Procedures for Determining Site-Specific Background

Report
CPANS – 2012 Spring Conference
Procedures for Determining Site-Specific
Background Conditions and Their Impact on Site
Remediation
April 24, 2012
Authors and Presenter
Authors
Anne G. Way
Tai. T. Wong
Yong Li
James G. Carss
Presenter
Anne Way, P. Chem.
[email protected]
O’Connor Associates - A Parsons Company
Calgary, Alberta
2
 What is background?
 Provincial and
Federal Guidance
 Methodology
 Picking locations
 Calculation
 What about outliers?
Outline
 Case studies
3
What is Background?
 Concentration of a substance in the environment
that can be attributed to natural sources
 Can also include anthropogenic sources, as long as they
are not specifically related to site activities
 Can vary regionally and locally based on the soil
and bedrock present
 So use of federal/regional background data for sitespecific remediation is generally not a good idea
4
Provincial and Federal Guidance
“Recommended” Procedures for Calculating Background
Source
Ontario
MOE
BC
MOE
AEW
Method Description
Soil
97.5 percentile
95 percentile
Outlier removal:
data ≥ Q3 + (1.5 x IQR)
data ≤ Q1 – (1.5 x IQR)
95 percentile
Outlier removal:
Groundwater
Reference
97.5 percentile
ON MOE 2011. Rationale for the Development of
Soil and Groundwater Standards for Use at
Contaminated Sites in Ontario
95 percentile
BC MOE 2004. Protocol 9 for Contaminated Sites
– Determining Background Groundwater Quality
BC MOE 2005. Technical Guidance on
Contaminated Sites #16, “Soil Sampling Guide
for Local Background Reference Sites
Not specified
AENV 2009. Subsoil Salinity Tool
data ≥ 2 x stdev + mean
USEPA
Other
95 % UPL (normal dist)
Outlier removal:
95 % UPL (normal dist)
Outlier removal:
Rosner or Dixon’s test
Rosner or Dixon’s test
95 percentile (non-parametric)
Outlier removal:
95th percentile (non-parametric)
Outlier removal:
Rosner or Dixon’s test
Rosner or Dixon’s test
mean + (3 x stdev)
mean + (3 x stdev)
Singh and Singh 2010. ProUCL Version 4.1.00
Technical Guide (Draft), EPA/600/R-07/041
5
Methodology - Locations
 Background locations need to match onsite conditions, but
are unaffected by anthropogenic activities
 As close in distance to site as possible
 Can use non-impacted onsite areas in SOME cases (EPA 2002)
 Not influenced by site activities (upgradient, up-wind, up-hill)
 Match geological strata represented by site characterization data
 Representative of range of soil samples to which they will be compared
(more than one area may be required)
 In most cases, this idealized background location does not
exist
EPA. 2002. Guidance for Comparing Background and Chemical Concentrations in Soil for CERCLA Sites. EPA 540-R-01-003, U.S. Environmental Protection Agency,
Office of Emergency and Remedial Response, Washington, DC.
6
Methodology - Locations
 Complications




Complex site history
Incomplete site characterization/conceptual site model
Minimal resources
Limited availability of background information
 e.g., Sites located within cities
 How many locations?
 The more the better!
Larger number of samples  more accurate estimate  lower error rates
7
Methodology - Calculations
 Histograms
 Assess shape of data
 Symmetric (normal distribution)
 Skewed (logarithmic, other)
 Assess spread of data
 Tightly clustered around a certain value?
 Stay within certain limits?
8
Methodology - Calculations
 Box and whisker plots
 Shows the shape, central tendency and variability of the data
 Useful for comparing several data sets
9
Methodology - Calculations
 Percentiles
 The nth percentile has n % of the data below it and
(100-n) % of the data above it
 Based on your current data set
 Will change with additional data
 Prediction Limits (PLs)
 The upper bound of the associated prediction limit (UPL)
 “about 95% of the time, or I am 95% confident that, the next
future observation taken will be less than X”
10
Methodology – What About Outliers?
 “An observation that does not conform to the pattern
established by other observations” (Hunt et al. 1981)
 An unavoidable problem
 Sources of outliers
 Recording, transcription, data-coding errors
 Calibration problems, unusual sampling conditions
 Manifestations of larger spatial or temporal variability than
expected
 e.g., small-scale variability within individual soil samples
 Indication of unsuspected contamination
Hunt, W.F., Jr., Akland, G., Cox, W., Curran, Frank,N., Goranson, S., Ross, P. Sauls, H., and Suggs, J. 1981. U.S. Environmental Protection Agency Intra-Agency Task Force
Report on Air Quality Indicators, EPA-450/4-81-015. Environmental Protection Agency, National Technical Information Service, Springfield, Va.
11
Methodology – What About Outliers?
 May be
a) True measurements of conditions on-site
b) An actual error
 Must identify which class the outlier falls into!
 Both outlier tests and a qualitative review of field and
laboratory data should be used to determine if the data
point should be eliminated from the data set
 Many different types of outlier analysis
12
Methodology – What About Outliers?
 EPA (2002) recommends 5 steps to treat outliers
1. Identify extreme values that may be potential outliers

