon limits the precision of measurement of changes in soil carbon and hence, the ability to detect changes. We analyzed variability of soil carbon by intensively sampling sites under different land management as a step toward developing efficient soil sampling designs. Sites were tilled cropland and a mixed deciduous forest in Tennessee, and old-growth and second-growth coniferous forest in western Washington, USA. Six soil cores within each of three microplots were taken as an initial sample and an additional six cores were taken to simulate resampling. Soil C variability was greater in Washington than in Tennessee, and greater in less disturbed than in more disturbed sites. Using this protocol, our data suggest that differences on the order of 2.0 Mg C ha^-1 could be detected by collection and analysis of cores from at least five (tilled) or two (forest) microplots in Tennessee. More spatial variability in the forested sites in Washington increased the minimum detectable difference, but these systems, consisting of low C content sandy soil with irregularly distributed pockets of organic C in buried logs, are likely to rank among the most spatially heterogeneous of systems. Our results clearly indicate that consistent intramicroplot differences at all sites will enable detection of much more modest changes if the same microplots are resampled.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
THE AMOUNT OF CARBON (C) contained in the soil is nearly as large as the amount in the atmosphere and terrestrial vegetation combined (Schimel, 1995). While much of the C in soil is stabilized due to aggregation, combination with minerals, or chemical recalcitrance (Oades, 1984), substantial amounts of C are transferred to and from the soil annually (Schimel, 1995). Human-induced changes in the balance between soil C inputs and outputs have led to large transfers of C from the soil to the atmosphere (Kern, 1994; Houghton et al., 1999), but these historic losses may be reversed, and atmospheric C sequestered, by encouraging processes that increase C inputs to soil and/or reduce losses from decomposition (Paustian et al., 1997b). The global potential for C sequestration in forest and cultivated soils may be considerable, potentially offsetting a large portion of CO₂ emitted to the atmosphere (Sampson et al., 2000).

Detecting changes in soil C brought about by changes in land management requires precise measurements, and a number of soil characteristics make this challenging. Soil C inputs and outputs are influenced by climate, vegetation, and soil physical characteristics, all of which vary spatially, leading to substantial spatial variability in soil C (Robertson et al., 1997; Conant and Paustian, 2002). Historical land use (i.e., more than 30 yr before present) can have a significant and persistent effect on soil C due to net loss of soil C through increases in decomposition, soil erosion, or loss of productive capacity. Finally, average C sequestration rates are small relative to the total amount of C in the soil (e.g., 0.5 Mg C ha^-1 yr^-1 versus 50.3 Mg C ha^-1 in top 0.3 m; National Soil Survey Laboratory, 1997; Conant et al., 2001), resulting in a small signal to noise ratio. While there are potential obstacles to detecting changes in soil C following changes in management, these can be overcome by using intensive, stratified, and/or highly replicated sampling schemes.

Evaluation of the confidence with which changes in soil C can be detected following changes in management is important for development of international treaties and emissions trading systems. The purpose of this project was to evaluate the efficacy of a sampling scheme developed to detect modest changes in soil C over time. We collected samples from sites representing four unique climatic–management combinations that ranged from relatively uniform (long-term cultivated site in TN) to highly spatially heterogeneous soil C distribution (sandy old-growth forest site with buried decayed logs in WA). Here we evaluate differences in soil C content within and between sites, and variation between simulated sampling times to investigate the implications of soil C spatial variability for change detection following changes in land use or land management. Specifically, we assessed effects of four variables on detecting soil C changes: (i) the spatial variability, (ii) bulking samples or analyzing separately, (iii) the number of sample units, and (iv) whether or not the same sample units are resampled in the future.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
Study Sites
Four study sites, characterizing diverse climate, vegetation, and management regimes (Table 1), were intensively sampled to evaluate soil C variability and our ability to detect changes in soil C due to changes in management. All sites were originally selected for use in analyses of management effects on soil C. Soil samples were collected at cultivated and forested sites in Tennessee (TN) and at two forested sites in Washington (WA). Conventionally tilled corn (Zea mays L.) was grown at the cultivated site in TN (35°57' N, 85°34' W) on Waynesboro loam (fine, kaolinitic, thermic Typic Paleudult) for at least 25 yr prior to sampling. The mixed hardwood forest site in TN (35°59' N, 85°35' W), located on Christian silt loam soil (fine, mixed, semiactive, mesic Typic Hapludult), has been continuously covered by mature mixed-hardwood forest for more than 50 yr. Individual trees have been cut and cattle have occasionally grazed the forested site.

