Environmental Forensics: Principles and Applications - Chapter 6 ppt

Intentional errors include:• Exaggerated vertical or horizontal scales • Selective data presentation • Data contouring manually and computer-generated • Color-coded data that obscure sou

1990, 1997) Intentional errors include:

• Exaggerated vertical or horizontal scales

• Selective data presentation

• Data contouring (manually and computer-generated)

• Color-coded data that obscure source areas

• Contaminant transport models based on biased data

When trial exhibits are exchanged, a concerted effort is required to validate theiraccuracy Obtain the underlying information such as chemical results, especially in

an electronic format, early in the discovery stage so that your expert witness and/orconfidential consultant can quickly review the underlying data used to produce thetrial exhibits Determining that a trial exhibit is scientifically accurate benefits allparties

6.2 EXAGGERATED VERTICAL

AND HORIZONTAL SCALES

Exhibit scales are often exaggerated, especially for geologic cross-sections and fencediagrams When portraying a relatively small vertical scale, such as shallow soil

contamination (<100 ft.), relative to a substantially larger horizontal scale (>1000 ft),exaggeration is a reasonable way to present the data Conversely, the depiction ofsubsurface contamination is skewed by excessively increasing the vertical scalerelative to the horizontal scale When vertical or horizontal scale exaggerationoccurs, it should be posted on the exhibit and described in the testimony so that theviewer is informed Plate 6.1* depicts the concentration of trichloroethylene (TCE)

in soil with a 1:1 and 1:10 vertical-to-horizontal scale (Morrison, 1998) The TCEdistribution is represented as an iso-surface for the purpose of depicting the volume

of contaminated soil While the respective horizontal-to-vertical ratios are accurate

in both versions, the perception regarding the extent of contamination is different.

Exhibits relying on this technique are routinely employed in cases that address thereasonableness of remediation costs When an exhibit prejudices the observers’perspective, prepare a rebuttal exhibit with a 1:1 vertical-to-horizontal scale with thesame data or decrease the three-dimensional area so that the exaggeration bias isreduced

6.3 SELECTIVE DATA PRESENTATION

It is the author’s experience that omission of selective data is common in mental exhibits Observed practices include:

environ-1 Data omission

2 Use of averages or mean data (i.e., obtaining the average of quarterly data, moving averages, geometric means, time series presentations; using averaged values, aver- aged value with standard variation, mean plus confidence interval, measured value plus the percentage of relative standard deviation or coefficient of variation, etc.) which results in an underestimation of contaminant concentrations and plume geometry

3 Selection and presentation of the higher or lower value from split samples

4 Creation of multi-chemical composite contour maps (i.e., combining all solvent measurements and reporting them as total volatile organic compounds [VOCs] rather than for each compounds) to mask source identification

5 Arbitrary elimination of anomalous data

6 Data presentation generated from imprecise or non-specific analytical methods

7 Data filtering to reduce or eliminate reported measurements

8 Aerial photo cropping

9 Arbitrary revisions to the original data

There is usually client reluctance to spend the money required for exhibitvalidation, especially when large data sets are used For large invalidated data sets(>1000 data entries), a 5 to 15% transcription error between laboratory data and thecomputer spreadsheets is common If the data entered onto the spreadsheet are doubleentered or cross-checked, this error is significantly reduced

* Plate 6.1 appears at the end of the chapter.

Validating a large data set (e.g., ≥100,000 entries) when the exhibit is based on

500 data points is unproductive Identification of transcription errors is cost tively performed by validating only those locations and compounds used in keyexhibits This strategy requires that the underlying chemical data sheets are quicklyaccessible once the exhibits are exchanged Once the data used to create an exhibit

effec-is validated, it effec-is used with the identical modeling and/or veffec-isualization software todetermine if the trial exhibit is reproducible If the animation software is proprietary,additional cost and time can result in purchasing or licensing the software from thecompany In addition, the software may require unique hardware as well as a personfluent with the software These hardware, software, and personnel issues should beresolved in advance of receiving the exhibits

Confirmation of the validity of data used to generate an exhibit may not bestraightforward Consider 100 split soil samples collected and tested for trichloro-ethylene Is it more appropriate to use the lowest, highest, or average of the twovalues in the exhibit or to plot all three? If a trial exhibit relies on averaged values

in some instances and alternates between high and low values for others, determine

if a pattern of intentional data bias exists A consistent method should be used andthe rationale for the selection clearly stated on the exhibit and/or testimony.Exhibits that rely upon the geometric mean of a data set are often encountered,

as water quality results are generally distributed geometrically in time and space Thegeometric mean is obtained by taking the log of multiple values, adding the log, andthen taking the anti-log of the averaged log values This technique tends to dampenthe impact of data outliers or individual anomalous values that may be important inidentifying contaminant sources Similarly, other statistical averaging techniques thatassume a normal distribution should be confirmed Minimization of biases due toconcentration averaging, geometric means, and mean values is accomplished byusing the actual values for a point in time This latter approach improves the validity

of the data interpretation, transport modeling, and ultimately the effectiveness of theremediation design (Martin-Hayden and Robbins, 1997)

