Title: Multiagency radiation survey and site investigation manual MARSSIM : Overview
1 Multi-agency radiation survey and site
investigation manual (MARSSIM) Overview
2MARSSIM Background
- Thousands of sites with known or potential
radioactive contamination - EPA, NRC, DOD and DOE have, use, or will use
dose- and risk-based regulations for site release - Different agencies use or have used different
guidance to demonstrate compliance - Sites vary - e.g., a single room in a building to
large weapons complex sites - Contamination can be uniform or there can be
small areas of elevated radioactivity, even after
remediation - Many site contaminant radionuclides are present
in background
3MARSSIM Timeline
- Developed by DOD, DOE, EPA, NRC
- Effort began in 1994
- Originally released December 1997
- Revision 1 released August 2000
- Additional updates released in June 2001 and
August 2002 - Additional updates discussed and approved yearly
at April Workgroup Meeting
4MARSSIM Scope
- Surface soils and building surfaces
- How to determine if regulatory release criteria
are met - How many measurements should be taken
- What measurement methods to use
- Scope does NOT include
- Selecting the release criterion
- Translating dose or risk into concentrations
- Groundwater or drinking water compliance
- Subsurface soil
- Building materials and release of components
- Evaluation of remedial alternatives
- Public involvement
5MARSSIM DQO/DQA Process
DCGL
Release criterion
Evaluate
Statistical Sampling Design and Analysis
Exposure pathways
Small area of elevated activity
Compare to
Radioactive Contaminant
Reference Area
Survey Unit
Survey Unit
Survey Unit
6What do we need to knowto release a site?
- 1) That the average concentration of residual
radioactivity in a survey unit results in a dose
less than the limit - and
- 2) That there are no hot spots that have been
missed
7Whats Old Whats New
- Dose Based
- Based on Contamination Potential
- Flexible, Integrated Scanning and Sampling
- Nonparametric Hypothesis Test
- Release Criterion
- Concentration/Activity Based
- Survey Unit Classification
- Affected vs Unaffected
- Survey Design
- Fixed Sampling Density
- Statistical Methodology
- Parametric Students t Confidence Interval
8Release Criterion
Dose Based
Derived ConcentrationGuideline LevelDCGL
Dose converted to a measurable quantity using
exposure pathway models Levels presented in terms
of ambient radiation, surface activity levels, or
soil activity concentrations
9Survey Unit Classification
Based on Contamination Potential
Class 1 Class 2 Class 3
Purpose
- Not all areas of the site will have the same
potential for residual contamination - so not all areas require the same level of survey
coverage to achieve an acceptable level of
confidence that the site satisfies the
established release criteria
10Integrated Survey Design
Flexible, Integrated Scanning and Sampling
- Direct Measurements
- field measurements or samples on a grid to
find the average level of contamination - together with
- Scanning
- to find Elevated Areas
11Nonparametric Statistics
- Wilcoxon Rank Sum Test
- ...when there is a background
- Sign Test ...when there isnt
- Do not assume the data are Gaussian
- (can handle skewed data)
- Can handle non-detects (less-thans)
- Not sensitive to outliers
- Work nearly as well as parametric tests even if
the data are normally-distributed
12Elevated Measurement Comparison
- To flag potential failures in the decommissioning
process over elevated areas within a survey unit. - How big is an elevated area?
- What concentration is considered elevated?
- An elevated area is one that has a
concentration which results in a dose
exceeding the release criterion.
13(No Transcript)
14(No Transcript)
15(No Transcript)
16(No Transcript)
17(No Transcript)
18MARSSIM WorkgroupDevelopment of Manual
Supplements
- Two supplements under development
- Multi-Agency Radiation Survey and Assessment of
Materials and Equipment (MARSAME) - Multi-Agency Radiation Survey and Assessment of
the Subsurface (MARSAS)
19Multi-Agency Radiation Survey and Assessment of
the Subsurface (MARSAS)
- PURPOSE
- To develop Final Status Survey design methods
and protocols that parallel MARSSIM for
performing final status surveys in subsurface
soils.
20 MARSAS Framework Questions Before the Workgroup
- How do you make the decision that the subsurface
is non-impacted? - How do you determine that remediation is
necessary? - How do you defend/justify leaving known or
suspected subsurface radioactive material
in-place?
21Active Research Areas
- Sampling Costs and analytical costs are higher
- How do we do more with less?
- Better Survey Design using site knowledge as a
guide (Bayesian) - Better Data Analysis using more sophisticated
statistics (geostatistics) - Dispersed plume versus discrete sources
- Elevated volume
- Cannot scan 100 in Class 1 How do we keep
confidence High and uncertainty Low? - Incorporating surrogate data into the decision
process (e.g., geophysical, hydrological data) - Incorporating professional judgment data into the
decision process quantitatively - HOW DO YOU MAKE IT USABLE?????
22Additional Information
Visit the MARSSIM website at http//www.epa.gov/r
adiation/marssim/
23Statistics test over the entire survey unit...
