CMRT Home > Projects > Geochemical Landscapes > Forum Comments

Geochemical Landscapes

Geochemical Landscapes Forum

Project Home Tasks Products Contacts Links

We welcome any comments and suggestions regarding the Geochemical Landscapes Project. The comments below have been submitted from several interested parties since the Soil Geochemistry Workshop.

Comments from Sam Bass, U.S. Army Corps of Engineers:

Here are my notes on Corps of Engineers' needs/wishes for a nationwide soil geochemistry data set.

Typically we must consider exposure risks in the top 0-6" or 0-12" of soil. Sometimes we must also consider risks in the upper 10 feet of soil, to allow for excavation of full basement and footers. Therefore the background data set should include samples from surface soils and at least one sample from around 5-6 feet in depth. The program we outlined last week (based on collecting samples from the A and C soil horizons) should cover this requirement.
Background values are used to determine if there has been a release at a site (i.e., if site concentrations exceed background concentrations). By CERCLA law we are not allowed to remediate a site below background concentrations. Therefore, cleanup levels at a site may be based on background.
We typically use the Shacklette and NURE data as a qualitative reality check for our site-specific data; as screening level data where no other data exists; or to complement site-specific data (to see if the ranges of each data set are comparable).
We usually collect background samples for each site; sometimes we collect background for an entire installation. This wasn't always the case, but is now pretty much SOP.
National data sets such as the proposed soil geochemistry study are indeed useful to USACE because of our nationwide mission.
USGS/NRCS are the best agencies to complete the proposed work because they are impartial and independent. Therefore the data generated by them is seen by regulatory agencies as unbiased and reliable.
We must ensure that the quantitation limits of the analytical methods used to evaluate organics in the soil samples are low enough to allow meaningful comparison to regulatory screening levels, i.e., they must be lower than the screening levels wherever possible.
A high-resolution soil geochemistry data set would be useful to identify regional and sub-regional trends in chemical distribution, soil types, and provenance, and relationships between these three parameters.
Products that would be most useful to USACE include the raw data sets in Excel or Access (to allow separate analysis of data near a site along with site-specific data); interpretive reports to help identify trends in the data (e.g., As trends in New England); and discussion of geochemical ratios / relationships between elements. Some work done by Shaw Environment and Infrastructure (formerly IT Corporation) has shown similar element:element ratios regardless of sample locality. In particular, V/Fe and Co/Mn ratios plot on a straight line when looking at samples from locations in GA, UT, FL, AL, and NC.
It would be good to include analysis for PCBs, PAHs (polynuclear aromatic hydrocarbons), pesticides, and radioactive constituents (specifically U, Th, and Ra), along with metals. It would also be helpful if the sample preparation and analysis methods were consistent with EPA SW-846 methods (e.g., Method 8270 for semi-volatile organics, which would include the PAHs). The aggressive digestions proposed at the meeting (including use of HF acid) may be overkill for our risk assessment needs since the digestion may liberate elements (metals) that typically would not be bioavailable. I understand the need for the HF digestion when doing whole-soil analysis, but it may be too much of a good thing for risk assessment. I shudder to think this may lead to collection of two samples - one for whole soil analysis, and one for risk assessment (bioavailability) analysis using the EPA sample prep methods. Alternatively, maybe one sample can be prepared using two different methods. This would allow some comparison between the different prep methods. Who knows, the sample prep methods discussed by Jim Crock may not be that different from the EPA methods. I'm just a geologist. The details of the lab bench work are outside my box.
A two-tier sampling approach may be appropriate. The first level is collected on a coarser grid and is used to identify regional trends. The second level is collected on a finer grid and is used to flesh out trends identified in the first-level sampling. Thanks for the opportunity to provide input. If you have any questions, give me a call or drop me a line.

Comments from Ken Sylvester, Inter-University Consortium for Political and Social Research (Univ. of Michigan):

I was struck by the contemporary nature of the instrument envisioned. The thoughts I expressed were related to the potential to integrate human and historical dimensions with the data set being designed.

