SCIENCE APPLICATIONS INTERNATIONAL CORPORATION1
Micro SWIFT/SPRAY (MSS)
Presentation to:Chemical Biological Information Systems Conference
January 2007
Tom Harris, C. Dougherty, J. Sontowski, SAIC ([email protected])
Jacques Moussafir, Julien Commanay, ARIA Technologies([email protected])
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION2
Agenda
Micro SWIFT/SPRAY Overview
Joint Urban 2003 Simulations• IOP2 and IOP8
MSS Integration in HPAC
Summary
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION3
Problem Statement
One of JSTO’s ‘Early Warning’ Thrusts:“CBRN Hazard Prediction in Complex Urban Terrain with Near-real Time Accuracy”
M&S challenges for urban simulations include:Extensive structural databases and interfacesComplex flow patternsHigh resolution modeling requirements
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION4
Micro SWIFT/SPRAY Solution
The wind field solver, MicroSWIFT, is an adaptation of SWIFT, a mature, well validated mass consistency model
MicroSWIFT is designed to handle urban topography, length scales, and flow features.
Particle transport is modeled using a Monte-Carlo Lagrangian technique, MicroSPRAY.
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION5
Micro SWIFT /SPRAY Elements
Exact representation of buildings
Initial estimate of 3D mean flow from SWIFT, driven by available meteorological data
Analytical corrections for flow about obstacles
Entire flow field is iteratively updated to satisfy mass consistency
Dispersion simulated using a Monte-Carlo Lagrangian technique
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION6
Urban SWIFT Wind Field Construction
Typical Objects - Building and Flow Zones:
DISPLACEMENT
CAVITY LINK
CAVITY LINK
CAVITY
WAKE
WIND
BUILDING
SKIMMING
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION7
Interpolated wind field
Adjusted wind field
Adjusted wind field
Micro SWIFT Steps
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION8
Primary MSS Validation Cases
U-Shaped Building (Wind tunnel data)
L-Shaped Building (Wind tunnel data)
Salt Lake City “Joint Urban 2000” (Field Experiment)
Oklahoma City “Joint Urban 2003” (Field Experiment)
Bologna, Italy (Field Experiment)
U-shaped Building (Kastner-Klein, et al)
Bologna traffic emissions study
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION9
Agenda
MSS Overview
Joint Urban 2003 Simulations• IOP2 and IOP8
MSS Integration in HPAC
Summary
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION10
JU 2003 Simulations
The JU2003 Field Data provides a unique database for the evaluation of urban dispersion models.
Selected IOPs are the basis for an intercomparison of diagnostic urban wind field models• See paper by John Hannan et al.
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION11
Urban Model Domain for JU2003
N
Domain ~ 1300 X 1300 m in north-south and east-west directions
IOP 8 Wind DirectionRelease Location
IOP 2 Wind Direction
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION12
JU2003 IOP Comparison
IOP2* (2 July 2003)• Westin release (Daytime)• SW winds• 3 CRs (5 g/s)
IOP8* (18 July 2003)• Westin release (Night)• Southerly winds• 3 CRs (3 g/s)
*IOP used for comparison of Rockle-based models
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION13
MSS Predictions for JU2003 IOP2
OKC JU2003 IOP2 Test ConditionsDate: July 2, 2003Release location: WestinContinuous releases/times examined:
- CR1/1000-1030CST- CR2/1200-1230CST- CR3/1400-1430CST
Met data input to MSS: PNNL SodarMSS computation domain (see slide)
- domain size: ~1km square- horizontal resolution: 4m X 4m
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION14
MSS Wind Speed at h = 8 m
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION15
Need for high-resolution in time in the MET input for MSS
11:1511:0010:4510:30 11:30
WindDirection
Time
: Data points as average over 15 minutes time step: Step-function representation: Linear interpolation on unshifted data points
: Linear interpolation on shifted data points
15 minutes time step : 900m displacement for U=1m/s
10:00 10:15
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION16
Long term averages for wind inputs:
• Can lead to unrealistic results, especially when the winds are variable over the 15 minute averaging period.
• Particularly important for microscale modeling.• SWIFT and MSWIFT within HPAC interpolate between solutions at
instantaneous times. Preferred approach is to assign wind average to mid-point of averaging interval and interpolate between interval mid-points.
• Realism is improved with shorter averaging periods.• MSS computational efficiency allows, and microscale in some
cases requires, real time calculations for averaging periods of 15 minutes or less.
