Professional Documents
Culture Documents
i,j,k
m
j
i
K
x
i
i3
g
0
(
0
)
1
x
j
(
ij
)
R
(u
i
u
i0
)
i3
where is the Exner function, m
j
is the momentum vector, is the subgrid
stress.
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory
Mesoscale Atmospheric Transport and Dispersion Modeling 5
The conservation equation of a scalar, , is used to calculate quantities such
as the concentration of contaminants in an Eulerian sense within the RAMS
model:
t
+
1
(m
j
)
x
j
m
j
x
j
= S
x
j
(
)
R
(
0
) (2)
where is replaced by for temperature, by q for moisture and by a concentration
of choice, such as c
CO2
.
Once the RAMS predicts the 3D evolution of the atmosphere, HYPACT can be
employed to supplement the Eulerian treatment of the contaminant transport
by using meteorological output from the RAMS to predict the location and con-
centration of uid particles in a Lagrangian sense using the following equations
for particle location (x
i
) and deterministic (u
i
) and random (u
i
) components of
its velocity:
x
i
(t +t) = x
i
(t) +[u
i
(t) +u
i
]t (3)
u
i
(t +t) = u
i
R
i
(t) +u
i
(t) (4)
While a purely Eulerian treatment for the calculation of contaminats concen-
trations using the RAMS is possible, the HYPACT module oers an alternative
method for estimating atmospheric transport and dispersion that is not restricted
by the spatial resolution of the RAMS model.
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory
Mesoscale Atmospheric Transport and Dispersion Modeling 6
B. HYSPLIT4 Model
The HYbrid Single Particle Lagrangian Integrated Trajectory Model version 4
(HYSPLIT) Model of NOAA Air Resources Laboratory (Draxler and Hess, 1997,
NOAA Tech. Note) is used as an alternative to RAMS/HYPACT in this work. It
has been widely used in the study of the transport of contaminants (e.g. Draxler
et. al, 1994, Atm. Env., 28, 21972210).
The dispersion of a pollutant is calculated by the HYSPLIT4 model by assuming
either a Gaussian or TopHat horizontal distribution within a pu, or from the
dispersal of a xed number of particles. A single released pu expands in the
horizontal until its size exceeds the grid size and then it splits into several pus.
In HYSPLIT4 either the pu or the particle approach can be used for horizontal
transport while only the particle approach is used for the vertical transport.
Similar to the RAMS/HYPACT model, HYSPLIT4 can be initialized with ob-
served and/or predicted meteorological data to reect the prevalent atmospheric
conditions when calculating the location and concentration of contaminants.
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory
Mesoscale Atmospheric Transport and Dispersion Modeling 7
NUMERICAL EXPERIMENTS:
A grid domain of (625x500x12) km
3
in (x,y,z) was selected for simulating the
emission and dispersion of contaminants released from a hypothetical non
buoyant point source located at ground level within SRS using both the
RAMS/HYPACT and the HYSPLIT4 models.
A total number of 125x100x40 grid points in (x,y,z) with a horizontal spacing
of 5 km in x and y, and a variable vertical grid spacing starting from 10 m above
the surface with a 1.2 grid stretch ratio up to 500 m maximum grid spacing at
the top of the model domain were used in both models.
Both models were initialized with meteorological observations available through
the NWS for the domain centered over the SRS region for the July 12, 1995
case study day.
A constant emission rate (of 5 g per second) of an inert contaminant was
assumed for a hypothetical point source located within the SRS during the
application of both models.
The model integrations were carried for 24 hours starting from 00 GMT of July
12 1995.
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory
Mesoscale Atmospheric Transport and Dispersion Modeling 8
RESULTS
Figure 1 shows the horizontal and vertical observed wind structure at 18:00
GMT (1:00 PM local time) as the driving force for the transport and dispersion
of released uid particles.
The general direction of uid particle trajectories (corresponding to inert con-
taminants) calculated by the HYPACT (gure 2a) and the HYSPLIT4 (gure
2b) models are similar; however, important dierences in terms of local varia-
tions of particle location are observed in the outputs of the models.
The horizontal extent of the high concentrations of the plume covering the SRS
region just above the surface predicted by the RAMS/HYPACT (gure 3a) is
about one third of the area predicted by the HYSPLIT4 model (gure 3b).
3D depictions of 0.3 g/m
3
isosurface taken from the outputs of both models
indicate that the vertical height that contaminants can penetrate is estimated
as 2 km by the RAMS model (gure 4a) while this height is estimated as 6 km
by the HYSPLIT4 model (gure 4b).