Experimental Design
Three microplots, each consisting of six regularly aligned soil cores (Fig. 1) , were established in early spring 1999 within each field and forest site. Our sampling scheme was based on that used by the Canadian Prairie Soil Carbon Balance Project (Ellert et al., 2001), which was designed to maximize the ability to detect changes in soil C over time by ensuring that exact sample locations can be relocated and resampled, limiting the confounding effect of horizontal variability of soil C. Microplots were always located on flat positions along ridge tops and oriented in the same direction; cores were collected from prearranged locations around the perimeter of each microplot (Fig. 1). The location of the northeastern-most core was determined with GPS, and a relocatable Scotchmark EMS magnetic ball marker (3M Corporation, Austin, TX) was buried more than 1 m deep to ensure our ability to relocate the exact sample locations for future resampling.

View larger version (22K): [in this window] [in a new window]   Fig. 1. Diagram illustrating design and orientation of all microplots used for collection of soil samples. Initial (gray) and resampling (black) cores are distinguished.

Each microplot was sampled twice to mimic "initial" sampling immediately followed by a "resampling" some time later (hereafter referred to as initial and resampling or Sample Times 1 and 2). Data within a microplot between these mock sample times were compared to test the precision of remeasurement. Additionally, all soil cores were analyzed individually to establish variability within and between microplots. Samples were then bulked by depth increment (see below) and by microplot to compare the benefits of analyzing individual versus composited samples. Results occasionally refer to specific site (cultivated, forested in TN and old growth and second growth in WA), sample time (1 or 2), and microplot (1, 2, or 3).

Sample Collection and Analysis
A Giddings (Fort Collins, CO) soil coring rig was used to collect 0.65-m-diameter soil cores to a depth of 0.3 m. All surface litter within a 0.18-m² area centered around the core was removed prior to core collection. Soil cores were separated into 0- to 0.1-, 0.1- to 0.2-, and 0.2- to 0.3-m depth increments for this analysis. Following collection, soil samples were placed in sealed plastic bags, returned to the laboratory, and weighed. Field moist soil samples were gently broken along the plane of least resistance to pass an 8-mm mesh sieve; visible root material was removed by hand-picking during 8-mm sieving. The 8-mm-sieved soil was air-dried, passed through a 2-mm mesh sieve, oven-dried at 60°C for 72 h, and ground to fine powder. Soil C concentrations were determined with a LECO (St. Joseph, MI) CHN-1000 analyzer. Since carbonates were not detected following acid addition, organic C was assumed equivalent to total C. Bulk density was calculated with sample volume and weight; sample weight was corrected for soil moisture and root and rock content.

Statistical Analyses
Planned comparison analysis of variance was used to test for differences in soil C between microplots, sampling times, and sites; Scheffe's test was used for comparison of means. The microplot was considered the sample unit, so number of sample units per site equaled three in all cases. All statistical analyses were performed in SAS (SAS Institute, 1985). Differences are considered significant at P < 0.05 and results are reported as means ± one standard deviation.

Along with minimum, maximum, and mean soil C values for each scale, coefficients of variation were calculated as an indication of soil C variability. The relationship between coefficient of variation and minimum detectable difference was derived from a formula designed to calculate the minimum number of samples required to detect a difference of a certain magnitude (Sokal and Rohlf, 1981). We calculated the minimum detectable difference as:

[1]

where is equal to minimum detectable change in soil C (as % of the mean), CV is the coefficient of variation, n is the number of samples collected during each sampling (i.e., n_initial = n_resampling = 3), and t values are the critical values for a one-tailed t test, a function of degrees of freedom (V), value ( = 0.05 or 0.10), and power (P, 0.8). This relationship was evaluated at all four sites to examine implications for change detection with resampling.

