2.2. The New Generation of Oil Spill Models
Broadly speaking, state-of-the-art oil spill models produce not only oil spill predictions, but also the assessment of ambiguity of such forecasts, which is crucial and urgent for up-to-date, beneficial, and cost-effective responses. This uncertainty in the forecasting of oil transport and transformation arises mostly from uncertainties in the input fields (errors in initial conditions, environmental data, and in the predictions of metocean models), internal model dynamics (e.g., numerical scheme, parameterization of transformation processes), and sparse observational data [
174,
175]. Due to this large number of uncertain sources introduced in oil spill models, ensemble forecasts are important to improve the quality of predictions (e.g., [
176,
177]). ASTM (American Society for Testing and Materials) [
178] has established a standard for oil spill models requiring uncertainty estimates for oil spill trajectory forecasts to support response operators. However, the methodology for uncertainty estimation is not well-established in oil spill models [
179] and presents a field of future research. Nevertheless, NOAA’s GNOME [
180,
181] model, for example, includes uncertainty algorithms regarding the perturbation of current and wind fields. De Dominicis et al. [
176] used an ensemble of metocean models to improve oil trajectory forecasts with MEDSLIK-II model. Liubartseva et al. [
179] introduced an uncertainty module in WITOIL Decision Support System, which includes MEDSLIK-II for oil spill forecasting, to automatically estimate prediction uncertainties related to the initial conditions of the spill, based on a parametric analysis methodology, employed in atmospheric pollution models [
182]. Oil spread probability maps are produced as an indication of predictions uncertainty.
In parallel, state-of-the-art oil spill models use satellite SAR images/data to identify potential oil slicks and implement spill and drifter surveillance to improve slick forecasting. In detail, existing oil spill remote sensing techniques are presented in the review papers of Fingas and Brown [
183,
184]. The attention of the scientific community has been focused on enhancing 4D predictions by simulating oil spills backward in time to track the slick to its source [
145]. These back-propagation approaches, when correlated with the operation of the AIS (automatic identification of ships) system, could track down the sources of world-wide oil spills.
Some of the most widely used oil spill models, capable of forecasting the trajectory and fate of surface and/or deep sea oil spills are: CDOG [
156,
172,
185,
186], OSCAR [
118,
119,
120], OSIS [
187], OILMAP [
122,
188], OILMAPDEEP [
188,
189,
190,
191], SIMAP [
38,
39,
192], TAMOC [
193,
194,
195], BLOSOM [
155], MOTHY [
196,
197,
198,
199], OILTOX [
200], MOHID [
201], POSEIDON OSM [
202], MEDSLIK [
124,
125,
174], GNOME [
180], OILTRANS [
203], OSERIT [
204], MEDSLIK-II [
89,
90], and OpenOil [
112,
205]. An analysis of these models is given in the following paragraphs.
CDOG (comprehensive deepwater oil and gas model) is a three-dimensional model, developed by Yapa and Li [
206] and modified by Zheng et al. [
186]. The model simulates the aspect of oil and gas released from deep water accidents [
172,
186]. Moreover, in CDOG, hydrate formation and disintegration, gas dissolution, non-ideal behavior of gas, and potential gas partition from the basic plume, in virtue of strong cross-currents, are connected to the jet/plume hydrodynamics and thermodynamics [
172]. CDOG includes unsteady-state 3D fluctuation of ambient currents, density stratification, salinity, and water temperature [
207]. Although CDOG has been implemented for response purposes, its main objective is research. Recently, the US government agencies (MMS (Minerals Management Service), NOAA) and oil companies have started using the CDOG model.
OILMAPDEEP (deep water oil spill model and analysis system) [
188,
189,
190,
191,
208] has been developed by Applied Science Associates (ASA) in order to estimate the fate and transport of subsea releases. OILMAPDEEP estimates the near-field plume characteristics and oil droplet size distributions for a specified release [
189,
208]. Oil droplet size distribution predictions are in accordance with the study of [
171,
209]. Moreover, the trap height and droplet size distribution are used as initial conditions for the SIMAP [
38,
39,
210], which computes the transport and fate oil processes in accordance with the near-field buoyant plume [
208,
210]. The model provides simulations of both the near-field and far-field environment [
127,
191,
208,
211]. Fate processes are included for both gas and oil, however, the details of the modeling algorithms are unpublished and rely on a database of oil chemical features, according to ASA. The model provides a subsurface dispersant treatment module and a Lagrangian particle tracking module, incorporating 2D and 3D hydrodynamic model flow fields [
190,
191]. Output data contain plan and section views of plume, in-water, and on-surface model forecasting [
189,
190,
191]. The model has global capacity and includes RPS (Rural Planning Services) ASA’s own GIS.
