Validation of the Nuclear Facility Accident (NFAC) Model

Metropolitan Edison station predictions versus observations for 03-29.
Metropolitan Edison station predictions versus observations for 03-29.


Oak Ridge National Laboratory (ORNL) researchers and developers who maintain the Nuclear Facility Accident (NFAC) incident source model for the Hazard Prediction and Assessment Capability (HPAC) completed two validation efforts.  First, an HPAC/NFAC model of the Three Mile Island (TMI) accident was compared to observed doses collected during the event.  Given the uncertainty and variance in observation data collected, the consistency in shape of by-bearing and –distance prediction-versus-observation curves indicates the model effectively represents the results of the TMI accident.  Pearson correlation coefficients (PCCs) of predicted versus observed values are strong for most of the Metropolitan Edison stations.

Second, activity vectors created by NFAC were compared to those produced by the Nuclear Regulatory Commission (NRC) Radiological Assessment System for Consequence Analysis (RASCAL) code for moderate and severe accidents at representative Boiling Water Reactor (BWR) and Pressurized Water Reactor (PWR) plants.  NFAC demonstrated a higher level of fidelity in two primary ways: a much larger range of isotopes, and inventory changes due to decay over the duration of each release.

Significance and Impact

The Defense Threat Reduction Agency (DTRA) requires all HPAC models and components to be validated as part of the acceptance process.  HPAC is the primary modeling tool used by the DTRA Technical Reachback Division, and NFAC is the component used to model accidents and releases at any nuclear facility in the world.  Analysts using HPAC/NFAC to model events must have confidence in the accuracy and effectiveness in the tool.  Further, as RASCAL is the primary tool provided by the NRC for modeling accidents at nuclear facilities, it is important to demonstrate to HPAC users that NFAC provides a more detailed and accurate representation of the releases associated with an accident.

Research Details

  • Thermoluminescent dosimeter (TLD) readings from 20 Metropolitan Edison stations and 37 NRC were recorded over the course of the TMI event.
  • The accident is modeled as a series of seven incidents defined as constant release rates in three time periods.
  • Releases are focused on three noble gases: 88Kr, 133Xe, and 135Xe.
  • Whereas HPAC/NFAC predictions match well with Metropolitan Edison station observations, results for NRC stations are inconsistent.  Uncertainty in the process for deriving exposure without background at NRC stations warrants their omission from any determinations.
  • The shape of by-bearing and –distance prediction and observation curves are consistent.
  • Generally, the model overpredicts exposure.
  • PCC values for predictions versus observations are good for most Metropolitan Edison stations, seven stations with very good values, and an additional five stations with good values.
  • NFAC creates activity vectors with a wider range of isotopes than those used by RASCAL.
  • NFAC performs decay during long-duration releases, thereby capturing daughter products that are ignored by RASCAL.


  • R.W. Lee, C.D. Sulfredge, “Validation of HPAC/NFAC Version 6.6 Nuclear Facility Incident Predictions at Three Mile Island Unit 2”, ORNL/SPR-2020/1756, November 2020.
  • R.W. Lee, C.D. Sulfredge, “Comparison of NFAC Version 6.6 and RASCAL 4.3.3”, ORNL/SPR-2020/1827, November 2020.


HPAC/NFAC predictions at Metropolitan Edison stations accurately represent observed doses.  By-bearing and -distance prediction-versus-observation curves are consistent.  PCCs of predicted versus observed doses are good or very good for most Metropolitan Edison stations.  NFAC activity vectors include a wider range of isotopes than those produced by RASCAL, and NFAC models decay during long-duration releases, thereby capturing daughter products ignored by RASCAL.

Last Updated: January 19, 2021 - 12:39 pm