Operational Oil Spill Modelling: From Science to Engineering Applications in the Presence of Uncertainty - Mathematical Modelling and Numerical Simulation of Oil Pollution Problems - page 97

evaluation methods that can be readily used to improve management and response
of oil spills. Herein we develop two major options in advancing oil spill modelling
with uncertainty: in Sect. 5.6 a systems-level approach is proposed for evaluating
real-time uncertainty at each level of modelling, and in Sect. 5.9 a probability-based
approach for oil spill modelling that could be used either as part of a systems-level
approach or on its own if the uncertainty in wind, wave, and hydrodynamics can be
apriori quantified.
Operational modelling of marine oil spill trajectories serves two key purposes:
(i) forecast the likely spill path for immediate mitigation and capture operations
[ 16 , 23 ], e.g. deployment of booms, skimmer boats, and dispersants; and (ii) hindcast
the likely impacted coastal shorelines, bays, and estuaries that might be affected
by escaping oil and hence require further monitoring (e.g. [ 26 ]). We are focused
on issues associated with operational modelling for the first purpose, where time
constraints require immediate application of available models that can be quickly
run immediately after a spill is reported.
Predictive oil spill models are inextricably linked to predictive numerical models
of atmospheric and oceanic dynamics [ 14 ]. In many places, operational numerical
models can provide the short-term (usually around the next 72-96h) forecasts for
ocean currents, wave conditions, and wind fields as input to oil spill fate and trans-
port models. These operational models integrate the conservation laws for mass and
momentum forward in time to provide physics-based (in contrast to statistics-based)
predictions of ocean and atmosphere dynamics. Due to the nonlinear nature of the
governing equations, there is a continuous need for acquisition of real-time ocean
data to validate, update, and adjust the model to better match reality. The complex
dynamics and limited predictability of coastal oceans has motivated development of
Coastal Observing Systems (COS), which are being implemented in many regions
(e.g. [ 30 , 40 , 44 , 51 ]). COS typicallymonitor physical, chemical and biological ocean
properties by combining remote monitoring (HF-Radar, satellite imagery) and in situ
devices (floats, drifters, gliders, moorings, etc.). This process, from the acquisition of
real data to the dissemination of ocean currents for diagnostic or prognostic purposes,
requires the combined efforts of basic and applied research in several engineering
fields as well as computer sciences, along with coordination and support from gov-
ernment mission agencies. The goals for all such systems are essentially the same:
fast, accurate and user-friendly tools with visual interfaces capable of providing the
information needed for making timely and well-founded decisions regarding coastal
protection, security and implementation of rapid, effective contingency plans [ 7 , 11 ].
Oil spills in coastal waters require rapid deployment of mitigation personnel and
equipment to the right places at the right time to maximize recovery and minimize
environmental damage. For spills sufficiently far offshore in good weather, distance
equates to time and emergency managers have the (relative) luxury of tracking actual
spill motion via aircraft, boats, and satellite. However, as weather turns foul or a
spill occurs close to shore (where response time is short), models provide a key
source of information for equipment deployment decisions. The value of model-
produced data for emergency managers depends on its timeliness and the reliability
of the predictions—both actual and perceived . Unfortunately, there is little guidance