Box plots, histograms
2. Apply statistical tests

Dixon’s, Rosner’s, others….
3. Review statistical outliers with qualitative field and
laboratory data

Decide on their class (true measurement or error)
4. Conduct data analysis with and without outliers
5. Document everything!
EPA. 2002. Guidance for Comparing Background and Chemical Concentrations in Soil for CERCLA Sites. EPA 540-R-01-003, U.S. Environmental Protection Agency, Office of
Emergency and Remedial Response, Washington, DC.
13
Case Study 1
 Site in Alberta
 Former oilfield facilities: well, pump jack, scrubber shack, storage
tanks, pipelines
 Currently agricultural land use
 Stratigraphy: silt and/or clayey silt with inter-bedded discontinuous
sand lenses
 Fine-grained, fine-textured soils
 Salinity related contaminants of concern in soil
 Tier 1 Salt Contamination Remediation Guidelines
(SCARG) evaluation
 Need to calculate background for EC and SAR
 Limited site characterization and budget
14
Case Study 1
Area of Potential
Impact
Background
locations
Well
Head
 Background locations
 Representative of unimpacted soil
 Collected near area of
potential salinity
impact
 Used to identify a soil
rating category
 Upper limit of the soil
rating category
becomes the
guideline
15
Case Study 1
Histograms not very useful in this case
16
Case Study 1
 Boxplots of background data
vs. Area of Potential Impact
(AOPI) Data
 EC and SAR data
distributions are very similar
in background vs. AOPI
 IQR is smaller in the AOPI since there
is more data
 Medians very similar
 No identified “potential outliers”
 Indicates that any elevated
EC/SAR located in the AOPI may
be natural and not-site related
 All depths included
17
Case Study 1
Electrical Conductivity (dS/m) 0 – 0.3 m 0.3 – 1 m 1 – 1.5 m > 1.5 m
Summary Stats (w/o outlier removal)
Number of Observations
6
7
7
22
 Calculation of
0.5
4.2
7.8
Background
0.14
5.9
2.4
 Sub-divided into
different depths
0.7
14.0
13.0
intervals
0.4
0.4
3.7
 USEPA UPL
Background Calculation Methods
method could not
0.7
13.4
12.5
be performed:
Average
0.4
Standard Deviation
0.14
Maximum
0.6
Minimum
0.3
ON MOE (97.5th Percentile)
0.6
BC MOE (95th Percentile)
0.6
0.7
12.8
12.0
AEW (95th percentile with outlier removal)
0.6
0.7
12.8
12.0
USEPA
95% UPL with outlier removal (normal dist)
97.5th Percentile with outlier removal
ID
0.6
ID
0.7
ID
13.4
ID
12.5
Other (average + 3 Stdev)
0.9
0.9
24.7
15.0
 Data not quite
normal
 Not enough data
(8-10 min)
18
Case Study 1
Electrical Conductivity (dS/m) 0 – 0.3 m 0.3 – 1 m 1 – 1.5 m > 1.5 m
Summary Stats (w/o outlier removal)
Number of Observations
7
6
6
22
 Calculation of
2.7
6.5
11.3
Background
3.26
5.3
4.5
 Sub-divided into
different depths
7.0
14.0
15.0
intervals
0.23
0.7
1.0
 USEPA UPL
Background Calculation Methods
method could not
7.0
13.5
15.0
be performed:
Average
1.07
Standard Deviation
2.1
Maximum
5.4
Minimum
0.16
ON MOE (97.5th Percentile)
4.8
BC MOE (95th Percentile)
4.1
7.0
12.9
15.0
AEW (95th percentile with outlier removal)
0.3
7.0
12.9
15.0
USEPA
95% UPL with outlier removal (normal dist)
97.5th Percentile with outlier removal
ID
0.3
ID
7.0
ID
13.5
ID
15.0
Other (average + 3 Stdev)
7.4
12.5
22.3
24.8
 Data not quite
normal
 Not enough data
(8-10 min)
19
Case Study 1
 Within each depth interval, background is calculated and
used to identify a soil rating category
 Upper limit of the soil rating category becomes the
guideline for that depth interval
Identified Soil Rating Category for SAR 0 – 0.3 m
0.3 – 1 m
1 – 1.5 m
> 1.5 m
Background Calculation Methods
ON MOE (97.5th Percentile)
Fair (8)
Fair (8)
Unsuitable (14)
Unsuitable (15)
BC MOE (95th Percentile)
Fair (8)
Fair (8)
Unsuitable (13)
Unsuitable (15)
AEW (95th percentile with outlier removal)
Good (4)
Fair (8)
