Workshop Agenda

1
Workshop Agenda
Model Overview Model history and features Computational method Trajectories versus concentration Code installation Model operation Example calculations Updating HYSPLIT Meteorological Data Data requirements Forecast data FTP access Analysis data FTP access Display grid domain Vertical profile Contour data Examples 1-5 Particle Trajectory Methods Trajectory computational method Trajectory example calculation trajectory model configuration Trajectory error Multiple trajectories Terrain height Meteorological analysis along a trajectory Vertical motion options Pollutant Plume Simulations Modeling particles or puffs Concentration prediction equations Turbulence equations Dispersion model configuration Defining multiple sources Simulations using emission grids Concentration and particle display options Converting concentration data to text files Example local scale dispersion calculation Special Topics Automated trajectory calculations Trajectory cluster analysis Concentration ensembles Chemistry conversion modules Pollutant deposition Source attribution using back trajectory analysis Source attribution using source-receptor matrices Source attribution functions GIS Shapefile output KML/KMZ output Customizing map labels Scripting for automated operations Extra Topics Modeling PM10 emissions from dust storms Restarting the model from a particle dump file
82
82
2
Modeling Particle Motion or Particle
Distributions (Puffs)

• To compute air concentrations its necessary to
  follow all the particles needed to represent the
  pollutant distribution in space and time. This
  can be done explicitly by following the
  trajectory of each particle, where a random
  component is added to the mean velocity (from the
  meteorological model), to define the dispersion
  of the pollutant cloud. 
• In the horizontal, the computations can be
  represented by the following equations
• Xfinal(t ?t) Xmean(t ?t) U'(t
  ?t)?t, where, U'(t ?t) R(?t) U'(t) U''(1 -
  R(?t)2)0.5, (horiz. turbulent velocity)
• R(?t) exp(-?t/TLu), (auto-correlation
  coefficient)
• TLu is the Lagrangian time scale
• U'' su?, where ? is a random number with
  mean of 0 and s of 1. 
• The computations can be simplified, if instead of
  modeling the motion of each particle, we compute
  the trajectory of the mean particle position and
  the particle distribution. The standard deviation
  of the particle distribution can be computed from
  all the particles,
•        ______ s2  (Xi-Xm)2
• or it can be computed without following
  individual particles by assuming a distribution
  shape (puff) and relationship to the local
  turbulence.  Many different formulations can be
  found in the literature. dsh/dt v2 su su
  (Ku / TLu)0.5
• These computations are set in the Advanced /
  Configuration Setup / Concentration menu, which
  modifies the SETUP.CFG file.

83
3
Modeling Particle Motion or Particle
Distributions (Puffs)

• Below, note the initial differences between the
  simulation using the 3D particle distribution
  (left) and the top-hat puff center position
  method (right). Without the random motion
  component, the top-hat puff positions follow a
  straight line during the initial few hours until
  vertical motions or horizontal divergence begins
  to act on the particles. In this particular case
  the primary reason for the sudden expansion of
  the puff-particles is that they have mixed to
  higher levels and we are seeing the differential
  horizontal advection acting upon the particles.

3D Particle Distribution
Top-hat Puff Center Positions
84
4
Modeling Particle Motion or Particle
Distributions (Puffs)

• The previous example showed a snapshot of the
  particle or puff center positions after 12
  hours.  Air concentrations are computed by
  summing each particles mass as it passes over
  the concentration grid.
• In the particle model mode, the concentration
  grid is treated as a matrix of cells, each with a
  volume defined by the grid dimensions.  Therefore
  the concentration is just the particle mass
  divided by the cell volume
• 3D Particle      ?C q(?x ?y
  ?z)-1 Top-Hat           ?C q(? r2
  ?z)-1 Gaussian        ?C q(2? sh2 ?z)-1 e-
  0.5x2/sh2
• In the puff model mode, the concentration grid is
  considered as a matrix of sampling points, such
  that the puff only contributes to the
  concentration if it passes over the sampling
  point.  In the puff calculation mode it is
  possible for a puff to pass between points and
  not be shown on the display
• Top-Hat           ?C q(? r2
  ?zp)-1 Gaussian        ?C q(2? sh2 ?zp)-1 e-
  0.5x2/sh2

85
5
Modeling Particle Motion or Particle
Distributions (Puffs)

• Shown below are the concentration patterns
  associated with the particle (left) and puff
  (right) distributions from the previous example. 
  Note that the puff distribution is smoother but
  also initially somewhat broader.  In this
  particular case, the horizontal puff growth
  equations give larger values than the particle
  expansion. The noisy particle distribution
  indicates that more particles than 5000 used are
  needed to better represent the horizontal
  distribution.

