U.S. EPA Contaminated Site Cleanup Information (CLU-IN)


U.S. Environmental Protection Agency
U.S. EPA Technology Innovation and Field Services Division

ITRC Internet-Based Training Course Questions Wiki

Incremental Sampling Methodology

ITRC Disclaimer, Privacy, and Usage PoliciesThe training page, with links to registration or archive, Links page, and feedback, for the Soil Sampling and Decision Making Using Incremental Sampling Methodology course is available at http://www.clu-in.org/conf/itrc/ism/

Remaining Simulcast Questions and Answers from Sept. 20, 2016

Question 1: On slide 99, it mentions that the Chebyshev is usually 10-45% higher than Student's-t with 3 replicates. Will more replicates reduce the difference between the two?

  • Trainer Name: (response)

Remaining Simulcast Questions and Answers from April 14, 2016

Question 1: Referring to slide 89, how can we properly correlate Field Sampling data with Lap-data results to find out the appropriate number of Incremental representative samples for each of them for the decision making? And if there are difference, which data we rely upon?

  • Trainer Name: (response)

Question 2: How does ISM address a site that may have portion with contamination, that discrete identifies the area, but ISM demonstrates concentrations are below action levels (due to having many clean samples and only one or very few samples that have contamination)?

  • Trainer Name: (response)

Question 3: How to present combining DUs and averaging them when risk assessors say they are already averaged and can not be averaged again.

  • Trainer Name: (response)

Question 4: Can we combine discrete sampling and ISM together, for example: use discrete sampling method to screen the Site; then use ISM to confirm the areas where the higher COPCs found?

  • Trainer Name: (response)

Remaining Simulcast Questions and Answers from April 12, 2016

Question 1: Slide 68: If you are doing a screening level investigation, as in an SI, and your result shows contamination present based on your IS sample, but below action levels. Are you done, or do you assume that since you used IS for screening, there may be a "hot spot" somewhere in the DU that needs addressing. The concentrations in the hotspot, mixed with remaining "clean" increments, resulted in the "hit" even if it came across as below action levels.

  • Robin Boyd: In that circumstance, you are done. Since you decided up front what your decision unit size should be, say evaluate risk to human health, there is no need for additional sampling. You must remember that concentration is mass dependent so once you decide on the proper size of your DUs, all you need to know is what the concentration is. If you look for so called hotspots, you will first have to define what that is, and secondly, you would have to find them all. It really is irrelevant and a holdover from discrete sampling which is done at the entirely wrong scale of sampling.

Question 2: Will you talk about cost-effective ways of collecting ISM samples at depth (generally greater than 1 foot deep)?

  • Robin Boyd: Obviously cost increases with depth, but I routinely do investigations to 5 to 10 feet bgs. If there is a need to know the concentration at deeper intervals, ISM can still be done. It really just depends on what quality of data you need. As you get deeper in the subsurface, screening data is often used to locate source areas for remediation. In that instance, you would likely want to do ISM as confirmation sampling.

Question 3: What is the difference between an increment and a replicate? Is a replicate just a duplicate sample using the same increments?

  • Robin Boyd: Numerous increments are collected and combined to make an incremental sample. A replicate sample is collected from completely different increment locations in the same DU.

Question 4: How exactly is ISM different from composite sampling? What are the advantages over composite sampling?

  • Robin Boyd: 1) a composite is a collection of grab samples with no rules to decide how much sample mass to collect or how to collect the sample. 2) incremental sampling uses the principles of sampling theory to determine the shape and size of each increment making up the sample. 3) ISM allows you to estimate your sampling error which is impossible with a composite sample owing to it being structurally biased. Incremental sampling takes heterogeneity into account while composite sampling has no such rules. In fact there are no rules for composite sampling. It is simply a collection of discrete samples collected without regard for sampling error. ISM also applies to the laboratory side of sampling, which is critical as well. ISM is based on science, while discrete sampling is judgmental and completely up to the person doing the sampling and therefore has poor reproducibility.

Remaining Simulcast Questions and Answers from November 10, 2015

Question 1: For core samples - what about the risk of cross contamination from surface soil along the outer edge of the core?

  • Trainer Name: (response)

Remaining Simulcast Questions and Answers from November 3, 2015

Question 1: Are there some types of contaminants that are distributed more homogeneously in soils than others, e.g. OCPs vs. metals?

  • Robin Boyd: Yes, some COPCs tend to be more evenly distributed, but it is heavily influenced by the release mechanism, site history, geology, etc.. PCBs and PAHs often have extremely high levels of heterogeneity. Pesticides can be relatively evenly distributed, largely owing to their intended use and method of application. But even pesticides often have mixing areas that can have quite high heterogeneity. As for metals, it will depend on the release mechanism. A firing range can have areas of quite high heterogeneity as well as areas that are relatively uniform. Once you collect your incremental samples in triplicate, you will likely see how much heterogeneity there is as indicated by your RSD/CV.

  • Deana Crumbling: Perhaps you are asking whether certain contaminants are inherently less likely to bind to soil particles? It certainly is possible. But other factors probably have more influence over short- and micro-scale heterogeneity than the type of contaminant. Factors such as how the contaminant was deposited and how long ago, rainfall, and how much binding capacity a certain soil has. Binding capacity is influenced by all kinds of things, such as soil pH, which minerals and how much are present, the type and amount of organic carbon present, and competition from nutrient ions (Ca, PO4, NO3, Mg, etc.) for binding sites.

Question 2: Is ISM sampling another way to reduce errors and decision making? Is ISM sampling known all over the states? ISM is a method to sampling?

  • (Trainer Name): (response)

Question 3: Slides 78-79. Do you find it curious that so many discreet samples were < AL in the areas subsequently determined to be > AL (blue-lined red areas in Slide 79)?

  • Robin Boyd: Not at all. This was a PCB site and PCBs often have quite high levels of heterogeneity. If you look at the skewed distribution of discrete samples as shown on slide 80, you will see that this is exactly what one would expect. Because of the skewed distribution, most of the discrete samples will be below the mean, which is exactly what was observed.