Finally, we evaluated the efficacy of resampling the same microplots versus future sampling of different microplots by comparing results from paired t tests (i.e., resampling the same microplots) with unpaired t tests (resampling at new microplots; Sokal and Rohlf, 1981). Initial measured soil C values from each microplot were compared with values 5 to 100% greater (at 5% increments) than initial values. One-thousand future soil C values were generated for each increment (i.e., every 5% step) by randomly selecting values from the distribution around the mean measurement for each microplot (based on standard deviation calculated from the six initial cores). The probability that a difference will be significant was evaluated for all four sites and for three levels of significance ( = 0.05, 0.10, and 0.25).

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
Soil Carbon Content
Average soil C for initial sampling and resampling were very similar for both the forested (29.8 ± 0.47 and 30.0 ± 4.35 Mg C ha^-1) and cultivated sites (18.2 ± 0.85 and 19.1 ± 1.22 Mg C ha^-1) in TN. Soil C content was significantly (P < 0.05) greater at the forested site than the cultivated site for both initial and resampled microplots (Fig. 2a) . The coefficient of variation for soil C collected from the cultivated microplots (5.3%) was larger than that at the forested microplots (4.5%).

View larger version (33K): [in this window] [in a new window]   Fig. 2. Soil C content (0–0.3 m, ±1 SD confidence interval) by depth increment for two sampling times at (a) cultivated and forested sites in Tennessee (TN), and (b) old-growth and second-growth sites in Washington (WA). The first term in the microplots is the sampling time (i.e., 1 = initial and 2 = resampling), and the second term represents the microplot number (i.e., 1–3).

Soil C content decreased with depth for all microplots at all four sites (Fig. 2) and for the vast majority of all soil cores at both sites. Most soil C was located in the top 0.1 m for all soils, ranging from 50.1 (TN cultivated site) to 61.4% (WA old-growth site) of total soil C. Between 25 and 32% of soil C was found in the 0.1- to 0.2-m increment and only 13 to 19% was found below 0.2 m. Soil C variability was not related to sample depth. Differences between microplots were relatively consistent through the soil profile, with the notable exception of WA Old Growth 1-2 and 2-2 (Fig. 2b).

Variability within and between Microplots
Soil C variability between microplots at the cultivated site in TN was minimal; soil C was not significantly different between any microplots at the cultivated site (Table 2). Core-to-core variability was also small at the cultivated site, with coefficient of variation within a microplot ranging from 6.4 to 20.7% and averaging 11%. Between-microplot variability was greater at the forested site where average soil C for one microplot (Forest 2-3) was significantly greater than for four of the others (Table 2). Greater variability was not strongly reflected in the coefficients of variation (range = 4.6–23.4%, averaging 14.6%) for soil C content of forest cores since mean soil C content was greater. There was limited correspondence between samples collected at the same microplot at different sample times; correlation coefficients across sampling times at the cultivated and forested sites in TN were 0.63 and 0.27, respectively. Combining all 12 cores from a microplot (i.e., initial and resampling cores) did not decrease the coefficient of variation within any of the microplots (i.e., 0 of 18 depth–microplot combinations), but did slightly increase CV for one microplot (Cultivated Microplot 3) for 0 to 0.3 m.

Change Detection
Based on measured coefficients of variation, changes in soil C that could be expected to occur over four years (2 Mg C ha^-1) following changes in land use or management in cultivated lands (e.g., 0.5 Mg C ha^-1 yr^-1; Paustian et al., 1997a) could be detected with = 0.05 by collecting samples from between five (initial sampling results) and nine microplots (resampling results; Fig. 3a) . Changes on the order of 2 Mg C ha^-1 could be detected in the forested system in TN by collection and analysis of samples from between two (initial sampling results) and 34 (resampling results; Fig. 3b) microplots. If our collection would have been limited to our initial sampling only, we would have concluded that differences of 2 Mg C ha^-1 could be detected with = 0.10 at both sites in TN with collection and analysis of soil cores from three microplots (Fig. 3).

