Oil spill hazard assessment using a reverse trajectory method for the Egadi marine protected area (Central Mediterranean Sea)

Oil spill hazard assessment using a reverse trajectory method for the Egadi marine protected area (Central Mediterranean Sea)

Marine Pollution Bulletin 84 (2014) 44–55 Contents lists available at ScienceDirect Marine Pollution Bulletin journal homepage: www.elsevier.com/loc...

7MB Sizes 0 Downloads 33 Views

Marine Pollution Bulletin 84 (2014) 44–55

Contents lists available at ScienceDirect

Marine Pollution Bulletin journal homepage: www.elsevier.com/locate/marpolbul

Oil spill hazard assessment using a reverse trajectory method for the Egadi marine protected area (Central Mediterranean Sea) Achille Ciappa a,⇑, Salvatore Costabile b a b

e-geos/ASI-Telespazio, via S. Cannizzaro 71, 00156 Roma, Italy Ministero dell’Ambiente e della Tutela del Territorio e del Mare, Via Cristoforo Colombo 44, 00147 Roma, Italy

a r t i c l e

i n f o

Article history: Available online 13 June 2014 Keywords: Oil pollution transport Marine protected area Egadi archipelago Lagrangian tracers Backward-in-time integration Receptor mode

a b s t r a c t The Egadi Marine Protected Area (MPA) on the western side of the Sicily Channel (Central Mediterranean) is exposed to a high risk of oil pollution from the tanker routes connecting the eastern and western basins of the Mediterranean Sea. Areas where an oil spill would do most damage, and thus where surveillance should be concentrated, are identified in this study by Lagrangian tracers tracked backwards in time from points along the MPA perimeter using data spanning six years from 2006 to 2011. Results indicate that the areas where oil surveillance would be most beneficial are segments of the tanker routes south of Sicily (highly frequented) and north of Sicily (scarcely frequented), both extending about 150 miles from November to March and 100 miles in the other months. The third route, close to the Tunisian shore, is the most frequented by oil tankers but the threat period is limited to November and December. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Oil tanker routes in the Mediterranean Sea connect pipeline terminals, refineries and offshore platforms concentrated along the coastal zone of Southern Europe, North Africa, the Middle East and the Black Sea, and account for more than 20% of global oil tanker traffic. A great threat to the marine environment of the Mediterranean Sea comes from oil drillings and illegal discharges (Alpers and Huhnerfuss, 1988; Pavlakis et al., 2001). According to the European Space Agency (1998), about 45% of the total oil pollution comes from regular spills along these routes while only 5% is caused by large accidental oil spills (Fingas, 2001). Satellite monitoring by Synthetic Aperture Radar (SAR) is an effective tool for discouraging the illegal practice of oil discharge (Brekke and Solberg, 2005). Satellite images, if acquired at the right time and place, provide early warning of dangerous spills and enable prosecution of the polluter, eventually acting as a deterrent (Ambjörn, 2008). Unfortunately, in the Mediterranean as in other highly frequented seas, the amount of data necessary to monitor the main tanker routes is very large and is beyond the capability of the current SAR constellations. More frequent surveillance is required in specific sea areas where the oil is quickly pushed to the coast by the prevailing winds and currents rather than dispersing offshore, or reaches a long portion of the coast rather than a short one. ⇑ Corresponding author. Tel.: +39 06 40793672; fax: +39 06 40796202. E-mail address: [email protected] (A. Ciappa). http://dx.doi.org/10.1016/j.marpolbul.2014.05.044 0025-326X/Ó 2014 Elsevier Ltd. All rights reserved.

The Marine Protected Area (MPA) of the Egadi archipelago, on the western tip of Sicily in Central Mediterranean (Fig. 1), is very close to the main tanker routes connecting the Western and Eastern basins of the Mediterranean Sea and for this reason is exposed to a high risk of oil pollution. A method for the identification of the risk areas, i.e. where the occurrence of an oil spill would do most damage, is proposed in this study and applied to the Egadi archipelago MPA. The intersection of these areas with the main tanker routes indicates where to concentrate the monitoring effort, for instance by monitoring satellite data. Recent increases in modelling capability have made possible the incorporation of marine dynamics in techniques for the reduction of environmental risks stemming from shipping offshore and coastal activities. Lagrangian trajectories have been used to investigate the drift pattern induced by currents in the Baltic Sea and in the Gulf of Finland with the aim of identifying safe routes, in order to minimize the probability of coastal pollution and/or to maximize the time before which adverse effects are seen at the coast (Soomere et al., 2010; Lu et al., 2012; Soomere et al., 2014). More specific to oil pollution, trajectory analysis has been used to investigate the risk posed by oil-dumping from ships along the German North Sea coast (Chrastansky and Callies, 2009), in the Gulf of Finland (Murawski and Woge Nielsen, 2013) and in the Strait of Bonifacio in the Mediterranean Sea (Olita et al., 2012) during reconstructions of atmospheric winds and marine currents spanning several years. A specific study on the current-driven risk of pollution for MPAs was carried out in the Gulf of Finland (Delpeche-Ellmann and Soomere, 2013). The common aspects of

