Roland R. Draxler


                Presented at the Tenth Session

                 Commission for Basic Systems

               World Meteorological Organization

                      Geneva, Switzerland


                         November 1992






     In the context of current agreements between NOAA and WMO, NOAA is prepared to provide diagnostic and forecast transport, dispersion, and deposition estimates for atmospheric releases of hazardous pollutants that may cross international political boundaries.  The primary region of coverage would include the United States, Canada, Mexico, and Central America (including the Caribbean).  The secondary region of coverage would encompass the entire Northern Hemisphere.  Southern Hemisphere support is also possible, but would require sufficient advance notification.


     Depending upon the details that are provided describing the emission source characteristics and species released, model output can include estimates of plume position, relative and actual air concentration, deposition amount, and dose.  The graphical output would be automatically distributed by facsimile to pre‑arranged lists of recipients or for those on the same computer network the output files can be directly transferred.






     The atmospheric pollution response system is made up of three components:  the models that generate the meteorological analysis and forecast fields, the model that computes the pollutant transport and dispersion from such information, and the software/hardware system that links them together in an operational procedure.  A brief description of each follows.


Meteorological Forecast Models


     The NOAA National Weather Service's National Meteorological Center (NMC) runs a series of computer analyses and forecasts (Petersen and Stackpole, 1989).  One of the primary operational systems is the Global Data Assimilation System (GDAS, Kanamitsu, 1989), which uses the spectral Medium Range Forecast model (MRF) for the forecast (Sela, 1980).  Another primary system is the Regional Analysis and Forecast System (RAFS), which uses the Nested Grid Model (NGM) for the forecast (Hoke et al., 1989).  Output from the RAFS covers North America.  For each run, observations are assimilated with "first guess" data fields (forecasts from the previous model run), and dynamic imbalances in the data are reduced, resulting in "analyzed" data fields.  Then the forecast is made.  The analyzed data provide a better representation of the real atmosphere than observations alone because of limitations in the observations.  Some of these limitations are due to measurement error or other instrument problems, and nonuniform spatial and temporal distributions of the observations.


     At NOAA's Air Resources Laboratory (ARL), the NMC model analysis and forecast products are used for air quality transport and dispersion modeling.  The RAFS/NGM is designed to provide numerical guidance over the United States, Canada, and Mexico out to 48 hours.  The GDAS/MRF provides global coverage twice‑daily out to 72 hours.  To facilitate the subsequent transport and dispersion calculations, all model fields, if not already available in that form, are converted to polar stereographic projections.  The NGM data are available hourly on a 91 km grid for the first 12 hours and then on a 180 km grid at 6 hour intervals out to 48 hours.  The fields are prepared both on the model sigma surfaces up to 5 km and on mandatory pressure surfaces through the rest of the atmosphere.  The MRF forecasts, are routinely available on a 381 km spaced grid at 6 hour intervals out to a 72 hour forecast.  With advance notification higher spatial resolution MRF data (80 km spacing on sigma surfaces) can be obtained if needed.


Transport and Dispersion Model


     The Hybrid Single‑Particle Lagrangian Integrated Trajectories (HY‑SPLIT; Draxler, 1992) model is routinely used for most transport and dispersion modeling studies conducted by NOAA/ARL. The HY‑SPLIT model has evolved in several stages, starting with a simple wind‑shear induced particle dispersion study (Draxler and Taylor, 1982), to which was added air concentration calculations with simple vertical mixing assumptions (Draxler, 1982), and with the further inclusion of the calculation of spatially and temporally varying vertical mixing coefficient profiles and vertical particle motions (Draxler, 1987).