The interpretation of non-detect (ND) results can skew the results of the data setused to create an exhibit A sample reported as ND can be interpreted as 0, as thevalue of the method detection limit, or as a value of one half the detection limit oromitted in the data set If the geometric mean of a data set is used, the centraltendency of the geometric mean will be significantly different if non-detects areexcluded vs if values equal to one half the method detection limit are used.For time-series data using single or averaged data (e.g., 10 years of quarterlygroundwater sampling data), graphing data from a single quarter or averaging valuesfor several quarters can skew the viewer’s perception if the chosen quarter(s) areanomalous relative to historical trends Combining 6 or 12 months of non-sequentialgroundwater data (e.g., annual, quarterly, and biannual sampling) for an aquifer with

a high velocity (e.g., >1000 ft per year) onto one exhibit when monitoring wells arespaced less than 1000 ft apart results in an unrepresentative portrayal of contaminantdistribution Creating a rebuttal exhibit depicting seasonal or more consistent histori-cal trends including anomalous sampling quarter data places the trial exhibit in amore balanced historical perspective

If sample integrity is suspected due to sampling bias, especially for volatilecompounds collected by multiple consultants, plotting the chemical results vs timeand labeling the tenure of the various consultants may identify whether this potentialexists Figure 6.1 illustrates trichloroethylene (TCE) concentrations in groundwatersamples collected from multiple wells by three consultants between 1991 and 1994.

In Figure 6.1, the TCE values for samples collected by Consultants A and B between

1991 and 1993 are smaller than the TCE concentrations from samples collected byConsultant C The higher TCE concentration collected by Consultant C may indicatethe use of different sampling equipment or procedures

Valid reasons exist for eliminating anomalous (e.g., outlier) values from a dataset used to create an exhibit; however, the presence of anomalous data may be theonly indication that the data are skewed and hence may be one of the most importantdata points in the population If data are omitted, the rationale should be prominentlyposted on the exhibit An example of omitted data is the presentation of changes ingroundwater flow direction via rose diagrams Figure 6.2 depicts the frequency of thegroundwater flow direction from quarterly monitoring reports The purpose of Figure6.2 is to demonstrate that a contaminant plume in groundwater is captured with agroundwater extraction system located downgradient of the source The groundwaterextraction system was designed to capture the contaminant plume when the ground-water flow direction is west to southwest (Figure 6.2A) Groundwater flow to thenorth results in the transport of contaminated groundwater beyond the capture zone

of the extraction system Figure 6.2A shows nine quarters of groundwater flow that

is predominately to the west to southwest Figure 6.2B is a rebuttal exhibit depictingall 13 quarters with the direction of flow alternating between the southwest andnortheast Figure 6.2A does not contain reference to the omitted data

FIGURE 6.1 TCE concentrations from five groundwater monitoring wells collected by three

consultants between 1991 and 1994.

While omission of anomalous data adverse to one’s position is usually obvious,subtle permutations are also encountered An example is a chemical or geologiccross-section A cross-section is a slice through the subsurface with informationintersected by the slice displayed two or three dimensionally A common cross-section manipulation is the inclusion or omission of data points not intersected on thecross-section Figure 6.3(3a) depicts a plan view of a cross-section (A-A ¢) that

intersects total petroleum hydrocarbon (TPH)-impacted soil Figure 6.3(3b),

how-ever, is the actual transect line reflecting the sampling points from which soil

chemistry was used in the cross-section In the case of the transect A-A¢ in Figure6.3(3b), data along the transect that were not used included locations S-EX7 and S4-3-PL Sampling locations within 5 ft of the A-A¢ transect from locations S1-EX3 and

S9-EX5 (see 3a) were also omitted Sample locations located 30 ft to the east 5-PL, S9-7-PL), however, were projected onto the A-A¢ transect in 3a and incorpo-

(S5-rated into the accompanying cross-section

Another example of data omission is the exclusion of non-CLP (Contract ratory Program) data Contract Laboratory Program data are the documentationrequired for sample testing associated with the Comprehensive Emergency CostRecovery Act (CERCLA) or Superfund and Resource Conservation and RecoveryAct (RCRA) investigations The primary components of this program include fieldand/or trip blanks, field duplicate sample results, and internal laboratory qualitycontrol results (e.g., matrix spikes, matrix spike duplicates, and laboratory methodblanks) Historical CLP and non-CLP (e.g., Phase I or II investigations) may not beavailable If non-CLP data are excluded in an exhibit, plot the CLP and non-CLP dataand compare the results If one component of the CLP documentation is unavailable

Labo-or has been violated (e.g., broken travel Labo-or field blank bottles) and is included in theexhibit, create the same graph or figure with and without the suspected data

FIGURE 6.2 Rose diagrams showing historical groundwater flow directions (From Morrison,

R., in Environmental Claims Journal, 11(1), 93–107, 1998 With permission.)