A. Ciappa, S. Costabile / Marine Pollution Bulletin 84 (2014) 44–55

45

Fig. 1. Egadi Marine Protected Area (MPA) in Central Mediterranean Sea and main tanker routes in December 2010. Tanker routes are plotted using Automatic Identification System (AIS) data (IMO codes 81 and 82) provided by courtesy of e-geos and the Italian Coast Guard (MARISS project). (Panel a): islands of Favignana (F), Marettimo (M) and Levanzo (L) within the MPA.

these studies are the use of long-term sequences of wind and current data and the forward-in-time tracking of Lagrangian particles, which are initially seeded at points located along the main sea routes or covering the entire sea area with a regular distribution. The method proposed in this study combines the use of long-term sequences of meteo-oceanographic data with the ‘receptor mode’ technique, where Lagrangian tracers are tracked backwards in time from known destinations (or receptors) located on the coast or in sensitive areas. As noted by Batchelder (2006), backward-in-time mode is more computationally efficient than forward-in-time tracking when the number of receptors (destinations) is significantly fewer than the number of potential sources, as is quite common in oil risk assessment studies. The backward-in-time approach in calculating fluid element trajectories is widely adopted in atmospheric science in order to identify the source of a pollutant or to estimate substance emissions from potential emission sites (Lin et al., 2003; Seibert and Frank, 2004; Stohl et al., 2012). In this context, the validity of backward-in-time Lagrangian stochastic dispersion models was discussed by Flesch and Wilson (1995), finding that the correspondence between forward and backward models is good in the case of turbulent diffusion notwithstanding the irreversibility of the process, on condition that any sort of deterministic bias to the direction of motion arising from turbulence nonhomogeneity is considered (the ‘‘well-mixed’’ condition in Thomson, 1987). As random walk is commonly used to simulate the physical process

of turbulent diffusion in the forward mode; random walk in the backward mode describes the uncertainty of the trajectory caused by turbulent diffusion and the result of backward-in-time integration provides probabilities of possible source locations. In the ocean, backward tracking has been used to investigate sources, destinations and transport pathways of plankton (Batchelder, 2006), and Isobe et al. (2009) investigated the possibility of relating source and receptor points by using a two-way particle tracking model that included forward- and backward-in-time integration modules. In oil-spill modeling, the ‘receptor mode’ option was implemented in the pioneer On-Scene Spill Model (Torgrimson, 1981; Galt and Payton, 1983) developed by the Hazardous Material Response Branch (HAZMAT) of National Oceanic and Atmospheric Administration (NOAA) and is still available in many commercial oil-spill models. It is based on Lagrangian particles tracked backwards in time for some days from a receptor site to multiple possible sources offshore, producing (i) a probability map, i.e. the probability that oil detected offshore reaches the receptor site, and (ii) a time-of-travel or arrival time map, i.e. the minimum time for the oil to reach the site. In the original formulation (see Galt and Payton, 1983), trajectories were simulated by using monthly or seasonal statistical distributions of the wind and current. It will be demonstrated that the use of the realistic sequence of wind and current data better accounts for the trajectory variability on a time scale of a few days. In this study, the ‘receptor mode’

46

A. Ciappa, S. Costabile / Marine Pollution Bulletin 84 (2014) 44–55

analysis is simultaneously performed for multiple receptor points equally spaced along the PMA perimeter. It will be shown that the resulting maps, joined together, provide information on the number of receptor points hit by the oil, that is, a measure of the extension of the potential damage caused by the oil detected offshore. The remainder of this paper is organized as follows. Meteooceanographic data used in this study and aspects of the implementation of the new method are described in Section 2. Results of the substitution of the statistical distributions of wind and current used in the traditional ‘receptor mode’ with realistic data are illustrated in Section 3, together with the final results obtained for the Egadi MPA. In Section 4, the position and shape of the sea areas around the Egadi MPA deserving more frequent oil surveillance are suggested and the main points emerging from data and simulations are discussed. Finally, Section 5 summarizes the novel results of this work. 2. Data and methods

Outputs from 2003 to 2005 were discarded due to the spin-up period of the model. The surface currents provided by the model are daily means spanning the period from 2006 to 2011. Wind data were extracted every 12 h from re-analysis dataset (Kalnay et al., 1996) provided by the National Centers for Environmental Prediction (NCEP) and Atmospheric Research (NCAR), because of the importance of sub-daily wind forcing in modeling the circulation of the Mediterranean Sea (Pierini and Simioli, 1998). In spite of the rather coarse spatial resolution (2.5°), the reliability of NCEP/NCAR winds in the Sicily Channel was verified by the substantial agreement of their kinetic energy (KE) with higher resolution winds (0.25°) shown in Fig. 2, the latter extracted from the European Centre for Medium-Range Weather Forecasts (ECMWF) re-analysis dataset during 2011. Winds and currents from 2006 to 2011 were interpolated in time and space during the calculation of the oil trajectories. The wind contribution to the oil transport is commonly estimated to be 3% of the wind speed (ASCE, 1996). The four cases of maximum transport towards the cardinal directions from 2006 to 2011 are illustrated in Fig. 3, and all occurred in the cold season.

