Operational Oil Spill Modelling: From Science to Engineering Applications in the Presence of Uncertainty - Mathematical Modelling and Numerical Simulation of Oil Pollution Problems - page 101

finer-scale motions that physically spread oil across wider areas, so some form of
dispersion (or diffusion) sub-model is required or the particle cloud will remain unre-
alistically compacted. For models including chemical evolution, the oil weathering
can affect the spill's interactions with wind, waves, currents, and dispersion, which
further change the spill response over time.
Oil spill models are available with several degrees of complexity for both physics
and chemistry. The simplest physics represent only movement of surface particles
driven by the 2-dimensional (2D) water surface velocity, wind drag, and waves.
Such models typically use statistics-based dispersion parameterizations (e.g. white
noise, Markov chains). More advanced physics models include 3D currents and
transport [ 53 ], along with more physics-based dispersion models (e.g. mechanical
spreading, particle breakup [ 43 ], Langmuir circulations [ 50 ]). Although 3D models
should theoretically be preferred, representing the vertical distribution of oil in the
near-surface water column remains a challenge: the vertical grid resolution in most
hydrodynamic models is relatively coarse and we lack the comprehensive data on
vertical oil dispersion under wave-breaking conditions that are necessary for coarse-
grid model parameterization. Indeed, it remains an open question as to whether
3D models are necessary for operational modelling or if forecast uncertainties will
dominate the 3D effects. Oil spill models with simpler physics are suitable where
confidence in the underlying geophysical forcing models is low, e.g. where a coarse
hydrodynamic model grid makes impossible to resolve the important velocity strain-
rates required by a physics-based dispersion model.
The simplest oil spill chemistry model is no model at all, which is appropriate
where uncertainty in the geophysical forcing dominates the results over short time
periods (for which weathering is less important). More advanced models include
effects of the type of oil along with processes such as dissolution, emulsification
and/or evaporation [ 12 ]. It should be noted that 3D transport and chemical evolution
for a deepwater blowout remains a scientific challenge due to the complex physics
of an oil/gas plume and transformations of the gas phase during ascent [ 13 ].
5.4 Sources of Uncertainty
From a science point of view, we seek to understand and minimize sources of error
and uncertainty in any modelling system. However, for emergency management we
need rapid answers to some practical questions: How good is this prediction and
can we rely on it for deploying response equipment? Before trying to quantify
uncertainty, it is useful to review the fundamental sources.
The major contributors to uncertainty in any modelling system fall into four
categories:
(i) structure of the model,
(ii) empirical parameters,
(iii) initial conditions,
(iv) boundary conditions.