3D Particle Distribution
Top-hat Puff Center Positions
86
6
Turbulence Equations
The method by which the meteorological data are
evaluated to determine the turbulent velocities,
used in either the puff or particle computation,
is set in the Advanced / Configuration Setup /
Concentration menu (below-left). Clicking on the
Configure the TURBULENCE method button produces
the menu given below-right.
87
7
Turbulence Equations

• Turbulence Computation Methods
• Standard velocity deformation - The default
  standard method computes the mixing using a
  diffusivity approach based upon vertical
  stability estimates and the horizontal wind field
  deformation
• Kz k wh z (1 - z/Zi)2
• Kh 2- 0.5(c ?)2 ?u/?y ?v/?x
• Short-range similarity (fluxes/profile) - In
  shorter range dispersion simulations (lt 100 km)
  the deformation parameterization used in
  conjunction with larger scale meteorological
  fields will not reflect the diurnal variations in
  horizontal turbulence. In these situations its
  desirable to use the short-range
  parameterizations in which the turbulent
  velocities are computed directly from the
  stability parameters, heat and momentum fluxes,
  if available, or derived from the wind and
  temperature profiles.  The user has the option of
  forcing the use of the profile to compute
  stability rather than the fluxes. This may be
  desirable, especially if the fluxes represent
  averages rather than instantaneous values. The
  boundary layer velocity variances are defined as
  a function of u, w, and Zi.  This method does
  not use the diffusivity and no assumptions are
  required about turbulent scales. For instance, in
  the stable/neutral boundary layer
• w'2 3.0 u2 (1 z/zi)3/2
• u'2  4.0 u2 (1 z/zi)3/2
• v'2   4.5 u2 (1 z/zi)3/2

88
8
Turbulence Equations

• Turbulence Computation Methods (Cont.)
• Input meteorological model TKE - If the turbulent
  kinetic energy (TKE) field is available from the
  meteorological model, then the velocity variances
  can be computed from its definition and the
  previous velocity variance equations to yield
  relationships with TKE
• E 0.5 (u2 v2 w2)
• w2 0.32 E,  u2 0.74 E,  v2 0.85 E
• u2   v2 0.36 w2
• TKE and mixed layer defined TKE The TKE can be
  used with the mixed layer depth as estimated from
  the TKE profile.
• Turbulence velocity variance - Some
  meteorological data sets may already contain the
  3-dimensional component velocity variances.  This
  would normally be the case for data that have
  been generated from local measurement programs.

The Puff Growth Computation Method section is
used to define either the Linear or Square Root
with time dispersion equation for the horizontal
growth rate of puffs. This option does not affect
particle dispersion. The user also has the
ability to set ratios of the vertical to the
horizontal turbulence for daytime and nighttime
in the Turbulence Aniosotropy Factors section
(click the Help button for more details).
89
9
Dispersion Model Configuration

• The control file (CONTROL) for dispersion
  simulations is configured from the Concentration
  / Concentration Setup menu tab.  The
  concentration setup layout is identical to the
  trajectory menu with the exception of an
  additional button to set the emissions,
  deposition, and concentration grid (top right).
• The Pollutant, Deposition and Grids setup button
  will bring up a submenu (lower right) with three
  options (Pollutant, Grids, Deposition).
• To make modifications, enter the number of
  pollutants to define in the Num box and then
  click on the Specie or Grid to access the
  next menu.
• The pollutant emission rate and deposition must
  be set for each pollutant. 
• Several independent concentration grids may be
  defined for each simulation. However, they may be
  nested in space or time, if desired. 
  Concentrations for each pollutant species are
  output on all grids.

