1. The Phantom 4 Manifesting 3H, 14C, 129I & 99Tc: Preliminary Findings for 129I & 99Tc
Lisa Edwards Sr. Program Manager, Chemistry & Radiation Management
EPRI
LLW Forum
April, 2015
Good afternoon, everyone. Thanks for the kind introduction.
Today, I am pleased to present some research the EPRI has conducted related to the phantom 4. The research that I present today along with the much more detailed greater body of research that accompanies it is expected to be published soon as a public document.
The document in its entirety is still under review and as such the information contained herein should be considered preliminary until such time as the technical review is completed and the document is published.

2. Quantifying the 3H, 14C, 99Tc & 129I Key Take Aways
Accurate quantification of these highly mobile nuclides is challenging,but important for correct performance assessment
There are better and more accurate methods to quantify and manifest the Phantom Four in reactor LLW:
3H follows moisture fraction and should not be scaled to other nuclides
14C method perhaps adequate: look a little harder
lowering required Table 1 LLD values by 10 times should resolve the 14C data (a further lowering is not practical)
99Tc and 129I should be scaled as real when non-detect result is received
Use of non-positive LLD values results in manifested values for 99Tc & 129I that are 100-1,000 times higher than actual -- could adversely impact that disposal site capacity
So just to get us started, I thought I would begin with the key take-aways of this presentation:
First, these four nuclides , tritium, carbon-14, Tc-99, and I-129 are required to be manifested and accurate accounting of them in the disposal site inventory is important because of their half life and/or water mobility.
Currently there are some short comings that often times are leading to the gross over estimation of these nuclides.
Tritium is sometimes being erroneously scaled to other nuclides, but in fact should NOT be scaled in this way. It instead should be quantified quantified based upon a the moisture fraction of the waste stream being characterized.
Carbon-14: EPRI research indicates that the current method is probably accurate and overestimation resulting from using the LLD values as real values could be mostly eliminated by looking just a bit harder. Current LLD values are defined at 10% of the Class A limit, if sampling lowered the LLD values to 1% of the class A values, most over estimation would be eliminated. This is the only time that I am going to reference these first two nuclides and the rest of the presentation will be devoted to our most recent work related to quantification of TC & I.
EPRI research indicates that sampling of these important nuclides should continue, but in the face of using a LLD (or non positive detection) as a real value which has been shown to introduce factor of and even greater error, an industry scaling factor should be used instead. The scaling factors I will recommend for consideration in this presentation were developed from a database of mass spec samples analyzed by PNNL. I will actually present the preliminary results of this research in the form of the actual scaling factors we are recommending and a brief overview of the technical basis used to derive those recommendations.

3. Over-Reporting of 99Tc and 129I in LLW
Multiple references have documented the positive bias in current reporting of these nuclides and the adverse impact on disposal site capacity, a few are listed below:
NUREG-1418 “Roles Report”, 1990
DOE/EH-0332P, LLW & MW Disposal During 1990, 1993
NUREG/CR-6567, LLW Classification, Characterization and Assessment, 2000
NCRP 152, LLW Performance Assessment, 2005
EPRI , LLW Disposal Practices, 2009
ML13260A075, EPRI Letter to Staff, August 1, 2013
There are better and more accurate methods to quantify and manifest 99Tc and 129I such as scaling generically as real when not detected by radiochemistry measurements
Without going through this list of references documenting the long-standing problem of over-reporting 99Tc and 129I on LLW manifests there remains a need to provide sufficient guidance that is easily implemented to resolve this problem and significantly improve manifesting accuracy in the process.