2.1. Meteo-oceanographic data 2.2. The adopted method of oil trajectory analysis The Egadi archipelago (Fig. 1), with a surface area of 208 square miles and a perimeter of about 46 miles, is the largest Italian Marine Protected Area (MPA) and includes the major islands of Favignana, Marettimo and Levanzo (Fig. 1, panel A). The MPA is very close to the main tanker routes connecting the Western and the Eastern basins of the Mediterranean Sea and to active oil extraction wells, with others awaiting a production license in the Sicily Channel (source: World Wildlife Fund). The surface circulation in the study area is characterized by the eastward pathway of the Atlantic Water (AW), flowing near the surface from the Western to the Eastern Mediterranean basin. Within the Sicily Channel, AW is advected by two main seasonal streams (Ciappa, 2009), the Atlantic Ionian Stream flowing south-eastward in summer close to the coast of Sicily (Robinson et al., 1999), and the Atlantic Tunisian Current, prevailing in winter and flowing southward over the Tunisian shelf (Pierini and Rubino, 2001; Béranger et al., 2004). Current measurements and hydrographic data indicate a strong seasonal variability within the Channel (Astraldi et al., 1996, 1998, 1999). Low salinity values very close to the surface in the cold period (winter and spring) suggest a significant eastward advection of AW, and the higher salt content of the surface waters in the other periods suggests a weakening of the eastward advection (Béranger et al., 2004). The surface currents in the study area were extracted from a numerical experiment in which the circulation in the whole Mediterranean Sea was simulated from 2003 to 2011 by using the Princeton Ocean Model (POM). This POM version (Blumberg and Mellor, 1987), similar to the model used in a general circulation study of the Mediterranean Sea (Zavatarelli and Mellor, 1995) except for the use of sub-daily wind forcing, improved spatial resolution and few minor modifications, was recently used to investigate other aspects of the upper circulation of the Mediterranean Sea (Ciappa, 2014). The spatial resolution of the model (regular grid equally spaced 8’ in latitude and longitude) is on the edge of the admitted range for solving mesoscale eddies, because the Rossby radius of deformation in the Mediterranean Sea varies from a few kilometers in winter to 12 km in the period of stratification (Grilli and Pinardi, 1998). As an additional measure adopted here, the calculation of the surface current was improved by the assimilation of daily maps of sea surface temperature (SST) in the upper two layers of the model. SST maps were derived from data acquired by the MODerate resolution Imaging Spectroradiometer sensors (MODIS) onboard the Terra and Aqua satellites (downloaded from http://oceancolor.gsfc.nasa.gov/) and post-processed in order to provide cloud-free daily mosaics covering the Mediterranean Sea.

The ‘receptor mode’ trajectory analysis (Galt and Payton, 1983) uses a particular target site as its starting point (receptor site) and proceeds to calculate where oil might come from to threaten this area. The technique is based on the Lagrangian approach and consists of tracking backwards in time for few days a large number of particles released at the receptor point by using the bi-dimensional and time-reversed Euler scheme. The movement and spreading of the oil is controlled by wind drift, current advection, spreading or diffusion, and ultimately weathering. Wind drift is simulated by an additional velocity that is a fraction of the wind speed (typically 3%), and current advection is given by the current vector because the oil is assumed to move with the water particle it is floating on. Spreading and diffusion are incorporated by random walk. Weathering processes, i.e. evaporative or decaying processes which have the effect of removing part of the oil from the active components of the spill, have not been incorporated into the trajectory analysis. Hence, the spilled oil is considered to be conservative and oil weathering is not included. In the traditional way, ‘receptor mode’ analysis was performed on a seasonal or monthly basis, depending on the local wind and current time scales. Typically, the statistical description of the wind was summarized in a wind histogram that was assumed uniform over the area, and the current field was assumed constant in time and was set to the mean values in the period (Torgrimson, 1981). The reversal path of the oil from the ‘receptor site’ to multiple sources offshore produces a probability map, i.e. the probability that the oil found in one of the multiple sources reaches the receptor site, and an arrival time map, i.e. the minimum time for the oil to reach the site. If oil is detected offshore, these maps provide useful information to logistics personnel for the activation of remedial procedures on the coast and are a natural resource when facing potentially hazardous activities such as a drilling rig or the management of a ship in distress. The method proposed in this study differs from the traditional ‘receptor mode’ in two aspects. The first is that it is applied to several receptor points at the same time, and probability and arrival time maps are produced for each receptor point. In this case study, 97 receptor points, at a spacing of 1 km, were selected along the perimeter of the Egadi MPA (shown in Fig. 1a). The use of multiple receptor points provides additional information about the potential extent of the damage, because in a sea area the number of overlapping plumes originating from the receptor points, which is equal to the number of receptor points potentially reached by

A. Ciappa, S. Costabile / Marine Pollution Bulletin 84 (2014) 44–55

47

Fig. 2. Comparison of the kinetic energy (KE) of NCEP and ECMWF wind datasets in the Sicily Channel during 2011.

