1. Waste Treatment Plant Project Adapting Dispersion Software to DOE Standard 3009 Jorge Schulz Thomas R. McDonnell Bechtel National, Inc. 2012 EFCOG Safety Analysis Workshop May 8, 2012

2. 22 Overview This paper describes work performed for the WTP project to enable use of MACCS2 version 2.5 to generate 95 th percentile overall site atmospheric dispersion factors (χ/Q) in accordance with the statistical treatment required by DOE-STD-3009-94 –The maximum sector 99.5 th percentile χ/Q was also determined by post-processing Appendix A, section A.3.3, of STD-3009 states that the 95 th percentile χ/Q should: –Account for variations in distance to the site boundary as a function of direction –Be consistent with the statistical treatment of χ/Q values described in regulatory position 3 of NRC Regulatory Guide 1.145 for the evaluation of consequences along the exclusion area boundary –Determine the distance to the site boundary in accordance with the procedure outlined in position 1.2 of Regulatory Guide 1.145 MACCS2 determines directionally independent χ/Qs at a specific distance; thus, there is no way of obtaining the direction-dependent 95 th percentile χ/Q without post-processing

3. 3 Site Boundary Distance Regulatory Guide 1.145, position 1.2, states: –“For each of the 16 sectors, the distance for...X/Q calculation should be... the nearest point on the building to the exclusion area boundary within a 45-degree sector centered on the compass direction of interest.” WTP has several large processing buildings that are clustered together on the site –The site boundary distances were determined from the building cluster perimeter to the closest points on the established public boundary at each compass sector The site boundary distances range from 9.1 to 13.9 km 3

4. 44 Combined Effects of Site Boundary Distance and Prevailing Wind Direction Although the SSW, SW boundary distance is the closest, prevailing winds are from the WSW MACCS2 does not take actual site distances into account MACCS2 “boundary”

5. 5 Meteorological Data Collection and Processing Previous meteorological data were 1-hour averages rounded to the nearest mph (~0.5 m/s), resulting in discrete steps when plotted as a cumulative distribution (red curve) Subsequently, ten years of 15-minute average met data were obtained and processed –The CCDF curve is significantly smoother (blue curve) –The processed 15-minute average data can be used for events involving short term accident release –This processed data was also used to generate hourly averages, which were then converted into the format used by MACCS2 5

6. 66 Software Validation Original V&V of MACCS2 was performed before the WTP project had fully implemented DOE-STD-3009-94 Subsequently, it was determined that the 95 th percentile χ/Qs produced by MACCS2 did not meet STD-3009 requirements –Required post-processing of MACCS2 expanded output to reflect actual site boundary distances and to be directional –Necessitated also V&Ving the calculation of individual hourly sequence χ/Qs that MACCS2 generates in the process of determining its 95 th percentile values A spreadsheet-based approach (using Microsoft ® Excel ® ) was developed and validated for calculating atmospheric dispersion factors based on the methodology given in NRC Regulatory Guide 1.145 –Basis for confirming that the 95 th percentile χ/Q results produced by MACCS2 ATMOS module meet the requirements for V&V in accordance with DOE Quality Assurance Order DOE O 414.1C and Safety Software Guide DOE G 414.1-4

7. 7 MACCS2 Runs MACCS2 expanded outputs were obtained using 10 years of site meteorological data and each spatial interval –Expanded output consists of individual χ/Q sequences for each hour –Each χ/Q sequence tracks the plume as it travels downwind from the release location –Spatial intervals cover intermediate distances and each unique site boundary distance Undepleted and depleted individual sequence χ/Qs were determined for both ground-level and elevated releases Dispersion Coefficients –The MACCS2 Gaussian model uses spatially-dependent dispersion parameters, σ y and σ z –For WTP, σ y and σ z values are input to MACCS2 as lookup tables using the Briggs Open Country parameterizations For ground-level releases, χ/Qs are calculated both with and without plume meander 7

8. 8 Post-Processing of MACCS2 Expanded Output The MACCS2 expanded outputs described in the previous slide were post-processed in an Excel spreadsheet –MACCS2 χ/Qs were imported into Excel worksheets and sorted by clockwise meteorological sector For each sector, the site boundary distance was entered and the 99.5 th percentile χ/Qs were determined The overall 95 th percentile χ/Q was determined by using the Excel function PERCENTILE on an array of hourly χ/Qs for each sector 8

9. 99 Comparison of MACCS2 Results to Post-Processed χ/Q Results The post-processed, 95 th percentile directionally-dependent, undepleted and depleted overall site boundary χ/Qs are compared below to those determined using the MACCS2 directionally-independent methodology –The post-processed χ/Qs ranged from 13 to 21% lower than those obtained by the MACCS2 method

10. 10 Identified MACCS2 Software Errors In the course of this effort, two software errors were discovered in MACCS2, which are described in the following slides –The first error was in the treatment of dispersion coefficients from a lookup table (freezing of σ y values) –The second involved the method of accounting for plume meander using the Regulatory Guide 1.145 model implementation An evaluation of these errors determined that they either produce conservative results or have no significant impact on existing WTP accident analyses

11. 11 σ y Freezing Error – Background An intermediate step in the MACCS2 computation process is determination of lateral and vertical dispersion coefficients (σ y and σ z ) –MACCS2 calculates the dispersion coefficients for each user-defined distance interval for all meteorological sequences MACCS2 tracks the plume travel downwind over time; thus, depending on the wind speed, meteorological conditions can change at some downwind distance interval –In those cases, MACCS2 calculates a virtual distance that would provide the same dispersion coefficients at the transition point for the current stability class 11 In this example, the plume expands in the first hour based on the stability class and wind speed of that hour. In the next hour, the stability class changes to a more stable condition, and the wind speed decreases to one half of the initial wind speed; therefore, the plume expands more slowly and takes longer to travel a similar distance.