G
using data taken on a random start grid.
24EMC is used to detect small areas...
primarily by scanning methods.
25The largest area with any real chance to be
missed by the grid ...
0.866 G2
is the grid area defined by neighboring sample
points.
26If the concentration equals the DCGL over the
entire survey unit...
the resulting dose is 25 mrem/y.
27If the residual radioactivity is confined to a
smaller area...
the concentration must be higher to give 25
mrem/y.
28The concentration needed in the grid area to give
25 mrem/y is the DCGLEMC
DCGLEMC (Area Factor) ? (DCGLW)
29Potential Elevated Areas
- The scanning MDC must be low enough to detect a
concentration of DCGLEMC.
If scanning MDC gt DCGLEMC, reduce the distance
between sample points, G, so that the area
0.866G2 is small enough and the area factor is
large enough that scan MDC lt DCGLEMC (Area
Factor) times (DCGLW )
30ExampleOutdoor Area Factors
Area (m2)
31Example Indoor Area Factors
32Graded Approach toFinal Status Survey Design
Effort Appropriate to Contamination Potential
GOALS
More efficient survey designs Accurate
remediation decisions
33Final Status SurveysRelationship to
Contamination Potential
- Class 1
- Systematic
- Sampling Grid
- Size determined by
- statistical tests and
- potential for
- elevated areas
- 100 Scans
- Area Limit
- 2,000 m2 outside
- 100 m2 inside
- Class 2
- Systematic
- Sampling Grid
- Size determined by
- statistical tests
- 10 - 50 Scans
- Area Limit
- 10,000 m2 outside
- 1,000 m2 inside
- Class 3
- Random Sample
- Size determined by
- statistical tests
- 10 Scans
- Area Unlimited
34Decision Making
Does the survey unit meet the release criteria or
not? Find out with measurements. What kind? How
many? Good, but not too good. Enough, but not
too many.
Decision Errors
- Failing a clean site
- Passing a dirty site
35Data Quality Objectives
Final Status Survey
36Decision Making Example
Does a soil sample contain 137Cs or
not? Facts Background is 10 cpm Sample Measures
15 cpm What do you decide?
More Facts What if 1? for background is 10
cpm? What if 1? for background is 1 cpm?
37Decision Making Example
As ? gets closer to the difference we want to
measure, ? 15 -10 5, What if 1? for
background is 3 cpm? What if 1? for background is
5 cpm? how do you decide?
38The Hypothesis Testing Framework is designed
specifically for such problems
The detection decision is determined by the LLD
Null Hypothesis H0 The Sample does NOT contain
the radionuclide
Alternative Hypothesis Ha The Sample does
contain the radionuclide
39Distribution of counts under null hypothesis
Distribution of counts under alternative
hypothesis
Critical Level Decision Trigger
Detection Limit
b
a
B
BLC
BLD
40The probability, a,of a false positive can be
decreased by increasing LC ... .but this will
increase the probability, b, of a false negative
Critical Level Decision Trigger
Detection Limit
b
a
B
BLC
BLD
41Decreasing probability both types of decision
errors. .requires a reduction in the standard
deviation, s, by increasing the counting time.
Critical Level Decision Trigger
Detection Limit
b
a
B
BLC
BLD
42The Site Release Hypothesis Testis essentially
the same!
- Null Hypothesis
- H0 The Sample does NOT contain the radionuclide
H0 The Site does NOT meet the guideline
- Alternative Hypothesis
- Ha The Sample does contain the radionuclide
Ha The Site does meet the guideline
43Release Limit
Distribution of concentration
under alternative hypothesis
Distribution of concentrations under null
hypothesis
Critical Level
a
b
44Decision Error Probability
1.0
If there were no measurement error and no
spacial variability...
Probability Site Will Fail
how many measurements would be needed?
LIMIT
0.0
Dose Rate
45Concept of Gray Area
a
b
46Survey Design
The number of measurements required is
determined by
- the accuracy desired in the statistical
hypothesis tests and - the need to discover any elevated areas
47The POWER of a test is defined as 1- b
It is the probability that the null hypothesis is
rejected when it is false
If the null hypothesis is not rejected when it is
false, you didnt look hard enough
How hard you look should depend on the
consequences of making an error.
48How to find the right Sample Size
- Signal to Noise ratio D/s
- DCGL
- Width of gray region, D
- Expected Measurement Variability, Standard
Deviation, s - Acceptable Error Rates
- a , b
49DQOs
STATE THE PROBLEM
- 20 year old facility
- building previously subject to potential
contamination with 137Cs and/or 60Co
50DQOs
IDENTIFY THE DECISION
- Can this building be released for unrestricted
use?