Fig. 3. Cases of maximum transport (to north, east, south and west in clockwise order) extracted from the run of the numerical model for the period 2006–2011. The adopted scale of the wind vectors allows visual comparison of the transports due to the wind (oil moving at 3% of wind speed) and the current.

the oil, is ultimately proportional to the length of the MPA perimeter (or coastal segment) hit by the oil. The second aspect is the substitution of the mean current field and wind histogram with the sequence of current and wind data extracted from a

multi-annual run of an ocean model. Advantages of the substitution are the possibility of accounting for the interaction of wind and current in general, and to preserve the short-term variability of the oil trajectories in particular. Indeed, when the Lagrangian

48

A. Ciappa, S. Costabile / Marine Pollution Bulletin 84 (2014) 44–55

Table 1 October wind histogram extracted from NCEP wind dataset from 2006–2011 near the Egadi PMA. Speed (m/s) dir (°)

1.0–3.5

3.5–6.5

6.5–10.5

0–30 30–60 60–90 90–120 120–150 150–180 180–210 210–240 240–270 270–300 300–330 330–360 Calm = 3.6

1.8 0.8 1.8 3.6 2.9 3.1 2.6 0.8 2.1 2.9 1.0 1.8

2.1 3.1 1.0 4.9 6.5 6.3 0.5 1.8 1.6 3.9 6.8 5.7

0.3 0.3 0.5 3.9 3.1 1.6

10.5–15.5

15.5–20.5

20.5–26.5

26.5–33.5

33.5–40.5

0.3 0.5 0.3 0.3 0.3

1.3 2.3 3.6 2.1

1.3 2.3 1.6

0.5 0.3

Table 2 October wind histogram from multi-decadal wind measurements in Trapani. Speed (m/s) dir (degrees)

1.0–3.5

3.5–6.5

6.5–10.5

10.5–15.5

15.5–20.5

20.5–26.5

0–30 30–60 60–90 90–120 120–150 150–180 180–210 210–240 240–270 270–300 300–330 330–360 Calm = 16

0.4 0.6 0.4 0.6 0.5 0.7 0.3 0.3 0.2 0.1 0.3 0.6

1.2 1.8 1.8 2.1 2.2 3.0 2.1 1.1 1.2 1.4 1.7 2.5

2.6 2.2 1.3 1.0 2.4 3.4 2.2 2.2 1.7 1.8 2.6 2.6

2.0 1.1 0.4 0.4 1.3 3.6 1.8 0.8 0.8 1.4 2.0 1.5

0.8 0.4 0.2 0.1 0.8 2.1 0.9 0.3 0.3 0.7 1.1 0.9

0.2 0.2 0.1

particles are tracked for a few days, the real wind and current fields could differ substantially from the fields depicted by wind statistics and mean current. The use of the long-term sequence of wind and current data implies that the Lagrangian particles are continuously released and back-tracked for a few days during the full period of the ocean model run, i.e. 6 years. The study area around the MPA was subdivided into a regular grid sufficiently large to include all the oil trajectories (about 500 km wide, with a spatial resolution of 250 m; not shown). Particles seeded at the 97 receptor points were tracked backwards in time for 5 days with a timestep of 5 min,

0.6 1.4 0.6 0.1 0.2 0.3 0.3 0.3

26.5–33.5

33.5–40.5

0.3 0.2 0.1

0.1 0.1

0.2 0.1

sufficiently short to reduce the truncation error in the Euler formula, and diffusion was simulated by random walk with coefficient Kh = 5 m2/s added to each trajectory at each timestep (Viikmäe et al., 2013). Oil trajectories were reported on the reference grid by storing the number and time of the particles in transit in each cell of the grid. Two particles per timestep were released at each receptor point, corresponding to an average of 18000 trajectories per month per point and to a total of 125 * 106 particles used in 97 receptor points from 2006 to 2011. Monthly probability and arrival time maps for each receptor point were obtained from about 6*18000 trajectories in 6 years.

Fig. 4. Mean surface current in October (panel a) and maximum transport cases (to north, east, south and west in clockwise order) in the October months from 2006 to 2011. The adopted scale of the wind vectors allows visual comparison of the transports due to the wind (oil moving at 3% of wind speed) and the current.

A. Ciappa, S. Costabile / Marine Pollution Bulletin 84 (2014) 44–55

49

Fig. 5. Results produced in October by the ‘receptor mode’ with mean current and wind statistics from 2006 to 2011 (upper row) and multi-decadal wind statistics (central row), and by the adopted method with realistic current and wind fields from 2006 to 2011 (lower row).

Merging together the maps produced for each receptor point, the final probability and arrival time maps were obtained by selecting, from among the 97 available, the maximum values of probability and the minimum arrival time respectively. Finally, the number of receptor points involved provides a third map related to the potential extent of the impact of the oil on the MPA. 3. Results 3.1. Comparison of the adopted method with the traditional ‘receptor mode’ analysis The two methods were compared for October using the same dataset (winds and currents extracted from the ocean model run) and the same diffusion (random walk) in order to verify the effect of the substitution of the statistical distribution of wind and mean current with the realistic sequence of wind and current fields.

