NOAA Technical Memorandum ERL ARL - 195 HYBRID SINGLE ...

1 NOAA Technical Memorandum ERL ARL - 195 HYBRID SINGLE -PARTICLE LAGRANGIAN INTEGRATED TRAJECTORIES(HY-SPLIT): VERSION -- USER'S guide AND MODEL DESCRIPTIONR oland R. DraxlerAir Resources LaboratorySilver Spring, MarylandLast Revision: June 1992iiNOTICEM ention of a commercial company or product does not constitute an endorsementby NOAA Environmental Research Laboratories. Use for publicity or advertisingpurposes of information from this publication concerning proprietary productsor the tests of such products is not Modeling system Validation 22. PROGRAM AND FILE Transport and 113. METEOROLOGICAL Upper Air and Surface NMC Gridded Archive and Forecast Special Observational Data Archive 154. HY-SPLIT OUTPUT Run-time Post-processing 205. SPECIAL Backward Vertical 236. MISCELLANEOUS Multiple Source Sulfur Nitrogen Particles, Deposition, and Radiological 247.

2 258. 25 Appendix A - Transport and DispersionAppendix B - HY-SPLIT Meteorological ProcessorAppendix C - Rawinsonde Data PreprocessorivAppendix D - Gridded Meteorological Data FormatAppendix E - Input File DescriptionAppendix F - Input File DescriptionAppendix G - Installation, Running, and File FormatsAppendix H - HYTEK Tektronix Terminal EmulatorAppendix I - HY-SPLIT Error CodesAppendix J - Verification StudiesvList of TablesPage1. Modeling system flowchart 5 List of Tables - Appendices1C. Functions for normalized profiles of momentum heat1D. Structure and orientation of meteorological grids2D. Availability of meteorological data Meteorological parameters and their organization Vertical levels available in each data set Grid structure of packaged data Statistical summary of ANATEX simulations of FiguresPage1 Sample display for the base case (test001)from the HYEDIT program.

3 62 Sample display from the HYPICK based file editor. 73 Sample display from the HYMAPS mouse based editor program. 84 Illustration of the ASCII screen displayduring a model simulation without run-time graphics (test001). 105 A snapshot map from a SINGLE release (test002)after 24th travel. 186 Trajectory map after 24-h travel time with positionsshown hourly (test003). 197 Run-time concentration map display (test004) showingthe 24-h average values. 208 Postprocessing trajectory output from TRAJPLOT (test005)is identical to that shown in Fig. 6. 209 Illustration from qualitative concentration display (test006).

4 2110 Illustration from the quantitative concentration display. 2211 Backward trajectory computation (test007) for theforward case shown in Fig. 6. 2212 Trajectory with vertical velocity computed fromthe divergence field (test008). 23viiList of Figures - AppendicesPage1A Normal probability distribution. The dashed horizontalline indicates the ordinate value at which the shadedarea equals the area above the line at . A-32A The distance (A) a particle is advected during a timestep. Three positions are interpolated so that theconcentrations can be computed at all grid points atthe interpolated (dashed) and initial and finalpositions (solid). A-43A Illustration of how a SINGLE particle Q0 splits dueto vertical diffusion into two particle Q2 and Q3. A-54A Illustration of how a SINGLE particle with radius Rsplits due to horizontal diffusion into four particleseach with radius R/2.

5 A-55A Illustration of when two particles are merged into asingle particle. It occurs when the difference betweenthe vectors is less than R. A-61D NGM data grid number 6. D-22D MRF northern hemisphere data grid number 2. D-23D MRF southern hemisphere data grid number 3. D-21J Three-month average measured concentration for releasesfrom both Glasgow and St. Cloud. J-12J Three-month average calculated concentration for releasefrom Glasgow and St. Cloud. J-13J Average measured concentration (fl/L) from all Average calculated concentration from all the CAPTEX tracer releases. J-35J The ratio of calculated to measured concentration pairssorted by cumulative percent for the CAPTEX and Trajectory using MRF data, comparable to the case shownwith NGM data in simulation test005.

