Abstract

Forest soils offer great potential for bioremediation, including the sequestration of atmospheric carbon.  However, the benefits of bioremediation with soils often are difficult to quantify due to the lack of a clear market value associated with forest management induced changes in soil carbon.  An additional factor that confounds the valuation of carbon remediation benefits is the inherent spatial and temporal variation of the soil chemical and physical properties responsible for carbon sequestration.  Policy makers are attempting to draft criteria for valuing carbon, and it is expected that tradable emissions permits will be used to manage global climate change issues related to atmospheric carbon.  However, many policy makers are neglecting to include soil carbon sequestration in market accounting systems.  This oversight will result in suboptimal allocation of carbon credits, and efforts need to be taken to prevent the exclusion of soil carbon as an atmospheric carbon sink.  Case studies in the Missouri Ozarks and in the Arkansas Western Gulf Coastal Plain illustrate the challenges of using soil map units for estimating both soil and plant carbon fixation.  A failure to address the effects of variation in these systems will result in suboptimal valuation of the forest soil resource, so societal benefits from carbon sequestration will not be maximized. 

Introduction

For more than a decade environmental scientists have researched the causes and effects of increased atmospheric carbon and the linkages to global climate change.  With increased knowledge about the interrelationship of human activities and climate change, governments have worked toward policies that are designed to mitigate human influences on climate, and the Kyoto Protocols were one of the first major efforts to create a worldwide consortium for managing atmospheric carbon levels.  Several challenges remain with regard to carbon management, however.  First, not all governments ascribe to the tenets outlined in the Kyoto Protocols, and the United States is one of the more significant non-participants.  Another obstacle to effective carbon management is the difficulty of identifying and quantifying all sources of carbon efflux and carbon sequestration.  Furthermore, in an effort to implement some form of “best possible” policy for carbon management, most current approaches to valuing carbon focus on plant biomass as the major sink for offsetting management induced increases in atmospheric carbon. 

Soil carbon sequestration does not receive the same consideration as above-ground biomass by policy makers at present, although some estimates of terrestrial soil carbon indicate that soils may account for three-quarters of all terrestrial carbon pools (Henderson, 1995).  Recent estimates of capacity and storage time suggest that soils have the potential for greater carbon storage over a longer timeframe than woody biomass (Lackner, 2003).  During a Southern Hardwood Forest Research Group meeting in March of 2003, Dr. Richard A. Birdsey, USFS Program Manager for the Northern Global Change Research Program, indicated in his keynote address that carbon accounting policies slated for completion in 2004 should perhaps omit the valuation of soils as a sink for atmospheric carbon.  Dr. Birdsey cited the multiple sources of soil carbon variability and measurement challenges as reasons for omitting soils from pending policy, which is a fair argument at present.  However, unless all forest resources and assets are quantified, including the valuation of major carbon sinks, socially suboptimal carbon management policies will result (Ficklin et al., 1996).  Additionally, the use of “bedding” procedures for site preparation during reforestation or afforestation have unknown effects on the long term storage of carbon in forest soils.  Quantifying the amount of carbon fixed in tree biomass without consideration of the influence of management techniques belowground cannot accurately estimate carbon sequestration.

The purpose of this paper is twofold.  First, we urge soil and other forest scientists to disseminate the large body of scientific knowledge on the capacity of forest soils to serve as a sink for atmospheric carbon.  Also, we want policy makers to become aware that forest soil carbon pools are not “no net change” sinks for carbon in forest lands that are actively managed.  To assure the inclusion of forest soils in carbon sequestration policies, better estimations of actual and potential soil carbon storage are needed.  Modeling approaches are one way to improve soil carbon accounting for certain forest systems, and some efforts already have been made to develop models for predicting soil carbon accumulation for sites that over time may warrant actual measurement of soil carbon for valuation purposes (Paul et al., 2003).  However, a better conceptual model that can be applied across many types of landscapes is needed before potential carbon storage can be predicted. 

Integrating such a conceptual model with existing soil map units may offer a long-tem approach for the inclusion of soil carbon valuation into forest carbon sequestration policies.  Past work on soil productivity indices may provide a useful framework for a potential carbon sequestration index (PCSI), since roots and root exudates are a major source of subsurface soil carbon enrichment.  If modeling of soil carbon is to succeed, then better process based knowledge rather than black-box knowledge is needed relative to the mechanisms of carbon fixation and efflux.  

Two objectives of this manuscript are:

1.       To explore the reasons that forest soils are not given appropriate consideration in current plans for the trading of carbon permits, including the issues of measurement challenges and misconceptions among policy makers about soil carbon pools.

2.       To examine two case studies of existing soil survey map units as a means of exploring options and challenges for modeling the potential quantity and value of forest soil carbon sequestration under different management scenarios.

A great deal of time and money has been spent mapping the soils of the United States and other parts of the world, but the utility of soil survey data for land management is limited by the resolution/ map unit size of the surveys.  The limitations of soil survey information for predicting site productivity are well known (Grigal, 1984).  Nevertheless, efforts to classify the potential carbon storage of soils across large land areas would be simplified if soil map units could be used to model soil carbon sequestration.  In this manuscript we first explore the status and deficiencies of current and proposed carbon policies.  To address the issue of gaps in carbon accounting, two case studies in two physiographic regions of the United States are presented to illustrate how variability within a given soil map unit results in variability in both above and belowground carbon storage.  If soil map unit data can be integrated into models to predict potential carbon sequestration, then the challenges observed in both case studies must be taken into consideration, and better conceptual and process models of how carbon cycles through various soil carbon pools should be developed.

Understanding potential carbon storage in both the soil and the woody biomass for an area such as this requires a better understanding of the mechanisms of plant growth, and rooting sufficiency is an area of research that may elucidate the variabilities observed on this site and others.  Topographic position has long been recognized as having profound influences on plant growth (Carmean, 1975), and soil survey map units generally follow topographic features.  Several studies have been conducted to evaluate site productivity based on the suitability of a soil medium for unimpeded root growth (Kiniry et al., 1983; Gale and Grigal, 1988; Henderson et al., 1990), but the components to include in these “productivity index” models are not yet clearly defined for all soils.  Perhaps an expansion of these types of investigations will provide information germane to the understanding of soil carbon enrichment, including contributions from rhizodeposition.  Studies of litterfall cannot adequately estimate the carbon added to a soil from plant growth, as the results of one of our case studies indicate.

