Page 3 - Assessment

D. Melaku Canu et al. / Marine Pollution Bulletin xxx (2015) xxx–xxx 3
commonly adopted assumptions of oil weathering (Guillen et al., than water) oil class was considered in the simulations. The sim-
2004; Abscal et al., 2010; Olita et al., 2012a; Cucco et al., 2012; ulations of oil spills from the 6 selected sites (Fig. 1) were
De Dominicis et al., 2013a,b). The oil spill sites were selected ran- performed considering the outflows as beginning at the water
domly within a portion of the Italian exploration and production surface.
areas within the SCH. Seasonal hazard maps were produced by sta-
tistical analysis of the trajectories belonging to a given season. 2.2. The simulation setup
Furthermore, a conservation prioritization example was devel-
oped by combining an evaluation of the oil slick hazard at the coast The described numerical tools use a finite element unstructured
and information on the natural and ecological value of the coastal mesh to represent the model domain. Integration of the different
area (Site of Community Importance, SCI, and Special Protection equation systems was carried out over an area covering the Sicily
Areas, SPA). Channel and part of the Thyrrenean Sea. A finite element grid
was generated to discretize the whole domain with more than
120,000 elements and 75,000 nodes, with a spatial resolution rang-
2. Methods
ing between 25 km for the areas far from the SCH and 100 m for
the areas of coastal Sicily and near the Archipelagos. In Fig. 1, the
We investigated the hazard and risk at the Sicilian Coast,
mesh is reported. The unstructured mesh facilitates the reproduc-
derived from surface spills in six areas, randomly chosen within tion of water circulation in the open sea and on a coastal scale
the zone of exploration and exploitation on the Italian continental simultaneously. Nevertheless, simulation of the large-scale
shelf (Fig. 1, R1–R6). The size of the area of interest, the complexity hydrodynamics and its effects on the coastal hydrodynamics
of the coastal geometry and the distance of the potential spill required the use of proper forcing and boundary conditions. A
sources from the coast require application of a multi-scale numeri- pre-existing oceanographic operational model (SCRM, Gaberšek
cal analysis. The hazard and risk valuation was tackled with a et al., 2007; Olita et al., 2012b) covering a much larger area
cross-platform approach combining a 3D hydrodynamic model (between approximately 31°N and 39°N and approximately 9°E
based on an unstructured mesh and capable of describing coastal and 16°E), with a regular 0.1 0.1 degree mesh, and constrained
areas with very high spatial resolution, a lagrangian trajectory
by assimilation of satellite data (along track sea level anomalies)
model to provide a prognostic prediction of oil spill trajectories was used to generate sea water level and 3D daily fields of water
in the study area, a simplified oil weathering model, and statistical
currents, temperature and salinity for the large domain. These data
analysis of trajectories. An overall description of the adopted were then used as boundary conditions for the higher resolution
method follows.
finite element model. This nesting procedure, combined with the
atmospheric forcing provided by the SKIRON high-resolution
2.1. The numerical model atmospheric numerical model (Kallos, 1998a–f), allows the hydro-
dynamic module SHYFEM to reproduce the water circulation in the
Hydrodynamic properties were computed by using SHYFEM, a area of interest (Cucco et al., 2012). The model run was carried out
3D oceanographic model based on the finite element method. to simulate surface water circulation for a 2-year period between
This method has been successfully applied in many previous stud- 2010 and 2011. The vertical discretization of the SHYFEM model
ies of waves, hydrodynamics and pollutant surface transport in domain consisted of 50 layers with a varying depth logarithmically
shallow waters and coastal seas (Cucco et al., 2012; Melaku Canu ranging from 0.5 m to approximately 1000 m.
et al., 2012; 2001).
SHYFEM uses finite elements for horizontal spatial discretiza- 2.3. Hazard assessment
tions, z-layers for vertical discretizations and a semi-implicit algo-
rithm for integration in time. The horizontal diffusion, baroclinic In accordance with the definition given by IPCS (2004), we refer
pressure gradient and advective terms in the momentum equation to hazard as the inherent property of a situation having the poten-
are treated explicitly. The Coriolis force and the barotropic pres- tial to cause adverse effects. In our case, we refer to the situation of
sure gradient terms in the momentum equation and the diver- an accidental spill of oil at sea which, depending on winds and cur-
gence term in the continuity equation are treated semi- rents, might reach the coast and negatively impact it. In this case,
implicitly, whereas the vertical stress terms and the friction term the hazard is a function of properties such as distance to the coast,
are implicitly computed for stability reasons. The model is meteorological and hydrodynamic conditions, and time spent by
unconditionally stable with respect to fast gravity waves, bottom the oil in the water before reaching the coast or before being
friction and Coriolis acceleration. A detailed description of the degraded or removed. Following the IPCS definition, the hazard
solved system of equations and adopted numerical treatments is assessment is meant to determine the possible adverse effects
reported in Umgiesser et al., 2004. (the presence of an oil slick) occurring when a system (in this case
The numerical model for simulating the behaviour of oil slicks the coastal area) is exposed to the accidental oil spill.
dispersed into the sea consists of a lagrangian module and a We assumed an equal probability of occurrence of an oil spill
weathering module both integrated into the hydrodynamic core. event for each day of the year and for each of the 6 hypothetical
The lagrangian model simulates the transport and dispersion of release areas. To obtain oil drift statistics representative of differ-
oil slicks into the sea by tracking the trajectories of a large number ent weather conditions, we performed an ensemble of 730 sim-
of particles instantaneously released at a given point, as a function ulations separately for each release point, each driven by a
of the current field, wind drift and turbulent dispersion. Each par- slightly different (1 day shifted) circulation field extracted from
ticle is assumed to represent a given amount of oil, depending on the 2-year period, and each tracking oil spill trajectory and trans-
the severity of the spill. The lagrangian module is integrated into formation for 10 days. Thus, the total number of simulated oil spill
the hydrodynamic model. The weathering module, integrated into events is 4380.
the lagrangian transport module, reproduces the oil evaporation Statistics were then extracted for each season and presented as
and shore stranding processes (detailed numerical treatments in averages of potentially relevant quantities, such as the following:
Cucco et al., 2012). Evaporation depends on oil density and
environmental properties. Based on the features of the oil currently (a) The % of oil slick that reaches the coast relative to the total
extracted in the SCH, a 15 API (i.e., a relatively heavy, but lighter released to the sea (Table 1, Total onshore, Onshore).

Please cite this article in press as: Melaku Canu, D., et al. Assessment of oil slick hazard and risk at vulnerable coastal sites. Mar. Pollut. Bull. (2015), http://
dx.doi.org/10.1016/j.marpolbul.2015.03.006