4. Scaling Factor Development - Approach
Sample data from NUREG / CR-6567 Table 7-8 for 99Tc and 129I spectroscopy measurements by PNNL* were evaluated
For 129I:
Direct measurements of 137Cs and 129I were transformed to their natural log
The log transformed data was checked for normality using three graphical depictions of the data
a histogram of the natural logarithm of the ratios of 129I/137Cs for frequency distribution
a quantile-quantile (Q-Q) plot showing the distribution of the data against the expected normal distribution
and a plot of residuals for randomness
Fuel conditions were considered by taking into account the Cs/Co ratio
Same approach applied with 99Tc in comparison to 137Cs & 60Co
A basic summary of the approach used to evaluate a proper treatment method of the data form PNNL and published in NUREG / CR 6567 is described in this slide.
As a general background, when this data was initially analyzed, a linear relationship didn’t appear to fit well. However, it should be understood that the apparent absence of a linear relationship is not uncommon with natural processes.  Similar to other scaling factor data that spans several orders of magnitude and appears non-linear in normal space may prove to be linear in log space if a normal distribution of the log transformed data is present --also known as a log-normal distribution.
To develop the log normal distribution, the direct measurements of the PNNL data and related ratios were transformed to their natural log values. The resulting data was checked for normality using 3 standard tests including a histogram, a Q-Q plot and a diagnostic residual plot. Once the data quality was determined to be acceptable, the data was evaluated for determination of scaling factors taking into account fuel condition as determined by the Cs/Co ratio.
So for I-129, the PNNL direct measurements of and Cs-137 and their relative ratio were transformed to their natural log values. The result was then validated using the 3 standard test just described.
*Pacific Northwest National Laboratory

5. Validity Checks for Log Transformed 129I Data
The validity checks in the form of the histogram, which depicts the frequency distribution of the natural logarithm of the ratios of 129I/137Cs exhibiting a good central tendency of the data tailing off on each side,
the Q-Q plot for the Ln 129I/Ln 137Cs data. When the data is normally distributed, observations should lie approximately on a straight line. Whereas, points will form a curve that deviates markedly from a straight line if the data is not normal. Outliers appear as points at the ends of the line, distanced from the bulk of the observations. The Q-Q plot for the Ln 129I/Ln 137Cs data in Figure 4-3 follows the expected line sufficiently to be considered normal. Therefore, the 129I/137Cs sample data is considered to be normally distributed in log space.
The last check of data is through a randomness test by plotting the residuals. A residual plot shows the difference between the measured values and the predicted values against the true values indicating disagreement between the data and the fitted model. The ideal residual plot should show a random scatter of points forming an approximately constant width band around zero. (32) The residual plot of the Ln 129I/Ln 137Cs data depicts good randomness of the data such that it may now be tested for acceptability as a scaling factor. While there are arguably approximately four outliers in the data set on either side of the band, they are purposely left in the evaluation to not overly influence the remainder of the evaluation.
Validity Checks Considered Satisfactory

6. 129I Mass Spec Data Evaluation
Raw PNNL 129I/137Cs power regression exhibits very good correlation and could reasonably be used as is.
The log transformed data validates the log-mean and log-mean dispersion provide good correlation as well.
The graphs here present the raw PNNL data for 129I over 137Cs on the left and the log transformed values on the right
The analysis of the PNNL 129I mass spec data focused primarily on two factors 1) how well is the fit of the relationships between the HTM and Key radionuclide and 2) what explains the change of the scaling factor with the change in key radionuclide concentration.
In the case of 129I/137Cs the raw regression of the data exhibits a good correlation that could be used as is under one of the IAEA suggested methods for non-linear regression. The decreasing change in the scaling factor with increasing 137Cs concentration is caused by the greater escape rate of cesium compared to iodine when fuel clad failures are present as compared cesium and iodine escape rates originating from tramp fuel only.
The red bounds in the log transformed graph represent a factor of ten deviation in log space. There is quite a bit more analysis behind the validation of this distribution that there isn’t time to go over here. Suffice it to say that what we have here is a reasonable dataset that exhibits central tendency near a factor of ten and that has been validated mathematically as likely present in these proportions in the RCS.