Status of Carbon Valuation and Trading Policies

We begin this discussion with the following observations: a) Carbon sequestration is a relatively new and rapidly growing area of science. Although research on soil carbon existed for decades, exploration of the potential role of this carbon in the context of global warming is a recent phenomenon. b) There has also been rapid development in the policy arena regarding global warming and carbon sequestration.  The Kyoto Protocol and the subsequent activities of the United Nations Framework Convention on Climate Change (UNFCCC) as well as the Intergovernmental Panel on Climate Change (IPCC) have established strategies aimed at mitigating the adverse effects of climate change due to greenhouse gas emissions. c) The interest in carbon sequestration is by no means limited to researchers and policy makers.  Private enterprises have demonstrated significant interest in carbon sequestration due to the prospect of capturing economic rents from carbon credits and reduction in pollution abatement costs.

These observations should collectively underscore the importance of this paper.  The general area of carbon sequestration is characterized by two conflicting phenomena.  On one hand, there is a considerable amount of interest among policy makers and business enterprises about soil carbon and its potential as carbon sink.  On the other hand, there remains substantial gap in the knowledge base, and impacts of policy alternatives are largely uncertain.  Additionally, many aspects of carbon policy remain vague.

Article 3.3 of the Kyoto Protocol contains provisions for forest carbon sinks.  The Article asserts that a reduction in net greenhouse gas emission can take place either by a reduction in the emission of these gases, or by increasing the rate at which these gases are removed from the atmosphere (Marland et al., 2001).  However, the Kyoto Protocol also stipulates that the latter method can only be employed in conjunction with the former, that is, a ratifying country could not fulfill all of its responsibilities solely by establishing carbon sinks.  Article 3.3 recognizes “afforestation” and “reforestation” as carbon sinks, without adequately defining these terms (Marland et al., 2001).  The issue of soils as a potential belowground carbon sink is not mentioned in the Protocol.  It is not clear if the Protocol even considered soil as part of a forest system.  However, a subsequent meeting of the IPCC did include soil as part of the forest system (Marland et al., 2001).  The Soil Science Society of America has also advocated the inclusion of soil in carbon sequestration (SSSA, 2001).  The Society of American Foresters, however, does not have a position statement on carbon sequestration to date.  Nevertheless, a majority of the efforts in land-use based reduction in greenhouse gases are aimed at aboveground biomass rather than belowground carbon sequestration.

Despite the vagueness, the Kyoto Protocol has certainly strengthened the argument for a “carbon market” where emission credits can be traded.  This is encouraging to industries interested in paying others for carbon credits (Marland et al., 2001).  As long as the cost of carbon credit is lower than the cost of emission reduction or pollution abatement, industries will be able to capture economic rents by trading carbon credits.  In fact, the Chicago Carbon Exchange, an open market for trade in carbon credits is currently being set up and is expected to be operational in the near future.  Nevertheless, vague policies and the current U.S. government’s decision to opt-out of the Protocol have created a level of uncertainty for forest landowners and businesses (Binkley et al., 2002).

Aboveground terrestrial ecosystems such as forests also possess a tremendous potential for carbon sequestration. As mentioned earlier, carbon sequestration policies to date appear to be geared toward forests. Much work, however, remains to be done in ironing out the specifics of this policy. Substantial private ownership of U.S. forests, 57% by individuals and Timberland Investment Management Organizations (TIMOs) and another 15% by forest industries, adds an interesting dimension to the policy arena (MacCleery, 1994). Economic incentives are crucial for these landowners to adjust their management practices to fit carbon sequestration policy needs (Binkley et al., 2002). Payment in exchange for carbon credits appears to be an intuitive and mutually beneficial solution. Contrary to earlier beliefs, the energy industry does not appear to have a widespread interest in investing in forestry projects (Binkley et al., 2002). Apparently, these industries have realized that the cost involved in branching out to a new venture that is foreign to their area of expertise is higher than paying someone to do it for them. This, in fact, opens new income opportunities for a profit-maximizing forest landowner, and including the soil carbon component would make forestry incentives even more attractive.

Section 1605(b) of the Energy Policy Act of 1992 initiated the Voluntary Reporting of Greenhouse Gases Program. This program, sometimes called a U.S. “alternative” to the Kyoto Protocol, records voluntary emission reduction or sequestration measures undertaken by energy industries. In 2001, a total of 228 entities reported undertaking 1,705 projects that resulted in 222 million metric tons of direct and 71 metric tons of indirect reductions, 8 million metric tons of carbon sequestration, and 15 million metric tons of other reductions (U.S. Department of Energy, 2003).

The issue of climate change policy has inspired a number of research publications. Majority of these research focused on policy or procedural aspects of carbon accounting in forests (Stainback and Alavalapati, 2002; Huang and Kronrad, 2001; Enzinger and Jeffs, 2000; Murray et al., 2000; Cathcart, 2000; Hoover et al., 2000), while others focused on topics such as impact of climate change policies on the U.S. pulp and paper industry (Ruth et al., 2000) and on biodiversity (Matthews et al., 2002). Stainback and Alavalapati (2002) performed an economic analysis of carbon sequestration through slash pine forests. The authors found that such objectives would increase the rotation age resulting in an increase in sawtimber supply and a consequent decrease in pulpwood supply. There was also an increase in the value of forest land. The authors predicted that such an increase would cause more land to be devoted to forests and would ultimately be beneficial to landowners. These results were largely similar to those of Enzinger and Jeffs (2000), based on an economic analysis of Australian forests as carbon sinks.

Huang and Kronrad (2001) investigated the amount of compensation necessary from utility companies to forest landowners in order to receive carbon sequestration credits. The authors found that in order to achieve the cost goal of $10 per ton for sequestered carbon set by the Department of Energy, use of current loblolly pine plantations would not be cost-effective. However, new plantations on currently unstocked land provided some promising results. A policy analysis by Murray et al. (2000) explored the potential relevance of the Kyoto Protocol for U.S. forests. The authors, while alluding to the possibility of significant changes in forest management, underscored the need for filling the information gaps and clearer definition and explanation of concepts used in the Protocol.

Ruth et al. (2000) analyzed the impacts of market-based climate change policies on the pulp and paper industry. A reduction in carbon emissions under different scenarios was found. This reduction was even higher in the presence of investment incentives. However, the authors cautioned that expected production increase was likely to be higher than the reduction in emission. Matthews et al. (2002) investigated the impacts of carbon sequestration policies on biodiversity. The authors examined bird populations in South Carolina, Maine and Wisconsin, assuming a conversion of 10 percent of the total agricultural land in each of the states. Their results indicated a relatively uniform decrease of farmland birds between 10.8 and 12.2 percent in each of the states while gain in forest bird species ranged from 0.3 percent (Maine) to 21.8 percent (Wisconsin). In terms of total bird species count, each of the three states experienced a net loss of birds.