Additionally, the ‘receptor mode’ was performed by using the statistical distribution of the wind resulting from multi-decadal measurements in the area. The ‘receptor mode’ analysis requires the monthly wind histogram and the mean current field. The first histogram was extracted from NCEP winds in the six October months from 2006 to 2011 and is shown in Table 1. The second histogram (downloaded from ‘Atlante Climatico dell’Areonautica Militare’ at http://www.meteoam.it/) was obtained from the existing database of wind measurements in Trapani from 1971 to 2000 and is shown in Table 2. The mean current field was obtained by averaging the surface currents simulated by the ocean model in the October months from 2006 to 2011, shown in Fig. 4 (panel a). The other four cases of maximum transport in October shown in the figure illustrate strength and orientation of the current pattern under the action of different winds. Both the adopted method and the ‘receptor mode’ analysis were applied to the 97 receptor points located on the MPA perimeter.

50

A. Ciappa, S. Costabile / Marine Pollution Bulletin 84 (2014) 44–55

Fig. 6. Monthly maps of minimum arrival time (panel A), maximum probability (panel B) and number of receptor points potentially impacted by the oil along the perimeter of the Egadi MPA (panel C) obtained using wind and current data from 2006 to 2011.

The maps of minimum arrival time, maximum probability and number of receptor points hit obtained by the ‘receptor mode’ in the two cases and by the adopted method are shown in Fig. 5. The two results obtained with the ‘receptor mode’ are similar (upper and central rows). Small differences are due to the stronger winds of the multi-decadal wind histogram (speed mostly between 3.5 and 15.5 m/s and peaks from S and SE; Table 2) with respect to the NCEP wind histogram (speed between 3.5 and 6.5 m/s with peaks from NW and SE; Table 1). In contrast, the results from the adopted method (lower row) are quite different from the ‘receptor mode’ results. The overall extent of the area covered by the trajectories is twice as large and the values of minimum arrival time and maximum probability are less concentrated around the MPA. These aspects, together with the random distribution of the number of receptor points hit, show that realistic oil trajectories over a period of a few days are characterized by a higher variability than the trajectories simulated by wind statistics and mean current field.

season (November to March) than in the warm season (April to October). From November to March the area extends north, west and south due to the higher intensity and variability of the wind and current fields. The shape of the area from May to August suggests that more stable winds and currents from NW and SE characterize the oil transport in these months. The monthly arrival time maps (Fig. 6A) show major differences on the 4- and 5-day contours, while the 1-day contour is constantly located at a distance of about 10 miles from the MPA. The monthly probability maps (Fig. 6B) are similar to the arrival time maps near the MPA, but differences emerge in the marginal area from November to March, i.e. the period of higher variability. The monthly maps of receptor points hit (Fig. 6C) are largely influenced by the oil transport variability. From November to March, large sea areas indicate potential impact with more than 60 receptor points (more than 62% of the MPA perimeter). The highest number of receptor points hit occurred in December, south-west of MPA, with a potential impact on more than 80% of the MPA perimeter.

3.2. Final results around the Egadi MPA 4. Discussion Results of the adopted method applied to the Egadi MPA are monthly maps of minimum arrival time (Fig. 6A), maximum probability (Fig. 6B) and number of receptor points hit (Fig. 6C). The extent of the area covered by the oil trajectories is larger in the cold

The cases of maximum transport shown in Figs. 3 and 4 could suggest that the wind is the main agent of oil transport in the study area. However, wind is more variable than sea current in

A. Ciappa, S. Costabile / Marine Pollution Bulletin 84 (2014) 44–55

51

Fig. 6 (continued)

direction and intensity, and a reversal in wind direction would negate the overall wind-induced transport in the period. This issue was investigated by considering the kinetic energy (KE) of wind and current means on 2-, 6- and 10-day scales, the former scaled such that the oil movement was 3% of the wind movement. Results illustrated in Fig. 7a for the year 2008 show that KE peaks in the 2-day means for both wind and current are reduced when periods of 6 or 10 days are considered, but the reduction is more evident for the wind than for the current. Focusing on the 6-day means, which is close to the back-tracking period chosen in this study, the wind acts as the prevailing component of the transport for most of the year but the contribution of the current is at least similar in periods of calm winds, as reported in similar experiments carried out in the Gulf of Finland (Murawski and Woge Nielsen, 2013). This suggestion is more valid inshore than offshore because in coastal areas and along the shelf the sea current is increased by the bathymetric constraint (Figs. 3 and 4). The plot shown in Fig. 7b, the difference between the KE of the transport (given by the sum of wind and current vectors) and the sum of wind and current KE, is proportional to the wind and current modules and to the cosine of the angle between them. Positive values indicate that wind and current push the oil in the same direction for most of the year.