View larger version (24K): [in this window] [in a new window]   Fig. 3. Relationship illustrating the magnitude of soil C change (0–0.3 m) that can be detected (minimum detectable difference; MDD) with 90% confidence by collecting a certain number of samples from (a) cultivated or (b) forested sites in Tennessee (TN).

View larger version (25K): [in this window] [in a new window]   Fig. 4. Relationship illustrating the magnitude of soil C change (0–0.3 m) that can be detected (minimum detectable difference; MDD) with 90% confidence by collecting a certain number of samples from (a) old-growth and (b) second-growth sites in Washington (WA).

View larger version (26K): [in this window] [in a new window]   Fig. 5. Diagram illustrating the relationship between magnitude of soil C change (percent) and the portion of simulated soil C changes in which the critical t value was exceeded for three confidence levels ( = 0.05, 0.10, and 0.25) using paired (P) and unpaired (NP) analyses. Soil C changes from 5 to 100% and were generated based on methods described in the text.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

Coefficients of variation for the cultivated site in TN ranged between 10 and 15%, coinciding with soil C and other soil organic matter parameters measured in intensively sampled fields elsewhere (Robertson et al., 1997; Bragato and Primavera, 1998; Garten and Wullschleger, 1999). Coefficients of variation for samples collected from forested sites in both TN and WA, however, were larger than in the TN cultivated soils; those for 0- to 0.3-m depths from WA were much larger, ranging between 60 and 126%, and were just as large when only the 0- to 0.1-m increment was considered (ranging from 45–180%). Spatial variability of soil C is often substantial in forest soils (Mollitor et al., 1980; Grigal et al., 1991; Mottonen et al., 1999) due to the heterogeneous nature of vegetation, microclimate, and soil physical properties (Saetre, 1999). Yet coefficients of variation for WA soils were much larger than those observed elsewhere (Mollitor et al., 1980; Johnson et al., 1990; Grigal et al., 1991; Cromack et al., 1999; Mottonen et al., 1999; Homann et al., 2001; and results from TN), leading to limited ability to detect changes in soil C over time compared with other work (Huntington et al., 1988; Johnson et al., 1990; Homann et al., 2001). Undoubtedly much of the observed variability was driven by inconsistent but frequent inclusion of buried decayed wood within the samples. Removal of undecayed wood from samples would decrease soil C variability, but would result in inaccurate estimates of total site C. Sample stratification across variables affecting soil C is an effective method to decrease variance in spatially heterogeneous ecosystems (Klironomos et al., 1999), but difficulties associated with stratification based on distribution of buried, decayed wood have not been fully explored. Wood decayed to the point of humification is not clearly distinguishable from other well-developed topsoil making it difficult to quantify the spatial distribution of buried wood. It may be possible to overcome the variability introduced by buried dead wood by stratifying cores into those containing visible decayed organic material and those that do not contain visible decayed organic material coupled with additional spatial sampling to establish the distribution of buried wood.

We tested a sampling scheme designed to maximize our ability to detect small changes in soil C over time on a range of systems. This range varied from the detection of differences that were straightforward, such as well-mixed soil that is spatially homogeneous (CV = 6% across entire cultivated site in TN), to differences that are likely to be very difficult to detect, such as a very homogeneous system (CV = 76% across entire WA old-growth site) with low C content, sandy soils containing irregularly spaced pockets with very high C content (buried logs). Based on coefficients of variation for initial sampling, this scheme worked well at the cultivated and forested sites in TN, similar to results from a number of pasture and forest sites in Virginia (Conant and Paustian, 2002), and at a variety of initially cultivated sites in Canada (B. McConkey, personal communication, 1999). This sampling method resulted in the ability to detect changes likely to occur over three to five years at those sites. Furthermore, our previous work (Conant and Paustian, 2002) suggests that six cores per microplot is adequate to represent the range of soil samples within more uniform sites sampled in that study. However, sample variability for the forested sites in WA suggests that this sampling scheme may not be ideal for all systems.

