Development and Custodial Organization:

For software technical support, please contact:

NARAC Contacts at LLNL:

Predicted evacuation and sheltering areas


Architecture and Operation

The following figure illustrates the major components in the NARAC modeling system. 
Primary atmospheric modeling tools



Dose and Effects Prediction

Output Products

Industrial chemical release
Radiological release


J.S. Nasstrom, G. Sugiyama, J.M. Leone, Jr. and D.L. Ermak. 1999. A Real-Time Atmospheric Dispersion Modeling System. 11th Joint Conference on the Applications of Air Pollution Meteorology with the Air and Waste Management Association, Long Beach, CA, 9-29-1999. UCRL-JC-135120.

John M. Leone, Jr., John S. Nasstrom, Denise M. Maddix, David J. Larson, Gayle Sugiyama and Donald L. Ermak.  2001.  Lagrangian Operational Dispersion Integrator (LODI) User's Guide Version 1.0.  UCRL-AM-212798. 

Nasstrom, J.S., G. Sugiyama, R. Baskett, S. Larsen, and M. Bradley.  2007.  The National Atmospheric Release Advisory Center (NARAC) Modeling and Decision Support System for Radiological and Nuclear Emergency Preparedness and Response.  International Journal of Emergency Management (IJEM), Vol. 4, No. 3, pp. 524-550 (2007).  UCRL-JRNL-211678-Rev2.

Gayle Sugiyama, John Nasstrom, Ron Baskett, Matthew Simpson.  2010.  National Atmospheric Release Advisory Center (NARAC) Capabilities for Homeland Security.  Fifth International Symposium on Computational Wind Engineering, Chapel Hill, NC.  May 23-27, 2010.  LLNL-CONF-425248.

The following are selected NARAC documentation products.  A more complete list is provided on the 

NARAC public web page

Software Quality Assurance Level: 
NARAC Documentation


Users should be mindful that the NARAC modeling system is only intended for emergency response, preparedness and training applications.  Like other non-safety-software modeling tools, the NARAC Web is not the appropriate tool for safety analyses or environmental planning hazard assessment (see DOE O151.1C FAQ answers on the EMI SIG web site for more information). 

NARAC provides graphical, mapped products using a predefined set of default model output contour values. These are either standard, default NARAC values, or pre-specified custom values for a particular site or user organization.  If other values are needed during an event, these must be requested by contacting NARAC on-duty staff.   A simulation may then need to be rerun (after communicating with NARAC staff) before meaningful results are plotted for the entire area of concern. 

Modeling results for simulations involving one or two hazardous substances and relatively short run simulations (a few hours or less) are typically delivered to users within about 5 to 10 minutes of the start of a simulation.  In many situations this delay is not significant, but for some applications (e.g., those involving rapidly evolving meteorological conditions, those focusing on close-in emergency response actions) this delay can represent a limitation, which users typically overcome by having the capability to locally run simpler faster running models.  Simulations involving the release of multiple contaminants (e.g., multiple radionuclides released in a single event), extended run durations, or the use of prognostic meteorological models can result in substantial increases in model run times.  Users who have not created predefined release scenarios that characterize the release situation of interest, should also factor in the time required to define release locations or multi-chemical source terms when considering how long it will take to initiate a model run.  The potential for multiple users to simultaneously request NARAC support during an emergency response event may result in additional delays in providing output products to users.  However, NARAC staff monitor requests, and have procedures to contact users to minimize conflicts and coordinate multiple requests for the same event. In addition, recent and planned upgrades to NARAC computing power are reducing the turnaround time for complex NARAC predictions.

There are significant advantages to NARAC’s centralize approach to modeling.  This includes the ability to offer sophisticated modeling capabilities that require high-performance computers and scientific/technical support that can only be provided at a specialized facility.  This centralized approach also has limitations.  One such limitation is the reliance on Internet-based communications to access and utilize the NARAC Web.  Internet disruptions can affect the ability of users to access the NARAC model system and receive products.  NARAC has alternative ways of receiving and delivering data and modeling results in the event of a communications disruption (e.g., via telephone communications or DOE Emergency Communication Network), although there is a potential for a loss of functionality and responsiveness when using these alternate communications methods.  