  • Deana Crumbling: To add to Robin's answer, there are 2 primary reasons for the observation that ISM results are frequently higher than discrete samples:
  1. There is a higher density of sampling across the DU than with discrete samples. If there is a "hot spot" present in the DU, 30 ISM increments have a much better chance of hitting it and incorporating its higher concentration into the average for the DU than taking 4 or 5 discrete samples which can totally miss it. Even if a high-concentration increment is mixed in with increments having much lower concentrations, the high increment is often enough to bring up the concentration for the ISM field sample to concentrations that indicate small areas of contamination could be present.
  2. Remember the section on micro-heterogeneity of soil? If laboratories scoop off the top of a segregated sample jar (which has most of the fines carrying high contamination at the bottom of the jar), the analytical result will be biased low. ISM subsampling fixes that problem.

Question 4: When replicates are discussed for statistics, is that a minimum of 3 replicates per decision unit or 3 replicates for the site? Or can it be both?

  • Robin Boyd: That's 3 replicates for a decision unit. The number of DUs that you collect replicates in depends on your site and your data quality objectives, but an absolute minimum is 10%. What I've found is that I usually wish I had budget for more DUs with replicates.

  • Deana Crumbling: And note that these must be "independent" replicates. That means 3 distinct 30-increment field samples must be collected from a single DU and all must cover the same area. Taking a single field sample and splitting it into 3 samples doesn't work. Those are not independent replicates because the splits "depend" on the concentration of the original field sample.

Question 5: For sample planning, if an ISM sample exceeds action level for analyte(s), do you recommend analyzing individual aliquots for the analyte(s) with exceedances to focus target site areas for remediation (e.g.)?

  • Robin Boyd: No. If you do your planning correctly, your DUs will be the correct size to evaluate whatever your sampling objectives are. The individual increments are not a probabilistic sample by themselves and therefore tell us nothing about any mass/volume other than what the lab analyzed. Now if you wish to refine and reduce the volume of soil requiring remediation, then you should simply resample with smaller DUs.

  • Deana Crumbling: Additionally, no part of an increment is retained for later analysis. Splitting increments would be another source of sampling error that would be immensely time consuming to control. Even if you did, you would still have the same problem of not knowing how much soil around it has the same concentration as the increment. Some ISM people prefer to go back and resample to find the dirty area when a DU fails. Others believe that more complicated ISM sampling designs can be used so that cleanups can be done without needing to revisit the site. Sometimes resampling is not an option for a particular project. The more complicated incremental designs employ the concept of ?sampling units?, SUs. The decision of which will be done (go back to resample or use an SU strategy) must be made during project planning before ever going out to the field. Deciding which strategy to build into the sampling design needs to take a number of project-related variables into consideration and balance the costs and benefits of each. If you wish to learn about that kind of design, contact me directly (crumbling.deana@epa.gov).

Question 6: How big is too big for a DU, i.e. when should it be split? Can your DU be too big (e.g., 400 acres)?

  • Robin Boyd: DU size is determined in your planning step and is based on your data quality objectives. There is no such thing as too big or too small a DU because it depends on your DQOs.

  • Deana Crumbling: There is no bright line. Much of our experience with ISM has been using DUs that run 1/4 to 1/2 acre or even smaller. But at least some DUs up to 1 acre can probably be handled by 30 increments, as long as triplicates are done (total of 90 increments) to give reasonably dense coverage by increments.

You are looking for a rationale for drawing boundaries around a portion of land that will be internally similar in contaminant concentrations throughout. If at all possible, you don't want to overlap a DU so that most is in a clean area and some is in a dirty area. The dirty area will win out (i.e., the ISM sample concentration will get pushed up near or exceeding your action level) and you will end up cleaning up more clean soil than you probably want to.

The CSM should help you with creating DUs: you need to understand the process of how contamination got there and how it could have moved around, especially for contamination deposited decades ago. A wind rose might help for some situations; a topographical map or GoogleEarth views help for other situations. If those things don't help, you may need to go out and walk the land to get a sense of natural divisions, like drainages (small and ephemeral streams) that create small watersheds. Other types of topography divides might also be present that could help decide what area looks like it would be an internally consistent "unit." Remember that a DU is an amount of soil that is covered by a single decision. Risk decisions use exposure units, and these become your DUs.

You can also use real-time field analytical methods (XRF, test kits, etc.) to separate cleaner from dirtier areas, which can help you draw DU boundaries. HOWEVER, you must also take short- and micro-scale heterogeneity into effect when collecting those data as well, or they will also give you misleading data. Even if incremental sample processing really isn't feasible for the site, you can use a high density of samples to keep short-scale heterogeneity under control when using test kits that use 10 or so grams of sample. Sieving at 10-mesh at least is highly recommended if you have a gravelly soil, but try to run the test on a whole increment to avoid sampling error by splitting the increment.

Because XRF sees such a tiny volume of soil when taking a reading, micro-scale heterogeneity can cause tremendous variability in XRF results (see Section 4.5 and the figure in 4.5.4 in the Hot Spot FAQ white paper in the Links for this webinar [ https://clu-in.org/conf/itrc/ISM/resource.cfm ]). Therefore simple in situ readings of unprepared soil are strongly not recommended. At the very least, "prep" an area of in situ soil by loosing and crushing the soil over an area of about 1 sq.ft. and to the depth you are interested in. Take no less than 5 XRF readings over that area, and see how well they agree with respect to the question: "Is this a higher concentration area or not." If you are looking only at Pb, or Pb and As (which are on the same XRF filter), readings of 15 sec each are adequate to get an idea of the magnitude of concentration. A quantitative reading is not needed. If your DU depth goes below 2 inches, it is also advisable to turn over your worked soil and take 5 more readings. This makes sure you are not missing something further down. Remember that XRF readings only penetrate soil to 1-2 mm. You can do transects made of these 1 sq.ft. areas to look for changes in concentration that could signal a likely DU boundary.

If nothing else suggests DU boundaries, you can go the "hot spot detection" route. Determine how big a hot spot would be that you (or a regulator) would not want to miss. Figure out the spacing of increments needed to detect that sized hot spot, while taking account of "false negative" increments (i.e., increments that fall within the boundaries of a hot spot but pick up cleaner soil than what is around it). Then the number of increments in a DU is proportional to the size of the DU. Balance the number of increments per field sample (field samples that are too big are very hard and expensive to process and subsample) against the number of field samples needing processing and analysis. Divvy up into DUs accordingly. Note that I skipped the details for how to "Figure out the increment spacing." I have some prepared information (that will not fit here) that I can email you if you'd want to pursue that option.