SIMAP (integrated oil spill impact oil system) [
38,
39,
208,
210] also developed by ASA provides simulations of the three-dimensional trajectory, fate, and transport, as well as biological effects and other impacts of spilled oil and fuels [
39,
192]. Moreover, the model may be run in both stochastic and deterministic modes and includes a buoyant plume transition stage to the far field. The model has a Lagrangian particle tracking module in the far field [
212]. It includes oil processes with specific model limitations, such as dissolution and sedimentation of oil, sinking, evaporation, dispersion, and spreading, complex oil and ice interaction, together with sediment and shoreline contamination. Some applications of the SIMAP model [
39,
192,
212] include the environmental impact assessment of oil spills, hindcast/forecast simulations, natural wealth damage evaluation, contingency planning, environmental risk assessment, and cost-effective study. SIMAP has been validated against data of more than 20 large spills, such as the Exxon Valdez [
38,
39,
213].
OSCAR (oil spill contingency and response model) is an advanced, three-dimensional model for planning and response to oil spills, developed by SINTEF [
37,
115]. It calculates the fate and effects of surface releases or blowout/buoyant plume of oil or gas [
121]. The chemical fates sub-model allows multiple separate pseudo-components, which are transported across all environmental segments [
37]. The transport and fate of oil spills at the surface are described not only by virtue of currents, winds, and turbulent diffusion, but also by means of oil-weathering algorithms, such as spreading, evaporation, natural dispersion, emulsification, dissolution, and volatilization. Moreover, in the water column, horizontal and vertical dispersion of entrained and dissolved hydrocarbons are represented via random walk approaches. Finally, the degradation and sedimentation processes of oil are described as first order decay process [
121]. Essential elements of the model are SINTEF’s data-based oil weathering model [
214,
215,
216], the three-dimensional oil trajectory and chemical fates model [
118], an oil spill battle model [
120], and exposure models for fish and ichthyoplankton [
119], birds, and sea mammals [
217]. Overall, OSCAR has been used in oil spill risk assessment, as well as in response planning and operations [
121]. The model has been applied for hindcast and forecast of accidental releases in locations such as the North Sea, the Baltic Sea, the Gulf of Mexico, and the Mediterranean basin [
121]. In the UK, OSCAR is routinely used for operational forecasting of oil spills, forced by ocean circulation models such as the U.S. Navy global hybrid coordinate ocean model (HYCOM) or the Copernicus system and wind forecasts from NOAA’s GFS (global forecast system) or CFS (climate forecast system).
OILMAP [
122] has been developed by ASA as well as SIMAP, and both of them share the same code base. However, OILMAP is a three-dimensional oil spill response and contingency planning model. It deals with both surface and subsurface hydrocarbon releases and provides algorithms for oil spreading, evaporation, emulsification, entrainment, and oil-shoreline, oil bed, and oil-ice interaction [
122,
218]. The stochastic module predicts an extensive number of trajectories from a single site for producing probability statistics [
218]. The distribution and mass balance of oil over time are simulated per type of oil spilled. The model has been applied in Dubai and Gulf region in 2006 [
219]. It is used operationally by Oil Spill Response Limited (OSRL) in United Kingdom.
TAMOC (Texas A&M oil spill calculator) [
193,
194,
195] is an open-source model, written in Python and Fortran, which simulates subsea oil spills and blowout plumes. Furthermore, its code is available for users in Github:
http://github.com/socolofs/tamoc. It computes near-field plume dynamics, dissolution, particle tracking, transport of oil droplets, and phase equilibrium of hydrocarbons and incorporates an all-inclusive fate module. An extensive report of the TAMOC model and its mathematical background and equations are mentioned in the works of Gros et al. [
193,
195]. The oil fate and transport are expressed according to the formulation of McGinnis et al. [
220] and as for jet and plume schemes, these are described by an integral model method [
170,
186,
221]. A key feature of this model is the combination among the extended hydrodynamic behavior and the dynamic equations of motion, such as plume and intrusion formation. Finally, TAMOC has been validated via several experimental studies of bubble plumes, such as [
221].