Another data omission example is the case of split samples from one laboratoryusing CLP procedures and a second data set with non-CLP documentation Plot thesplit CLP and non-CLP sample data collectively and individually to determine ifsignificant differences in interpretation occur If the non-CLP data are significantlydissimilar, the non-CLP data can be used for a different purpose (e.g., qualitatively

vs quantitative) or weighed differently For example, the CLP data may be used forrisk assessment purposes or to provide a quantitative measure of the volume of soilexceeding a clean-up concentration The combined CLP and non-CLP data can beused to establish the boundaries of the contamination

Determining the reasonableness of an analytical method relied upon to create anexhibit may be required In Plate 6.2,* 24 soil samples from a soil excavation are splitinto three discrete samples, with each sample forwarded to an analytical laboratoryand tested for total petroleum hydrocarbons as gasoline When each data set iscontoured, different contaminant source areas as well as volumes above a remediationconcentration of 100 mg/kg occur A plan view of the contours from the Method 3data depicts three source areas, while Method 1 and 2 data indicate two source areas.The exhibit relying on the Method 3 data or an average of the three data sets willresult in significantly different interpretations of the distribution of the TPH in thesoil The data set selected for the exhibit influences the interpretation regarding thelocation of TPH contamination The solution is to perform an analysis of the repre-sentativeness or accuracy of each analytical method to determine which data set ismost representative In Plate 6.2, Method 1 introduced false positive readings, whilethe methanol extract used in Method 2 was less effective in contaminant removal than

* Plate 6.2 appears at the end of the chapter.

FIGURE 6.3 Plan view of cross-section transects A-A¢ (From Morrison, R., in tal Claims Journal, 11(1), 93–107, 1998 With permission.)

Environmen-Method 3 For this soil type and contaminant, Environmen-Method 3 is the most representativedata set.

A variation of the Plate 6.2 example is reliance on a testing technique such asEPA Standard Method 418.1 to detect total petroleum hydrocarbons (TPHs) in soilsamples used to guide the excavation of hydrocarbon-impacted soil EPA Method418.1 is a non-chromatographic technique and detects the presence of biogeniccompounds in the soil (i.e., peat, pine needles, organic matter) resulting in false-positive measurements (George, 1992; Zemo et al., 1995) The author has observedcases when EPA Method 418.1 is used to define where to excavate, until theexcavation is inhibited by the presence of a building or road The consultant thenchanges to an analytical method that does not introduce a false bias (i.e., EPA Method8015) Testing using EPA Method 8015 results in non-detect sample measurementsand becomes the basis for halting the excavation Whether the original excavationusing EPA Method 418.1 was warranted becomes not only a source of contention butalso affects the reliability of an exhibit combining test results using EPA Methods418.1 and 8015

Figure 6.4 is a cross-section of a soil excavation where soil samples were testedfor TPHs using EPA Standard Methods 418.1 and 8015 Soil samples collectedwithin the interior of the excavation were tested via EPA Method 418.1, while EPAMethod 8015 was selected for confirmation soil sampling along the excavationperimeter The potential implications of this observation are that over-excavationprobably occurred and that the consultant may have intentionally relied upon thefalse-positive bias results inherent with EPA Method 418.1 to excavate non-petro-leum-contaminated soil as a means to generate income EPA Method 8015 was thenused to halt the excavation, in this case when its proximity to subsurface pipingpresented significant complications to continued excavation Once the excavated soil

FIGURE 6.4 Excavation cross-section using EPA Methods 418.1 and 8015.

is remediated or co-mingled with other petroleum-impacted soil, it becomes lematic whether subsequent test results of these excavated soils can determine if theoriginal EPA Method 418.1 results were valid.

prob-Figure 6.5 depicts a plan view of excavated gasoline-contaminated soil Theorganic-rich subsurface soils provided consistent false-positive measurements whenusing EPA Standard Method 418.1 Once the excavation proceeded close to a coolingtower and manufacturing building, the consultant switched to soil analysis using EPAStandard Method 8015 which resulted in non-detect sample results Excavation nearthe surface structures then ceased The distribution of analytical methods used for soilanalysis relative to the above-ground structures in Figure 6.6 suggests an intent tocreate non-detect boundaries in areas in which extensive shoring was required