NARAC faces challenges as it continually evolves to provide more sophisticated and accurate modeling capabilities for its large and diverse user community.  Because NARAC addresses such a wide range of hazards and facilities, its products must support a very broad range of users.  On one end of the user spectrum are those with negligible atmospheric dispersion modeling know-how and on the other end of the user spectrum are experts in dispersion modeling. The NARAC Web interface provides a simple and quick method for users to get initial predictions with minimal information in an emergency.  However, more detailed predictions and analysis, requiring more information or more expertise, must be done by staff at NARAC’s operations center at LLNL.  Staff at the NARAC operations center can control a large number of input parameters not directly adjustable by those using the NARAC Web, including parameters that may be required to do detailed calculations.  An example of an adjustment that can only be made by NARAC staff is the incorporation of available field monitoring data to refine model predictions.



The NARAC Web allows the user to select from a variety of output products depending on the type of release and the nature of the hazardous materials.  This includes an array of graphical products.  Examples are shown in the figures below.

For chemical releases, NARAC air concentration results are typically plotted using standard chemical exposure Protective Action Criteria (PAC) levels. For chemical and biological agents,  incapacitating or lethal concentration levels can be plotted.

Radiological dose can be calculated for several pathways: internal dose due to inhalation, external dose due to air submersion (“cloud shine”) and external dose due to ground exposure (“ground shine”).  The NARAC modeling system can account for weathering effects on ground deposition (which decreases groundshine dose over time), and the contribution of ground contamination resuspension to airborne contamination (which, for example, can increase inhalation and cloudshine dose). NARAC radiological dose predictions are typically plotted using standard EPA/DHS Protective Action Guide dose levels, for example, for guidance on sheltering, evacuation and relocation decisions. Higher, acute radiological dose levels can also be plotted using different probability levels for fatalities or injuries.

LODI uses a coordinate system with a continuous terrain representation at the lower boundary. The model supports nested grids and variable grid resolution in both the horizontal and vertical directions.  LODI contains several source models including point, line, area, and volume source geometries and buoyancy/momentum driven sources.  It can model both instantaneous and continuous sources.  Source characteristics such as emission rates, location, and geometry can vary in time.  Mean air concentration, time-integrated air concentration, peak mean air concentration, ground deposition, and time-integrated ground exposure (needed for “ground shine” radioactive dose estimates) can be output from LODI.

Turbulent dispersion is modeled via a random walk method that uses atmospheric eddy diffusivity values to parameterize the effects of turbulent motions that are unresolved by the gridded mean winds.  Radioactive decay and production, first-order chemical reactions, bio-agent decay, wet deposition and dry deposition can be simulated.

LODI is the primary dispersion model used in the NARAC modeling system.  This model solves the three-dimensional advection and diffusion equation using a Lagrangian stochastic, Monte Carlo method that calculates possible trajectories of fluid “particles” in a turbulent flow. Particles are created at their modeled source location with an appropriate amount of contaminant mass based upon prescribed mass emission rates.  These computational particles can also be given total density and diameters, sampled from an input aerosol size distribution, which are used to calculate gravitational settling and deposition.  Initial particle positions are assigned by sampling the spatial distribution based on the geometry of the source.  A large number of independent particle trajectories are calculated by moving particles in response to the various processes, such as mean wind advection, gravitational settling and turbulent dispersion, as represented within the simulation.  The mean contaminant air concentration is estimated from the spatial distribution of the particles at a particular time.  Two key processes are advection by the mean wind and dispersion by turbulent motions.  To calculate the mean wind advection, three dimensional gridded mean wind fields from the ADAPT model are used as input to LODI.

ADAPT produces non-divergent (mass-consistent) winds by minimal adjustment of input fields derived from observational data.  This adjustment is applied to both horizontal and vertical winds.  The meteorological fields produced by ADAPT are highly dependent on the density and distribution of measurements, the complexity of terrain, and the proper parameterization of the atmospheric structure to represent physical processes that are not directly modeled.  ADAPT’s mass-consistent wind algorithm minimally adjusts the winds to add stability dependent steerage around topographical features.

ADAPT provides a selection of approaches to process meteorological data.  The model incorporates a number of interpolation and extrapolation techniques, including both direct and iterative solvers, and atmospheric parameterizations.  ADAPT’s capabilities include the ability to integrate surface data and upper air observations to calculate three dimensional wind fields and turbulence parameters which can be used to drive dispersion modeling.