6 J-57J Air concentration estimates using MRF data, similar tothe NGM case shown in Fig. 10 (test006). J-6 HYBRID SINGLE -PARTICLE LAGRANGIAN INTEGRATED TRAJECTORIES(HY-SPLIT): VERSION -- USER'S guide AND MODEL DESCRIPTIONR oland R. DraxlerABSTRACT. The algorithms and equations used in the calculation oflong-range pollutant transport and dispersion are presented from themeteorological data sources through the calculation of airconcentrations. The model calculation methods are a HYBRID betweenEulerian and Lagrangian approaches. A singlepollutant particlerepresents the initial source. Advection and diffusion calculationsare made in a Lagrangian framework. However, meteorological inputdata can either be gridded from rawinsonde observations or archivedfrom other sources, such as the analyses fields or forecast outputsfrom Eulerian primitive equation models. As the dispersion of theinitial particle spreads it into regions of different wind directionor speed, the SINGLE particle is divided into multiple particles toprovide a more accurate representation of the complex flow concentrations are calculated on a fixed three dimensional gridby integrating all particle masses over the sampling time.

7 The modelcode can be run on a personal computer. Calculations consist ofsimple trajectories from a SINGLE source to complex multiple sourceemissions. A flow chart of the calculations aswell as severalexamples are given. The code is structured so that concentrationcalculations or simple trajectory (forward or backward) calculationscan be performed on sigma (terrain following) or Modeling System OverviewThe development of the HY-SPLIT long-range transport calculation techniquehas evolved in several stages, from a simple wind-shear induced particledispersion study (Draxler and Taylor, 1982), to the inclusion of airconcentration calculations with only a day/night mixing assumption (Draxler,1982), to the calculation of vertical mixing coefficient profiles and verticalparticle motions (Draxler, 1987). The modeling technique has manysimilarities to particle-in-cell (PIC) methods (Lange, 1978) with the primaryexception that one starts with only a SINGLE particle that may split into asmany particles as needed to describe the pollutant distribution.

8 Thisapproach requires special numerical techniques to limit the number ofparticles. The equations and calculational methods are discussed in much moredetail in Appendix each particle has an associated horizontal and vertical diffusivecomponent, similar to a "puff" model, the subsequent particle divisionsdetermine the spatial and temporal concentration distributions, rather thanthe diffusion about any one particle. With PIC methods the initialconcentration distribution is defined by the distribution of particles, but atlater times there may be too few particles in adjacent cells to define theconcentration field without assuming a particle probability distribution. Incontrast with HY-SPLIT, near the source region, the concentrations are mostsensitive to the assumed distribution about the SINGLE emitted particle, whileat further downwind distances the concentrations are most sensitive to thesubsequent particle most transport and dispersion models, the limiting feature to theaccuracy of the calculations is the spatial and temporal resolution of themeteorological data.

9 In previous model versions rawinsonde observations, withlimited spatial (400 km) and temporal density (every 12 hr), were the onlydata source. The rawinsonde data were then gridded on a user-selected polarstereographic grid by a preprocessor program (see Appendix C). This was themethod used when the dispersion model calculations were compared with the airconcentrations measured during the Cross Appalachian Tracer Experiment(CAPTEX) (Draxler, 1987; Draxler and Stunder, 1988). The current HY-SPLIT version permits the use of meteorological input data from other sources, inparticular meteorological data fields archived from the NationalMeteorological Center's (NMC) Nested Grid Model (NGM - Phillips, 1975) overthe continental (90 to 180 km every 1 to 2 hr) or data archived from theinitialization fields of NMC's Medium Range Forecast Model (MRF) for globalcoverage (381 km every 6 h). These data are discussed in much more detail inAppendix D.

10 An explanation of how the HY-SPLIT code converts the differentdata sources into a common structure can be found in Appendix Validation StudiesOne of the more important aspects of any transport and dispersion code isunderstanding the causes of uncertainty and the potential errors in theresulting calculations. A major long-range validation experiment wasconducted during January through March of 1987, Across North America TracerExperiment (ANATEX - see Draxler et al., 1991). Inert tracers were releasedfrom two sources routinely every 2 days and daily averaged samples werecollected at distances of up to 3000 km from the tracer sources. Enhanced 4per day rawinsonde observations as well as high resolution NGM model outputwere available for the three month experimental period. A detailed evaluationof several long-range transport models, including HY-SPLIT, using the ANATEX data was conducted by Clark and Cohn (1990).Other validation studies during ANATEX (Draxler, 1990) compared the NGMdata with meteorological observations on a 500 m tower.