There are currently efforts underway to improve the historical county boundary maps of the United States from 1790 to the present, that will dramatically improve our ability to employ historical census information in a spatial design resolved at the continental scale. http://www.nhgis.org/ What I hoped to suggest to the design team in Denver was that if legacy effects (evidence of anthropogenic effects in the past) were too complex in terms of soil sampling and analysis, that the scale of the sample design be appropriate for integration with land use history and other human dimensions that have been measured since 1790 in varying degrees of detail. Ecologists and social scientists will be increasingly interested in the soil geochemistry data as an outcome of long term change, at least at the regional scale that Clemens addressed, to see if there are statistical relationships between land use change and current background levels.

The spatially random hexagonal sample design that Bob Garrett set out on Thursday morning satisfies the crucial continental scale requirement, is elegant, allows for some nesting, and most importantly, does not require the use of prior knowledge that is practically non-existent for most of Canada and Mexico.

If this is to be a continental scale effort, and one that can incorporate human dimensions, and be repeatable in the future, the spatially random design is crucial.

Comments from Mark Evans, Agency for Toxic Substances and Disease Registry:

In lieu of providing detailed comments on the workgroup summaries I would like to offer some general observations about the workshop and the proposed continental scale sampling program. I thoroughly enjoyed the workshop and got a lot of valuable information out of many of the presentations. However, I see little potential value in the sampling program as proposed. As proposed, it will re-do prior sampling programs, but add relatively little value to the existing data sets (other than additional non-spatially resolved data points).

The general approach appears to be typical of many, if not most, sampling programs, which is to collect data and then try to determine how it can be used. In order to be useful, I believe this program must make a much larger effort to define the questions which the data may answer. This program would be well served by a very ambitious DQO effort. In order for a DQO process to work, there must be a well defined target audience for the data and a set of specific questions and required statistical confidence. I believe that if you do define a target audience and set of questions, you will find a very valuable role for this program.

During the course of the design group discussions, Randy Maddelena (sp?) offered a very good suggestion to employ Bayesian statistics. This would entail using the existing data as prior knowledge and using the proposed sampling to test and refine that knowledge. Anything less will just provide another, spatially limited data set. I believe that a detailed evaluation of the existing data should also occur before additional sampling takes place. The existing data should be used to develop a series of hypotheses which may be regional or local in scale (detailed evaluation of this data would make an excellent dissertation topic).

During the course of the discussion in the Sample Design Breakout Group, Bob Garrett indicated that the purpose of the sampling would be to develop continental scale average concentrations. Ultimately, other than a few academics, I believe there is very little support for a continental scale program. While there may be a large audience for such numbers, I really doubt it.

As I mentioned to you in Denver, I find this proposed program very interesting and would actively participate in helping to develop a set of useful data quality objectives.

Mark Evans

Additional Comments from Mark Evans

Considering that Forestry Service and NRCS already have programs addressing their needs, it seems that human health and risk assessment are the most significant remaining audiences. That may be my own, limited, worldview.

If, however, you decide that the health/risk audience is a primary target, I would suggest convening a series of workshops in each EPA region to help define the study questions and help you structure data collection in ways that will answer specific data needs. This effort should occur after you have done the initial data evaluation of the Shacklette data sets and generated some preliminary hypotheses. This will also allow you to tie in to much of the site specific background data that states and EPA have already collected.

I realize that at the workshop the idea of integrating data was shot down by the lab guys, however, I have worked with a lot of environmental data sets and sampling error is always so much greater than analytical error that I think those limitations can be overcome. The lab guys tend to believe that a measured value of 5 actually means 5 instead of some probability distribution around 5. There is a great deal of information in a large number of imprecise data points. Also, there are so many dimensions to the problem related to land use and alterations that tapping into other data sets is probably the only way to approach that complexity.

