http://home.comcast.net/~brooks50/Sampling.htm
Summary: The effect of using homogenized soil samples on the LEFPC PHAs is examined. Due to the presence of a relatively clean surface layer the resulting PHA is mildly conservative.
Much has been made of the nature of the LEFPC soil samples[1] which were used to establish the Public Health Assessment (PHA). It is claimed by some that the true hazard is higher because the level in the homogenized sample is lower than the level in the maximally contaminated layer. This is true but is it really germane to the question? Below we examine the problem.
Scenarios and Sampling
Scenarios and sampling protocols are interdependent, and both are dependent on the purpose of the scenario. Even sampling which not intended to evaluate an exposure pathway has a purpose which dictates a proper sampling protocol. A scenario (and its data can be either conservative or non-conservative, although the latter is to be strongly avoided. A conservative scenario is on which overestimates the exposure for one of two reasons (or both): 1) The estimated comparison value or minimal risk level (MRL) is to high, i.e., larger than the correct value, and 2) The estimated exposure dose to the individual is too large compared to the actual value. The exposure dose is a site-specific value but the MRL can be applied to several sites when appropriate.
Sampling Protocols
What is the proper soil sampling protocol? It depends amongst other things what the sample results are going to be used to determine. The many possible purposes of soil samples in a remediation program include at least the following:
|
Purpose of
sampling |
Sampling
Protocol |
Comments |
|
To locate the horizontal position of the contaminant |
Patterned random, deep core, homogenized |
Random within patterned grid areas Statistical conclusions questionable |
|
To locate the depth(s) of the contaminant |
Random, deep core, stratified |
Statistical conclusions valid |
|
To estimate the hazard to excavation workers |
Random, deep core, homogenized |
Statistical conclusions valid Core to limit of excavation. |
|
Estimate worse case scenario |
Selected locations, deep core, stratified |
Selection on some basis to produced biased samples Statistical conclusions questionable |
|
Estimate long term public health effects on small children |
Random, top 3"-4" core, homogenized |
Limit of small child digging Statistical conclusions valid |
|
To locate the horizontal position of the contaminant |
Uniform spacing along uniformly spaced transects |
Statistical conclusions questionable |
The point being made is that the proper sampling protocol is dependent on the purpose of the samples and one size does not fit all although sometimes one sampling protocol will meet the needs of two or more scenarios. On occasion, a less-than-perfect sampling protocol will satisfy the needs of a scenario if the error made is conservative and reasonably small.
Other considerations are the cost of the sampling and subsequent analysis of the samples, and the existence of proper analytical methods.
LEFPC Floodplain Soil Sampling Protocols
The DOE/EPA Protocol
The purpose of the DOE/EPA floodplain sampling protocol was to determine the locations in the floodplain which had elevated total mercury levels for the purpose of excavation or other treatment. The protocol selected was 16" homogenized, equally spaced samples along equally spaced transects along the LEFPC. Given all of the factors, including the number of samples (in the thousands) and the costs of each analysis, the protocol is not unreasonable. Real time analysis of field samples during excavation was expected to provide detailed information permitting the cleanup crews to follow deposits to their end to effect a thorough removal to the remediation goal. Such a field test for mercury was developed.
After the transect sampling, an SAIC geologist interpolated between transects taking into account the topography of the floodplain and the flow of the flooded stream. The resulting patterns of mercury deposition seemed a bit conservative but obviously had been in a responsible manner.
In addition to
the homogenized 16" samples a few samples from relatively high mercury
areas were segmented to ascertain the soil mercury vertical profile. There
appeared to be two periods of major contaminated soil releases separated by
relatively clean soil deposition. These indicated that the mercury profile
comprised: 1) a surface layer about 3" thick of relatively clean soil <
50 ppm), 2) a layer about 5" thick of soil containing about 1600 ppm
mercury,
3) a second relatively clean, ~3" layer (<50 ppm) and 4) a second
relatively contaminated, (~4") layer (~ 1600 ppm) followed by clean soil.
This profile is roughly consistent with known periods of high rainfall and
extreme flooding of the Creek. It is important to note that the surface layer
is relatively clean by comparison to the total mercury level. It has been 35-40
years since significant levels of mercury have been seen in LEFPC.