Consistent intermicroplot differences between sample times at the WA old-growth site (r = 0.99) suggest that smaller changes in soil C may be detected if original values are used as a covariate in analysis of samples collected some time in the future. Saffigna et al. (1989) and Homann et al. (2001) demonstrated that accounting for initial spatial variability using this method may reduce CV for differences in soil C and N by as much as 50%. If future changes in soil C in response to management are consistent across microplots, then sampling multiple microplots in the future could enable detection of more modest changes in soil C. Statistically accounting for initial spatial variability at the TN sites only slightly increased ability to detect changes in soil C at all levels, but the differences in WA were dramatic. Results from this sampling design demonstrate that resampling the same microplots in the future greatly enhances statistical power, particularly in systems that are more spatially variable, and suggest that changes could be detected as much as eight years earlier (assuming soil C changes at 0.5 Mg C ha^-1 yr^-1) if the same microplots are resampled. It should be noted that these results assume that soil C content will increase uniformly within a microplot and that soil C increases are directly related to initial soil C content. Violation of these assumptions will alter the results of the statistical investigation evaluating utility of resampling at the same microplots. It seems unlikely that these assumptions will be universally true; indeed, it is often assumed that soil C stabilization capacity is inversely related to C content (e.g., in tilled agricultural systems; Paustian et al., 1997b). One other factor may have influenced the results presented here. Future soil C values were generated using values randomly selected from a new normal distribution, based on microplot mean and standard deviation, increased by a fixed portion (5%, 10%, etc.). Though the soil C values were normally distributed for all microplots included in this analysis, the standard deviations in some cases were so large that negative values were occasionally generated for one of the WA old-growth microplots. Negative values were eliminated in less than 2.5% of cases, but slightly skewed the results in favor of greater increases in soil C.

Decreasing statistical confidence (i.e., increasing acceptable risk of Type I error) slightly increased ability to detect differences in soil C with any particular number of microplots. The amount of change required for detection with collection and analysis of samples from three microplots decreased by 16.8%, or 0.18 (cultivated) or 0.32 (forest) Mg C ha^-1. Therefore, using = 0.10 instead of = 0.05 would enable detection of changes approximately one year earlier. At the WA sites, an identical change in statistical confidence will lead to a similar proportional change in sensitivity amounting to increased sensitivity of between 7.7 (second growth) and 22.7 (old growth) Mg C ha^-1. Although decreasing statistical confidence levels substantially increase sensitivity in relative terms, minimum detectable changes are still very large (33 and 96 Mg C ha^-1 for second growth and old growth, respectively) and are unlikely to be detected using this sampling design. If the same microplots are resampled, increasing acceptable Type I error (from = 0.05 to = 0.10 or from = 0.10 to = 0.25) decreases duration required between sample periods before detection is possible by approximately four years for the forested site in TN.

Changes in soil C likely to occur due to changes in agricultural or forest management can be detected using this or a variety of more traditional methods for collection and analysis of soil samples. However, the minimum detectable difference is inversely related to the number of samples required (and hence cost). The variables controlling the makeup of this relationship are not uniform from site to site and limit the applicability of any one sampling scheme. But combining multiple cores in regularly aligned microplots, and sampling the same microplots in the future appears to be broadly useful. Our results suggest that analysis of samples bulked by microplot and depth increment, which considerably reduces processing time, accurately represent the information generated by analyzing each core individually, but information about spatial variability, which can be critical in answering some questions, is greatly reduced. Following proper statistical techniques can substantially decrease the duration required before changes can be detected and increasing the amount of Type I error acceptable can lead to further decreases. Verification of changes in soil C content are achievable, but careful consideration of the system of interest and desired output will lead to more effective use of resources.