I think that fundamentally, you must decide whether the project should be at a continental scale, or should develop a standard protocol that can be applied to any scale. If the primary audience is human health/risk, the spatial component needs to focus on local and regional scales with the continental scale addressed secondarily by integration of smaller scale projects. It also might be very valuable to ask someone with knowledge of Bayesian statistics to compile a 2-3 page summary of how they could be applied to the program. I am not exactly sure how the Bayesian models would apply to spatial patterns, etc. but it sure would be fun to find out.

Comments from Mike Amacher, U.S. Forest Service:

Further Thoughts on the Soil Geochemical Survey

The question of scale: Scale questions seemed to occupy the thoughts of many at the Soil Geochemical Survey Workshop. Many wanted to develop a continental-scale geochemical survey, while others wanted to focus on smaller scale projects to address human health issues and to develop a pilot project. It seems to me that if a continental-scale geochemical survey is one of the main objectives of this project, then it should be designed with that in mind from the beginning. There are a myriad of past and current local-scale geochemical surveys to address site cleanup issues and more localized human health issues, but these are usually too narrowly focused to serve as a template for a landscape-scale project. One can always implement a landscape-scale project at a smaller scale as a pilot and build in stages (I tend to favor this approach), but the overall project design should have the continental-scale objective firmly in place at the outset. I think it would be easier to design in local-scale projects as part of an overall continental-scale project, then to try to scale up an existing local project. I recommend a phased approach beginning with a pilot project at some local scale and then build from there, but the continental-scale design is the driver from the beginning. The forest soil quality indicator was piloted for about a decade before it was moved into the FIA program, but national implementation was designed into the project from the start.
Sampling design: There seemed to be two overall schools of thought on the sample design. One group tended to favor a probabilistic approach, which mandated a grid-based system, so that unbiased population estimates could be obtained. Another group tended to favor a more deterministic approach and advocated a stratified design based on prior knowledge. Since determining the status of the soil geochemical landscape was one of the stated main objectives of the proposed project, the best way to achieve this is by determining unbiased population estimates and a grid-based design is a powerful tool to achieve those ends. The problem with a stratified design is that there are simply too many ways to stratify the sampling design that may lead the project into too narrow a focus. Some of the more popular stratifications discussed included soil taxonomic units, ecoregions, and hydrologic units, but without demonstrated knowledge that these actually influence the trace element status of the broader geochemical landscape, this approach may not lead to an accurate assessment of that landscape. Better to use a design which assumes no a priori knowledge of the landscape. Two of the powerful attributes of the grid-based approach are that unbiased population estimates can be obtained, which will provide an accurate snapshot of the true geochemical landscape, and that post data-collection stratification prior to data analysis is still possible and indeed is encouraged for hypothesis testing. With a well-designed sampling scheme based on a grid across most if not all the main stratifications (ecoregions, soil map units, hydrologic units, etc.), enough data can be obtained to test these stratifications, but they don't drive the sampling scheme.

Another powerful feature of a grid-base system is that it can be adjusted to fit any scale ranging from a landscape-scale continental project to a localized scale to address a particular local geoenvironmental issue. For example, sampling hexagons can be very large at distances removed from population centers or can be very small and numerous at population centers. This is the basis for the sampling grid used for the FIA ozone indicator. Also, hexagons can be sampled along climate transects, or any other type of transect. Transect sampling was advocated by the topical studies group.