Unsuitable (13)
Unsuitable (15)
USEPA
95% UPL with outlier removal (normal dist)
97.5th Percentile with outlier removal
ID
Good (4)
ID
Fair (8)
ID
Unsuitable (14)
ID
Unsuitable (15)
Other (average + 3 Stdev)
Fair (8)
Unsuitable (13)
Unsuitable (22)
Unsuitable (25)
20
Case Study 1
Different excavation area depending on background calculation method
Excavation required
based on BC, MOE,
AEW, USEPA and
“other” method
Additional excavation
required based on
AEW and USEPA
method
21
Case Study 2
 Former fertilizer facility in Manitoba
 Fertilizer contaminants in soil and groundwater
 Native soil profile: inter-layered silt and clay to
4.4 mbg, fractured bedrock below
 Groundwater in overburden: 1 mbg
 Depth to groundwater in bedrock: 11 to 14 mbg
 Groundwater ingestion pathway a concern
 Water wells within 500 m of the site
 No aquitard between the impacted zone within
overburden and the underlying bedrock aquifer
22
Case Study 2
 Calculation of background nitrate in groundwater required
 Possible non-site related anthropogenic sources from residential
septic tanks located upgradient of site
 Suspect that background nitrate is greater than the CCME drinking
water standard of 10 mg/L
 Required a soil clean-up criteria to delineate site-related nitrate
impacts
 Calculated based on background groundwater criteria
 Complications
 Choosing appropriate background locations
 Potential seasonality of groundwater concentrations
 Difficulty in determining groundwater flow direction
23
Case Study 2
Background locations in blue
500 m
radius
line
 Representative of un-impacted soils
 Takes into account potential, non-site
related sources of nitrate
 Collected upgradient from site
Downgradient locations in green
Site
24
Case Study 2
No significant seasonal fluctuations
Site data has different distribution than background data
25
Case Study 2
 All data, fall and spring
distributions are very similar to
each other in both background
and site data
 Negligible seasonal variability
 Background data distributions
different than Site data
distributions
 IQR is larger in the site data
 More identified “potential outliers”
on site
 Indicates that elevated nitrate
concentrations onsite are siterelated (above background)
26
Case Study 2
Nitrate (mg/L)
All Seasons
Summary Stats (w/o outlier removal)
Number of Observations
53
Average
7.2
Standard Deviation
3.6
Maximum
20.0
Minimum
1.1
Background Calculations
ON MOE (97.5th Percentile)
15.7
BC MOE (95th Percentile)
13.2
AEW (95th percentile with outlier removal)
11.0
USEPA
95% UPL with outlier removal (normal dist)
97.5th Percentile with outlier removal
NA
15.7
Other (average + 3 Stdev)
18.0
 Background calculation
method comparison
 Since no significant
seasonal fluctuations, only
calculated for all seasons
 All site-specific
backgrounds are above
CCME drinking water
standard of 10 mg/L
27
Case Study 2
Nitrate Groundwater (mg/L)
Soil (mg/kg)
Background Calculations
ON MOE (97.5 Percentile)
15.7
3.4
BC MOE (95 Percentile)
13.2
2.8
AEW (95 percentile with outlier removal)
11.0
2.4
USEPA
95% UPL with outlier removal (normal dist)
97.5 Percentile with outlier removal
NA
15.7
3.4
Other (average + 3 Stdev)
18.0
3.9
Partitioning calculation for site-specific soil criteria
Using standard CCME parameters
C soil  C water

 K


OC
f OC
 W  H ' a

b




28
Case Study 2
Different excavation area depending on background calculation method
29
Conclusions
 BC MOE and ON MOE methods are very similar
 Methods including outlier analysis yield different results than
those that don’t
 Take care in the identification and treatment of outliers
 Other methods (e.g, average + 3 x stdev) often yield
background values that are above the data maximum
 Use at your own discretion
30

similar documents