These studies are just some examples of what is likely to be a plethora of research in the near future. As the knowledge base in the carbon sequestration area grows and policies become more well-defined, researchers are sure to focus more of their attention in this area.

Materials and Methods

In addition to conducting secondary data collection on the current status and prospectus of carbon sequestration policies, two case studies are provided to illustrate the challenges and benefits of using soil map units in estimating or modeling soil carbon sequestration.  The first study was conducted in the highly dissected landscapes of the Ozark Highlands of southern Missouri.  The second study was performed in the relatively flat coastal plain region of southern Arkansas.  Both investigations were done within areas mapped as a single soil map unit, and the variation of soil carbon and plant biomass was distinct across both study areas. 

Case study one:

Upland Ozark forest soil:  site characterization and map unit description

The original objectives of this project were to evaluate the temporal and spatial variation of soil organic carbon, and study plots were established in one soil series in an effort to address a portion of the spatial variability component.  Upon implementation of the research, it became clear that below the litter layer dramatic differences in soil properties still would present challenges to the measurement of spatial soil carbon variability.  Importantly, the soils in this study area are in or near a state of equilibrium carbon flux, as the area has not been managed for approximately 75 years.  In areas with intensive management plot measurements of soil carbon contents are of limited value, due to dramatic fluxes in carbon following disturbance (Korner, 2003).

Previous research on soil organic carbon in terrestrial ecosystems often focused on only the upper few centimeters of soil (Hseih, 1992).  This shallow sampling is a constraint to our understanding of the dynamic nature of soils as living bodies (Hammer, 1998), and the inferential limitations associated with shallow sampling often are overlooked (Stone and Kalisz, 1991).  Ideally, the entire plant rooting depth should be sampled to evaluate changes over time, but the depth of sampling in Ozark forest soils often is limited by the presence of roots and rocks.  By utilizing the cutting strength of a post-hole digger with a fixed-volume sampling tube, many of the limitations of fixed-volume sampling in rocky soils are overcome (Ponder and Alley, 1997).  However, sampling sandy and very rocky soils is still problematic, and powered samplers often compact the soil sample, resulting in a biased observation of bulk density (Freckmann and Baumann, 1937; Ficklin, 1997).  Furthermore, the power-auger is most manageable with cores that are approximately 30cm in length.  While 30cm does not reflect the entire plant rooting depth for a forest, it is an improvement over studies that evaluate a 15cm depth or less.  The Ozark study area and the design for measuring soil carbon quantity and variation within a given soil map unit were selected within the context of these sampling constraints. 

Soil cores were extracted using a two-person power auger technique with a ten- centimeter diameter sampling tube.  For consistency with other auger sampling projects in the Missouri Ozarks, samples were taken to a depth of 0.3m when possible (Ficklin, 1997; Ponder et. al, 1995).  Following extraction, each core was packaged in a PVC tube, and both ends of the tube were sealed with duct tape.  Cores were subsequently stored in the laboratory under refrigeration at approximately 10^º C to minimize microbial soil organic carbon oxidation.

As with limited sampling depth in previous studies, interpretation of carbon and nutrient data in the scientific literature are often convoluted by a failure to define the sampling scale (Gardner, 1998).  The three types of landforms within the study area were delineated based on the criteria outlined by Hammer (1997; 1998).  Previous research in this landscape illustrated that representative transects were difficult or impossible to establish, due to microrelief and multiple landforms (Hammer, 1997); therefore, a stratification of plots by landform was deemed appropriate for this investigation.  Additionally, the physical characteristics of the soils within the study area were expected to define the compressibility of the soil, so soils with a range of physical characteristics were sampled with the landform stratification so that bulk density estimation and measurement error could be evaluated.

The holes from which each core came were excavated and horizon thicknesses were recorded for comparison with the thickness of horizons of the core.  The observed thicknesses of each horizon were used to calculate the best estimate of the actual bulk density of the soils in the field.  Since core sampling is a fixed-volume approach to density measurement, the thickness of each horizon determines the volume for a measured mass of soil.  Errors in bulk density measurement present one challenge to the quantification of forest soil carbon. 

To assess the temporal variability of soil physical and chemical properties, it is necessary to address both temporal and spatial variability concomitantly (Hammer, 1986).  The study was conducted within the Missouri Ozark Forest Ecosystem Project (MOFEP) Site #8, which is an upland oak-pine forested tract with large variability in soil physical and chemical characteristics across short distances (Meinert, et al., 1997; Hammer, 1997).  This variability may be attributed to the concave and convex landforms associated with the upland forest landscape (Udawatta and Hammer, 1995).  Despite the presence of several landforms within the landscape of the study area, all the soils within the study area were mapped as a Clarksville series Loamy-skeletal, siliceous, mesic Typic Paleudult (Butler, 1990).  The accomplishment of the objectives of this case study project required consideration of both the characteristics of the soil populations (landforms) and the vegetation populations (overstory species).  Stratification by landform within the soil map unit was deemed the most effective approach for addressing the spatial variability within the map unit. 

Furthermore, the landform stratification permitted a range of hydrologic, structural and textural characteristics that was needed for the determination of soil carbon and bulk density magnitudes and variabilities across the soil map unit.  A randomized block design was utilized with replication for two years.  The blocks or strata for the field design correspond with landforms (structural bench, shoulder, and sink hole), seasons (summer, fall, winter, and spring), and horizons (A, AB [EB], BA [BE], and Bt).  The experiment was replicated for two years (1997-1998 and 1998-1999). 

Temporal variability was assessed by sampling every three months for two years (a total of eight sampling periods).  Sampling times corresponded roughly with the four seasons.  During each sampling period, six soil cores 10.1 centimeters in diameter were extracted to a depth of 30 centimeters from each of the aforementioned landforms (18 cores per period * 4 seasons * 2 years = 144 cores total).  Different plot areas were established for year one and year two to minimize the effect of prior sampling disturbance on bulk density and other physical and chemical soil characteristics. 

Spatial variability of soil density was addressed by limiting our sampling area to Roubidoux geologic substrata.  It also was necessary to have an unmanaged/ non-harvest area, so that anthropogenic influences could be minimized.  It was determined that Missouri Ozark Forest Ecosystem Project (MOFEP) Site #8 in Carter County, Missouri, would meet these criteria.  Within this area the scale of sampling was set to 1/10^th- acre plots in landforms of low, medium, and high coarse fragment contents.  The classification of plots as low, medium, or high rock content was based on the geomorphic features of the selected landforms.  This sampling intensity resulted in each core representing an area roughly 11 ft² or about 3.35 m².