90
10
Dispersion Model Configuration

• Definition of Pollutant
• An arbitrary 4-character field identifies each
  pollutant.
• The Emission rate is mass units per hour. The
  actual mass unit is not specified, so for
  instance, if the units are kg, then concentration
  output will be in kg/m3.  Any unit is acceptable,
  however some chemical conversion modules require
  specific units.
• The Hours of emission may be defined in
  fractional hours.
• The pollutant Release start can be set to any
  time at or after the start of the simulation.  As
  is true for all time units, zeros default to the
  simulation start time in the main menu.  Zero for
  the month and non-zero values for day and hour
  cause those values to be treated as relative to
  the simulation start time.

91
11
Dispersion Model Configuration

• Definition of Concentration Grid
• Each concentration grid must be defined. 
• Zeros for the grid center default to the source
  location. 
• The grid spacing is especially important in
  concentration computations in determining the
  cell size (particles) or sampling resolution
  (puffs). 
• When multiple levels are defined, each height
  represents the top of the cell (particles) or
  actual height (puffs). 
• The averaging time (Avg) starts at the sampling
  start time for the hours/minutes specified in the
  output interval. 
• Snapshot concentrations (Now) are defined as the
  average over one time-step at the time interval
  specified. Max will save the maximum
  concentration at each grid point over the
  duration of the output interval.

92
12
Example Dispersion Calculation

• Run the dispersion model using these settings
• Source 28.50N, 80.70W _at_ 10.0 m
• Meteorology NAM 12 km
• Emission 12 hrs beginning 1200 UTC on 19 Dec
  2005
• Output Snapshot after 12 hrs between the ground
  and 100 m-agl
• Top-hat-horizontal, particle-vertical
• 5000 particles
• Run Standard Model

93
13
Example Dispersion Calculation

• Results
• Change the map background file from arlmap to
  floridamap in the Concentration Display menu and
  display the results.
• The resulting graphic should be the same as that
  shown (right).

The floridamap file and other high resolution map
backgrounds can be downloaded from the NOAA ARL
website at http//www.arl.noaa.gov/ready/hysp_uti
l.html
94
14
Example Dispersion Calculation

• All HYSPLIT simulations generate a text MESSAGE
  file, which contains diagnostic information about
  the calculation.  Use the View MESSAGES link from
  the Advanced menu tab to view the last MESSAGE
  file. In this case (below), at the end of the
  simulation, 12.00014 units of mass were still on
  the domain. The vertical mass distribution showed
  more than 80 of the mass to be within 1000 m of
  the ground.  The vertical mass distribution is
  computed independently of the vertical
  concentration grid.

95
15
Defining Multiple Sources

• Now run the dispersion model for 2 sources
  using these settings
• Source1 28.50N, 80.70W _at_ 10.0 m
• Source 2 28.0N, -80.0W _at_ 10.0 m
• Meteorology NAM 12 km
• Emission 6 hrs beginning 1200 UTC on 19 Dec 2005
• Output 6 hr average concentration between the
  ground and 100 m-agl
• Top-hat-horizontal, particle-vertical
• 5000 particles
• Run Standard Model

96
16
Defining Multiple Sources

• A second source added at location 28.0N and 80.0W
  results in two adjacent, almost identical
  plumes. 
• Note that the emission rate of 1 unit per hour is
  applied to each source individually.

97
17
Defining Multiple Sources

• The emission rate can be set for each source by
  including that information after the release
  height in the Starting Location Setup menu. (A
  fifth field can be added that sets an initial
  plume area in square-meters, but is only valid
  for puff simulations.)
• In the example shown here (top right), the
  emission rate of the second source has been
  increased to 10 units/hr
• The concentrations in the second plume (right)
  have increased by the same amount as the emission
  increase (10).
• To display the same concentration levels as the
  last graphic, change the output contour levels in
  UserSet to 1.0E-091.0E-101.0E-111.0E-12

98
18
Simulations using Emissions Grids

• Option 1 Matrix Approach
• An emission matrix is defined using three
  locations the first two represent the lower left
  and upper right grid corners, respectively, and
  the third represents the grid spacing.
• Example to run  Start sources every 0.5 degrees
  between the grid corners (27.0, -82.0) and (30.0,
  -79.0). Leave all other parameters the same as
  the last Florida example, however increase the
  maximum number of particles to at least 50,000 in
  the Advanced / Configuration Setup /
  Concentration menu.
• Run the model from the Concentration / Special
  Simulations / Run Matrix menu option.
• Prior to running the model, the CONTROL file is
  redefined with 49 starting locations (right).