12. 12 σ y Freezing Error Description If the virtual distance exceeds the dispersion coefficient look-up table (e.g., for stable atmospheric conditions), the last σ z is used until the meteorology changes again Typically, the σ z limit is reached, while the σ y is not –i.e., the plume does not expand in the vertical direction, but can continue to expand in the horizontal direction However, once the value of σ z reaches its limit, two distance intervals later MACCS2 v. 2.5 also freezes the value of σ y –This should not occur because the limit of the σ y look-up table is not reached –χ/Q values are also frozen 12 MACCS2 Expanded Output Extract

13. 13 Effect of σ y Freezing Error Freezing σ y rather than allowing it to grow with distance, results in a sequence χ/Q that is higher than it should be, which is conservative –For WTP meteorology, the error typically occurs in about 4% to 8% of the sequences The reduction in a sequence χ/Q after correction would be no more than about 10-20% Therefore, for a sequence to have an impact on the 95 th percentile χ/Q, the uncorrected χ/Q value would have to be about 1 to 20% higher than the 95 th percentile χ/Q Initial assessment determined the number of sequences affected out of 10 years of hourly sequences (87,600 sequences) –5.3% of 87,600 sequences were affected by the error –Further analysis was required 13

14. 14 Impact Assessment of σ y Freezing Error Next, 175 randomly-selected sequences were reviewed to determine the impact –If freezing of σ y occurred after the site boundary distance, the sequence had no impact –If freezing of σ y occurred before the site boundary distance, the σ y and χ/Q values were recalculated correctly for each distance following where the freezing occurred –The new χ/Q values were reviewed to determine if the changed χ/Q would shift below the overall 95 th percentile χ/Q  Only sequences where the uncorrected value exceeds the 95th percentile and the corrected value is less than the 95 th percentile can cause the corrected 95th percentile to shift 98 sequences occurred before the site boundary distance, requiring further analysis –The adjusted χ/Q values for the 98 sequences were, on the average, about 11% lower than the unadjusted χ/Q values 14 Results –Only one sequence resulted in shifting from above the 95 th percentile χ/Q to below (about 0.6% of the total sample) –Four sequences resulted in shifting from above the 99.5 th percentile χ/Q to below, but none shifted the maximum sector 99.5 th percentile χ/Q

15. 15 Meander Correction Factor Error MACCS2 has an option to use the Regulatory Guide 1.145 plume meander model, which multiplies σ y by a correction factor that depends on stability class and wind speed –For wind speeds between 2 and 6 m∙s -1, the correction factor decreases as a straight line in a log-log scale from the maximum value (for stability classes D - F) to 1 However, MACCS2 v. 2.5 incorrectly uses a linear interpolation –This can over-estimate the meander correction factor by up to 1.5 times –Because χ/Q is inversely proportional to the dispersion coefficient, this error produces non-conservative individual sequence χ/Qs 15

16. 16 Assessment of Meander Correction Factor Interpolation Error A sample of 100 χ/Q sequences for a typical year was investigated –Chosen randomly by the starting day and hour –39% of the calculated site boundary sequence χ/Qs were affected by this error  Of these, the maximum discrepancy was 0.05% to 5.5% non-conservative The site-wide 95 th percentile χ/Q is 1.52 × 10 -5 s∙m -3 –The maximum site boundary χ/Q value for these affected sequences is 8.85 × 10 -6 s∙m -3 –The site-wide 95 th percentile χ/Q is a factor of 1.7 higher than the maximum χ/Q value for the affected sequences This means that none of the affected sequences could affect the 95 th percentile χ/Q because they are all more than the maximum deviation of 5.5% below the 95 th percentile value –This is expected because, typically, the wind speed that produces the 95 th percentile χ/Q is less than 2 m∙s -1 –For these wind speeds, MACCS2 v. 2.5 correctly implements the Regulatory Guide 1.145 meander model (no interpolation required) 16

17. 17 Conclusions If MACCS2 is used to determine atmospheric dispersion factors at the public boundary, in order to meet DOE-STD-3009-94, post-processing of the MACCS expanded output sequence χ/Qs is required –to account for variations in distance to the site boundary as a function of direction and –to be consistent with the statistical treatment of χ/Q values of position 3 of NRC Regulatory Guide 1.145 The current version 2.5 of MACCS2 has two software errors that affect the horizontal dispersion coefficient –For WTP, the net impact of these errors was determined to be essentially negligible –Other sites may need to perform similar evaluations 17 The work described in this paper was performed for the U.S. Department of Energy (DOE) under contract DE-AC27-01RV14136, to support the Hanford Tank Waste Treatment and Immobilization Plant (WTP) Project.