Question 7: If a site has multiple decision units which are similar in historic use and anticipated contaminant levels they should, in theory, be similar sizes, have similar numbers of increments, and have similar increment spacing. If the size of one DU is constrained due to site features, for example a building in the way, should the number of increments be decreased (potentially to less than 30) or should the spacing of the increments be decreased for that DU?

  • Robin Boyd: Good question. As indicated in question 6 above, DU size is determined by your DQOs and once that size has been determined, the DUs should remain close in size for a given sampling objective. The number of increments should be 30 or more for all DUs. If you have different numbers of increments, you will have different sampling errors, but you will still have an estimate of the mean for that DU. And if you collect replicates, you have an estimate of your sampling error. One thing you have to remember is that each step of the incremental sampling process is done to minimize specific errors.

  • Deana Crumbling: Do not decrease the number of increments below 30. The reason for 30 increments is based on statistical theory and simulation, not the area of the DU. You should decrease the spacing to accommodate the smaller area.

Question 8: If a DU is very linear (along a set of railroad tracks or a road), would it be appropriate to collect increments down the centerline, and replicate increments alone each side, or should all replicates be collected from the centerline?

  • Robin Boyd: Increment locations should be based on either a systematic random pattern or a stratified random pattern (random within a grid, see slide 105). Replicates must be collected from independent increment locations. In your example, you would not want your increments collected in a straight line since this would not be random.

  • Deana Crumbling: The purpose of replicates is to see how reproducible the results from your sampling design is under the same conditions (over the same DU area, same number of increments, etc.). An ISM sample from each side does not cover the same area that the center ISM sample does; therefore these are not replicates.

Question 9: Was a golf course a good choice for testing ISM? At a golf course the Arsenic was applied as uniformly as possible.

  • (Trainer Name): (response)

Question 10: How much should geology used to determine decision units?

  • Robin Boyd: That depends on the site and your sampling objectives/DQOs. Changes in geology can often be used to determine the boundaries for DUs as a result of the different retention properties of different soils. If part of your site is composed of mostly clayey soils and another part is sand and gravel, then DUs could likely mimic the geology and you would probably not want a DU to cross the boundary.

  • Deana Crumbling: Lots! Geology and topography can be very important elements of the conceptual site model (CSM) that help determine areas that are more or less "internally consistent," and help keep clean and dirty areas separate, especially for older sites.

Question 11: Can I use side by side box and whisker plots from Pro UCL instead of side by side dot plots for comparing site data to background?

  • Kelly Black: Yes. Side-by-side box plots ("box and whisker plots") are a great way to look at site data and background and visually compare them. However, with ISM data where only very few samples were collected (e.g., three from each of the site and the reference area), side-by-side dot plots may make the comparison more visually apparent. For larger datasets, where visually summarizing the data is helpful, box plots are generally used.

Remaining Simulcast Questions and Answers from May 14, 2015

Question 1: Do you recommend milling for mixed wastes of PCB, PAH's and metals?

  • Mark Bruce: Generally milling is done for particulate based contaminants such as explosives or metals. There is less data that demonstrates improvements for organic analytes such as PCBs and PAHs when milled.Splitting the sample in half right before the milling step and comparing milled and unmilled results is recommended when there is a potential but uncertain benefit.

Question 2: Do you need to decon a new bucket before using?

  • Mark Bruce: Decontamination of new sample containers depends on the manufacturing processes of producing the container, how it was stored and the contaminants of interest.

Question 3: What type of lab processing would be appropriate to best relate to potential exposures of exacavator employees working on a site with naturally occurring asbestos? The employees are not processing the soil during the work process (eg no rock crusher or processing equipment used), but they excavate with backhoes and excavators.

  • (Trainer Name): (response)

Question 4: Is there one concise document which i could use to "convince" team members that this is the way soil sampling should be conducted?

  • (Tim Frederick): (response)This is a fairly concise factsheet: http://www.epa.gov/esd/factsheets/csutpg.pdf Incremental sampling methods are based on the theories of Pierre Gy.

Remaining Simulcast Questions and Answers from May 7, 2015

Question 1: Slide 66. It's not clear what is identified as the DU. Are each area A, B, and C separate DUs? Could you then have multiple DUs at a site? If so, would you need to have separate objectives for each DU or could multiple DUs meet several objectives?

  • Roger Brewer: This was just an example, assumed that you knew ahead of time the exact nature of the soil contamination. This of course is rarely known. In the field smallish (e.g., few 100ft2 to few 1,000ft2) DUs would be designated across the suspect area and tested, with "Perimeter DUs" place around the "Source Area DUs" in what are anticipated to be clean areas.

Question 2: When calculating a 95%UCL based on discrete sample data, how do you know if the set of sample data provided is in fact representative of the targeted area? Isn't this the advantage of using ISM replicate samples?

  • Trainer Name: (response)

Question 3: Generally speaking how much does ISM cost vs a similar discrete sampling method.

  • Roger Brewer: In the long run it can save significantly on the investigation and cleanup of contaminated sites, since the data you collect are more reliable. Upfront planning and proper DU designation allows a site to be reliably invested in one or two mobilizations. ISM is especially important for remediation, since highly reliable confirmation samples can be collected up front. The biggest expense in site investigations is typically the need to remobilize to collect yet more discrete samples, carry out additional excavation. Discrete soil samples can be highly unreliable and in particular ensure that you don't miss anything. Cost-wise the latter is especially important. As we discuss with developers, what's the cost of getting sued after you sold a supposedly clean property to someone that later turned out to be contaminated?

Question 4: Do you suggest using incremental sampling for background sampling?

  • Roger Brewer: Definitely, but keep in mind that most existing data are likely to be discretes. Keep in mind the scale of likely variability, since ISM gives you the mean. It's possible that the mean for a DU is similar to the upperbound, natural background. The NRCS uses a 1m2 "DU" for their work, with (as I recall) a 9-point composite sample being collected. They also typically grind their samples, which can result in higher metal concentrations that incomplete extraction of unground samples.

Question 5: How do you calculate the 95UCL if your triplicate sample results are nonparametrically distributed? The T-test is only for normally distributed Is the highest value the 95UCL?