The ADAPT model is designed to access and utilize all available atmospheric data to build three-dimensional gridded meteorological fields.  There are two broad classes of meteorological data that ADAPT can utilize: observational data and gridded fields.  Observational data includes local and regional meteorological measurements, forecast soundings, and user-generated observations.  ADAPT divides observational data into three categories: surface, tower, and upper air.  Surface data consist of measurements at a single near-ground height.  Tower data may contain measurements at a single elevated height or at multiple levels, the lowest of which may be at or near the surface.  Upper air soundings provide multi-level data with the lowest levels located in the planetary boundary layer.  In contrast, gridded fields are analyses or forecasts either acquired from external sources or generated by numerical models (e.g., WRF) that are run by federal agencies (e.g., NOAA) or in-house at LLNL/NARAC.

The primary atmospheric modeling tools are show in the figure below.  Two of the core components are the Atmospheric Data Assimilation and Parameterization Techniques (ADAPT) model and the Lagrangian Operational Dispersion Integrator (LODI) model. 

The NARAC modeling system offers authorized users access to a sophisticated suite of three-dimensional models, including tools for estimating meteorological fields (using observed and forecast data), atmospheric dispersion, and health and environmental consequences.   Users are granted remote access to NARAC models through the NARAC Web.  Using the NARAC Web interface, users can select from several different types of release scenarios.  This includes point and line sources, stack emissions, fires, explosions, chemical spills, spray releases, and nuclear detonations.  Users have the ability to specify the time and location of the atmospheric release, the type and amount of the hazardous materials involved in the event, the type of atmospheric data to use (i.e., data acquired by NARAC or data input by the user), and release properties (e.g., the fraction of hazardous material that becomes airborne, properties particular to the type of release).  Users can access scenarios they have previously defined for their site to streamline the initialization of model runs.  Actual model simulations are performed at the NARAC facility and modeling results (including graphical products) are displayed for the user via the NARAC Web.  The results of most three-dimensional NARAC simulations are made available to the user in 5 to 15 minutes of the start of modeling.  Model simulation times vary greatly based on the nature of the simulation, the number of different hazardous materials released to the atmosphere, and the duration of the simulation.  Longer or more complex simulations can take additional time to complete. 

Other federal, state and local emergency response agencies involved in emergency response to airborne hazards.

Department of Homeland Security’s Interagency Modeling and Atmospheric Assessment Center (IMAAC)

Department of Defense sites

Department of Energy (DOE), including emergency response teams and EOCs at DOE headquarters, DOE field sites, national laboratories, and other facilities

NARAC's large and varied set of users include emergency response personnel affiliated with the following:

NARAC is located in a facility at the Lawrence Livermore National Laboratory (LLNL) in Livermore, California.  The facility features an operations center with redundant computer systems and uninterruptible/backup power.  The NARAC facility is staffed by a team of research and operational personnel with expertise in atmospheric science, operational meteorology, numerical modeling, computer science, software engineering, geographical information systems, computer graphics, and hazardous material properties and effects.  The facility is staffed during normal business hours (in the Pacific Time zone) and staff members are on call around-the-clock in the event of an emergency response incident.  The staff members provide scientific and technical support, facility maintenance, training on NARAC tools, quality assurance assessments of model runs, and detailed analysis of atmospheric releases.  The staff can also refine NARAC modeling results by incorporating field measurements and other supplementary data as model input.  

The National Atmospheric Release Advisory Center (NARAC) is a national support and resource center for planning, real-time assessment, emergency response, and detailed studies of incidents involving the atmospheric release of hazardous materials.  It provides a modeling system, and related tools and services, which are used to estimate the potential spread of radiological, nuclear, chemical, or biological materials that may be emitted to the atmosphere in an accidental or intentional release (including malicious events).  A primary objective of NARAC is to provide atmospheric modeling and consequence assessment predictions in time for emergency managers to help guide effective protective actions, as needed, to safeguard the health and safety of people in affected areas.  

Gayle Sugiyama, Program Leader, 


John Nasstrom, Deputy Program Leader, 


Shawn Larsen, Systems Development Manager

, 925-423-9617,

Brenda Pobanz, Operations Manager

, 925-422-1823,