Distributional diagnostics were performed on all variables considered for inclusion in the final analyses.  Although the large number of observations might suggest that the data should approximate normal, based on the central limit theorem (Snedecor and Cochran, 1989); stem and leaf plots and histograms revealed long right tails for some variables, including:  bulk density difference (compaction from sampling) and soil organic carbon concentration.  Soil textural and water content data exhibited no significant deviations from normal. 

A Shapiro-Wilk test for non-normality provided no evidence of significant non-normality for the bulk density difference variable, despite the apparently skewed right tail of the distribution; therefore, no transformation was necessary for this variable.  However, for data analysis a square root transformation was performed on the “bulk density difference” variable.  Both stem and leaf and histogram plots better approximated normal with this transformation.  Based on the Shapiro-Wilk test results, soil organic carbon concentrations did require a log₁₀ transformation to correct non-normality.  This suggests that the large sample size was not sufficient for the data to approximate normal in accordance with the central limit theorem with a sample size greater than 250 observations.  All statistical analyses, including ANOVAs, were performed using standard SAS analytical procedures (SAS, 2003).

Laboratory Analyses

Each core was described in the laboratory according to standardized procedures (USDA et al., 1996), and approximately four genetic horizons were delineated for each core.  The typical horizon sequence in soil cores was A, AB, BA, and Bt; however, description of pit horizons at the end of the study often included an E or transitional EB, BE, or EB/BE.  This apparent discrepancy resulted from the relatively thin zone of eluviation and the mixing of transitional horizons during power-auger sampling.  Delineating soil core horizons facilitates the estimation of horizon thickness necessary for determining sample volume, but accurate description of genetic horizons is difficult from cores taken with the power-auger.

Each sample was air-dried and passed through a 2mm sieve prior to chemical and physical analyses.  Bulk density was measured using an adjustment for coarse fragment content.  The volume of coarse fragments was determined from the mass of fragment and the mean density of fragments (g cm^-3) as determined by fluid displacement.  Antecedent water content was determined by drying the samples at 105^º C for 24 hours.  Both the difference in horizon thickness and antecedent water content were anticipated to be significant variables for modeling and estimating the adjusted bulk density of the soil.

The soil organic carbon content of each horizon was measured using a Leco C-144 carbon determinator for organic carbon combustion (Nelson and Sommers, 1996).  Reference standards were measured after every ten samples to assure that the carbon determinator maintained calibration for accuracy within approximately 0.05% of actual organic carbon content.  Nitrogen was determined using a Leco nitrogen determinator based on the procedure developed by Wong and Kemp (1977). 

Soil texture also was expected to be a significant predictor of core compaction, so particle size determination was performed using the sedimentation/ pipette method based on Stoke's Law (USDA, 1996).  The USDA size classification was used:  five categories of sand size fractions were measured:  0.05 to 0.106mm, 0.106 to 0.25mm, 0.25 to 0.5mm, 0.5mm to 1mm, and 1 to 2mm).  Similarly, two silt-size fractions were measured (0.002 to 0.02mm and 0.02 to 0.05mm), and particles smaller than 0.002mm in effective diameter were classified as clay-sized.  Coarse fragments larger than two millimeters were further separated into three size classes:  2mm to 6.3 mm, 6.3 mm to 19 mm, and everything greater than 19 mm.

The determination of cation exchange capacity (CEC) and exchangeable bases was accomplished by a sum of cations method using a sodium hydroxide digestion of samples leached with 1N ammonium acetate at a pH of 7 with a Centurion extractor. Total cation exchange capacity was determined by titration, and the contribution of base cations to CEC was measured by atomic absorption and atomic emission spectrometry, following dilution with lanthanum to prevent fixation of Ca and Mg by phosphate.  Exchangeable acidity was measured as a component of total CEC, using the BaCl₂ Triethanolamine IV extraction procedure (USDA, 1996).

Case study two: 

Arkansas coastal plain tightly-spaced experimental plantation:  site characterization and experimental design

The original objective of this project was to evaluate spacing and fertilization treatments in a “miniature plantation” setting similar to other research by Amateis and others (2003).  This approach may be helpful for understanding forest stand dynamics in a relatively short timeframe, compared with natural stands.  During the course of the experiment, variation in the rate of tree growth in parts of the study area was visually evident.  Such aboveground variation can be addressed through appropriate sampling/ inventory techniques, but the reason for the variation can be attributed largely to the variation in soil properties.  The nearly flat areas of the coastal plain provide a stark contrast to the soils in the Ozark Highlands, so comparing the variation within a soil map unit in these two geographic areas provides an illustration of the challenges of using map units as components in future soil carbon estimation models. 

The study site used in the coastal plain project is located just outside of Monticello, Arkansas, and is part of University of Arkansas-Monticello School of Forest Resources’ research forest. The study area, predominately juvenile loblolly pine (Pinus taeda L.) and sweetgum (Liquidambar styraciflua L.), was cleared and tilled prior to planting. The soil in this study area is mapped as a Calloway series soil with a Fine-silty, mixed, thermic Glossaquic Fragiudalf taxonomic classification (Larance et al., 1976). A hardpan is present at about 35 cm of depth. Ditches were installed around the study area to allow drainage during the wet winters common to this area.  This tightly-spaced experimental plantation study area is approximately 35 by 20 meters in size, nearly flat, with one convex area approximately 40cm higher than the surrounding area.  The apparent homogeneity of this area made observations of variability in both soil properties and plant biomass germane to our subsequent discussion of variation within soil map units.  Prior to the installation of the study, 24 soil samples were taken systematically across the study area, so that variability in tree growth could be considered within the context of soil fertility.  Soil analyses were conducted using the same protocols as in the Ozark forest study.

This tightly-spaced experimental plantation study utilized genetically improved loblolly pine (Pinus taeda L.) seeds were stratified and germinated during the 2000-2001 dormant season. The seeds were germinated in Jiffy® peat pellets. At approximately 2 months of age, the seedlings (in the pellets) were planted at the research site.  Seedlings were planted in plots (experimental units) consisting of a factorial arrangement of four within row spacings (10.2, 20.3, 30.5, and 50.8 cm) and four between row spacings (10.2, 20.3, 30.5, and 50.8 cm) resulting in 16 spacing combinations (or plot sizes) for use in the experiment. Each plot consisted of nine rows of nine seedlings, with the outer two rows in all directions serving as buffer trees. Therefore, just the interior 25 trees were used as observational units and averaged by plot for subsequent analysis.