  • Trainer Name: (response)

Question 6: Also, is the RSD the same thing as the Coefficient of Variation?

  • Trainer Name: (response)

Question 7: When you say the ISM DU must be small for acute toxicity (slide 21), does that also apply to relatively high (acute) exposures?

  • Roger Brewer: Very few chemicals have acute toxicity factors for ingestion. If they were available, resulting screening level would likely be orders of magnitude higher than those used for chronic exposure. From a risk standpoint it is generally assumed that contamination in soil will not pose hypothetical, acute risk if chronic risk is addressed, since if the exposure area is properly characterized hypothetical, sample size (e.g., 10 gram) "acute hot spots" within a DU would also cause chronic screening levels to be exceeded. It is incorrect to say that ISM "isn't applicable" when toxicity factors and soil screening levels for acute toxicity are in fact available and evaluation of this concern is an objective of the investigation. An acute exposure corresponds to the incidental ingestion of 10 grams of soil (default for pica child) to possible a few hundred grams (i.e., exposure over 14 days = 1-g x 14 = 140 grams). This is the size and volume of your "DU" to evaluate acute exposure. ISM is simply an approach to subsample a DU volume of soil, since in most cases submittal of the entire, targeted volume of soil to the lab for analysis is not practical. ISM subsampling of the DU isn't necessary in the case of the hypothetical, tiny DUs for acute toxicity. It isn't that ISM in the field isn't applicable -- it isn't needed. The "sample" you send to the lab is then entire DU. ISM-type processing and subsampling in the lab is still necessary to get representative data, unless the entire mass of the sample can be extracted.

Question 8: How can ISM be applied for petroleum hydrocarbon sampling?

  • Roger Brewer: The is discussed in both Alaska and Hawaii ISM ("MIS") guidance. Petroleum is treated the sample as any other chemical. As will be discussed in Part 2, increments for ISM samples to be tested for VOCs (including TPHg) are placed in methanol in the field. The lab then tests an extract of the methanol. For semi-VOCs, including TPHd and some PAHs where there is a concern about loss during sample drying and sieving, we recommend that the ISM sample be collected in the normal manner (typically placed in a 1-2gal freezer bag), chilled and then ISM subsampled at the lab for the collection of an aliquot for analysis immediately after it is spread out for drying. Non-VOCs are tested for after normal drying and sieving of the sample.

Question 9: I understand the concept of a DU but not how one decides the size to use when out on the field collecting samples

  • Roger Brewer: The designation of DU will become very apparent with experience. You've always be doing it -- "This is a suspected spill area" "This is an exposure area of concern", etc. Hawaii's guidance provides more examples. This will be expanded in updates due out this summer.

Question 10: When should ISM not be used?

  • Roger Brewer: It's important to start from the beginning and flip this question around: "Given what we know now and the unreliability of the data, when (if ever) should discrete samples be used (at more than a gross screening level)."

Question 11: Other than saving money on analytical costs, is there a reason for not analyzing ISM incremental samples separately and then averaging the results together, as long as you keep in mind any one result could be result of small-scale heterogeneity?

  • Roger Brewer: Small-scale heterogeneity negates the usefulness of analyzing individual increments from the beginning. Our recent field study demonstrated orders of magnitude differences in contaminant concentrations around grid points at the scale of a traditional, discrete soil sample. This has significant implications in their use in environmental investigations. Also keep in mind that the objective of an investigation is always to estimate the mean contaminant concentration for a targeted volume of soil. The sample for a targeted DU area must be representative, i.e., incorporating a representative proportion of smaller-scale "hot" and "cold" spots within the DU to estimate a representative mean.

Question 12: pertaining to extrapolation of results to unsampled areas. This does not seem like a valid justifiable process. Aren't the extrapolated data often shown to be wildly misleading?

  • Roger Brewer: We were discussing the extrapolation of ISM replicate data (i.e., the estimated precision of the data) to DUs where only a single ISM sample was collected. Replicates are primarily collected to tell you how representative your ISM sampling approach is. With experience, you'll find that it is unnecessary to collect replicates in every DU. It's also time consuming and expensive. It may impractical to test every DU for very large areas, e.g., former ag fields being redeveloped into residential house lots. In this case statistical approaches can be used to test a subset of the DUs (e.g., 5,000ft2 lot areas) for a pre-specified degree of confidence. This is discussed in the Hawaii guidance (google Hawaii Technical Guidance Manual). For example, if target action levels are not exceeded in 59, randomly selected, hypothetical lots then you can be 95% certain that action levels will not be exceeded in 95% of the lots that were not tested. You'll also get a better feel for the number of increments required to get reproducible data. In our experience the default minimum of 30 increments is adequate when random, small-scale variability is low (e.g., sprayed pesticides). We now recommend at least 75 increments for scenarios where small-scale heterogeneity is anticipated to be very high; PCBs is a prime example.

Remaining Simulcast Questions and Answers from February 5, 2015

Question 1: Are comparisons to background concentrations that are 95% UCLs themselves valid? For example, we are collecting ISM soil samples and analyzing for pesticides in a region where most, if not all, of the surrounding area has been extensively farmed. Access to "clean" Reference location is not possible

  • Tim Frederick: Usually a 95%UCL is calculated for use in a risk assessment or for comparing to an action level. The ITRC Tech Reg recommends graphical analysis of the actual ISM results with the actual background data results. It is not necessary or recommended to calculate a 95%UCL for background comparisons.

Question 2: Referencing slide 9, with sites containing numerous DU's (15+), how would one be recommended to approach sampling.

  • Trainer Name: (response)

Question 3: I am curious as to what the level of effort will be to collect 10 Decision Units (10 samples) as compared to 10 soil samples using traditional discrete sampling. What will be the lab cost differential will be for sample prep and analysis versus traditional analysis?

  • Mark Bruce: The lab costs associated with ISM processing are highly dependent on the options selected. They generally range from $50-200 per sample in addition to the regular analytical costs.

Question 4: What type of sand is best to use for metals equipment blanks?

  • Mark Bruce: Sand is not a good general purpose equipment blank when testing for metals, because it already contains many metals. If a short metals list is to be determined (e.g. Pb only), it might be possible to find sand that has a very low concentration of the target analyte. There is no solid material that has soil like characteristics and zero metal content.