One of the arguments advanced against grid-based sampling is that you don't know the spatial integrity of the sample. If you want a representative sample of a spatial element on the landscape, then the entire spatial element needs to be sampled (a composite approach works well here). This argument dissolves, however, if the spatial element being sampled has no influence on trace element content. Since there is already a large sample library of soils sampled by pedon, it is possible to test the influence of soil classification and soil forming processes on the trace element content of soils by analyzing all the pedon-based soil samples. If, as I suspect, there is as much variation in trace element content within a soil series as across soil series (or any other taxonomic unit), then there is little utility to stratifying the soil landscape by soil classification prior to sampling. Better to implement a grid-based system and obtain unbiased population estimates. Any spatial element on the landscape, whether it be a hexagon, a soil taxonomic map unit, an ecoregion, or hydrologic unit can be sampled in such a way as to reduce localized variation if so desired via subsampling and/or compositing. I would caution against over sampling at the localized scale, however, if the goal is to obtain a landscape-scale snapshot of soil geochemical status. I think it is almost always better to sample more hexagons than to sub-sample within a hexagon if that is your sampling unit.
Depth-based sampling: Some advocate depth-based sampling, whereas others want to sample by soil horizon. Certainly, if one of the goals is to study the effects of soil formation on trace element content, then sampling by soil horizon would be necessary, but this brings with it a host of logistical problems. Not all soil horizons are found in all soils, which makes implementation of a continental-scale sampling problematic. Furthermore, many soil horizons are weakly expressed if present and require more specialized training to recognize. Additionally, considerable judgment is exercised in delineating horizons to sample. Also, sampling by soil horizon usually requires digging soil pits, which is very time consuming and labor intensive. During piloting of the soil indicator in the Forest Health Monitoring (FHM) program, we sampled the mineral soil by horizon. This led to numerous horizon recognition problems because trained soil scientists could not be on all field crews. Having soil scientists on all field crews is impossible to implement at the national level because it is too cost prohibitive. Because not all soil horizons are present everywhere, are weakly expressed in many areas, and occur at variable depths, we abandoned sampling mineral soils by horizon and went to depth-based sampling. We still sample the forest floor (O horizon) separately from the mineral soil, but we no longer sample the mineral soil by horizon. This allows the crews to focus more on obtaining a clean separation between the forest floor and the underlying mineral soil. If no forest floor is present, then, of course, it is not sampled.

A recent carbon re-measurement study by the Forest Health Monitoring program (check their website for availability of this report) showed that different results are obtained for horizon sampling vs depth sampling. This is not surprising since C is very much depth stratified in soils. However, for purposes of status and trend sampling, it really doesn't matter which approach you adopt as long as it is consistent. For this reason plus the logistics, we went with a depth-based approach. Another key finding of this study was that spatial variability (as expected) was very much scale dependent. That is, the least amount of variability was found within subplots, variability among subplots was the next largest in magnitude, and variability among plots was the largest. For this reason, I think it is better to add more hexagons and do less sub-sampling or at least composite at the local scale.

The sampling scheme recommended by the sampling group at the workshop is a combination depth and horizon-based sampling protocol which included the O horizon (where present), 0-5 cm top of the mineral soil layer, composite of the surface mineral layer, best-expressed B horizon, and parent material. Sampling the deeper layers would almost certainly require digging soil pits with accompanying time and cost requirements. This is necessary if the parent material is to be sampled. Many sections of the continent will not have B horizons, expressed or otherwise. If the project is confined to sampling the top 20 to 30 cm of mineral soil (in addition to O horizon if present), then soil core sampling will be much faster and cheaper than sampling the parent material from soil pits.
Analytes: If it is possible to implement a full analytical suite, I recommend using lithium metaborate fusion followed by ICP-AES analysis to determine the macro-elements, four-acid digestion followed by ICP-MS for most trace elements, and hydride-AAS for the hydride elements. Analysis for persistent organic compounds would be a big plus to the program. I would like to see some ancillary analysis included in the program including soil bulk density, water content, coarse fragment content, particle-size analysis, soil pH, total organic and inorganic carbon, total nitrogen, exchangeable cations (Na, K, Mg, Ca, Al and cation exchange capacity), extractable metal oxides and associated elements, and soil mineralogical analysis by XRD.
Project name: I'd like to suggest that an alternate name for this project be considered. The name "National-Scale Soil Geochemical Survey" may be too technical and may not capture the imagination of the public and funding entities. I would suggest something like the "North American Soil Quality Census". The North American part of the name immediately telegraphs that this is a continental-scale project. Although the terms quality and census tend to be somewhat value-laden and nebulous, the FIA program has enjoyed considerable success by promoting itself as the nation's forest resource census. A census implies enumeration and is something most understand. Also, most people have a grasp of what the term "water quality" implies so taking a page from their book and substituting soil for water implies that this would be a large scale assessment of the quality or health of the soils of North America with a geochemical emphasis.