The three monthly maps shown in Fig. 6A, B and C have been combined in order to produce a single map of the area deserving accurate oil surveillance around the MPA. The three maps were arbitrarily combined using homogeneous thresholds of probability (0.17%) and number of receptor points (44), obtained at the arrival time of 72 h as averages of the monthly distributions plotted in Fig. 8. Hence, all the points satisfying at least one option between the arrival time shorter than 72 h, probability higher than 0.17% and more than 44 receptor points hit have been selected to compose a monthly risk map. Basing on their similarity, monthly risk maps have been grouped together in Fig. 9 from January to March (a), April to June (b), July to October (c) and November and December (d). Tanker routes (AIS data, IMO codes 81 and 82) in December 2010 and July 2010 have been superimposed in winter (a and d) and summer (b and c) respectively, but are similar. The tanker traffic is distributed between two highly frequented routes passing through the Sicily Channel south of MPA (routes A and C in Fig. 9) and one, less frequented, north of MPA (route B). The route segment defined by the area A is the most critical because it is a source of threat for the MPA all year round and, furthermore, is highly frequented. This segment is approximately 150 miles long from November to March and significantly shorter from April to October (about 100 miles). The route segment B, shorter from April

52

A. Ciappa, S. Costabile / Marine Pollution Bulletin 84 (2014) 44–55

Fig. 6 (continued)

Fig. 7. Panel a: kinetic energy (KE) of the mean current and wind in 2-, 6- and 10-day bins during 2008 in the study area. Panel b: the value plotted, proportional to the cosine of the angle between the wind and the current, shows that for most of the year wind and current move the oil in the same direction.

to October as segment A, also deserves surveillance all year long but is less frequented. Finally, the route segment C is the most frequented but the threat to the MPA is limited to the months of November and December.

The main results emerging from this study are summarized in the following points. First, the use of realistic wind and current fields commonly adopted in operational oil spill forecasting systems in the last decade is mandatory in backward-in-time

A. Ciappa, S. Costabile / Marine Pollution Bulletin 84 (2014) 44–55

53

Fig. 8. Monthly distributions of maximum probability and number of receptor points per interval of the arrival time. The mean values of maximum percentage (0.17%) and number of receptor points (44) at the arrival time of 72 h (3 days) were arbitrarily chosen as thresholds for obtaining monthly risk maps around the MPA.

Fig. 9. Extension of the risk areas around the Egadi MPA and AIS data showing the main tanker routes (AIS data in December 2010 in a and d; in July 2010 in b and c). Segments A and B of the tanker routes (less extensive westwards from April to October) require surveillance all year long, segment C in November and December.

54

A. Ciappa, S. Costabile / Marine Pollution Bulletin 84 (2014) 44–55

trajectory tracking as well, because statistical distributions of wind and current are not adequate to establish a sound statistical basis for quantifying the likely oil trajectories. The results of the comparative test effected in this study show that, in the short term (5 days), part of the oil trajectory variability is filtered out if statistical distributions of the wind and mean current fields are used. As consequence, considering that multi-annual sequences of wind and current data are produced by numerical models, the second point is that results of backward trajectory analysis can be improved if based on data available at higher spatial resolution. As an example, results of this study based on sub-daily wind fields at 2.5° resolution and daily current data at 8’ resolution can be greatly improved by data produced by higher resolution models. The third point is a source of uncertainty arising from the fact that oil trajectory statistics are extracted from a limited number of years, and adequate statistics were reached after 5 years in the Gulf of Finland (Soomere et al., 2014). A general improvement coming from the evolution of the modeling techniques and the increasing quality of the data collected year after year will be the possibility of managing longer datasets. As the final point, the choice of the backward-in-time tracking done in this study is a valid and computationally efficient alternative to the use of the forward-in-time technique. Indeed, release points uniformly covering the 3  3° domain of Fig. 6 (spaced at 1 min) used in forward tracking would be about 300 times the number of receptor points used in this study. 5. Conclusions Over the coming years traffic densities in the Mediterranean will continue to grow, and new scientifically based tools for adequate response and coastal management are especially needed (Jordi et al., 2006). The increased capability for forecasting current and wind patterns on the one hand and the development of new techniques of trajectory analysis on the other has recently allowed identification of safe routes in the Baltic Sea and in the Gulf of Finland, in order to minimize the probability of coastal pollution and/ or to maximize the time within which adverse impacts reach the coast, and to investigate the potential damage caused by oil dumping from ships in the German North Sea, in the Gulf of Finland and in the Strait of Bonifacio in the Mediterranean Sea. In this study, the risk of oil pollution was investigated around the MPA of the Egadi archipelago at the entrance to the Sicily Channel. The area is exposed to a high risk of oil pollution caused by the proximity of the main oil tanker routes connecting western and eastern basins of the Mediterranean Sea. Oil trajectories have been simulated by Lagrangian tracers tracked for 5 days backwards in time, starting from release points located along the MPA perimeter, for a period spanning six years from 2006 to 2011 using sub-daily wind and daily current data produced by an ocean model. 6. Results of this study indicate that the transport in the area on a time scale of 5 days is mainly due to the wind except during calm periods, when the current prevails, and that wind and current usually push the oil in the same direction. The oil risk area around the MPA is NW–SE oriented, extending NW about 100 miles offshore in the cold season and 60 miles in the warm season, and about 40 miles to SE. The area extends for about 60 miles along the SW–NE axis but grows to the SW in November and December, grazing the Tunisian coast. Based on the oil-spill surveillance provided by satellites but considering the present limitations of coverage and frequency of acquisitions, the aim of this study is to identify specific areas on which to concentrate the monitoring effort, given