Question 5: Can you supply a good reference on quantifying error and a reference on using the rule of thumb of 30-35% CV as being an upper level of acceptance?

Question 6: How many field samples should be taken within each sample area? I am not talking about the samples analyzed in the lab, but the soils samples taken from the field to the lab.

  • Tim Frederick: The ISM document recommends a minimum of 30-increments (holes in the ground) that are then combined into a single sample that is sent to the lab (replciate). The ISM document recommends a minimum of three samples (replicates)sent to the lab for each decision unit (area sampled).

Question 7: What is the best way of accurately splitting a Geoprobe-type macrocore sample into thirds?

  • Trainer Name: (response)

Question 8: Please further describe issues for shipping bottles containing methanol if courier is not available.

  • Mark Bruce: It is also possible to ship containers with more than 30 mL of methanol as hazardous goods. This requires special training and packaging. A third option is to use smaller containers (e.g. 40 mL VOC vials with 5-10 mL methanol for each 5-10 g increment) to facilitate commercial shipping. Once at the lab, representative aliquots of methanol from each container can be composited to form a single methanol extract that represents the entire decision unit.

Question 9: Can ISM be used at sites with urban fill/urban soils? These are some of the most heterogeneous soils that exist. Can you comment on any experiences with urban fill sites and ISM postives/negatives? (I would imagine soil processing could be challenge.)

  • Trainer Name: (response)

Question 10: How do you address large contamnated sites that cover many acres of land?

  • Trainer Name: (response)

Remaining Simulcast Questions and Answers from February 3, 2015

Question 1: Can you comment on what implications ISM has for site owners that claim hot spot excavation (based on discrete grab samples) has resulted in a property that will meet remediation standards for redevelopment and/or off-site soil disposal characterization criteria?

  • Robin Boyd The only way to know what the concentration is for a given exposure unit size suitable for the selected land use is to collect an incremental sample. It is possible that the site is clean but one has no defensible data to back that decision up. In Hawaii, the regulators will accept discrete samples in a situation like you describe as long as confirmation samples are collected using incremental sampling. As indicated in the training, it is inappropriate to remove small hotspots based on discrete samples and call it clean. That is not scientifically defensible.

Question 2: How many samples should be taken within the DU's perimeter per square foot of surface area of the DU, if the history of the property isn't available?

  • Robin Boyd The number of increments in a DU should be 30 or more to control the heterogeneity. If your replicate data has a coefficient of variation much greater than 0.35, it is telling you that you have not adequately captured the heterogeneity. Remember that the number of increments is not related to the size of the DU but rather the amount of heterogeneity in the DU.

Question 3: If soil contains larger stones/gravel bigger than expected contamination is it advisable to screen them out?

  • Robin Boyd This will depend on your data quality objectives. In most instances, everything larger than 2mm is screened out to reduce the size of sample needed to get representative data, since the sample mass required is related to the cube of the 85% grain size. This tends to bias the result high since most larger grains have less contaminant mass associated with them.

Question 4: Can you elaborate on the relative costs of ISM sampling compared to discrete sampling?

  • Robin Boyd: In many cases, the cost is comparable to a discrete design with a realistic number of samples. I have not done a study showing the cost at a given site, however I did have a site that was switched from an incremental sampling plan to a discrete one at the planning phase and the discrete plan cost the client an extra $150,000.

Question 5: Can a software program like Visual Sample Plan be integrated to determine sampling plots into an ISM plan?

  • Trainer Name: (response)

Question 6: Has ISM been used for injury assessment sampling?

  • Trainer Name: (response)

Question 7: Why cannot ISM be used to determine the site background concentration if the site only contains one DU?

  • Robin Boyd: Excellent question. It's not that ISM cannot be used to determine site background, its just that it is inappropriate to try and compare a background value based on discrete samples to incremental data. It is perfectly acceptable to determine a site-specific background using ISM sample(s) and then compare your ISM results from your investigation to that background value.

Question 8: If the ISM result in one DU is much higher than other DUs (above action limit), will that entire DU be the area for remedial action or further sampling is neccessary to fine the boundary of remediaiton?

  • Robin Boyd: You are correct in that the entire DU needs to be remediated in some way, or additional sampling can be done to reduce the volume requiring remediation. If additional sampling is selected, the size of your DUs will be based on a cost benefit analysis comparing the cost of remediation versus the cost of additional sampling.

Remaining Simulcast Questions and Answers from November 6, 2014

Question 1: in large sites (about 1500 km2), how I can use ISM for baseline characterization, in order to have good representative results?

  • Jay Clausen: For a large site it is necessary to decide on your objective for the project and how that relates to the site conceptual model. If you are concerned with a small impact within your 1,500 km2 area then the size of your DU needs to be scaled to that activity. If it some larger activity such as uniform pesticide action on a golf course then your DU can be larger. A key component of the conceptual model is the amount of heterogeneity anticipated. An anticipated large heterogenity is going to drive smaller spacing between increments and a smaller DU.

  • Roger Brewer: We routinely screen very large, former agricultural lands for residual pesticides using ISM (or MIS in Hawaii). Google Hawaii HEER Technical Guidance Manual and refer to Sections 3-5. For initial screening we split area into at least 15 DUs, taking into account soil types, drainages, crop history, land use history, etc. (e.g., 1,500km2 area = 15 approximately 100km2 DUs). Suspect "hot areas" "not "spots!") are tested separately from the larger area with separate DUs. I suggest 100-increment samples, with triplicates in at least 10% of the DUs, or in this example at least two. (Note that more detailed sampling is required for sites over 100 acres if they want a "concurrence" letter from out office, normally ISM data for a minimum of 59, randomly located, 5,000ft2, hypothetical residential lots.) If an experienced field team should be able to do this in three days for open areas. We've pre-cleared grid pathways through large, heavily overgrown sites (bushhogs, bull dozers, by hand, etc.) with surveyors marking increment pts for us ahead of time. Makes the field work a lot easier. We also do something similar for large, former munitions practice ranges. One was close to 100K acres. The Army Corps has lots of ISM experience at these types of sites also.

Question 2: Do you recommend ISM for excavation sidewall confirmation sampling and if so, how would these samples be collected?

