Overview of Existing Modeling Platforms
Scott Socolofsky Coastal and Ocean Engineering
Texas A&M University
Berkeley, California November 27, 2012
Purposes of Models Response.
Where to go to find the spilled oil How to make decisions about response actions (e.g., to apply
subsurface dispersants or not)
Planning. Designing mitigation strategies before an accident and understanding the risks of an accident
Biological Effects Modeling. To quantify exposure and predict effects of an event, either in real time or forensically.
NRDA. Hindcasting and quantifying impacts for an historic event
Science. To explore the physics of buoyanct, multi-component, multiphase flows. Usually for the purpose of improving models’ ability to address the other model purposes.
Existing Modeling Platforms GNOME – General NOAA Operational Modeling Environment,
NOAA
OSCAR – Oil Spill Contingency and Response, SINTEF
SIMAP & OILMAPDEEP – Integrated Oil Spill Impact Model System & Deepwater Oil Spill Model, ASA
OSIS – Oil Spill Information System, BMT-ARGOSS
ERO3S – EPA Research Object-Oriented Oil Spill, U.S. EPA
OSRA – Oil Spill Risk Analysis, BOEM
CDOG – Comprehensive Deepwater Oil and Gas Model, Yapa et al., Clarkson University
Other Literature Models (e.g., TAMU, MIT)
Initial Conditions Initial Droplet Size Distribution
Prescribed Empirical – based on correlations with Weber number (Hinze 1955)
and modified Weber number (Wang & Calabrese 1986)
Population balance models accounting for breakup and coalescence dynamically (e.g., Bandara & Yapa 2011)
Zone of Flow Establishment (ZFE) Start plume models with finite width and velocity Froude number conditions (e.g., Wuest et al. 1992)
Plume Models Integral (one-dimensional) approached based on self-similarity
and the entrainment hypothesis
Two main types: Lagrangian or Eulerian
Well-documented in the literature: CDOG – Papers by Yapa et al. DeepBlow – Papers by Johansen et al. MIT/SIMP – Papers by Adams et al. and Socolofsky et al.
Must account for possibility of separation between bubbles and droplets due to crossflow and/or stratification
Track oil from the ZFE to an intrusion or surface expression
Must provide initial conditions to Lagrangian particle tracking models
Separation by Crossflow Employed in Lagrangian models (DeepBlow and CDOG)
Eject bubbles when their trajectory due to their slip velocity is too steep to stay within the plume
Separation by Stratification Employed in Eulerian models (MIT, TAMU, lake aeration models)
Eject entrained fluid when its momentum flux dies to zero
Boundary Conditions Currents, temperature, and salinity profiles
Provided by general circulation models (e.g., Gulf of Mexico) SABGOM – South Atlantic Bight and Gulf of Mexico Nowcast/
Forecase System (He et al. at NCSU) HYCOM + NCODA – Hybrid Coordinate Ocean Model and Navy
Coupled Ocean Data Assimilation Model (National Ocean Partnership Program)
Other ROMS – Regional Ocean Modeling System
Usually linked by data assimilation to satellite data to get loop current correct
Better predictions as surface (via data assimilation) than at depth
Typical grid resolution is of order kilometers
Typical temporal resolution is of order seconds (calculated) and hours / days (reported)
Lagrangian Particle Tracking Models Track fluid elements that may contain bubbles, drops, dissolved
hydrocarbons
Tracking equation includes advection (needs CFD model output) and diffusion (stochastic random walk)
Advantage is the subgrid nature of the model that avoids excessive numerical dispersion
Can track millions of particles in multiple CFD model domains
Challenge to define initial conditions of fluid element and to convert properties along a trajectory to concentration
Expensive to compute complex chemistry
LTRANS, CMS or built-in routine
Fate Modeling Predict rise velocity, dissolution, biodegradation, emulsion
formation, hydrates, surface weathering, etc.
Include non-ideal effects (pressure) and mixture chemistry
Variations of the Ranz-Marshall Equation
Mass transfer coefficient depends on hydrodynamics and diffusion coefficient. See e.g., Clift et al. (1979).
Solubility depends on non-ideal equation of state. Use industry software such as MultiFlash by InfoChem.
Surface Oil Models Predict evolution of surface and near-surface oil
Dilution and mixing of surfacing oil (and gas)
Advection by currents (with significant wind forcing) and spreading
Evolution processes include Evaporation and volatilization Emulsification Wave mixing and effect of dispersant application Effects of burning and clean-up
Easier problem to study experimentally (can use large atmospheric tanks)
More data available during a spill
Therefore, better understood than subsurface oil transport
Exposure and Effects Convert oil presence (gas, liquid and dissolved) into exposure
Predict region above a threshold toxicological endpoint Predict dose by integrating concentration over time
Relate exposure to effects Track populations of species of interest Predict mortality due to acute toxic effects Predict movement of hydrocarbons through the food chain
Track regions of direct contact (usually for birds, mammals, and other wildlife)
Usually stochastic and risk based
Of high economic (e.g., fisheries) and health (e.g., responders) importance; therefore, key aspect of response
(ASA SIMAP Documentation)
Status of Existing Platforms Initial droplet size – prescribed or poorly validated equations
Plume model – account for gas dissolution, some oil dissolution, arresting by stratification (but not peeling) and crossflow separation
Circulation model – couple to some general circulation model(s) using various time-steps of model coupling
Lagrangian particle tracking – track oil, gas, and dissolved hydrocarbons; oil usually degrades in first-order die-off
Dissolution – wide variability of dissolution models; predictive quality is linked to 1.) validation data and 2.) performance of equation of state
Surface Models – most validated to good datasets
Effects – see biological community for details
Improvements for Existing Models Initial droplet size – ongoing experiments; need for large-scale
experimental validation data
Plume model – include detrainment and subsequent plume stage formation; need data on downdraught plume in crossflow
Circulation model – ability to link to multiple models
Lagrangian particle tracking – show strong sensitivity to CFD model spatial and temporal scale; need data on chaotic mixing
Dissolution – oil solubility is important for biological effects modeling; need reliable and fast equations of state
Surface models – most models are mature
Effects – As computing power increases, biological community modeling can expand; currently follows box-model approach