Issues regarding the averaging periods and HPAC/MSS inputs
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION17
IOP2 Concentration Contours
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION18
IOP2 Concentration Contours
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION19
Vertical Concentration Profiles
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION20
Comparison of Concentration Predictions with Sampler Data
IOP2 at 1600-1615, MSS vs OBS (Close to Source)
0
1
2
3
4
5
6
7
BANK1GRD B_M_SE B_M_SW MAIN_MED B_M_NE
STATION
Log1
0(PP
T+1)
Log10(OBS)Log10(MSSsw)Log10(MSSiw)
IO P2 at 1600-1615, M SS vs O BS (South of Source)
0
1
2
3
4
5
6
CC_2L CC_GRD S_B_NW WESTIN_G REN_NW
STATION
Log1
0(PP
T+1)
Log10(O BS)Log10(MSSsw)Log10(MSSiw)
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION21
Comparison of Concentration Predictions with Sampler Data
IOP2 at 1600-1615, M SS vs OBS (North from Park)
0
1
2
3
4
5
6
43 44 45 46 54 55 56 64 65
940 945 946 950 954 963 964 965PARKROOF
P_B_SW
STATION
Log1
0(P
PT+1
)
Log10(OBS)Log10(MSSsw)Log10(MSSiw)
IOP2 at 1600-1615, MSS vs OBS (West of Source)
0
1
1
2
2
3
3
4
940 CT_ROOF MAIN_TOP OKL_ROOF R_M_NE R_M_NW R_M_SE CCP_GRD CCP_TOP
STATION
Log1
0(PP
T+1)
Log10(OBS)Log10(MSSsw)LOg10(MSSis)
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION22
MSS Simulations for JU2003 IOP8
OKC JU2003 IOP8 Test ConditionsDate: July 18, 2003Release location: WestinMet data input to MSS: PNNL SodarMSS computation domain (see slide)
- domain size: 1km square- horizontal resolution: 4m X 4m
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION23
IOP8 Concentration Contours
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION24
Comparisons with Selected Tracer Instruments
LOG10(PPT) MSS vs MEAS for IOP8 CR2 - 15 min avg (MEAS at 0610)
0
1
2
3
4
5
6
7
MAIN_G
RDMAIN
_TOP
PKROOF34CT_R
OOFB_M
_NW
CCP_TOP
B_M_S
WBANK1G
RDOKL_
GRDCCP_G
RDBANK1_
SWBANK1_
RFPKROOF12WESTIN
_R
B_M_S
EOKL_
ROOFWESTIN
_GPKROOF47
REN_NW
Station
LOG
10(P
PT)
LOG10MSS
LOG10MEAS
•Data indicates upstream advection (e.g. at Westin_G and Ren_NW) that MSS does not predict.•Sensitive to relative position of reattachment and release points.
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION25
Velocity Vectors along Broadway for IOP8
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION26
Velocity Vectors along Broadway
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION27
Agenda
MSS Overview
Joint Urban 2003 Simulations• IOP2 and IOP8
MSS Integration in HPAC
Summary
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION28
MSS – HPAC Interface MSS Input Provided by HPAC
Domain – 640 m2 to 1 km2 @ 3, 4, or 5 m resolution, centered on the release locationMeteorology – profiles extracted from the larger domain SWIFT mass consistent wind field, which can be based on:• Observations – surface and/or upper air• Gridded output of NWP models (in MEDOC format)• Historical data (AFCCC climatology)• Fixed winds
Terrain/land use – interpolation of the HPAC DTED level 0 (1 km) and LandScan databasesUrban structures – triangulation of Shapefiles obtained from GEDIS (Geographic and Environmental Database Information System)Release – type, location, amount, duration, material properties
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION29
MSS – HPAC Interface MSS Output Provided to HPAC
Puffs – created by aggregating particles as they leave the MSS domain, at specified synchronization times
Plot quantities inside the MSS domain• Surface dosage• Surface concentration• Concentration slices• Vertically integrated concentration
3D wind and turbulence fields at specified times3D concentration and surface dosage fields at specified timesConcentration values at specified locations (samplers) at specified times
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION30
HPAC 5.00- Oklahoma City
Met: PNNL profile; Domain: 1 km @ 4 m
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION31
MSS: New Orleans
Met: fixed winds (MSWIFT at 0 min); Domain: .8 km @ 5 m; Runtime: 5 min.
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION32
MSS/HPAC Predictions
IOP2_CR1 - 1615
0
1
2
3
4
5
6
7
43 44 45 46 53 54 55 56 63 64 65 66 940
945
946
950
954
955
956
963
964
965
PARKROOF
BANK1GRD
B_M_S
EB_M
_SW
CC_2L
CC_GRD
CT_ROOF
MAIN_M
EDMAIN
_TOP
OKL_ROOF
R_M_N
ER_M
_NW
R_M_S
ES_B
_NW
WESTIN
_GB_M
_NE
CCP_GRD
CCP_TOP
REN_NW
P_B_S
W
Station
Log(
1+PP
T)
OBSMSS in HPAC
IOP2 CR1 t=1615
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION33
MSS applications in France
CEA-DAM is the military division of the French Atomic Energy Commission.(Commissariat à l’Energie Atomique – Direction des Applications Militaires)
Acquired MSS for Urban Emergency Response (Paris, other major cities).Develops a centralized operational system.Decided to fund the parallelization of the MSS code.
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION34
MSS applications in PARISRelease in the City Center : Châtelet – Les Halles
Courtesy of CEA-DAM Dr. Patrick ARMAND
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION35
MSS applications in PARIS
Release in the City Center : Place de la Concorde
Courtesy of CEA-DAM Dr. Patrick ARMAND
Elysée : French President’s Residence
US Embassy in Paris
SCIENCE APPLICATIONS INTERNATIONAL CORPORATION36
Summary
Micro SWIFT/SPRAY (MSS) is operational in HPAC 5.0Current focus on validation• JU 2003 and international studies
Comparisons with JU 2003 data indicate some successes, and some areas requiring more attention• Correct modeling of advection (recirculation,
channeling, vertical transport, etc.) in urban terrain may be more important than turbulent diffusion.
• Results are highly dependent on MET input.• Collaboration between Röckle-based modelers is
proceeding and demonstrating benefits.