  • Jay Clausen: ISM is great approach for sidewall sampling. The approach I have seen is that each sidewall and floor is a separate DU. Within each DU a minumum of 30 increments are collected. The depth of penetration of the sidewall and floor is up to you.

  • Roger Brewer: We do the same. This is also very useful for VOCs. At least 30, 5g plugs are collected from targeted sidewall and floor areas (could be a specific depth horizon of the sidewall) and placed in an amber jar with a pre-measured amount of methanol (based on planned # of increments). Use SIM analysis to get better reporting limits for a limited list of target VOCs. Collect replicates by offsetting the initial grid points (not from immediately adjacent points).

Question 3: Has this method for sampling proven to be more cost effective when working with subsurface decision units than grab sampling? It just sounds expensive compared to grab sampling..... Could it be that the whole process is more cost effective because of the ISM results

  • Jay Clausen: ISM subsurface sampling is defintately more involved and costly compared to grab samples from a single core. That additional sample cost needs to be weighed against the potential of making a wrong decision with the grab sample data.

  • Roger Brewer: Correct - You get what you pay for. "Bad" discrete subsurface data is the bane of failed confirmation samples for excavations and in situ remediation. Thirty "core increments" per subsurface DU are getting routine in Hawaii for subsurface investigations, with soil split up into targeted layers based on screening borings or knowledge of the site history. This isn't always necessary or cost-effect, however, depending on the targeted COPC and your field objectives. An experienced push-rig operator can get around 300ft of core a day for a pre-toned site in non-rocky soils.

Single borehole "DUs" can also be useful, for example to estimate the depth of contaminated soil (e.g., lead-contaminated fill material) at a single point, same as geologists have always done (i.e., making a decision based on a single boring). Split the core up into targeted layers rather than collect discrete samples from targeted depth points however. Send the entire targeted layer to the lab for ISM processing and testing. Subsample using 5g plugs in methanol for VOCs.

A smaller number of borings (even one) can also be useful for estimating the lateral extent of contamination, e.g., at a UST site, same as also done in the past. Just be aware of the limitations (and use DU layers, not discrete depth points). This might work reasonably well for a gasoline UST site, but we see orders of magnitude variation in side-by-side discrete samples (and even cores) for lead, PCBs, etc., so don't trust a single boring.

Question 4: Can ISM be applied to a large property (500-acre) of similar land use as one DU per depth, using 30 equidistant sampling locations without extrapolating to provide useful information for making decisions acceptable to regulators?

  • Jay Clausen: ISM can be applied to a large property but a single ISM DU would be not appropriate just as a couple discrete samples are not appropriate. The size of the DU is driven by the intended use of the data and the conceptual model.

  • Roger Brewer: See above. A single DU actually can be appropriate in some circumstances, depending on what your question is. For example, a developer (or more likely the bank financing the loan) might want to make a quick each of several potential, for former agricultural field sites as cheaply as possible before selecting one to proceed with more detailed sampling. The question is "What is the average concentration of residual pesticides in this field as a whole?" That's perfectly reasonable. We would probably recommend 100-increment triplicate samples. We would push for the 15 DU approach noted above for better resolution, however, especially if they want a letter of concurrence from our office.

Question 5: why does all the guidance use 95%UCL to compare with the Action Limit? Action limit is a threhold, while 95%UCL is an estimate of mean. I thought it would be more reasonable to use 95%UTL (upper tolerance limit)

  • Tim Frederick: This is not specific to ISM samples, but the purpose of the 95%UCL of the mean statistic is to provide a conservatibe estimate of a central value for comparison with action levels/screening levels. "Tolerance limits define the range of data that fall within a specified percentage with a specified level of confidence." ref: http://goo.gl/9nBx83

Remaining Simulcast Questions and Answers from November 4, 2014

Question 1: Is this methodology (ISM) cheaper than discrete sampling?

  • Annette Dietz: The part 2 training discusses cost information and provides an example. In general, the field sampling and lab processing costs are higher for ISM and analytical costs are lower because fewer samples are analyzed. However, many project-specific factors can affect the actual numbers. Of course, this type of comparison only considers the costs for the field and lab components; it is much harder to evaluate the cost of making a bad decision. Sometimes "cheaper" samples can cost a project more in the long run.

  • Roger Brewer: You get what you pay for. After ten years of "MIS" in Hawaii I don't know a single consultant who would risk their reputation on discrete sample data, except in the rare case where you have data from dozens of sample points for a single area that you are trying to make a decision on. The questions consultants, responsible parties and their attorneys have to ask include: "How much would a lawsuit cost me five years from now (frivolous or not?"; "How much will it cost me if I have to pay this consultant to keep coming back to over excavate what was supposed to be clean soil?; "How much will it cost me (e.g., a developer) if we discover more contamination than expected after I already have construction contracts in place and redevelopment of the site is underway?"

Question 2: The Mill technique would volatilize organics.

  • Annette Dietz: If VOCs are the contaminant of interest, then samples should not be milled. This will be discussed in the Part 2 training, but in general for VOCs the increments would be preserved in methanol in the field rather than being mixed or ground. With ISM it is very important to identify the contaminants to be analyzed in advance so you can plan the sample collection and lab processing accordingly. See ISM document Sections 5.4.2 and 6.2.1 for specific information on VOC field sampling and lab processing.

Question 3: In your experience, how many labs are up-to-speed on subsampling sample jars in the lab? Presumably some labs are better at it than others. How is one to know?

  • Robin Boyd: There are a few that I'm aware of. Mark Bruce can tell you which Test America lab is qualified. Also, APPL in California does a good job for me. The easiest check to see if the lab is doing a good job subsampling is to have them do lab replicates, at least 3. If they don't have good coefficient of variance/RSD, i.e., less than 0.35/35%, then there may be a problem.

  • Roger Brewer: A number of both small and large labs on the west coast used by consultants in Hawaii are up-to-speed on ISM processing. This is rather telling, especially in places like California where ISM is just getting started. Several consultants I know quietly get more reliable, ISM data for their clients and for themselves, even though the regulators don't require or understand it. Ask for the labs SOP for ISM processing and talk to the person in the lab who actually does it. Also ask how much they charge. They typical cost per sample is $75-$125 per sample. If it's "already included" in the normal price then you can bet that they aren't doing it.