     In HY‑SPLIT, the algorithms and equations used in the calculation of long‑range pollutant transport and dispersion are a hybrid between Eulerian and Lagrangian approaches.  A single pollutant particle represents the initial source.  Advection and diffusion calculations are made in a Lagrangian framework.  However, meteorological input information is from gridded fields such as the analysis fields or forecast outputs from Eulerian primitive equation models.  As the dispersion of the initial particle spreads it into regions of different wind direction or speed, the single particle is divided into multiple particles to provide a more accurate representation of the complex flow field.  Air concentrations are calculated on a fixed three dimensional grid by integrating all particle masses over the sampling time.  Calculations consist of simple trajectories from a single source to consideration of complex emissions from several sources.  The code is structured so that concentration calculations or simple trajectory (forward or backward) calculations can be performed using sigma (terrain following) or pressure coordinates. 


System Linkage


     The modeling systems are linked at the ARL site.  An IBM RS\6000 workstation automatically transfers the most current analysis and forecast information from the NMC mainframe site through a T1 ethernet link.  The latest forecast and analysis fields going back several days are maintained on the workstation.  A graphical menu system permits the display of any meteorological field or the ability to run the transport and dispersion model and display of the subsequent output. 


     The transport and dispersion code has some minimal input requirements, such as the latitude and longitude of the pollution source and the time and duration of the emission.  Additional information such as emission amounts or species would permit calculation of air concentrations as well as deposition.  A facsimile connected directly to the workstation automatically sends the graphical output to designated recipients. 


     The response system is initiated by notifying ARL directly during working hours, or after‑hours by calling a dedicated phone number that is automatically routed to the "on‑ duty" ARL meteorologist.  The on‑duty staff member would have access to a terminal and the capability to access the ARL RS/6000 to initiate the system.  The dispersion program would be started and results automatically disseminated.  In the case of continuous on‑going emergencies, the dispersion modeling capability could be initiated automatically when new NMC forecasts become available.




     The response system ‑‑ current automatic forecast downloading, dispersion calculation, and facsimile dissemination system ‑‑ was first routinely placed in operation during the Kuwait oil fire plume emergency.  Between May and November 1991, twice‑daily transport and dispersion forecasts were calculated and distributed automatically by telephone facsimile to the WMO, affected Gulf nations, and other national weather services. 


     The "on‑duty" telephone routing notification system, initially set up for the U.S. Nuclear Regulatory Commission (NRC), is routinely tested several times during the year.  To date there have been no NRC emergencies that have required an ARL response.


     A comparable dial‑in configuration has been designed and tested to support aircraft operations during a volcanic ash hazards alert, associated with volcanic eruptions.  A transport and dispersion model customized for volcanic ash particle distributions (VAFTAD; Heffter and Stunder, 1992) resides on the ARL workstation and accesses the same meteorological data base and graphical  dissemination system.  



     There are several aspects to the verification issue.  First is the accuracy of the meteorological information predicted by the meteorological model.  This is a fairly universal and frequently addressed issue for all forecast centers, as well as NMC.  Accuracy varies with the model, season, synoptic regime, and geographical region.  Papers discussing these issues can, for instance, be found in the April 1989 (No. 4) issue of Weather and Forecasting as well as in more recent conference proceedings such as the 9th American Meteorological Society Conference on Numerical Weather Prediction (14‑18 October 1991, Denver, Colorado).  With respect to volcanic eruptions, trajectory forecasts have compared favorably with ash observations made using satellites (Heffter et al., 1990).