MOTHY (modèle océanique de transport d’hydrocarbures), developed by Météo-France [
197], is a 3D Lagrangian pollutant drift model predicting the fate and transport of oil slicks on the ocean surface. MOTHY has been operational since 1994 and it has been used and validated during major real oil spill incidents, such as the Erika [
197,
222] and the Prestige [
223]. The mixed layer is expressed via a combination of a shallow water model relative to the wind and the atmospheric pressure, in cooperation with a well-described turbulent viscosity model, while hydrodynamics are provided by CMEMS (Copernicus Marine Service) Med MFC (Marine Fisheries Commission) models and wind forcing provided by European Center for Medium-Range Weather Forecasts (ECMWF) [
224]. National, higher-resolution ocean forecasting systems nested in CMEMS Med MFC are used in several cases to resolve coastal scale processes in various areas of the Mediterranean. The water column is described by a continuous profile from surface to bottom [
224,
225]. Turbulent diffusion is modeled via a three-dimensional random walk scheme [
226]. This oil spill model provides some additional capacities: beaching, sedimentation, and backtracking, while pollutants can be either oil or floating objects [
225].
OILTOX is a Lagrangian oil spill model [
200] adapted to the Black Sea environment. It includes hundreds of oil types that are transported via the Black Sea and their fundamental physical-chemical features. This model simulates oil transport and fate according to [
200,
227] in five phases: oil-on-surface, oil-in-water, oil-on-bottom, oil-on-suspended sediments, and oil-at-shoreline. The model incorporates the basic transport and weathering processes, such as spreading by virtue of gravity and surface tension, advection due to wind and surface currents, evaporation, emulsification, oil-shore interaction, wave entrainment, resurfacing of entrained oil, and sedimentation [
200]. Moreover, the model incorporates horizontal and vertical turbulent diffusion processes, which are represented by means of a random walk method.
The MOHID (Modelo Hidrodinâmico) Lagrangian oil spill model [
201] is a sub-model of the MOHID water modeling system [
228], developed by the Technical University of Lisbon. The movement of the tracers is caused by the surface flow field, the atmospheric winds, the spreading velocity from the dispersion module, and a randomly produced velocity via a random walk approach. MOHID Lagrangian transport module includes the following features: oil transport in water column, sedimentation, and beaching; oil weathering processes such as evaporation, dispersion, entrainment, sedimentation, dissolution, emulsification, and dispersion; and Eulerian concentration result.
The POSEIDON OSM is an oil spill model generated by the Hellenic Centre for Marine Research (HCMR), implemented and operational in the Aegean and Ionian Seas ([
202,
229] since 2000. It is a completely 3D oil spill model with the capacity not only to predict the transport, spreading, and weathering of the oil particles in the 3D space, but also to provide various oil weathering processes, such as evaporation, emulsification, beaching, and sedimentation [
174,
225]. Oil advection and dispersion is illustrated via a vast number of particles, each of which expresses a group of oil droplets of similar size and composition [
174]. Oil transport is calculated using two modules: the circulation module and the wind generated wave module [
174,
225]. Moreover, the horizontal movement considering advection and the vertical transport of the oil are described through the results of the POSEIDON ocean forecasting system [
225]. Stokes drift is also provided by the coupled wave model of POSEIDON ocean forecasting system [
174,
225]. Recently, some additional characteristics were integrated in the model via a dedicated web-based application (
https://poseidon.hcmr.gr/components/forecasting-components/oil-spill-model [
230]), where the user can determine the parameters of a real or hypothetical scenario and submit this event to the system, receiving the model output in Google Earth file format for a more real-time geospatial simulation.