Two fertilization levels (fertilized and non-fertilized) are also present within the experiment as a split plot factor. Therefore, a complete replicate consists of 16 spacing combinations and 2 fertilization levels totaling 32 plots. Three replicates were installed, so a total of 96 plots were installed and a total 7,776 seedlings were planted in the spring of 2001.  All plots were irrigated during the 2001 growing season, and periodically treated with herbicide (sulfometuron methyl, 140 grams ha^-1, and metsulfuron methyl, 35 grams ha^-1) to control competition during the 2001 and 2002 growing seasons. Additionally, the plots were periodically treated with permethrin (290 ml ha^-1) to control Nantucket tip moth (Rhyacionia frustrana [Comstock]) during the 2002 growing season.

The study was first measured in January 2002, one growing season into the experiment and prior to fertilization (200 kg ha^-1 of a 11-40-6 fertilizer in late February 2002 and 300 kg ha^-1 of the same in February 2003 ), and has been measured five times after fertilization, in April, June, and September of 2002 and January and April of 2003.  Attributes measured include the vigor (alive or dead), total height (nearest cm), height to base live crown (nearest cm), and root collar or basal diameter (nearest 0.1mm) of each seedling.  The results from the April 2003 measuring period are reported herein. 

Results and Discussion

Sequestration of carbon in forest soils begins with the fixation of atmospheric carbon by the photosynthesizing autotrophs in the forest.  The ability and capacity of plants to fix carbon is a function of many soil properties, and the prediction or estimation of biomass production has been approached from both “black-box” (site index) and process-based (productivity index) techniques.  If data for existing soil map units are to be used in future models of potential carbon sequestration, then it is critical to identify the measurable factors that are the most highly correlated with plant growth and subsequent soil carbon enrichment.  The Ozark case study provides some insights into the variability of soil properties within a soil map unit, and the coastal plain study illustrates how the variability within a soil results in variable aboveground biomass production. 

Case study one:  upland Ozark forest soil

The soil physical and chemical properties differed most between the sinkhole (low near surface rock content) landforms and the shoulder and structural bench landforms (medium and high near surface rock contents).  Differences between the medium and high rock content landforms were less striking within the soil profiles (Figure 1), although a stoneline was visible at the surface of the high rock content bench landforms, and no stoneline was apparent at the surface of the medium rock content shoulder landforms.  While the data suggest that the medium rock content landforms were stonier (Table 1), many of the surface rocks in the high rock content landforms were larger than 10cm in diameter.  Accordingly, these large fragments did not fit into the sample tube.  Rock content is an important factor to consider when determining the bulk density of soil, and quantifying soil organic carbon requires accurate density measurements.  All bulk density measurements in this study have been adjusted for rock/ coarse fragment content.



Figure 1. Soil profiles from each landform of the upland Ozark forest soil.

Errors in bulk density measurements by horizon and landform are listed in Table 2.  Predictably, sampling compaction and measurement error tended to increase as core depth increased.  Measurement error ranged from 5% to nearly 18% with no clear predictability for the compressibility of the soil.  These results indicate that multiple samples must be taken to achieve a “best estimate” of the true bulk density of soil in the field. 

Table 1.  Rock volume across landforms by horizon.

Measurements of horizon thicknesses in the field also are essential, if the magnitude of density estimation error is to be known.  An important point to consider is that the clay mineralogy on this Ozark forest site consisted primarily of kaolinitic clays.  In areas with smectitic clays, variation in bulk density with changes in soil water content presents another challenge to density estimation.

Table 2.  Bulk density estimation error by horizon and landform.

 

Once the best estimates of bulk density were determined, the mass of soil organic carbon was calculated for each landform and horizon (Table 3).  Due to the variation in horizon thicknesses across each of the landforms, the mean thickness for each horizon was calculated for each landform, and carbon contents were calculated in increments that approximated 0-15cm and 15-30cm increments.  While the subsurface quantity of soil carbon varied little across landforms, the surface carbon (mineral soil) varied by as much as 4 Mg per hectare.  Using the landform stratification within this soil map unit, the best estimate of soil carbon for this soil could be calculated by “weighting” the proportion of the map unit that consists of each of the landforms measured.  Estimating the area of each landform was beyond the scope of this project, so no estimates of the mean carbon content in this soil is provided.   

Table 3.  Quantity of soil organic carbon by depth and landform.

 

A comparison of surface carbon and subsurface carbon was performed to address one of the objectives of this study.  For each of the landforms, A and AB horizons were designated as surface horizons, and all BA and Bt horizons were designated as subsurface horizons.  If surface organic carbon concentrations were truly surrogates for subsurface carbon concentrations, as is implied in studies that sample only to a depth of 15cm, then a strong association should exist between the surface and subsurface soils.  With a total of 40 observations a general linear model procedure was run using horizon as an independent variable to predict soil organic carbon concentration.  Only 56% of the variance by horizon was explained (R²=0.559), which indicated that surface carbon was a poor surrogate for subsurface carbon.  Accordingly, litterfall inputs are not adequate for estimating soil carbon enrichment.  Additional soil cores or pits would be required to better evaluate the association of surface and subsurface carbon for soils other than those within this study area.

Although it was hypothesized a priori that coarse fragments and organic carbon would be negatively correlated, the data from this project showed a positive association between rocks and carbon similar to results reported by Schaetzl (1991).  The mineralogy of the coarse fragments was predominantly SiO₂, which is generally considered chemically inert.  The mechanism by which rocks and organic carbon interact to slow carbon oxidation is not clear, but some aspect of mineralogy, hydrology, or both appear to protect interact to protect carbon from microbial or other oxidative decomposition.  Conceptual models of soil carbon sequestration potential should incorporate oxidation potential, so that good estimates of carbon storage or efflux are achieved.

An ANOVA of the texture data for each plot yielded some interesting results.  The three landforms were significantly different (p<0.001) for clay content, and both the high rock and low rock plot were significantly different from each other.  The high and medium rock landforms were not significantly different.  The ranking of clay content by plot from lowest to highest was medium rock, low rock, then high rock.  Differences among landforms also were observed for silt content (p<0.0001), with high rock/low rock and low rock/ medium rock comparisons significantly different.  The ranking of silt content by plot was medium rock, high rock, then low rock.  Sand content also was significantly different among landforms (p<0.0001), with high/medium and low/medium comparisons significantly different.  The ranking of sand content by plot was low rock, high rock, then medium rock content landforms.