     A second aspect to the verification that is particularly relevant for transport and dispersion studies relates to how well the analyzed fields actually represent the flow of the pollutant.  The evaluation may be impacted by smaller scale features that are not represented in the coarser gridded data.  This aspect has been addressed with respect to the NMC meteorological data fields through sensitivity studies and comparison with experimental tracer data.  Draxler (1990) compared the NGM data with observations on a 500 m tower.  The results showed that on average the NGM accurately predicted the diurnal variations of wind, temperatures, and mixing. Trajectory accuracy from calculations using HY‑SPLIT was estimated (Draxler, 1991) from back trajectories calculated from aircraft observations of the tracer plume position.  Transport distances varied from about 50 to 300 km from the source in the 30 trials with aircraft observations.  Trajectory error ranged from 20% to 30% of the travel distance.  Slower transport in more homogeneous zonal flow regimes resulted in some of the smallest errors of around 15% of the travel distance.  High resolution temporal and spatial meteorological data are not always available when the data source is from the MRF.  The effect of lower resolution data on trajectory accuracy (Rolph and Draxler, 1990) was tested by degrading the high resolution NGM data in space and time and calculating the difference in those trajectories from the base case calculated with the high resolution data.  When the NGM data were degraded to a level comparable with the MRF data, the trajectory error after 96h travel increased by about 50%.  A comparable verification study (McQueen and Draxler, 1992) of MRF forecast trajectories compared with the Kuwait oil fire smoke plume, as observed from the polar orbitor, indicated relative trajectory errors of about 15% to 30% of the travel distance.


     The last aspect of verification, and perhaps one of the more important to any transport and dispersion code, is understanding the causes of uncertainty and the potential errors in the resulting concentration calculations.  A major long‑range verification experiment was conducted during January through March of 1987, the Across North America Tracer Experiment (ANATEX ‑ see Draxler et al., 1991).  Inert tracers were released routinely from two sources every 2+ days and daily averaged samples were collected at distances of up to 3000 km from the tracer sources.  Enhanced 4‑ per‑day rawinsonde observations as well as high resolution NGM model output were available for the three month experimental period.  A detailed evaluation of several long‑range transport models, including HY‑SPLIT, using the ANATEX data was conducted by Clark and Cohn (1990).  They found, with some exceptions, that most models tested had comparable performace statistics.  Further evaluation (Rolph et al., 1992) of the HY‑SPLIT concentration predictions was performed using sulfur dioxide emissions, transformation, and wet and dry deposition.  Calculations of air concentration and wet sulfate deposition were generally within a factor of two of the measurements over most of the Eastern U.S.; in this application, the NGM winds and precipitation predictions were used.





     A mesoscale modeling capability for finer resolution predictions is available on the ARL computing system.  The Regional Atmospheric Modeling System (RAMS; Walko and Trembak, 1991) runs routinely on the workstation.  Currently it is used in a diagnostic mode, constrained by the NGM initialization, producing archive data sets at 40 km and 10 km resolution over the eastern U.S. for a variety of research studies.  In the event of an emergency it could be initialized from one of the NGM or MRF forecast fields to produce a mesoscale forecast consistent with that of the larger scale synoptic model.  The advantage of this approach would be a more realistic simulation of flow fields, such as lake‑ or sea‑ breezes, that would not be available from the larger scale models, but could be highly relevant to pollutants released into that environment.  If available, additional local meteorological observations could be assimilated into this system.


     Depending upon the geographic location, and given sufficient advance notification, it is possible to produce more detailed information from the MRF forecast, with finer spatial and vertical resolution.  Further, NMC experimental mesocale models (ETA) could be initiated over the region of interest.




Clark, T.L. and Cohn R.D., 1990:  The Across North America Tracer Experiment (ANATEX) Model Evaluation Study, EPA/600/3‑90/051, June, U.S. Environmental Protection Agency, Research Triangle Park, N.C.


Draxler, R., 1982:  Measuring and modeling the transport and dispersion of Kr‑85 1500 km from a point source.  Atmos. Environ., 16, 2763‑2776.


Draxler, R., 1987:  Sensitivity of a trajectory model to the spatial and temporal resolution of the meteorological data during CAPTEX.  J. Clim. Appl. Meteorol., 26, 1577‑1588.


Draxler, R., 1990:  The calculation of low‑level winds from the archived data of a regional primitive equation forecast model.  J. Appl. Meteorol., 29, 240‑248.


Draxler, R., 1991:  The accuracy of trajectories during ANATEX calculated using dynamic model analyses versus rawinsonde  observations.  J. Appl. Meteorol., 30, 1446‑1467.