7. Preliminary Results – 129I
129I should be scaled to 137Cs
A scaling factor of 2.0E-07 should scale within a factor of 10 with a 80% confidence factor
This value is the most conservative and it bounds the “all fuel clad integrity” data by a factor of ~2 and the “poor fuel clad integrity” data by a factor of ~5.
The use of detection limit (non- positive) values has been shown to be incorrect by a factor of ,000 or more
So, the preliminary conclusion of this research is that I should be scaled to cesium.
If a scaling factor of 1.2E-7 is used, based on this data set, you would be within a factor of 10 at the 80% confidence level.
Although a slightly better confidence level could be achieved thru the use of one scaling factor for when the Cs/Co ratio is <10 and another scaling factor when the Cs/Co ratio is >10, the recommended value for the entire data set is more conservative that either of the alternatives.
I expect the this analysis to be subjected to debate and independent scrutiny as all good science is. I recognize that a larger data set would be more ideal and that higher confidence level would be more appealing, however, I am comfortable in making this preliminary recommendation based on its alignment with other calculations that are described in the work, but for the sake of time not presented today.
In addition, if the aim here is for accurate reporting, one must consider the potential shortcomings of this approach in comparison to the shortcomings of using LLD values which are known to grossly over report the presence of Iodine.

8. 99Tc Mass Spec Data – Preliminary Results Log Transformed Data
Two scaling factors for 99Tc emerged from this analysis
99Tc is scaled to the activated corrosion product 60Co when 137Cs/60Co <10
99Tc is scaled to the fission product 137Cs when 137Cs/60Co >10
These results suggest that the production mechanism for 99Tc under high fuel integrity conditions is dominated by activation of 99Mo, while fuel leaks have a greater influence when the leak is large enough to result in a 137Cs/60Co >10
The same data transformation to the natural log values and validty tests were conducted for the Tc-99 data.
In this analysis, what we found was that 99Tc from fission is not significantly influential until fuel clad integrity is poor as gauged by 137Cs to 60Co ratio of between 10 and 100 and what emerges is two scaling factors that are applied based on a the measure of fuel clad integrity by using the 137Cs to 60Co ratio in the waste.
Here we show two plots of log transformed data plots supporting the two 99Tc scaling factors. Again, the red bounds in the log transformed graphs represent a factor of ten deviation in log space.
As with the 129I analysis, there is quite a bit more work behind the validation of these distributions and the results are reasonable datasets that again exhibit central tendencies near a factor of ten and that have been validated mathematically as likely in the RCS under varying fuel clad integrity conditions.

9. Scaling Factor Conclusion - Preliminary Proposed Indirect Method
The PNNL mass spec datasets for 99Tc and 129I in LLW:
Provide a reasonable basis for generic scaling factors
Proposed Indirect Method
Perform analysis for 99Tc and 129I in waste to required sensitivity (1% Table 1 A-Priori at a minimum)
When 99Tc and/or 129I results are at the detection limit (LLD) in lieu of using the detection limit value - use a scaling factor from this research to calculate a value as applicable and treat as a positive measurement:
129I should be scaled to 137Cs
99Tc should be scaled to 60Co when the 137Cs/ 60Co <10
99Tc should be scaled to 137Cs when the 137Cs/ 60Co >10
In conclusion, the PNNL mass spec measurements make sense and demonstrate good fit that form the basis for reasonable scaling factors with a sufficient level of confidence.
While additional mass spec analyses are welcome and will always serve to refine the central tendency of the datasets - additional mass spec analyses are really not necessary to use the existing data generically in lieu of using LLD values.
Generic scaling factors were conservatively derived from this EPRI analysis that were mathematically validated. These generic scaling factors provide a reasonable assurance of accuracy and are far better used than an LLD result that is know to be high by a factor of 100 to 1,000 or even 10,000 times actual.
What EPRI proposes is that NRC licensees that choose to use the scaling factors from this work continue to perform radiochemistry analysis for 99Tc and 129I at a minimum in accordance with the sensitivity requirements of the 1983 BTP but when the results of those measurements do not detect any activity or LLD, a generic scaling factor be used from this work in lieu of manifesting an LLD value as real.