Comments from Kent Carlson, Ligia Mora-Applegate, Tom Angus:
State-based Considerations of Soil Geochemistry

Kent Carlson, Maryland Department of the Environment; Ligia Mora-Applegate, Florida Department of Environmental Protection; Tom Angus, Massachusetts Department of Environmental Protection

How is Soil Geochemistry Utilized by our Agencies?

Soil geochemistry data generated by federal, state, or local interests are utilized in a variety of manners by State agencies. These uses can be lumped into three categories depending on the nature of their uses: Reactive, Proactive, and Outreach.

Reactive - Description of soil geochemistry in reaction to an incident - One of the primary uses for geochemical information is to reactively describe "background" conditions at hazardous waste sites. Clarification of site-specific geochemical conditions enables the creation of remediation goals or cleanup standards to which contaminated site geochemistry can be compared. Background geochemical conditions generated from reactive investigations can also be utilized to understand regional backgrounds. On a broader scale, such data can also form the basis of reactive human health-based risk assessments. With both uses, comparison of source geochemistry to "natural background" or risk-based background values ensures that potential human health implications are fully considered.
Proactive - Description of soil geochemistry prior to an incident - Extant geochemical information is utilized when considering terrestrial placement of waste from industrial, academic, agricultural, and municipal processes, road building, and dredge projects. As with reactive settings, concentration-based geochemical data from sources and/or sites ensures that human health-based concerns resulting from placement are considered. Geochemical information is also utilized when determining beneficial uses for the reuse of waste materials (http://www.mass.gov/dep/recycle/laws/310cmr19.htm#060).
Outreach - Geochemical information is also utilized in public outreach efforts. Stakeholder inquiries are referred to Shacklett data and/or soil geochemistry experts in academic, state, or federal positions. The lack of a central repository for geochemical information for many states, however, makes communication of local data challenging.

What Sources of Geochemical Data Do States Utilize?

States utilize a variety of geochemical sources to address a multitude of issues. These sources can be grouped into three general categories:

Hazardous Waste Site Investigations - Soil geochemistry from State or Federal investigations are utilized by states to address site background conditions. States have also extrapolated this data to estimate regional conditions (http://www.mde.state.md.us/assets/document/hazcleanup_Aug2001.pdf [PDF file, 1.9 MB]).
Shacklett (1984) Database - The States interviewed primarily use the Shacklett data set (http://tin.er.usgs.gov/ussoils/) for stakeholder outreach and to validate other background soil geochemistry data. The limited numbers of samples precludes its use entirely in a few States.
Special Studies - Special geochemical studies range from complete statewide soil geochemistry surveys (http://www.hinkleycenter.com/publications/back_conc_flsoils_99-7.pdf [PDF file, 3.5 MB]) to special urban geochemistry studies that target arsenic or lead (http://lqma.ifas.ufl.edu/PUBLICATION-subject.html#Arsenic%20background%20concentrations%20in%20Florida%20soils). States also utilize compiled data from a variety of sources to generate background soil geochemistries for metals and Polycyclic Aromatic Hydrocarbons in natural soil and soils contaminated with wood and coal ash (http://www.mass.gov/dep/cleanup/laws/backtu.pdf [PDF file, 116 KB]).

Desired Products from a High Resolution Geochemical Analysis of Soils in the U.S.

Specific products from the high-resolution geochemical soil analysis can be integrated into State's current regulatory systems. These products can be lumped into three categories depending on the nature of their utility: Data, Analysis, and Risk Assessment.

Data

Raw data sets and complete documentation (analytical methods, detection limits, quantitation limits, sampling protocols, exact locations, etc.). Access to raw data is critical in order to adapt it to our state-specific purposes. Also, budgetary constraints may preclude access to GIS systems. 2) GIS data files. 3) Interpolated data sets (provide data distributions for potential inclusion in Monte Carlo analyses). 4) Publications on data limitations and use. 5) Identification of a central repository(ies) for soil samples and geochemistry data.