Many of the chemical properties of the soils also differed.  Salt pH (CaCl₂) varied significantly among landforms (p=0.041), but only the low/medium comparison was significantly different at a 5% level.  The ranking of salt pH was low rock, high rock, then medium rock.  Calcium, sodium, and soil organic carbon contents were not significantly different among the landforms, with p-values of 0.155, 0.990, and 0.377, respectively.  Magnesium content did differ among landforms (p=0.023), and the comparison of high and medium rock content landforms was significantly different.  The ranking of magnesium by plot was medium rock, high rock, then low rock.  Potassium varied significantly among the landforms (p<0.0001), primarily due to the significant difference between the low and medium rock content landforms.  No other comparisons were significant.  The ranking of potassium by plot was medium rock, high rock, then low rock.  Cation exchange capacity (CEC) determined from an ammonium acetate extraction was significantly different among landforms (p<0.0001), and the comparisons of high/medium and low/medium were significantly different.  The ranking of CEC by plot was medium rock, low rock, then high rock.  Interestingly the rank sums for the high rock and low rock landforms were 357.0 and 343.5, respectively- remarkably similar.  Nitrogen content also varied significantly by landform (p=0.002), with comparisons between high/medium and low/medium landforms significantly different.  The ranking of nitrogen by plot was medium rock, low rock, then high rock.

Based on these analyses, it is apparent that the three landforms are significantly different on the basis of both physical and chemical data, and these differences could be important components of a model designed to estimate soil carbon sequestration potential.  However, there is no clear trend that suggests that one landform is consistently lower or higher in all of the indicator variable categories.  The number of significant multiple comparisons by plot does suggest that the medium rock content landforms were more dissimilar from the high and low landforms than the low and high landforms differed from each other.  One possible explanation would be that the low rock content landforms were aggrading in materials from the erosional surfaces that were used as high rock content landforms. 

The overstory vegetation between areas of low and high rock contents were as variable as the hydrology and chemistry of the underlying soil strata from which the vegetation grew.  Basal area for each plot area varied from 67 ft² to 137 ft² per acre, but there is no discernable association between plot areas and basal area.  Since each plot area is only 1/10^th acre in size, the expansion factor of ten may provide a skewed estimate of stand density.  It would be inappropriate to extrapolate the basal area for the entire site from this vegetation inventory, since a full inventory cruise was not performed for the area.  Species composition and the number of trees per acre varied the most between plot areas, and this variability is an indication of the underlying differences in chemistry and hydrology between the landforms.

Examination of the overstory composition reveals an association between landforms with high surface rock content and the presence of shortleaf pine (Pinus echinata Mill.).  Reciprocally, the absence of pine was observed in areas designated as low rock content plot areas.  Grabner et al., (1997) observed differences in herbaceous vegetation on the MOFEP sites, including the site considered in this investigation.  The authors asserted that vegetation was associated with differences in geology, landform, and soils across the study areas. 

While similar patterns of overstory vegetation by landform were observed, the analyses of the soils of these landforms offered little explanation for the compositional difference on the basis of texture and chemistry alone.  The similarity of CEC between the high and low rock content landforms is one example of apparent contradiction between the potential to support pine versus hardwoods.  Moreover, the salt pH was actually lower in the low rock content landforms than in the high rock content landforms, indicating more acidic conditions under herbaceous vegetation than under pine vegetation, when the entire profile is considered. 

Anecdotal historical evidence suggests that the high rock content landforms were used for hay production in the past, so the effects of prior land use could account for the relationships of pH, landforms, and vegetation.  Stone (1975) noted that management practices may be as important as species for the observed changes in soil pH and productivity over time.  This position is supported further by the findings of France et al., (1989) that the buffering capacity (CEC) of a soil is a key to the rate of change in soil pH under various types of vegetation.  It is clear that vegetation and soils have reciprocal influences on their respective physical and chemical properties, and several studies have examined the influences of leaf chemistry on decomposition rates, changes in pH, and rates of podzolization (Miles, 1985; Brand, et al., 1986; Cote and Fyles, 1993).  However, the complexity of soil-plant interactions depends upon too many factors to make broad generalizations about vegetation composition from a “snapshot in time” of soil properties.  All of these interactions through time contribute to the difficulty of forest soil carbon measurement and subsequent valuation of forest management practices.

Case study two:  Arkansas coastal plain tightly-spaced experimental plantation

Since the purpose of this study was to evaluate the effects of spacing and fertilization on tree growth, extensive soil chemical and physical data are not available.  However, the variation in antecedent soil chemical properties was not statistically significant in the 24 soil samples systematically extracted across the study area.  Hydrologic differences across the study area were the most notable influences on plant growth.  Microscale differences in elevation resulted in periodic ponding in some portions of the area with other portions only meters away that were dry enough to result in drought stress on the trees. 

Given the small area on which this project was installed and the absence of significant variability in soil fertility, the variation in tree growth across all treatments and replications was striking.  An analysis of the coefficients of variation for all surviving trees in each spacing and fertilization combination based on the April 2003 measurements is evidence of the influence of soils on plant growth.  Basal diameter (mm), and total height (cm) were highly variable, both within and between replicates (Figures 2 and 3).  Figure 2 depicts the average of the 32 plot-level coefficients of variation for each response by replicate.   Much within-plot variation is evident.  The coefficients of variation for treatment combinations within replicate one are the greatest, and the variation is readily visible in the field.  Figure 3 depicts coefficients of variation of the replicate-level average responses.  These values indicate the level of variation present between the replicates.  The amount of variation present at the plot-level as well as the replicate level is surprising, given the small area encompassed by the study.

Figure 2.  Mean coefficients of variation by replicate for three measures of tree growth:  basal diameter (B-D), total height (T-H), and height to live crown (H-L-C)

Although the soil in this plantation does have a restrictive layer at approximately 35cm, according to the soil survey, the depth to the pan was variable across the plot.  As a result, rooting restrictions differ from plot to plot and between replicates.  From the perspective of aboveground biomass measurement, variability of stand density and growth can be addressed by proper sampling stratification.  However, belowground root biomass and soil organic carbon are not as easily measured.  Site index values reported in the soil survey are of limited value for assessing potential productivity on scales below that of the soil map unit area. 

Figure 3.     Coefficients of variation across replicates for three measures of tree growth:  basal diameter (B-D, mm), total height (T-H, cm), and height to live crown (H-L-C, cm)

Accounting for variation in soil and plant carbon on areas as small as this research site probably is beyond the ability of any model applied to landscape scale carbon estimates.  The use of small areas, similar to this study area, for determining which factors are key determinants of potential carbon sequestration may be one approach to defining components that are keystones for an improved conceptual model.  Before any empirical model can be employed to estimate belowground carbon pools and their fluxes, a better conceptual model is needed. 