99
19
Simulations using Emissions Grids

• The result (top right) shows 49 plumes over a
  uniform 0.5 degree grid.
• To make the graphic less noisy, from the
  Concentration Display menu, turn off the source
  location labeling and remove the black contour
  lines from the graphic by setting the contour
  outlines to none
• Execute the display to create a considerably
  simplified graphic (below right).

100
20
Simulations using Emissions Grids

• Option 2 Emissions File Approach
• Another approach is to define an emissions file,
  which tabulates hourly emission rates by
  location. 
• An emission.txt file defines the grid to which
  those emissions data will be accumulated and is
  located in the working directory. 
• A simulation using this approach for the same
  case (right) is almost identical to the previous
  matrix run. 
• The CONTROL file must define the lower left and
  upper right corners of the desired emission
  domain, which may be larger or smaller than the
  data available. 
• Due to the continuous emissions at many
  locations, these simulations may require the
  maximum number of particles be increased from the
  default value as was done for the matrix run.
• Information on the format of the emissions file
  and the emission text file can be found in the
  HYSPLIT User's Guide under Advanced / Special
  Topics (S441).

101
21
Concentration and Particle Display Options
Now, we will look at the particle distributions
for the Florida case for various source terms.

• Setup the following run
• Delete the emission.txt and emission.asc files
  from the working directory if used previously.
• Source1 28.50N, 80.70W _at_ 10.0 m
• Meteorology NAM 12 km
• Emission 6 hrs beginning 1200 UTC on 19 Dec 2005
• Output snapshot at 6 hours between the ground
  and 100 m-agl
• 3D particle horizontal and vertical
• 500 particles
• Dump the particles after 6 hours (right)
• Run Standard Model

102
22
Concentration and Particle Display Options

• Results
• Turn back on source labeling and color contour
  outlines in the Concentration Display menu and
  execute the display.
• The resulting graphic should be the same as that
  shown (right).
• The concentration output clearly shows a noisy
  pattern indicating too few particles were defined
  to adequately represent the plume.

103
23
Concentration and Particle Display Options

• Setup the following runs
• Rerun the last case, but use 5,000 and 50,000
  particles.
• Make sure the maximum number of particles is
  greater than 50,000.
• The particles are beginning to better define the
  plume, but at the expense of longer computational
  time.

5,000 Particles
50,000 Particles
104
24
Concentration and Particle Display Options

• Results
• To speed up the run without loosing the plume
  structure, change the type of run from a 3D
  particle to a top-hat-horizontal
  particle-vertical and reduce the number of
  particles to 2500 (500 is still not enough to
  resolve the plume.
• The resulting plume (right) covers a similar
  footprint as the 50,000 3D particle run.

105
25
Concentration and Particle Display Options

• Particle Display
• In addition to the standard display of particle
  concentrations, individual particle positions can
  also be displayed on a map.
• The Concentration / Display Options / Particle
  menu (right) has options to show snapshot
  particle distributions, assuming that the
  particle dump option was set in the Advanced /
  Configuration Setup / Concentration menu before
  running the particle simulation.
• Horizontal, vertical, and cross-sectional views
  are available.
• Other options include color-coding the particles
  by mass size (Mass Sizing), by height (Color
  Scale) or output as a shapefile (GIS).

26
Concentration and Particle Display Options

• Particle Display
• Rerun the 50,000 3D particle simulation to
  produce a PARDUMP particle dump file.
• Then choose the particle vertical cross-section
  with the color Scale option checked.
• As seen in the graphic (right), the center line
  of the cross-section is drawn automatically based
  upon the particle distribution.
• The particles toward the west are at a higher
  level than those to the east.

27
Concentration and Particle Display Options

• Pointer Select Concentration Display
• Another display option is to view the
  concentration values directly on the grid without
  any interpolation through the Concentration /
  Display Options / Pointer Select tab.
• This option will draw the entire concentration
  domain as defined in the concentration grid setup
  menu. The grid span may need to be reduced to
  zoom in on the area of interest.
• Click on the initial map domain image with the
  right mouse button to display the concentrations
  (right). In this case the full 30 x 30 degree
  concentration grid defined previously covers an
  area much larger than the plume.