GNOME (general NOAA operational modeling environment) is an oil spill model that predicts the fate and transport of pollutants and oil movement caused by winds, currents, tides, and spreading [
180]. GNOME was developed by the Hazardous Materials Response Division and it is an open-source model, freely available in Github:
https://github.com/NOAA-ORR-ERD [
231]. The model is publicly available for use by the broader academic, response, and oil spill planning communities. GNOME provides the following elements: 3D particle transport, able to work with virtually any hydrodynamic model and measured field data, 1, 2 or 3rd order Runga-Kutta algorithm, with droplet rise velocity depending on density and droplet size; “leeway” wind surface transport: randomly adjustable with various user-adjustable values, providing a configurable downward spread; open-source code; backward running; oil weathering algorithms from the integrated open source ADIOS oil database, which is currently getting updated, with a beta version available for testing at:
https://adios-stage.orr.noaa.gov [
232]; sea ice interaction according to ice concentration and velocity; shoreline interaction (beaching) with configurable half-life based re-float; includes the TAMOC deep-water blowout model; comprehensive script for stochastic analysis and other batch processing; configurable for use on other drifting objects, such as for SAR and marine debris; integrated response options calculator (ROC) to evaluate the performance of spill response systems such as skimming, burning, and application of chemical dispersant; the PyGNOME, a Python setting, to build the web GNOME interface to the model, that performs batch processing and testing; and includes a GIS system for the model outputs visualization. In addition, GNOME is extremely configurable and tunable to adjust to field conditions and it can be driven via numerous data: measured point data, met models, and hydro models with a variety of meshes (structured, triangular). Finally, it has been used to support spill response for oil spills in the USA for almost twenty years. As GNOME can be integrated with any ocean circulation and meteorological model providing forecasts at different file formats, as well as observational data, NOAA has developed the GNOME Operational Oceanographic Data Server (
https://gnome.orr.noaa.gov/goods) [
233], a publicly available system to provide access to all the driver models and data sources available. Another important feature of the operational use of GNOME is the assimilation of available observations of oil spill locations in each forecasting cycle. Model parameters are fine-tuned to match the observations, and subsequently a new forecasting cycle and analyses are produced. Observational data assimilation improves the accuracy of forecasts for response authorities.
The OILTRANS particle transport model [
203] is based on the LTRANS v.2 particle transport model, developed by North et al. [
234]. The oil fates module of OILTRANS simulates the transport, fate, and oil weathering processes coupled to state-of-the-art operational metocean model [
203]. The model provides the oil fate processes of spreading, advection, diffusion, evaporation, emulsification, and dispersion in order to estimate the horizontal movement of surface oil slick, the vertical entrainment of oil into the water column and the oil mass balance [
203]. OILTRANS can be applied in any ocean or coastal field. The minimum data required are: bathymetric data, tidal current, and wind fields, together with information on the location, quantity, and type of spilled oil [
235,
236]. OILTRANS has been used, for example, for an accidental release in the Celtic Sea in February 2009 [
203].
OSERIT, oil spill evaluation and response integrated tool, is an oil spill model that is capable of predicting the 3D drift and the fate of an oil spill at the surface and into the water column [
237]. It contributes to the forecasting service of EMSA CleanSeaNet and has been used in the North Sea. The Lagrangian module expresses the independent movement of each parcel due to the winds, currents, and waves. Furthermore, OSERIT contains the buoyancy effect, turbulent diffusive transport, vertical dispersion of oil from surface to the water column, horizontal spreading, and beaching [
237]. Moreover, it is able to calculate the drift of chemically dispersed oil and forecasts oil weathering processes, such as evaporation and emulsification, and their effects on oil features. Biodegradation and oil sedimentation are not included in OSERIT. The oil database of OSERIT is based on the oil types included in the ADIOS database [
54,
238].
BLOSOM (blowout and spill occurrence model) has been developed by the National Energy Technology Laboratory (NETL) of the USA. The model is written in Java programming language [
239], and it is an extensive, open-source modeling suite that displaces the fate and transport of both subsurface oil blowouts and surface spills [
240]. Moreover, this model is developed to predict offshore oil spills resulting from deep water (>150 m) and ultra-deepwater (>1500 m) well blowouts [
155]. The model simulates oil spills from source to final fate and degradation stage. BLOSOM is flexible in its construction and utility from using it for basic particle tracking to applying advanced weathering modules and modules for jet/plume modeling [
155]. BLOSOM supports risk evaluation and provides a comprehensive tool for response planning. It is designed to handle deep-water blowouts, such as Deepwater Horizon [
241]. The jet plume element of BLOSOM has been assessed via experimental studies, which took place in the North Sea [
242,
243]. BLOSOM integrates various oil types from the ADIOS oil library [
25].