ML19093B816
| ML19093B816 | |
| Person / Time | |
|---|---|
| Issue date: | 02/12/2019 |
| From: | Bixler N, Clayton D, Keith Compton NRC/RES/DSA, Sandia, US Dept of Energy, National Nuclear Security Admin |
| To: | Office of Nuclear Regulatory Research |
| Shared Package | |
| ML19093B814 | List: |
| References | |
| Download: ML19093B816 (15) | |
Text
2/12/2019 Benchmarking of MACCS Gaussian Plume Segment Model against HYSPLIT dispersion results Daniel J. Clayton1, Nathan E. Bixler1 and Keith Compton2 1Sandia National Laboratories, 2U.S. Nuclear Regulatory Commission Presented at the 31st Annual Regulatory Information Conference Bethesda, MD, USA, March 14, 2019 Sandia National Laboratories is a multi-mission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC., a wholly owned subsidiary of Honeywell International, Inc., for the U.S. Department of Energys National Nuclear Security Administration under contract DE-NA-0003525.
SAND2019-0795C 1
2/12/2019 Outline MACCS/HYSPLIT Integration of HYSPLIT atmospheric transport results into MACCS Benchmark study Benchmarking results Summary 2
2
2/12/2019 MACCS The MELCOR Accident Consequence Code System (MACCS) is used to perform probabilistic assessments of health, societal Gaussian and economic impacts. plume &
HYSPLIT MACCS models an atmospheric release of radionuclides from a severe accident (e.g.
plume rise, dispersion, dry and wet deposition).
Analysis estimates the health effects from:
inhalation, cloudshine, groundshine, skin deposition, and ingestion (e.g. water, milk, meat, crops), as well as costs and societal impacts associated with protective actions to reduce exposure 3
3
2/12/2019 HYSPLIT HYbrid SingleParticle Lagrangian Integrated Trajectory (HYSPLIT) performs complex dispersion and deposition simulations Initially created through joint effort by two agencies National Oceanic and Atmospheric Administration & Australia's Bureau of Meteorology Ongoing enhancements and features provided by the HYSPLIT development team and community.
Wide variety of high quality, publicly available meteorological data Dispersion of pollutant calculations Puff model Particle model Hybrid puff/particle models Air and ground concentration calculations 4
4
2/12/2019 MACCSHYSPLIT Coupling Process HYSPLIT results are sent to MACCS through input files instead of direct incorporation HYSPLIT developments/updates from NOAA easily incorporated No need to update MACCS with HYSPLIT update Preprocessors to generate files Air and ground concentrations as a function of time Local meteorology for given site 5
5
2/12/2019 Implementation Test Cases Test cases were created to verify the implementation of the HYSPLIT/MACCS coupling based on the MACCS2 Sample Problem A Compared results using Gaussian plume and HYSPLIT models Inputs were chosen to cause the models to behave similarly, given the correct implementation Modified Inputs Constant weather > Weather sampling from meteorological database Single, fixed deposition velocity > Ten aerosol sizes, each with own deposition velocity calculated from internal HYSPLIT model Single plume segment > Multiple plume segments 6
6
2/12/2019 Benchmark Cases Five representative sites Site A Large river valley Site B Central midwestern plain Site C Dry western region Site D Atlantic coast with potential for sea breezes Site E Southeast river valley influenced by Bermuda high Source term
- 1 NUREG1150 historic (puff release followed by a longer duration tail)
- 2 SOARCA ShortTerm Station Blackout (more delayed and prolonged)
General evacuation scheme Modeled with multiple evacuating cohorts Weather Sampling Weather bin sampling (777935 samples)
Mean, 95th and 5th percentiles 7
7
2/12/2019 Meteorology Wind Direction North American Mesoscale 12-km meteorological database - 2008 Wind Speed 8
8
2/12/2019 Source Terms Source Term #1 NUREG1150 historic (puff release followed by a longer duration tail)
Source Term #2 SOARCA Short Term Station Blackout (more delayed and prolonged) 9 9
2/12/2019 Benchmark Output Metrics Peak (around the compass) timeintegrated air concentration
(/Q, s/m3)
Peak ground deposition (D/Q, 1/m2)
Normalized peak dose (unitless)
Normalized regional (within a circular area near the point of release), populationweighted early fatality risk (unitless)
Normalized regional population doses (unitless)
Normalized regional, populationweighted latent cancer fatality risk (unitless)
Normalized land areas that exceed various levels of contamination (unitless)
Normalized total regional economic losses (unitless) 10 10
2/12/2019 Model Result Comparisons (Large River Valley, NUREG1150 Source Term, HYSPLIT/Gaussian)
Compare model results over distance Mean values over all weather trials 5th and 95th percentiles over weather trials Peak Quantities Integrated Quantities 11 11
2/12/2019 Benchmark Comparison Summary The level of agreement between the Gaussian and HYSPLIT ATD model results was very good for the five sites, two source terms, and all output quantities that were evaluated Most output results were within a factor of two when comparing the Gaussian and HYSPLIT ATD model results The trends as a function of distance for the output quantities also agree well between the two models The 5th and 95th percentile results also follow the same trends for the Gaussian and HYSPLIT ATD models For the most part, these results are also within a factor of two for the two models More frequently exceed a factoroftwo separation, as expected 12 12
2/12/2019 Sensitivity Cases Sensitivity studies arose from discussions during the benchmark analyses Analysis performed for Site A HYSPLIT model Meteorological data source Gaussian plume model Meteorological data source Stability class determination method Dispersion model 13 13
2/12/2019 Computational Cost HYSPLIT Preprocessing
~31,000 processor hours total per site Total disk space per site = 500600 GB Converted files an additional 200 GB per site MACCS NUREG1150 source term Run on one Windows processor Gaussian ADT model, ~2 processor minutes per site HYSPLIT ADT model, ~20 processor hours per site (600 times longer)
SOARCA, ShortTerm Station Blackout (STSBO)
Run on one Windows processor Gaussian ADT model, ~30 processor minutes per site HYSPLIT ADT model, ~130 processor hours per site (260 times longer) 14 14
2/12/2019 Summary HYSPLIT option supplements Gaussian plume segment option in MACCS Benchmark analysis performed over wide set of metrics and distances Five sites Two source terms Compared statistics showing mean and 5th95th percentile range of consequences over one years worth of weather data Good agreement between the Gaussian and HYSPLIT analyses for the sites and source terms studied Need to balance the need for higher fidelity models with associated higher computational costs Peer review of implementation has been initiated 15 15