The ATSDR Protocol (See The
400 Level for more on the ATSDR Protocol.)
The ATSDR, having the purpose of estimating public health effects to residential children, normally adopts a different sampling protocol: the surface layer (3'-4" cores). (ATSDR will consider other sample depths or protocols if circumstances warrant other exposure scenarios.) This is what ATSDR would have preferred to have done based on their past professional experience. This surface layer is believed to be the origin of wind blown dust inhaled or adsorbed through the skin as well as being the maximum depth to which young children will dig (ingestion route) especially in the hard clay soils of East Tennessee. Also most tilling of soil does not go deeper than this.
The ATSDR is now faced with a dilemma: use the several thousand, DOE/EPA samples or repeat the sampling using the surface layer. To do the latter will be very costly and time consuming; they chose to use the DOE/EPA samples. The impact of this decision on the ATSDR PHA must be considered.
There are two cases to consider:
|
Relative
Soil Mercury Levels |
Effect on
the Applicable Scenario |
|
The mercury level in the homogenized sample is less than the Maximum Level in the core sample |
This affects the "worst case scenario" making its risk estimate too low |
|
The mercury level in the homogenized sample is greater than the level in the relatively clean surface layer |
This affects the "reasonable case scenario" making its risk estimate too high and thus conservative. |
Conclusions
ATSDR bases its PHAs on mildly conservative data and scenarios and not on the most conservative data and scenario. To do otherwise would be to base public health actions on highly improbable scenarios which would never be realized in real life. The money spent would be better spent remedying high risk scenarios. If one considers ATSDR's long established policies and analyses this site properly, the conclusion is "ATSDR did it correctly".
Alternative Scenarios
There have been some suggestions that the scenario comprising the digging in the floodplain for a residential building will expose the worker to the strata with higher levels of mercury. But how many people will build a residence in an area that is frequently flooded with 1-4 feet of water. The same goes for vegetable gardens; the families that tried on 102 Wiltshire Drive found that 4 feet of flowing water is not good for green beans and other crops. Further, workers will not spend years digging in the floodplain and his/her exposure will be intermittent and limited. The land is good for pasture and that's about all. Exposure scenarios must be probable before they are suitable for public health decisions. To be a probable scenario for public health purposes the scenario must involve a significant number of the general population, not the worker population in workplace scenarios. The exposure path must not only be possible but also probable; it must also deliver a significant contaminant dose to the population. The allegation that an exposure scenario is important without some attempt to analyze its path and impact is not very productive.
The general criteria for a significant public health scenario are:
1) There must be a detectable presence of a contaminant in the environment.
2) The contaminant must have a documented, observable, adverse medical effect in humans.
3) There must be a significant exposed, human population or subpopulation.
4) There must be one or more documented exposure pathways which deliver a predictable dose to the subpopulation.
5) There must be sufficient data, human or animal, to establish a supportable comparison dose to humans at which no adverse effect occurs. OR, if a risk approach is used, the risk response function must be known.
6) If a sta6istical model is used, the above information must be available as unbiased distribution functions where appropriate.
If any one of the first four criteria is not met the dose due to the scenario is zero. If one or more of the criteria are minimally met then the dose will be small or very small. The significance of a scenario cannot be determined except by the evaluation of the exposure path.
Preliminary Screening and Follow-up
Scenarios
Accurate scenario evaluation is neither easy nor cheap; scenarios are complex and good data is missing and world is contaminated by natural processes; Failure to find an exposure when one exists can be deadly, the over estimation of all risks will destroy the earth's surface and bankrupt a nation. Great care and experience are needed. The assumptions and data used to evaluate a scenario are clearly a part of that scenario for they determine the outcome as surely as does the assumptions that give rise to the physical description of the scenario and its exposure pathway. A new set of data with the same physical scenario is a new scenario. Scenarios fall into two main classes based on their use and conservatism:
1. Preliminary screening scenarios which are intended to eliminate any candidate contaminants from further consideration by using deliberate, conservative assumptions and data for the pathway evaluation. Any contaminant found significant in the screening scenario will be further evaluated in a more accurate follow-up scenario. The screening scenario is usually evaluated with the data at-hand or readily available from the literature. If such a scenario
2. Follow-up scenarios are those evaluated with the best assumptions and data which are still conservative enough to ensure that false negative results will not be probable. This usually requires a more thorough search for data to support the estimation of a comparison value AND the exposure dose. If data does not exist then new measurements may be required. Under many circumstances empirical body levels may be measured on the human population thus eliminating questions about the exposure pathway, If final scenarios are too conservative, there will be needless remediation and public concerns.