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
Detecting changes in soil C with changes in land use (between forest, pasture, or cultivation) or land management (e.g., changes in tillage, grazing, or harvesting practices) is complicated by the size of the changes relative to the total amount of soil C and by the spatial variability of C within the soil. These problems can be overcome by collection, preparation, and analysis of many soil samples, but this is very time-consuming and expensive. New more portable methods for faster analysis of soil C samples (Cremers et al., 2001) or analyses that are more sensitive to change (i.e., the ability to detect recently added C) may improve our ability to accurately detect modest changes in soil C over shorter periods of time. For the time being, intensive sampling schemes that enable future resampling in the same area seem best-suited to change detection.

	ACKNOWLEDGMENTS

 
We wish to thank Greg Brann, Charles Parris, and Jerry Prater for assistance locating field sites and collecting samples in Tennessee. Thanks also to Emus Holland and Dan Holman for allowing us to collect samples from their properties. Jim Zumbrunnen and Stephen Ogle provided statistical advice. Three anonymous reviewers suggested improvements for this manuscript. This research was funded by the USEPA and a grant from the Bradley Fund for the Environment to the Environmental Resources Trust and Colorado State University.

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 

• Bragato, G., and F. Primavera. 1998. Manuring and soil type influence on spatial variation of soil organic matter properties. Soil Sci. Soc. Am. J. 62:1313–1319.[Abstract/Free Full Text]
• Conant, R.T., and K. Paustian. 2002. Spatial variability of soil organic carbon: Implications for detecting change at different scales. Environ. Pollut. 116:S127–S135.
• Conant, R.T., K. Paustian, and E.T. Elliott. 2001. Grassland management and conversion into grassland: Effects on soil carbon. Ecol. Appl. 11:343–355.
• Cremers, D.A., M.H. Ebinger, D.D. Breshears, P.J. Unkefer, S.A. Kammerdiener, M.J. Ferris, K.M. Catlett, and J.R. Brown. 2001. Measuring total soil carbon with laser-induced breakdown spectroscopy (LIBS). J. Environ. Qual. 30:2202–2206.[Abstract/Free Full Text]
• Cromack, K., R.E. Miller, O.T. Helgerson, R.B. Smith, and H.W. Anderson. 1999. Soil carbon and nutrients in a coastal Oregon douglas-fir plantation with red alder. Soil Sci. Soc. Am. J. 63:232–239.[Abstract/Free Full Text]
• Ellert, B.H., H.H. Janzen, and B.G. McConkey. 2001. Measuring and comparing soil carbon storage. p. 131–146. In R. Lal, J.M. Kimble, R.F. Follet, and B.A. Stewart (ed.) Assessment methods for soil carbon. CRC Press, Boca Raton, FL.
• Garten, C.T., and S.D. Wullschleger. 1999. Soil carbon inventories under a bioenergy crop (switchgrass): Measurement limitations. J. Environ. Qual. 28:1359–1365.[Abstract/Free Full Text]
• Goldin, A. 1992. Soil survey of the Snoqualmie Pass area, parts of King and Pierce counties, Washington. USDA Natural Resour. Conserv. Serv., Washington, DC.
• Grigal, D.F., R.E. McRoberts, and L.F. Ohmann. 1991. Spatial variation in chemical properties of forest floor and surface mineral soil in the North Central United States. Soil Sci. 151:282–290.
• Homann, P.S., B.T. Bormann, and J.R. Boyle. 2001. Detecting treatment differences in soil carbon and nitrogen resulting from forest manipulations. Soil Sci. Soc. Am. J. 65:463–469.[Abstract/Free Full Text]
• Houghton, R.A., J.L. Hackler, and K.T. Lawrence. 1999. The U.S. carbon budget: Contributions from land-use change. Science 285:574–578.[Abstract/Free Full Text]