Analysis

Publications on population distributions (normal, lognormal, etc.) for geochemical constituents within each geographic province or ecoregion as applicable. 2) Publications on the statistical evaluation of new geochemical data versus "background" geochemical information generated in other investigations (CERCLA/NPL/Federal Facility investigations, State/local investigations). 3) Publications that report trend analyses for geochemical constituents in soils.

Risk Assessment

Publications on the relative bioavailabilities for various geochemical constituents for children and/or adults (or develop a bioavailability translator). Partnerships with toxicology testing facilities may have to be explored to fulfill this product request. 2) Publications on the bioavailability/bioaccumulative potential for plants in soils with different geochemical consistencies.

Specific Needs or Recommendations for a High Resolution Geochemical Analysis of Soils in the U.S.

Consideration of the recommendations below would maximize State-based utility of products created by a high-resolution geochemical soil analysis. Many of these proposed modifications require substantial contemplation prior to sampling (scale of sampling, location of sampling, sample parameters), while others require ongoing analysis and sampling modification.

Collect/analyze data with consideration of watershed or ecoregion (or sub region) boundaries (application in TMDL, water quality analyses).
Coordinate geochemical sampling within major soil types or highly erodable soils (SSURGO, STATSGO) or bedrock types.
Use adaptive sampling routines to ensure adequate coverage for robust spatial interpolations.
When sampling, analyze physical soil parameters (bulk density, soil texture, and soil organic carbon) to enable calculation of watershed/area specific Soil Screening Levels for Health Risk Assessment.
Utilize analytical methods that are acceptable to states.
Consider methods to reduce within sample heterogeneity.
Minimize non-detects by utilizing analytical methods with lower Levels of Quantitation or larger soil sample volumes.
Consider methods to assess speciation data for certain metals.
Develop a strategy for including PCB congeners and dioxins in urban and suburban areas.
Develop criteria that exclude sampling local sites that may contain "non-background" concentrations of analytes (dredge disposal, waste fill, sludge application, incinerators, power plants, railroad beds, roadsides, potentially contaminated sites).
Evaluate agricultural samples separately from other "natural background" sites.
Coordinate pesticide/organic analysis selection with state pesticide usage surveys or state-based contaminants of concern, USGS NAWQA work, and/or the EPA Endocrine Disruptor Screening Program.
Consider sampling that will generate a three dimensional geochemical map (sample at different horizons/layers).
Include multiple samples at a particular location to get an estimate of variability.
Collect surface samples (upper 6 inches or less) to evaluate long-term and short-term atmospheric transport and deposition of chemical (particularly organics and mercury).
Critically document physical conditions of the sample area and potential sources of contamination.

Comments from Oliver Chadwick, University of California, Santa Barbara:

Dust. We now know that the upper portion of soil is strongly influenced by dust (mineral and aerosol). One end member in the dust influence continuum are soils forming in Loess (in which case dust is the parent material). Moving away from the loess zones, soil are composed of weathering local parent material with weathering dust superimposed. There are few places on earth that are not influenced by dust additions, in Hawaii for instance we can easily identify continental dust input by measuring quartz that is not present in the underlying parent material. Hawaii receives 2 to 3 orders of magnitude less dust than most continental areas and yet dust dominates the surface 10 to 30 cm of stable, pre-Holocene soils there. In steep, unstable terrain, erosion will preferentially strip off the surface layers and hence minimize the impact of dust relative to the underlying parent material. It is clear that a sampling regime based on surface horizons only will give a skewed view of the geochemistry. We are of course keenly interested in the composition of surface horizons. Lead for instance is often highly elevated in these horizons. To a first order we can estimate whether elements are derived in situ by comparing the surface with subsurface horizons and especially with C horizons.
Soil Dilation/Collapse. Element concentration differences can occur due to dilution or concentration in situ as well as through actual addition or loss of the element in question. Thus leaching losses of mobile elements will lead to residual enrichment of less mobile elements and addition of elements carried in dust in different proportions than in the underlying parent material composition can confuse our understanding of specific element behavior. One way to sort these things out is to quantify soil dilation/collapse using relatively immobile elements as index elements measured on the horizons of interest and on the supposed parent material. Differences in bulk density are important here as well. These kind of analyses can make the difference between thinking that a surface horizon has been losing a trace element or gaining it. We have found Zr, Nb, and Ta to be relatively immobile in soil environments, although Zr seems to me subject to chelation than the other two.
Tracers of element provenance. Pierre Biscaye at Lamont Dougherty, Columbia University has done some very nice work fingerprinting the source of dust in the Greenland Ice Sheet by using Sr and Nd isotopes. Their ratios plotted against each other provide nice separation for different source areas. We have used these approaches with success as well and I believe we could do similar things with the soil geochemistry samples. I imagine that running isotopes on TIMS machines is beyond the scope of what you wish to accomplish in your survey, but making the samples available for researchers who could run specialized isotope analyses would be very useful. We are in the early stages of developing Si isotopes as tracer in the terrestrial Si cycle, and I could readily imagine using samples in a survey mode to better understand the full range of isotope fractionation.
Selective extractions. We have found that the ammonium oxalate extraction tells provides a good break between elements that have undergone ligand exchange with the non-crystalline soil material and elements that have been incorporated into more resistant crystalline phases. Also it is becoming evident that in different soil Orders different mechanisms are responsible for binding carbon in soils. Thus in Andisols we find that the oxalate extractable components play the most important role in binding carbon (as measured by reduced carbon turnover using C14 analyses of organic matter), whereas in Spodosols, Alfisols, and Ultisols it seems to be the pyrophosphate extractable phase that is doing the binding of carbon (essentially this process is one of highly electropositive ions saturating the active sites on organic matter, thus lowering the sites available for enzymatic attach). Clearly a long-term goal for soil studies needs to be defining the maximum carbon stabilization capacity for soils having different fundamental chemistry (Soil Orders, etc.).
Sampling by Profile. I realize that sampling soil profiles requires more work both in the field and in the lab and that there will be a trade off in terms of number of sites versus intensity at site. There are critical trade offs that need to be thought about. Soil profile sampling is the ideal because we can do so much more in terms of data interpretation. Much of what is described in the above points depend on being able to compare A, B and C horizons. Also sometimes even with those horizons identified we do not have a good sense of what the parent material really is, so it is important to apply good field sense during the sampling process and to specifically ask the question. On the other hand, there are real problems with investing a tremendous amount of work on a soil profile that may not be representative of the general area around the site. Soil fertility people often will increase their sample support by sampling a number of sites and compositing the sample. When making regional interpretations this can be a critical step. The ideal would be to dig four or five pits in the general area of the sample location, id the horizons and then composite horizons. Probably too much work but this whole problem should be thought about.
Sampling design. Ordinarily I would claim that stratified random sampling is the way to go, but that leaves a major headache about all participants agreeing what are the most critical elements of the landscape to sample. And how can you sample enough major landforms within physiographic regions to effectively scale up? So it may be better to sample on a grid with careful notation of landscape position and the apply geostatistical approaches to the mapping process. It is not clear to me whether the project could collect enough samples to develop useful variograms. We have been working on these problems but don't have final answers on the best approaches. I imagine that you will be considering the USDA MLRAs as a possible sampling base.

Comments from Tony Olsen, USEPA

An alternative to a systematic grid that also allows intensification in subregions if necessary is a spatially-balanced survey design based on methodology of Stevens, D. L., Jr. and A. R. Olsen (2004). "Spatially-balanced sampling of natural resources." Journal of American Statistical Association 99(465): 262-278. Software to select such a sample is also available. spatial input is in form of shapefiles. This does not take advantage of a linkage with FIA. It does make it easier to focus sampling in areas where soil changes may be at a different scale that other areas.

Return to top of this page

Mineral Resources	Eastern / Central / Western / Alaska / Minerals Information
	Crustal Imaging & Characterization / Spatial Data