It may be warranted to retain an analytical chemist to reconstruct the validity ofthe test method(s) used to direct a soil excavation The chemist can identify databelieved to be unreliable which should be omitted from an exhibit Conversely, if noquality assurance analysis is performed, both parties may erroneously assume that thedetection of a particular compound is correctly identified It is the author’s experi-ence that in the case of gas chromatography/mass spectrophotometry (GC/MS), it isnot unusual to find that 5 to 10% of the compounds are misidentified, especially ifthe interpretations are not manually examined

FIGURE 6.5 Plan view of soil excavation and selective use of EPA Standard Methods 418.1

and 8015.

Data filtering is the revision or omission of data based on identification of theremoved data as anomalous and/or non-representative An example is the detection

of 20 parts per billion (ppb) of TCE in a rinsate sample collected from a groundwaterbailer The bailer is subsequently used to collect a groundwater sample that results

in a reading of 24 ppb The data (24 ppb) are omitted from the data set based on aconcentration of 4 ppb (lower than the maximum contaminant level of 5 ppb) viasubtracting the equipment blank value from the measured groundwater sample.Another example of data filtering is assigning a new detection limit at five timesthe contamination level detected in the rinsate sample The new detection limit istherefore 5 ¥ 20 = 100 ppb The detection of 24 ppb in the groundwater sample is now

regarded as non-detect, as are trichloroethylene concentrations up to 100 ppb Thismethod results in significant data omissions

Data filtering may be represented as justified through re-sampling For example,monitoring wells may be re-sampled immediately after contamination is detected orre-sampled several times until contamination is not detected The non-detect sample

is then reported in the quarterly groundwater monitoring report and relied upon forthe trial exhibit Another technique is repeated groundwater sampling at the samelocation using a cone penetrometer test (CPT) rig or less quantitative technology (soilgas), with the re-sampling occurring days, months, or years after the original results

to confirm the use of the non-detection measurements shown on a trial exhibit.For data sets where measurements are omitted, the major difficulty often lies inidentifying the omissions For large data sets (>1000 entries), omissions may not beapparent without a thorough review Another difficulty is the testing of split samples

by multiple samples, with only those sample results supportive of a particularposition being reported One technique for identifying data omissions is to aggres-sively pursue any electronic databases kept by the consulting firm or facility operator.Another option is to subpoena the original laboratory sheets and create a separatedatabase

Aerial photo cropping is a technique that can remove undesirable information.Figure 6.6 shows two versions of an aerial photo of a tank farm in 1925 — uncroppedand cropped; the cropped version deletes a tank under construction in the upper leftcorner Be aware that when a person selects an aerial photo from a repository ordealer, the portion that is selected for the hard copy is usually a subset of the original,usually due to the scale of the parent aerial photograph When ordering aerialphotography, a number of scales and coverage dates are available It is the author’sexperience that all of the coverage dates are rarely ordered This can result in theomission of aerial photo information if the opposing side obtains copies duringdiscovery and relies on these rather than independently obtaining their own aerialphotographs

When forensically evaluating a trial exhibit, examine all the underlying tional information, especially field and laboratory notes Figure 6.7 depicts a fieldand final soil-boring log contained in an environmental report The field log depicts

founda-a 3-ft zone of contfounda-aminfounda-ation, while the finfounda-al log contfounda-ained in the environmentfounda-alreport shows a contaminant zone that is 7 ft thick The final boring log was used withother boring logs to estimate the volume of contamination and associated remediation

costs While the difference between the field and final boring log is small (ª4 feet),

this difference extrapolated over a large area results in substantial differences incontaminant distribution and associated remediation costs

displays showing regions of elevated concentrations Contouring can identify taminant source areas and is useful for designing remedial programs Whether two

con-or three dimensional, contouring fcon-orms the foundation fcon-or most environmental its depicting chemical/spatial information (Joseph, 1996) Contour reliability is afunction of the following items:

exhib-• Data density

• Mathematical contouring method

• Nature of the contaminant (a pure phase liquid vs a dissolved phase contaminant)

• Site-specific geologic environment that controls contaminant movement ment in fractured bedrock vs in a uniform fine sand)

(move-The last item is important, as geologic environments may be encountered (e.g., ahighly heterogeneous aquifer) for which contouring of a dissolved contaminant may

be misrepresented by concentration contouring In cases such as groundwater tions and LNAPL thickness on groundwater, however, contouring may be moreappropriate in the same setting

eleva-FIGURE 6.7 Field and final soil boring logs (From Morrison, R., in Environmental Claims

Journal, 11(1), 93–107, 1998 With permission.)