Question 4: Its not unusual for a sample from a UST release in a sandy environment to have a strong odor but low chemistry results. The professional suspects problems but the chemistry does not test out

  • Robin Boyd: This situation can have three primary causes. Either your field sample is non-representative, the subsample the lab collects is non-representative, or losses due to volatilization are biasing your data. It could also be that you have very little actual contamination, but the contaminant has a low odor threshold such that it is smelled at low concentrations.

  • Roger Brewer: Expanding on Robin's notes... Labs only test five grams of soil for VOCs, about the amount that would fit into a plastic soda bottle cap. In the field, if you look closely enough, you'll note that petroleum contamination in soil is rarely if ever uniform; there are lots of anastomosing "hot spots" and "cold spots." This is why discrete soil sample data is not recommended for vapor intrusion studies, and why the mass of VOCs present is often vastly underestimate for in situ remediation actions. You need to designated targeted core intervals for testing and place as many 5g plus of soil from the each interval in methanol in the field to get decent data. Another potential problem is that you are testing for BTEX in a largely BTEX-free release. What you are smelling in the field are C5-C12 aliphatic compounds, not BTEXN. Hawaii recently completed a field study demonstrating that non-specific, aliphatic compounds typically drive vapor intrusion risks at sites, not benzene, etc. The average concentration of benzene in gasoline has also been dramatically reduced over the past 5-10 years. Testing for benzene at these sites is like declaring a shark-infested lagoon to be clear of alligators.

Question 5: Where your DU overlaps both impacted and nonimpacted areas how do you account for dilution. Multipal subsamples that would be nondetect could lessen your ability to even see a constitute

  • Robin Boyd: If you are concerned about dilution, your decision unit is probably not the correct size. Are you really interested in the concentration of every 30-gram mass at your site, or are you interested in the average concentration over some volume? By the way, all concentrations are averages. The only thing that changes is the volume/mass over which it is representative. And for a discrete sample, that mass is 30 grams for an organic analysis and usually 1 to 5 grams for metals. Screening criteria are not intended for comparison to individual discrete samples.

  • Roger Brewer: Right - Set you DU to the maximum size that you are willing to make a decision on from the start. You are also "heating up" the "cold spots" with ISM. That's the goal - estimate the average for a targeted area. If your DU is of acceptable size, your samples are representative and the mean is below the screening level then you are done. The problem with discrete samples is that you will inevitably hit a "spot" of clean soil in an otherwise contaminated area as you move outward from the main source area. This is clearly demonstrated in a field study of discrete sample error recently completed in Hawaii (google Hawaii Technical Guidance Manual What's New). If it can be done cheaply then screening of large areas with large numbers of discrete samples to ID potential areas of higher or lower contamination can be useful. We use a field XRF to do this for suspect arsenic and lead sites. Most of the time it isn't necessary, however, as long as you keep your DUs to an adequately small size (e.g., few thousand ft2 max). Keep in mind that you aren't looking for the mythical "maximum" concentration within a targeted area (always 1,000,000ppm if you could test down to the microscopic level).

Question 6: How are ISM results not skewed by collection of clean samples from areas at locations outside the impacted area within the DU when the actual extent of impact was not previously delineated?

  • Robin Boyd: See question 5 response above.

Question 7: Perhaps this will be covered later, but the methodology of collecting multiple samples within a decision unit seems limited to surficial contamination in pliant material. If samples are collected through an impermeable surface, such as a street, at depths greater than 5 feet, the drilling costs appear to make the methodology impractical. Is there a way to mitigate this problem and implement ISM in a cost effective manner?

  • Robin Boyd: You are correct the cost goes up the deeper the samples need to be collected. But the same is true of discrete sampling. First, it is not as expensive as you might think, and second, do you want representative data or not. Other alternatives for locating deeper subsurface source areas can be used as well such as the qualitative data from a MIP or LIF tool. Once remediation is complete, then you should collect an incremental sample from that source area.

Question 8: How do you use traditional action levels that are based on grab samples with decision units and incremental sampling?

  • Robin Boyd: See the end of the response to question 5 above. Action levels are based on risk and risk is based on average concentrations over an exposure unit, say a residential lot. The regulations were not intended to be compared to individual discrete samples. Hopefully one of the regulators on the team will chime in as well.

  • Roger Brewer: I've been publishing screening levels as a regulator for twenty years, including the "ESLs" in California, the action levels we use in Hawaii and some as a private consultant. Screening/action levels were never intended for comparison to individual, discrete soil samples. They are intended to be applied to the mean concentration of a contaminant within a defined area and volume of soil. We rather blindly did this anyway before we started to realize how significant small-scale variability of contaminant concentrations in soil can actually be. This was a light bulb that explained why our discrete sample data so often appeared to miss contamination when the soil was retested. ISM (MIS)helps us start at a scale large enough to address problems with small-scale heterogeneity and small enough to address the questions that we are actually asking about a site.

Question 9: How small a decision unit is too small to be effective use of ISM?

  • Robin Boyd: There are only practical limitations on DU size. The size of the DU depends on your data quality objectives.

  • Roger Brewer: The smallest DU I've seen is one-square foot. These were "source area" DUs for an illegal dump site that included scattered car batteries. The dump was discovered after a brush fire scorched the area. One DU was place at each former battery location, around ten or so in total as I recall. The upper few inches of soil from each spot was removed and disposed of. They then collected a 30-increment ISM sample from each of the locations. Personally I though this was overkill but the property owner wanted to document that the former location of each batter was clean. That was their "DQO" question - "Is the concentration of lead in soil at each of the former battery locations below the state action level?"; even though we wouldn't apply the action level to such a small area.

Question 10: Is statistical analysis and UCL calculation required if ISM is implemented at any given site?

  • Robin Boyd: Sometimes yes, sometimes no. It depends on the regulator for your site.

Question 11: On slide 94 is it also true (conversely) that there is a 67% chance that you would overestimate the mean by 90% - looking at the first line only, CV=1?

  • (trainer name): (response)

Question 12: How do increment sampling deal when soils are impacted with volatiles? (mass loss?)

  • Annette Dietz: This will be discussed in the Part 2 training, but in general if VOCs are the main concern, the soil increments would be preserved in methanol in the field rather than being mixed or ground. It is very important to identify the contaminants to be analyzed in advance and plan the sample collection and lab processing accordingly. See ISM document Sections 5.4.2 and 6.2.1 for specific information on VOC field sampling and lab processing.