28
Converting Concentration Data to Text Files

• The concentration output file is in a binary
  format, however there are several options
  available through the Concentration / Utility
  Programs menu that can be used to convert the
  concentration data to other formats.
• First, prepare a multi-time period output file by
  setting up a simulation as in the previous
  example, but with the following changes
• Top-hat-horizontal particle-vertical,
• No particle dump interval (0),
• 12 hour simulation,
• 12 hour continuous release,
• 500 particles, and
• 1 hour average concentrations.
• Check the Fix-Exp box in the Display menu to keep
  the contours constant and you may need to change
  the name of the output file from partplot to
  concplot.
• After displaying the Postscript output, create an
  animated gif image by using the Concentration /
  Utility Programs / Convert Postscript menu by
  checking the animate box in the Postscript
  Conversion menu.
• The continuous emission plume moves southwest and
  then more to the south near the end of the
  simulation.

29
Converting Concentration Data to Text Files

• Time Series Data Extraction
• Next, select the Concentration / Utility Programs
  / Grid to Station menu (right).
• Select a point downwind in the plume (27.7N,
  81.0W),
• Give it a unique Integer ID (2781),
• Set the Concentration Multiplier to 1.0,
• and choose a Log Ordinate scale.
• Click Extract Data and an ASCII con2stn.txt file
  will be created with the concentration values
  interpolated to that location.  (An input file
  with the station locations must be created for
  multiple locations).
• Selecting the Display Time Series Yes button
  results in the creation of a time series plot
  (right). In this case the peak concentration
  occurred at 1600 UTC on 19 December 2005.

30
Converting Concentration Data to Text Files

• The Concentration / Utility Programs / Convert to
  ASCII menu will convert every non-zero grid point
  value to its ASCII equivalent, writing the output
  to one file per time period unless you specify
  Single File.
• Files are labeled according to the name of the
  binary file, Julian day, and hour of the sampling
  period.
• See the contents of this file for the output from
  the first time period.
• This file can useful when importing the data into
  other mapping applications.

31
Example Local Scale Dispersion Calculation

• HYSPLIT can be configured for applications such
  as emergency response, when the scale of the
  simulation is on the order of 10-30 km.
• For this example, set up the run as shown below
  for Washington, D.C.
• 38.880N 77.027W _at_10m and 100m,
• 1200 UTC 19 December 2005,
• NAM 12 km forecast data,
• 1-hr emission and simulation,
• 1-hr average concentration after 1 hr,
• Lat/lon Grid Resolution of 0.001 degrees, and
• Grid Span of 1.0 degrees lat/lon.
• Using the Advanced / Configuration Setup /
  Concentration menu, set
• 3-D particle horizontal and vertical method,
• 1000 particles,
• 10000 maximum number of particles, and
• Short-range similarity (fluxes) method for the
  turbulence computation.
• Run Standard Model

32
Example Local Scale Dispersion Calculation

• After running the model, set the Concentration
  Display menu to the following (right)
• Output File concplot
• Number of rings to 4 every 10 km
• Map background to countymap
• Zoom to 100
• Dyn-Exp contours
• Turn on the contour outlines 
• Turn on the Google Earth option

33
Example Local Scale Dispersion Calculation

• The resulting plume (below left) produces a very
  narrow plume moving southeast into Maryland over
  the 1 hour period.
• As will be discussed later, the Google Earth file
  (HYSPLITconc.kmz) was created and allows the
  emergency manager to overlay the plume with other
  geographic features (below right). This file can
  be provided directly to the emergency manager.

34
Example Local Scale Dispersion Calculation

• Now assume the release was very small and only
  lasted 15 minutes. Use the Concentration setup /
  Pollutant, Deposition and Grids setup menu
  (right) to define a 15 minute (0.25h) release of
  one unit of mass. Note that since the release
  rate required is per hour, you will need to
  multiply the mass by 4 in this case.
• Also, change the averaging period to 10 minutes
  over the 1 hour simulation, which can be defined
  in the Concentration Grid Setup menu.
• Run the model and create an animated GIF (right).
  To keep the contours from changing as the
  concentrations decrease, you may want to fix the
  contours by checking Fix-Exp box in the
  Concentration Display menu.