Draxler, R., and A.D. Taylor, 1982:  Horizontal dispersion parameters for long‑range transport modeling.  J. Appl. Meteorol., 21, 367‑372.


Draxler, R., R. Dietz, R.J. Lagomarsino, and G. Start, 1991:  Across North America Tracer Experiment (ANATEX): Sampling and Analysis, Atm. Environ. 25(A), 2815‑2836.


Draxler, R., 1992:  Hybrid single‑particle Lagrangian integrated trajectories (HY‑SPLIT): Version 3.0 ‑‑ user's guide and model description, NOAA Technical Memo ERL ARL‑195, June, National Technical Information Service, Springfield VA.


Heffter, J.L., B.J.B. Stunder, G.D. Rolph, 1990:  Long‑range forecast trajectories of volcanic ash from Redoubt volcano eruptions.  Bull. Amer. Meteorol. Soc., 71, 1731‑1738.


Heffter, J.L. and B.J.B. Stunder, 1992:  Volcanic Ash Forecast Transport and Dispersion (VFTAD) Model, submitted to Weather and Forecasting.


Hoke, J.E., N. A. Phillips, G.J. DiMego, J.J. Tuccillo, and J.G. Sela, 1989:  The Regional Analysis and Forecast System of the National Meteorological Center, Weather and Forecasting, 4, 323‑334.


Kanamitsu, M., 1989:  Description of the NMC Global Data Assimilation and Forecast System,  Weather and Forecasting, 4, 335‑342.


McQueen, J.T. and R.R. Draxler, 1992:  Evaluation of model back trajectories of the Kuwait oil fires smoke plume by utilizing digital satellite data.  To be submitted to Atm. Environ.


Petersen, R.A. and J.D. Stackpole, 1989:  Overview of the NMC Production Suite, Weather and Forecasting, 4, 313‑322.


Rolph, G.D., and R.R. Draxler, 1990:  Sensitivity of three dimensional trajectories to the spatial and temporal densities of the wind field,  J. Appl. Meteorol., 29, 1043‑1054.


Rolph, G.D, R.R. Draxler, and R.G. dePena, 1992:  Modeling sulfur concentrations and depositons in the United States during ANATEX, Atm. Environ., 26A, 73‑93.


Sela, J.G., 1980:  Spectral modeling at the National Meteorological Center, Mon. Wea. Rev., 108, 1279‑1292.


Walko, R.L. and C.J. Tremback, 1991:  RAMS the regional atmospheric modeling system, ASTer, Inc., Ft. PO Box 466, Fort Collins, CO

                APPENDIX - Sample Model Outputs


The two example outputs shown in this section are based upon a hypothetical reactor accident located near the U.S. Canadian border.  A single short-time-period release near ground-level of radioactive material of unknown amount is assumed to have occurred on August 11 at 0000 UTC.  In the first example, illustrated in Fig. 1, a simple trajectory is used to represent the center-line of the forecast path of the material from the time of the accident to 48 hours in the future.  The source location is represented by the * symbol and the center position each 6 hours along the trajectory path is shown by the + symbol.  The map is labelled on top with the time of the release.  The rectangle near the bottom of the display shows the vertical projection (hPa) of the trajectory with time, in this case showing little vertical motion.  The calculation is based upon the forecast wind fields from the Nested Grid Model (NGM).


In the second example illustrated in Fig. 2, the 12-h average air concentrations at a level of 10-m are given for the period between August 12 at 0000 UTC (81200) and August 12 at 1200 UTC.  This period represents calculations using the NGM forecasts out to 36 hours.  Because the actual amount of material released was unknown at the time of the calculation, a unit emission rate was used for the source term.  Air concentrations are given by a 3 digit code representing the mantissa and negative exponent, such that 615 would represent the number 6 x 10-15.  Contour lines are drawn at approximately factor of two intervals.  Concentration units will depend upon the units of the emission.  This map will of course be output at consecutive intervals, showing the plume position every 6 or 12 hours, out to the maximum 48 forecast.




Email me the full report