by the intersection of the three main tanker routes around the MPA with the areas of greatest threat of oil pollution. The route following the northern coast of Sicily is the least frequented one but requires monitoring all year long. The route following the southern coast of Sicily is the most critical because it is highly frequented, very close to the MPA and a source of threat all year long. The third route, close to the Tunisian shore, is the most frequented but the threat is limited to the months of November and December. The oil trajectory method illustrated in this study, based on the ‘receptor mode’ trajectory analysis, consists of tracking backwards in time Lagrangian elements from the receptor points to multiple sources offshore. This technique is a valid alternative to the forward-in-time integration in computational terms. Indeed, the seeded points lie on the coast or around the MPA rather than offshore, and the number of Lagrangian elements tracked is largely reduced. Acknowledgements Grateful acknowledgement is made to the institutions that in different countries provide free access to the models and data used in this study: the Ocean Model group at Princeton University (POM), National Center for Environmental Prediction (NCEP) and National Center for Atmospheric Research (NCAR), the Oceancolor team of National Aeronautics and Space Administration (NASA). A special thank to Alex Cooper for the revision work. The authors are grateful to the Editor and the reviewers for their constructive support that improved the manuscript. References Alpers, W., Huhnerfuss, H., 1988. Radar signatures of oil films floating on the sea surface and the Marangoni effect. J. Geophys. Res. 93, 3642–3648. Ambjörn C., 2008. Seatrack Web forecasts and backtracking of oil spills, an efficient tool to find illegal spills using AIS. in: 2008 IEEE/OES US/EU-Baltic International Symposium, IEEE, pp. 168–176. ASCE Task Committee on Modeling Oil Spills, 1996. State-of-the-art review of modeling transport and fate oil spills. J. Hydraulic Eng. 122 (11), 594–609. Astraldi, M., Gasparini, G.P., Sparnocchia, S., Moretti, M., Sansone, E., 1996. The characteristics of water masses and the water transport in the Sicily Strait at long timescales. Bulletin de l’Institut Océanographique (Monaco) 17, 95–115. Astraldi, M., Gasparini, G.P., Sparnocchia, S., 1998. Water masses and seasonal hydrographic conditions in the Sardinia–Sicily–Tunisia region. Rapport de la Commission Internationale de la Mer Méditerranée 35, 1998. Astraldi, M., Balopoulos, S., Candela, J., Font, J., Gacic, M., Gasparini, G.P., Manca, B., Theocharis, A., Tintoré, J., 1999. The role of straits and channels in understanding the characteristics of Mediterranean circulation. Prog. Oceanogr. 44 (1–3), 65–108. Batchelder, H.P., 2006. Forward-in-time-/backward-in-time-trajectory (FITT/BITT) modeling of particles and organisms in the coastal ocean. J. Atmos. Oceanic Technol. 23 (5), 727–741. Béranger, K., Mortier, L., Gasparini, G.P., Gervasio, L., Astraldi, M., Crépon, M., 2004. The dynamics of the Sicily Strait: a comprehensive study from observations and models. Deep-Sea Res. II 51, 440–441. Blumberg, A.F., Mellor, GL, 1987. A description of a three-dimensional coastal ocean circulation model. In: Heaps, N. (Ed.), Three-dimensional Coastal Ocean Models. American Geophysical Union, 208 pp. Brekke, C., Solberg, A.H.S., 2005. Oil spill detection by remote sensing. Remote Sens. Environ. 95, 1–13. Chrastansky, A., Callies, U., 2009. Model-based long-term reconstruction of weather driven variations of chronic oil pollution along the German North Sea coast. Mar. Pollut. Bull. 58, 967–975. Ciappa, A.C., 2009. Surface circulation patterns in the Sicily Channel and Ionian Sea as revealed by MODIS chlorophyll images from 2003 to 2007. Cont. Shelf Res. 29 (17), 2099–2109. Ciappa, A.C., 2014. The controversial path of Atlantic Water in the Eastern Mediterranean. Prog. Oceanogr. 123, 74–83. Delpeche-Ellmann, N.C., Soomere, T., 2013. Investigating the Marine Protected Areas most at risk of current-driven pollution in the Gulf of Finland, the Baltic Sea, using a Lagrangian transport model. Mar. Pollut. Bull. 67 (1–2), 121–129. European Space Agency, 1998. Oil Pollution Monitoring. ESA brochure: ERS and its applications – marine, BR-128, 1. Fingas, M., 2001. The Basic of Oil Spill Cleanup. CrC Press, Lewis Publisher, Taylor & Francis Group. Flesch, T.K., Wilson, J.D., 1995. Backward-time Lagrangian stochastic dispersion models and their application to estimate gaseous emissions. J. Appl. Meteor. 34, 1320–1332.

