The 3rd International Workshop on Next Generation Climate Models for Advanced High Performance Computing Facilities Tokyo, Japan, March 28 -30, 2001 Performance and Verification of the Oil Spill Modeling System for the Sea of Japan Sergey M. VARLAMOV Institute of Environmental Sciences, RISSHO University E-mail: [email protected]Jong-Hwan YOON Research Institute for Applied Mechanics, Kyushu University
20
Embed
The 3rd International Workshop on Next Generation Climate Models for Advanced High Performance Computing Facilities Tokyo, Japan, March 28 -30, 2001 Performance.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
The 3rd International Workshop on Next Generation Climate Models for
Advanced High Performance Computing Facilities
Tokyo, Japan, March 28 -30, 2001
Performance and Verification
of the Oil Spill Modeling System
for the Sea of JapanSergey M. VARLAMOV
Institute of Environmental Sciences, RISSHO University
Oil spill simulations system (Varlamov et. al., 2000)
Oil spill modelOil spill model**
Spill Transport and Weathering
Particles Tracking Model Graphical analysisGraphical analysis
subsystemsubsystem
Ocean Ocean CirculationCirculation
ModelsModels
Meteorological Meteorological subsystemsubsystem
Oil propertyOil propertydatadata
EnvironmenEnvironmentaltal
datadata
Ocean Ocean currentscurrents
datadata
MeteorologicMeteorologicalal
datadata
Oil spill Oil spill simulationssimulations
Ocean circulation models
• At nowadays the ocean circulation models is a part of system
Reasons: no products similar to meteorological analysisand forecasts, i.e. 3D currents, temperature, salinity, wavesfield are available ...
• OCM’s spatial resolution and model domain defines the scale and applicability area of the oil spill analysis and prediction system
although the simplified version with the local wind drift currents model could be applied for any area covered by meteorological forecasts and with the land-sea mask defined (GTOPO30 for the global version).
Ocean circulation models
3-D primitive equations Modular Ocean Model (GFDL MOM)- 1/6° grid (1218 km horizontal resolution)- 19 vertical levels (step 15 m in upper layers; up to 600 m below
1500 m)- rigid grid approximation
wind drift shear local current model + 2-D shallow water nonlinear model (2.5 dimensional model)
- variable vertical step (from 1 m in upper sea layer)- 1/12° grid (69 km horizontal resolution)- 16 tidal constituents are used at the open boundaries etc.
Oil spill analysis and prediction system was tested with two types of the Japan Sea ocean circulation models:
Outline of Oil Spill Fate and Trajectory Model (year 2001 version)-1
Particles tracking model:
Spill is presented as an ensemble of particles each characterized by its properties that includes
- position on plane and the depth- diameter- density- viscosity- fraction of water in oil (emulsion)- fraction evaporated- fraction [biochemically] degradated etc.
Outline of Oil Spill Fate and Trajectory Model (year 2001 version)-2
Depending on position, model oil particles are
- transported by wind (at the sea surface)- transported by 3D sea currents- experience buoyancy effect- 3D random diffusion (wind and current dependant)- oil–coastal line interaction- evaporation at the sea surface (Fingas, 1996) - emulsification at the sea surface (Mackay et al, 1980)- biochemical degradation- [oil-sea bottom interaction]
Outline of Oil Spill Fate and Trajectory Model (year 2001 version)-3
Other model properties:
- universal 2D horizontal bicubic spline interpolation from arbitrary meteorological, bottom topography, and ocean circulation model grids (with masking of land surface meteorological data)
- initial droplet diameters distribution in accordance with observed dispersion droplet size distribution (Delvigne and Sweeney, 1988), modified by weathering processes, both “permanent” surface slick and oil dispersed into the water column are simulated
- modification of oil particles density and viscosity as result of emulsification, evaporation and biochemical degradation
- assimilation of observed oil slick position for long-term spill simulation
Technical Characteristics and the Improvement of System Performance Model properties:
- programming languages: C (system routines and meteorological GRIB data decoding software) and FORTRAN-90 (ocean circulation and spill models)
- OpenMP directives are used for the parallelization of ocean circulation and spill physical core modules
- computational effects on the parameterization of physical processes was analyzed and fixed (buoyancy-vertical diffusion-particles diameter-integration time step problem)
Routine oil spill simulation and prediction system parameters
Meteorological analysis and forecasts data are received once a day. Total 3.7 Mb are received, and at 18:00 UTC (local night) it usually takes approximately 7 - 8 minutes
SW OCM is started and run the currents simulation and prediction for 126 h (5 days). With 10s time step it takes 1 h 03 min on 667 MHz Alpha processor (GFDL MOM 1/12° - 25 min)
Oil spill simulation could be run by operator and for 5 days with 10000 droplets it will takeless then 14 min