The maximum concentration level estimated by the HYSPLIT4 model is about
one fth of the amount estimated by the RAMS/HYPACT model.
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory
Mesoscale Atmospheric Transport and Dispersion Modeling 9
Figure 1: Horizontal and vertical cross sections of wind elds available from
NWS used for the case study at time 18:00 GMT (1:00 PM local time).
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory
Mesoscale Atmospheric Transport and Dispersion Modeling 10
Figures 2a, b: Trajectories of contaminants predicted by the the
RAMS/HYPACT (a) and the HYSPLIT4 (b) model at 18:00 GMT (1:00 PM
local time) at 10 m above the ground level. [An mpeg le (meso.mpg, see
below) is included that depicts the transport of uid particles released at ground
level within the SRS.]
(a) (b)
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory
Mesoscale Atmospheric Transport and Dispersion Modeling 11
Figures 3a, b: Spatial pattern of gaseous contaminant concentrations predicted
by the RAMS/HYPACT (a) and the HYSPLIT4 (b) model at 18:00 GMT (1:00
PM local time) at 10 m above the ground level.
(a) (b)
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory
Mesoscale Atmospheric Transport and Dispersion Modeling 12
Figures 4a, b: 3D concentration elds of the plumes predicted by the RAMS
model (a) and the HYSPLIT4 (b) model at 18:00 GMT (1:00 PM local time).
(a) (b)
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory
Mesoscale Atmospheric Transport and Dispersion Modeling 13
DISCUSSION
The accidental release of contaminants from DOE sites is possible, and infor-
mation for the potential location and concentration levels of contaminants is
important for emergency preparedness.
There are numerous dispersion/transport models available to the scientic com-
munity with varying degree of sophistication regarding the atmospheric physics
built in. Evaluation using eld measurements is needed in order to decide which
methods are expected to give more reliable predictions.
While the particle trajectories calculated by both models considered here are
generally similar (gure 2), discrepancies in particle locations (concentration
patterns) are large.
Figure 3 suggests that the Gaussian treatment of the dispersion within the
HYSPLIT4 model spreads the contaminants much faster than the Eulerian cal-
culations used in the RAMS model.
The vertical penetration level of the plume that is predicted by the HYSPLIT4
model is much higher than that predicted by the RAMS/HYPACT model. How-
ever, since there was not any weather disturbance during the case study day,
inert and non-buoyant contaminants are not expected to penetrate into the
higher altitudes beyond the top of the ABL, which is typically 2 km. Therefore,
the RAMS/HYPACT predictions appear to be more reliable.
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory
Mesoscale Atmospheric Transport and Dispersion Modeling 14
The execution time of the HYSPLIT4 is much smaller (less than half an hour)
compared to that required by the RAMS/ HYPACT (of the order of two weeks
on a Sun-Ultra 1 Creater 3-D computer). Emergency planning requires realistic
predictions for the spatial concentration patterns in as a short time period as
possible, and the short execution time of the HYSPLIT4 model would make
the HYSPLIT4 model a preferred model for such use. However, HYSPLIT4
model appears to overestimate both the horizontal and vertical extent of plume
coverage. This result also aects the 3-D spatial concentration patterns of the
plume.
Comparisons of the results of the RAMS/HYPACT with those of the HYSPLIT4
model can be used to improve the model physics within the HYSPLIT4 model
in estimating the spatial patterns of contaminants.
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory
Mesoscale Atmospheric Transport and Dispersion Modeling 15
FUTURE PLANS
An independent evaluation of the results of both models will be performed using
the data obtained from tracer experiments such as MATS eld study conducted
at SRS (Weber and Berman, 1992, Nuclear Safety, v.33, page 58).
Additional data from eld studies for other geographical locations, such as
DOPPTEX (Thuillier, R. H., 1991, Pacic Gas and Electric Company, San
Ramon California) conducted in Los Angeles, will be selected for further poten-
tial evaluation of the dierent modeling approaches.
Upcoming work will also consider the eects of buoyant plumes such as those
originating from res or from heavy gases that can accidentally be released from
DOE sites.
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory
Mesoscale Atmospheric Transport and Dispersion Modeling 16
ACKNOWLEDGEMENTS
This work has been supported by the CRESP (Consortium for Risk Evaluation
with Stakeholder Participation) project, which is funded by the U.S. Department
of Energy (DOE) under cooperative agreement DEFC0195EW55084.
CRESP/EOHSI-Exposure Assessment Computational Chemodynamics Laboratory