A. Ciappa, S. Costabile / Marine Pollution Bulletin 84 (2014) 44–55 Galt, J.A., Payton, D.L., 1983. The use of receptor mode trajectory analysis techniques for contingency planning. In: International Oil Spill Conference Proceedings: February 1983, vol. 1983, no. 1, pp. 307–311. Grilli, F., Pinardi, N., 1998. The Computation of Rossby Radii of Deformation for the Mediterranean Sea. MTP News No. 6, March 1998. Isobe, A., Kako, S., Chang, P., Matsuno, T., 2009. Two-way particle-tracking model for specifying sources of drifting objects: application to the east China Sea shelf. J. Atmos. Oceanic Technol. 26 (8), 1672–1682. Jordi, A., Ferrer, M.I., Vizoso, G., Orfila, A., Basterretxea, G., Casas, B., Àlvarez, A., Roig, D., Garau, B., Martínez, M., Fernández, V., Fornés, A., Ruiz, M., Fornós, J.J., Balaguer, P., Duarte, C.M., Rodríguez, I., Alvarez, E., Onken, R., Orfila, P., Tintoré, J., 2006. Scientific management of Mediterranean coastal zone: A hybrid ocean forecasting system for oil spill and search and rescue operations. Mar. Pollut. Bull. 53, 361–368. Kalnay et al., 1996. The NCEP/NCAR 40-years reanalysis project. Bull. Am. Meteorol. Soc. 77, 436–470. Lin, J.C., Gerbig, C., Wofsy, S.C., Andrews, A.E., Daube, B.C., Davis, K.J., Grainger, C.A., 2003. A near-field tool for simulating the upstream influence of atmospheric observations: the Stochastic Time-Inverted Lagrangian Transport (STILT) model. J. Geophys. Res. 108, 4493. http://dx.doi.org/10.1029/2002JD003161. Lu, X., Soomere, T., Stanev, E.V., Murawski, J., 2012. Event driven approach for the identification of the environmentally safe fairway in the south-western Baltic Sea and Kattegat. Ocean Dyn. 62, 815–829. Murawski, J., Woge Nielsen, J., 2013. Applications of an oil drift and fate model for fairway design. In: Soomere, T., Quak, E. (Eds.), Preventive Methods for Coastal Protection. Springer, Cham, pp. 367–415. Olita, A., Cucco, A., Simeone, S., Ribotti, A., Fazioli, L., Sorgente, B., Sorgente, R., 2012. Oil spill hazard and risk assessment for the shorelines of a Mediterranean coastal archipelago. Ocean Coast. Manag. 57, 44–52. Pavlakis, P., Tarchi, D., Sieber, A.J., 2001. On the monitoring of illicit vessel discharges using spaceborne SAR remote sensing – a reconnaissance study in the Mediterranean Sea. Annales des Télécommun. 56, 700–718.

55

Pierini, S., Rubino, A., 2001. Modeling the oceanic circulation in the area of the Strait of Sicily: the remotely-forced dynamics. J. Phys. Oceanogr. 31, 1397–1412. Pierini, S., Simioli, A., 1998. A wind-driven circulation model of the Tyrrhenian Sea area. J. Mar. Syst. 18, 161–178. Robinson, A.J., Sellshop, J., Warn-Warnas, A., Leslie, W.J., Lozano, C.J., Halley Jr., P.J., Anderson, L.A., Lermusiaux, P.F.J., 1999. The Atlantic Ionian Stream. J. Mar. Syst. 20, 129–156. Seibert, P., Frank, A., 2004. Source-receptor matrix calculation with a Lagrangian particle dispersion model in backward mode. Atmos. Chem. Phys. 4, 51–63. Soomere, T., Viikmäe, B., Delpeche, N., Myrberg, K., 2010. Towards identification of areas of reduced risk in the Gulf of Finland, the Baltic Sea. Proc. Estonian Acad. Sci. 59 (2), 156–165. Soomere, T., Döös, K., Lehmann, A., Markus Meier, H.E., Murawski, J., Myrberg, K., Stanev, E., 2014. The potential of current- and wind-driven transport for environmental management of the Baltic Sea. AMBIO 43, 94–104. http:// dx.doi.org/10.1007/s13280-013-0486-. Stohl, A., Seibert, P., Wotawa, G., Arnold, D., Burkhart, J.F., Eckhardt, S., Tapia, C., Vargas, A., Yasunari, T.J., 2012. Xenon-133 and caesium-137 releases into the atmosphere from the Fukushima Dai-ichi nuclear power plant: determination of source term, atmospheric dispersion, and deposition. Atmos. Chem. Phys. 12, 2313–2343. Thomson, D.J., 1987. Criteria for the selection of stochastic models of particle trajectories in turbulent flows. J. Fluid Mech. 180, 529–556. Torgrimson, G.M., 1981. A comprehensive model for oil spill simulation. In: Proceedings of the Oil Spill Conference. American Petroleum Institute, Washington DC, pp. 423–428. Viikmäe, B., Torsvik, T., Soomere, T., 2013. Impact of horizontal eddy diffusivity on Lagrangian statistics for coastal pollution from a major marine fairway. Ocean Dyn. 63 (5), 589–597. Zavatarelli, M., Mellor, G.L., 1995. A numerical study of the Mediterranean Sea circulation. J. Phys. Oceanogr. 25, 1384–1414.