51DQOs
IDENTIFY THE INPUTS
- DCGL from level one screening dose model
- 1120 dpm/100 cm2 for 137Cs
- 1260 dpm/100 cm2 for 60Co
- 100 s direct total b/g measurements with 16 cm2
10.1 efficient GM counter - DCGL is 30 GM counts per 100s
- background about 60 /- 10 GM counts per 100 s
52DQOs
DEFINE BOUNDARIES
- One Class 1 Survey Unit
- 5 m by 6 m room
- 60Co seen in portable HPGe spectrum
- No contamination above 2 m or ceiling
- Survey Unit area 61 m2
- Reference area is in a concrete building nearby
- No history of radioactivity ever being used in it
- 60Co not seen in portable HPGe spectrum
53DQOs
DEVELOP DECISION RULE
- A Wilcoxon Rank Sum test will be used to compare
Survey Unit and Reference Area - The null hypothesis The median difference in
counts between them exceeds the DCGL - The DCGL set at 1120 dpm/100 cm2
- (conservatively covers both Cs and Co)
- An elevated measurement comparison will be based
on the DCGLEMC
54DQOs
SPECIFY LIMITS ON DECISION ERRORS
- Type I Error That a Survey Unit with over 1120
dpm/100 cm2 is released - alpha 0.05 - Type II Error That a Survey Unit with under 560
dpm/100 cm2 is not released - beta 0.05
55DQOs
OPTIMIZE THE DESIGN
- Consider Sampling Plan Alternatives
- Radionuclide specific measurements
- See effects of the choices of a, b, D, s on the
number of samples required
56Sample Size
- Width of gray region, D 15 GM counts
- Standard Deviation, s 10 GM counts
- D/s 1.5 gives Pr 0.856
- a b 0.05 gives Z 1- a Z 1- b 1.65
- N (1.65 1.65)2 / 3(0.856-0.5)2 28.6
57Sample Size
DQOs
- DCGL is 30 GM counts per 100s
- background about 60 /- 10 GM cp(100)s
- Width of gray region, D 15 GM cp(100)s
- Standard Deviation, s 10 GM cp(100)s
- D/s 1.5
- for a b 0.05, MARSSIM table gives
- N 18
58Sample Size
DQOs
59Prospective Power
Probability Survey Unit is Released
True Contamination Level (GM Counts)
60Questions?
61Sample Grid
- For a 0.05 , b 0.025, LBGR 20
- need 39 measurements each in the Survey Unit and
Reference Area - Then the sample spacing on a triangular grid is
- L 61 m2 / (0.866N) 1/2 1.3 m
62Elevated Measurement Comparison
- Sample spacing on the triangular grid is 1.3 m
- Grid Area is 0.866 (1.3)2 1.5 m2
- Area Factor 8 for both 60 Co and 137 Cs
- DCGLEMC is thus 240 GM counts
63Grid Layout in Indoor Survey Unit
64Grid Layout in Indoor Reference Area
65Elevated Measurement Comparison
- Suppose scan sensitivity is 300 GM counts?
- Must be able to detect the DCGLEMC by scanning.
- The area factor needed is 300/ DCGLW 300/30
10. - Work the calculation backwards!
66Elevated Measurement Comparison
- For an area factor of 10,
- the Grid Area is about 1.0 m2
From the Grid Area find the sample spacing 0.866
(L)2 1.0 m2 so L 1.0 m2 / (0.866) 1/2
1.07 m
67Number of Samples
- L 1.07 m
- Grid Area 1.0 m2
- Survey Unit Area is 61 m2
- Thus 61 m2 / 1.0 m2 61 Samples are needed
In Class 1 areas, the scanning sensitivity can
determine the sample size required!
68Data Quality Assessment
Data Quality Assessment (DQA) is the scientific
and statistical evaluation of data to determine
if the data are of the right type, quality, and
quantity to support their intended use (EPA
QA/G-9). There are five steps in the DQA process
- (1) Review the DQOs and sampling design.
- (2) Conduct a preliminary data review.
- (3) Select the statistical test.
- (4) Verify the assumptions of the statistical
test. - (5) Perform the statistical test.
69Data Quality Assessment
- Tools
- Posting Plots
- Summary Stats
- Histograms
- Quantile Plots
- Retrospective Power Curve
70Count Rates in Indoor Survey Unit
posting plot
71Count Rates in Indoor Reference Area
posting plot
72Summary Statistics
Reference Survey Area Unit Samples 40
50 Average 58 88 Std Dev
10 92 Median 59 58
73Histogram of Survey Meter Readings Indoors
74Wilcoxson Rank Sum Test
- The sum of the ranks of the adjusted reference
area measurements, 2432, - is greater than the critical value 2023.
- Thus, the null hypothesis that the Survey Unit as
a whole uniformly exceeds the release criterion)
is rejected.
75Count Rates in Indoor Survey Unit
Area exceeds 240
76Elevated Measurement Comparison
- DCGL30
- One area of 1.5 m2 (area factor 8) exceeds 240 GM
counts - An area of 6 m2 (area factor less than 3)
averages about 230 - A 24 m2 area (area factor 1.2) averages about
130
Survey Unit fails EMC