Question 13: On slide 84 if 30 borings are ideal, and you have four DU's based on depth, how do you combine the soil from 30 samples on one DU to get a representative sample? It seems like just scraping a little bit off each core would introduce more error....

  • Robin Boyd: We typically use something like a terracore sampler to collect 20 to 30 grams over the depth interval of your DU in each core. These increments get combined to give a sample mass of 600 to 900 grams for 30 increments. Depending on my site, I often collect 50 increments per DU.

Question 14: Phil, Can you please address how 3 replicates allows statistical calculation of a 95% UCL of the mean using ISM data, but additional replicates are needed to achieve 95% coverage for the 95% UCL?

  • (trainer name): (response)

Question 15: Can you use traditional action levels based on grab sampling techniques with ISM decision units and ISM methodology? How small a decision unit can be used for this technique to still be effective?

  • (trainer name): (response)

Remaining Simulcast Questions and Answers from Sept. 16, 2014

Question 1: For large, mixed-use range sites (small arms, rockets, bombs, etc.), is there value in separating decision units based on the amounts of munitions debris present on the surface? Should a grid be laid down regardless of presence of debris? We're looking at explosives and metals.

  • (trainer name): (response)

Question 2: For volatile sampling, would multiple terracore (or equivalent) samples be collected for each of 30 samples in each replicate? If so, how many is recommmended. Are En Core samplers acceptable?

  • Mark Bruce: (The recommended minimum number of increments (5-10 g soil plug)per ISM sample is 30. A larger number of increments can be used. En Core samplers are acceptable. That would allow the field crew to ship the increments back to the lab without methanol handling in the field or transport.)

Question 3: What kind of bias may be introduced if all incremental samples are placed into a 2 gallon ziplock bag, homogenized by physical shaking, then pouring or scooping a portion of that to submit to the lab (usually <2/3 of a quart sized ziplock).

  • (trainer name): (response)

Question 4: a lab dried and ground the samples within the holding time but left the samples without subsampling and extracting too long. The samples were then analyzed and the data submitted. The issue was discovered later. The laboratory maintains that once the samples were dried they were preserved and then the clock for extraction didn t start until subsamples were collected. do you ageee?

  • Mark Bruce: (ISM sample processing does not officially change normal holding time requirements. Reducing the moisture content of the soil probably does slow biological degradation of the analytes. Some informal stability studies did show much longer than stability of some explosives in air dried soils than reflected in current holding time requirements.)

Question 5: Speaking of hotspots, is ISM applicable for largely unknown/uncharacterized sites?

Question 6: Could we use percent relative standard deviation determine the representiveness of the subsample and replicate subsample from the same DU

  • (trainer name): (response)

Question 7: how can ISM be used to delineate boundary of a hot spot if you don't know where it is before you start..

  • (trainer name): (response)

Question 8: couldnt higher standard deviation with discrete samples be related to areas of the DU that have higher contaminant concentrations that you cannot see using the ism?

  • (trainer name): (response)

Question 9: How do I convince our risk assessor that we're not going to miss "hot spots" implementing ISM?

  • (trainer name): (response)

Question 10: If you are collecting samples for the purpose of delineation (not a risk assessment), is it necessary to collect the 3 replicates?

  • (trainer name): (response)

Question 11: If you are sampling at depth and have cores from which to collect samples, can you collect for VOCs as well as for other analytes from the same core? on Slide 33, it looks like the VOC sampling would make it difficult to collect a core wedge from the same core. Does that mean you can only use a core slice?

  • (trainer name): (response)

Question 12: has an ISM approach been used for XRF field screening for making decisions about the need for additional samples to be sent to the lab?

  • (trainer name): (response)

Question 13: Michael - When combining in-situ XRF with ISM sampling for metals in soils, how early in the process do you try to do your XRF field work to guide your decision unit designs?

  • Michael Stroh: (We have used XRF in a variety of ways on different projects. On some larger projects, where it is economically feasible, we have conducted pilot studies to obtain initial estimates of concentration and variability, and then used that data to guide our subsequent full scale sampling design. The free software program Visual Sampling Plan has been very useful for us in that regard, and it now includes an incremental sampling module http://vsp.pnnl.gov/. For a project we are working on right now where we will be assessing over 100 DUs in remote areas, we will be using a dynamic work plan. There we have developed a decision tree for screening out DUs that are clearly clean using in-situ readings and then only collecting ISM samples in DUs where the decision is ambiguous.)

Remaining Simulcast Questions and Answers from Sept. 9, 2014

Question 1: What sample volume is recommended for each subsample?

  • Robin Boyd: It depends on the grain size of the soil being sampled and the desired fundamental error. If you need a reference for the equation, it is located in the Hawaii TGM. For soils less than 2 mm, we typically collect 20 to 30 grams per increment.

  • Deana Crumbling: See the EPA guidance ?RCRA Waste Sampling Draft Technical Guidance: which is online at http://www.epa.gov/osw/hazard/testmethods/sw846/samp_guid.htm See Section 6.3.1 "Determining the Optimal Mass of a Sample" (page 97) and see also Appendix D, which supplies more details, along with a chart. The equations are drawn from Gy sampling theory (see discussion of Gy theory in the ITRC ISM document and in the above EPA guidance (page 90). As Robin said, the minimum mass for a soil sample depends on the maximum particle size of the material you are sampling and the amount of ?fundamental? sampling error (covered in references above) you are willing to tolerate. The same principle applies whether you are taking an increment from the field or an analytical subsample from a field sample. When taking an analytical subsample from a soil sample, the maximum particle size should be just under 2 mm. (The material going through a 10-mesh (2 mm) sieve is the definition of ?soil? for chemistry testing purposes.) The generalized equation for achieving 16% RSD (as the minimum fundamental error that can be expected) is:

Mass (of analytical sample in grams) = 1000 x d^3 , where d is the diameter of the largest particle in cm (NOT mm!)

Using that equation, the minimum analytical mass for soil having a max particle size of 2 mm (0.2 cm) is 8 grams. So with an 8-gram analytical sample and 16%RSD fundamental error, the best data precision that can consistently be expected is an RPD of 22%. This means tha

Top of Page