• Huntington, T.G., D.F. Ryan, and S.P. Hamburg. 1988. Estimating soil nitrogen and carbon pools in a northern hardwood forest ecosystem. Soil Sci. Soc. Am. J. 52:1162–1167.[Abstract/Free Full Text]
• Johnson, C.E., A.H. Johnson, and T.G. Huntington. 1990. Sample size requirements for the determination of changes in soil nutrient pools. Soil Sci. 150:637–644.
• Kern, J.S. 1994. Spatial patterns of soil organic matter in the contiguous United States. Soil Sci. Soc. Am. J. 58:439–455.[Abstract/Free Full Text]
• Klironomos, J.N., M.C. Rillig, and M.F. Allen. 1999. Designing belowground field experiments with the help of semi-variance and power analyses. Appl. Soil Ecol. 12:227–238.
• Mollitor, A.V., A.L. Leaf, and L.A. Morris. 1980. Forest soil variability on Northeastern flood plains. Soil Sci. Soc. Am. J. 44:617–620.[Abstract/Free Full Text]
• Mottonen, M., E. Jarvinen, T.J. Hokkanen, T. Kuuluvaninen, and R. Ohtonen. 1999. Spatial distribution of soil erosterol in the organic layer of a mature Scots pine (Pinus sylvestris L.) forest. Soil Biol. Biochem. 31:503–516.
• National Soil Survey Laboratory. 1997. National Soil Survey Laboratory characterization data. USDA Natural Resour. Conserv. Serv., Lincoln, NE.
• Oades, J.M. 1984. Soil organic matter and structural stability: Mechanisms and implications for management. Plant Soil 76:319–377.
• Paustian, K., O. Andren, H.H. Janzen, R. Lal, P. Smith, G. Tian, et al. 1997a. Agricultural soils as a sink to mitigate CO₂ emissions. Soil Use Manage. 13:230–244.
• Paustian, K., H.P. Collins, and E.A. Paul. 1997b. Management controls on soil carbon. p. 15–49. In E.A. Paul, K. Paustian, E.T. Elliott, and C.V. Cole (ed.) Soil organic matter in temperate agroecosystems. CRC Press, Boca Raton, FL.
• Robertson, G.P., K.M. Klingensmith, M.J. Klug, E.A. Paul, J.R. Crum, and B.G. Ellis. 1997. Soil resources, microbial activity, and primary production across an agricultural ecosystem. Ecol. Appl. 7:158–170.
• Saetre, P. 1999. Spatial patterns of ground vegetation, soil microbial biomass and activity in a mixed spruce–birch stand. Ecography 22:183–192.
• Saffigna, P.G., D.S. Powlson, P.C. Brookes, and G.A. Thomas. 1989. Influence of sorghum residues and tillage on soil organic matter and soil microbial biomass in an Australian vertisol. Soil Biol. Biochem. 21:759–765.
• Sampson, R.N., R.J. Scholes, C. Cerri, L. Erda, D.O. Hall, M. Handa, et al. 2000. Additional human-induced activities. Article 3.4. p. 180–281. In R.T. Watson, I.R. Noble, B. Bolin, N.H. Ravindranath, D.J. Verardo, and D.J. Dokken (ed.) Land use, land-use change, and forestry. Cambridge Univ. Press, Cambridge.
• SAS Institute. 1985. SAS user's guide: Statistics. Version 5 ed. SAS Inst., Cary, NC.
• Schimel, D.S. 1995. Terrestrial ecosystems and the carbon cycle. Global Change Biol. 1:77–91.
• Sokal, R.R., and F.J. Rohlf. 1981. Biometry. W.H. Freeman and Co., New York.

This article has been cited by other articles:

				C. M. Flint, R. B. Harrison, B. D. Strahm, and A. B. Adams Nitrogen Leaching from Douglas-fir Forests after Urea Fertilization J. Environ. Qual., August 8, 2008; 37(5): 1781 - 1788. [Abstract] [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (14) Citing Articles via Google Scholar Google Scholar Articles by Conant, R. T. Articles by Paustian, K. Search for Related Content PubMed PubMed Citation Articles by Conant, R. T. Articles by Paustian, K. GeoRef GeoRef Citation Agricola Articles by Conant, R. T. Articles by Paustian, K. Related Collections Soil Organic Matter Experiment Design Soil Analysis Spatial Distribution