Conclusions

Opportunities for Improving Forest Carbon Accounting

Many factors influence the capacity of a forest soil to sequester carbon, and an equally large number of factors inhibit our ability to accurately measure the quantity of carbon sequestered or the rate of carbon flux.  From the current developments in carbon accounting policy, it is clear that more effort is needed to assure that the value of forest soils as a significant sink for soil carbon is not overlooked.  To achieve this end, scientists must be proactive in policy-making, and we must continue to improve our understanding of carbon dynamics in terrestrial, particularly forest, ecosystems.  The case studies presented in this manuscript highlight some of the challenges to applying our “coarse resolution” knowledge of carbon fluxes to areas as small as soil map units.  However, with a better understanding of the processes underlying carbon fluxes, improvements in our knowledge resolution will improve.

Under a multiple-use forest management scenario, both commodity and non-commodity [intrinsic] values must be considered.  If carbon sequestration becomes a societal priority, and the conference at Kyoto, Japan, was a first attempt to control atmospheric CO₂, then many forested areas of marginal value for timber or other agricultural productivity may be better managed as pools for atmospheric carbon storage.  Such use would serve to maximize the net societal benefit from land management.

From a traditional forest production context, maintaining or increasing soil organic carbon offers a means to improve the fertility of many forested lands.  The valuation of carbon sequestration is another benefit that forest landowners may soon realize from their forest assets.  Landscape-scale deficiencies in essential nutrients that influence biomass production and carbon storage can better be identified with multiple samples from landforms within a given soil map unit.  It is hoped that this information will improve our understanding of many forest soil chemical and physical properties in addition to soil organic carbon dynamics.

The evaluation of surface carbon as a surrogate for subsurface carbon concentrations indicated that surface concentrations of carbon are poor predictors of subsurface carbon.  Only 56% of the variability in subsurface carbon was explained by surface carbon, based on the data derived from the soil pit samples.  The relatively shallow 30cm sampling depth for soil cores made comparisons of surface and subsurface carbon inappropriate, since the difference in surface and subsurface depth was approximately 15cm for most observations.  Better modeling of soil carbon enrichment throughout the range of tree rooting depth is needed to estimate subsurface carbon sequestration following reforestation or afforestation activities.  Better information on the types and fates of organic compounds in the rhizosphere also is needed, so that the persistence of carbon following management activities can be better predicted.

In order to develop and implement policies related to carbon sequestration in forest ecosystems, several things should be done to integrate science into policy:

1. Carbon sequestration policies need to be refined to clearly define appropriate terminology and enumerate the role of soils in carbon sequestration valuation.

2. Although accounting for belowground carbon is a challenge, the soil carbon pool should not be ignored since it offers great potential for a significant role in carbon policies.  Modeling appears to be the most promising approach to overcoming measurement challenges.

3. More research needs to be conducted on the mechanisms of soil carbon enrichment, so that process level understanding can be achieved.  The types of research needed include studies of plant species-specific propensities for soil carbon enrichment, mechanisms by which carbon forms change in the soil due to microbial activity, and identification of soil properties that best predict sequestration of carbon in soils.

While it is clear from the case studies presented, as well as other investigations, that broad generalizations about soil carbon based on soil map units would be inaccurate, it may be possible to utilize some soil survey data within a revised conceptual model based on rooting sufficiency.  If such a “potential carbon sequestration index” can be derived for soil map units, then policy makers will be given the scientific backing that is needed for equitable and efficient policy implementation.  Moreover, since previous work already has been done to estimate potential plant biomass production based on rooting sufficiencies, it may be possible to develop models or indices that estimate both above and belowground carbon sequestration for a given soil map unit.  While such a model may be difficult or cost-prohibitive to develop, the benefits that could be realized from such a scientific and policy-making tool would be profound.

Acknowledgements

The authors would like to extend their sincerest appreciation to the many research technicians and graduate students that assisted with field research.  Additionally, we wish to thank the Missouri Department of Conservation, the Arkansas Agricultural Experiment Station Research Incentive Program, and the Arkansas Forest Resources Center for providing funding for this endeavor.  We also extend our thanks to the International Paper Company for donating hybrid loblolly seed. 

References

Amateis, R.A., M. Sharma and H.E. Burkhart. 2003. Using miniature scale plantations as experimental tools for assessing sustainability issues. Canadian Journal of Forest Research. 33:450-454.

Binkley, C., D. Brand, Z. Harkin, G. Bull, N. Ravindranath, M. Obersteiner, S. Nilsson, Y. Yamagata, and M. Krott. 2002. Carbon sink by forest sector—options and needs for implementation. Forest Policy and Economics 4: 65-77.

Brand, D.G., P. Kehoe and M. Connors. 1986. Coniferous afforestation leads to soil acidification in central Ontario. Can. J. For. Res. 16:1389-1391.

Butler, E.R. 1990. Soil survey of Carter County, Missouri, northern and eastern parts. USDA-SCS and Forest Service in cooperation with the Missouri Agricultural Experiment Station.  U.S. Government Printing Office. Washington, D.C.

Carmean, W.H. 1975. Forest site quality evaluation in the United States. Adv. Agron. 27:209-269.

Cathcart, J. 2000. Carbon sequestration: a working example in Oregon. J. For. 98(9): 32-37.

Cote, B. and J.W. Fyles. 1993. Nutrient concentration and acid-base status of leaf litter of tree species characteristic of the hardwood forest of southern Quebec. Can. J. For. Res. 24:192-196.

Enzinger, S., and C. Jeffs. 2000. Economics of forests as carbon sinks: an Australian perspective. J. For. Econ. 6(3): 227-249.

Ficklin, R.L., E.G. Dunn and J.P. Dwyer. 1996. Ecosystem management on public lands: an application of optimal externality to timber production.  Journal of Environmental Management 46(4):395-402.

Ficklin, R.L. 1997. Reduction of timber production externalities by using an alternative harvesting technique to minimize soil and stand damage. Masters Thesis. University of Missouri, Columbia. pp.145.

France, E.A., D. Binkley and D. Valentine. 1989. Soil chemistry changes after 27 years under four tree species in southern Ontario. Can. J. For. Res. 19:1648-1650.

Freckmann, W. and H. Baumann. 1937. Zu den grundfragen des wasserhaushalts imboden und seiner erforschung.  Bodenk. U. Pflanzenernahr. 2:127-168.

Gale, M.R. and D.F. Grigal. 1988. Development of a soil- and root-based productivity index model for trembling aspen.  In: Proceedings of forest growth:  Process modeling in response to environmental stress.  IUFRO Symposium. Gulf Shores, Alabama.

Gardner, R.H. 1998. Pattern, process, and the analysis of spatial scales. P. 17-34. In Peterson, David L. and Thomas V. Parker (ed.) Ecological Scale- Complexity in Ecological Systems/ Theory and Applications. Columbia University Press.

Grigal, D.F. 1984. Shortcomings of soil surveys for forest management. Pp. 148-166. In: Bockheim, J.G., ed., Forest land classification symposium. Department of Soil Science. University of Wisconsin. Madison, WI.

Hammer, R.D. 1986. Soil variability, soil water, and forest tree growth on three forested landtypes on the Mid-Cumberland Plateau.  Ph.D. Diss., Univ. of Tennessee, Knoxville (Diss. Abstr. 8708772).

Hammer, R.D. 1997. Missouri Ozark forest soils: perspectives and realities. In: Proceedings of the Missouri Ozark Forest Ecosystem Project Symposium. USDA Forest Service, North Central Forest Experiment Station. General Technical Report NC-193: 106-121.

Hammer, R.D. 1998. Space and time in the soil landscape: The ill-defined ecological universe. P. 104-140. In Peterson, David L. and Thomas V. Parker (ed.) Ecological Scale- Complexity in Ecological Systems/ Theory and Applications. Columbia University Press.

Henderson, G.S., R.D. Hammer and D.F. Grigal. 1990. Can measurable soil properties be integrated into a framework for characterizing forest productivity? In: Gessel, S.P., ed., Sustained productivity of forest soils. Forestry Publications, Faculty of Forestry, University of British Columbia, Canada.

Henderson, G.S. 1995. Soil organic matter: a link between forest management and productivity. In: Bigham, J.M and J.M. Bartels (eds.), Carbon forms and functions in forest soils. Soil Science Society of America.  Madison, WI.  Pp. 419-435.

Hoover, C., R. Birdsey, L. Heath, and S. Stout. 2000. How to estimate carbon sequestration on small forest tracts. J. For. 98(9): 13-19.

Hsieh, Y. 1992. Pool size and mean age of stable soil organic carbon in cropland. Soil Sci. Am. J. 56:460-464.

Kiniry, L.N., C.L. Scrivner and M.E. Keener. 1983. A soil productivity index based upon predicted water depletion and root growth. Missouri Agri. Expt. Sta. Res. Bull. 1051. Columbia, MO.

Korner, C. 2003. Slow in, rapid out-carbon flux studies and Kyoto targets. Science 300:1242-1243.

Lackner, K.S. 2003. A guide to CO₂ sequestration. Science 300(5626):1677-1678.

Larance, F.C., H.V. Gill and C.L. Fultz. Soil survey of Drew County, Arkansas.  USDA-SCS in cooperation with the Arkansas Agricultural Experiment Station.  U.S. Government Printing Office.  Washington, D.C.

Marland, G., B. McCarl, and U. Schneider. 2001. Soil carbon policy and economics. Climate Change 51: 101-117.

Matthews, S., R. O’Connor, and A. Plantinga. 2002. Quantifying the impacts on biodiversity of policies for carbon sequestration in forests. Eco. Econ. 40: 71-87.

MacCleery, D. 1994. American Forests: A History of Resiliency and Recovery. Forest History Society Issues Series. Durham, NC. Pp. 58. 

Meinert, D., T. Nigh, and J. Kabrick. 1997. Landforms, geology, and soils of the MOFEP study area. In: Proceedings of the Missouri Ozark Forest Ecosystem Project Symposium. USDA Forest Service, North Central Forest Experiment Station. General Technical Report NC-193: 56-68.

Miles, J. 1985. The pedogenic effects of different species and vegetation types and the implications of succession. J. Soil Sci. 36:571-584.

Murray, B., S. Prisley, R. Birdsey, and R. Sampson. 2000. Carbon sinks in the Kyoto Protocol: potential relevance for U.S. forests. J. For. 98(9): 6-11.

Nelson, D.W. and L.E. Sommers. 1996. Total carbon, organic carbon, and organic matter. In: D.L. Sparks, et al. (eds.) Methods of Soil Analysis, Part 3- Chemical Methods. Soil Science Society of America- Book series number 5:961-1010.

Paul, K.I., P.J. Polglase and G.P. Richards. 2003. Predicted change in soil carbon following afforestation or reforestation, and analysis of controlling factors by linking a C accounting model (CAMFor) to models of forest growth (3PG), litter decomposition (GENDEC) and soil C turnover (RothC). For. Ecol. And Mgmt. 177:485-501.

Ponder Jr., F. and N. M. Mikkelson. 1995. Characteristics of a long-term forest soil productivity research site in Missouri. Proceedings of the 10^th Central Hardwood Forest Conference. USDA Forest Service general Technical Report NE-197. pp. 272-280.

Ponder, F., Jr. and D.E. Alley. 1997. Soil sampler for rocky soils. USDA For. Serv., North Central Forest experiment Station, Res. Note NC-371. 5 pgs. 

Ruth, M., B. Davidsdottir, and S.Laitner. 2000. Impacts of market-based climate change policies on the U.S. pulp and paper industry. Energy Policy 28: 259-270.

SAS Institute Inc. 2003. User’s Guide for SAS Software Navigator. Cary, NC, USA.

Schaetzl, R.J. 1991. A lithosequence of soils in extremely gravelly, dolomitic parent material, Bois Blanc Island, Lake Huron. Geoderma 48:305-320.

Snedecor, G.W. and W.G. Cochran. 1989. Statistical Methods. Iowa State University Press. Ames, Iowa. Pp. 503.

Soil Science Society of America. 2001. Carbon sequestration: position of the Soil Science Society of America.  Burras, L., chair of Ad Hoc Committee S893.  Approved by SSSA Board of Directors October 25, 2001.  

Stainback G., and J. Alavalapati. 2002. Economic analysis of slash pine forest carbon sequestration in the southern U.S. J. For. Econ. 8: 105-117.

Stone, E.L. 1975. Effects of species on nutrient cycles and soil change. Phil. Trans. R. Soc. Lond. B. 271:149-162.

Stone, E.L. and P.J. Kalisz. 1991. On the maximum extent of tree roots. For. Ecol. Manag. 46:59-102.

Udawatta, R.P., and R.D. Hammer. 1995. Variation in soil organic carbon due to surface shape and topography in forested watersheds. Agronomy Abstracts 1995. Annual Meetings. St. Louis, Missouri. Oct 29 to Nov. 3, 1995. p 307.

USDA, NRCS, National Soil Survey Center. 1996. Soil Survey Laboratory Methods Manual. Soil Survey Investigations Report No. 42 version 3.0. pp. 693.

U.S. Department of Energy. 2003. Voluntary Reporting of Greenhouse Gases 2001. DOE/EIA-0608(2001). Washington, DC.

Wong, H.K.T. and A.L.W. Kemp. 1977. The determination of nitrogen in sediments using an induction furnace. Soil Sci. 124:1-4.