Journal of Environmental Management 166 (2016) 357e373

Contents lists available at ScienceDirect

Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman

Research article

Platform of integrated tools to support environmental studies and management of dredging activities Alessandra Feola a, *, Iolanda Lisi b, Andrea Salmeri b, Francesco Venti b, Andrea Pedroncini c, Massimo Gabellini b, Elena Romano b a b c

ISPRA e Institute for Environmental Protection and Research, Loc. Brondolo, 30015 Chioggia, Italy ISPRA e Institute for Environmental Protection and Research, Via Brancati 60, 00144 Rome, Italy DHI, Via degli Operai 40, 16149 Genova, Italy

a r t i c l e i n f o

a b s t r a c t

Article history: Received 18 March 2015 Received in revised form 8 October 2015 Accepted 12 October 2015 Available online xxx

Dredging activities can cause environmental impacts due to, among other, the increase of the Suspended Solid Concentration (SSC) and their subsequent dispersion and deposition (DEP) far from the dredging point. The dynamics of the resulting dredging plume can strongly differ in spatial and temporal evolution. This evolution, for both conventional mechanical and hydraulic dredges, depends on the different mechanisms of sediment release in water column and the site-specific environmental conditions. Several numerical models are currently in use to simulate the dredging plume dynamics. Model results can be analysed to study dispersion and advection processes at different depths and distances from the dredging source. Usually, scenarios with frequent and extreme meteomarine conditions are chosen and extreme values of parameters (i.e. maximum intensity or total duration) are evaluated for environmental assessment. This paper presents a flexible, consistent and integrated methodological approach. Statistical parameters and indexes are derived from the analysis of SSC and DEP simulated time-series to numerically estimate their spatial (vertical and horizontal) and seasonal variability, thereby allowing a comparison of the effects of hydraulic and mechanical dredges. Events that exceed defined thresholds are described in term of magnitude, duration and frequency. A new integrated index combining these parameters, SSCnum, is proposed for environmental assessment. Maps representing the proposed parameters allow direct comparison of effects due to different (mechanical and hydraulic) dredges at progressive distances from the dredging zone. Results can contribute towards identification and assessment of the potential environmental effects of a proposed dredging project. A suitable evaluation of alternative technical choices, appropriate mitigation, management and monitoring measure is allowed in this framework. Environmental Risk Assessment and Decision Support Systems (DSS) may take advantage of the proposed tool. The approach is applied to a hypothetical dredging project in the Augusta Harbour (Eastern coast of Sicily IslandeItaly). © 2015 Elsevier Ltd. All rights reserved.

Keywords: 3D numerical modelling Dredging plume mapping Magnitude, duration and frequency of SSC variation Environmental effects Dredging management

1. Introduction Dredging activities are commonly used in coastal areas to maintain or improve the designed depth of navigation channels or basins, for the creation or the improvement of facilities, and to carefully remove and relocate contaminated materials. These activities involve the processes of removing sediments from the bottom and subsequently relocating elsewhere. According to the

* Corresponding author. E-mail address: [email protected] (A. Feola). http://dx.doi.org/10.1016/j.jenvman.2015.10.022 0301-4797/© 2015 Elsevier Ltd. All rights reserved.

working principles for these processes, dredges may be divided into two broad categories (EPA, 1993; OMOE, 1994; IADC, 1998; USACE, 2003; Anchor Environmental C.A. L.P., 2003; Eisma, 2006): mechanical (grab or clamshell and backhoe) and hydraulic (stationary and cutter suction) dredges. The increase of the Suspended Solid Concentration (SSC) during dredging operations and the subsequent deposition of sediments (DEP), transported as a dredging plume, are considered a prominent environmental issue. In recent years, increasing attention has been paid to reduce any physical, chemical and biological changes related to the sediment resuspension and pollutants (if any) dispersion (Christensen et al., 2001; Wilber and Clark, 2001; HR

358

A. Feola et al. / Journal of Environmental Management 166 (2016) 357e373

Wallingford Ltd & Dredging Research Ltd, 2003). Tighter controls, in the form of strict regulations, and proper enforcement on monitoring and mitigating measures, help to prevent or minimize adverse impacts (Erftemeijer and Lewis, 2006). Specific operating precautions are now successfully used to minimize the sediment release due to “conventional” dredges, and a number of newer “environmental” dredges are specifically designed to carefully remove contaminated materials (Palermo and Averett, 2003). Site and operational conditions affect the suspended rate (sediment loss rate) close to the dredging sources, and the resulting plumes are complex in terms of spatial distribution and temporal evolution (USACE, 2003; Hayes et al., 2000; Palermo and Averett, 2003; Bridges et al., 2008; Palermo et al., 2008; IADC, 1998; Hayes and Wu, 2001; Burt et al., 2000). For projects that involve the handling of sediments, a detailed Environmental Impact Assessment (EIA, Directive 2014/52/UE) should be carried out to determine the potential environmental impacts, to evaluate technical alternatives and design appropriate mitigation, management and monitoring measures. In the absence of local legislation and guidelines, well-established international guidelines, aimed to support environmental studies during these activities, are available. Most of these guidelines include the use of numerical modelling as a valuable tool (Jouon et al., 2006; Edwards et al., 2006; PIANC, 2010; EPA, 2011). Models can help to support environmental studies before dredging programs begin, and interpretation of results can help to optimize environmental objectives while maintaining desired production rates (Savioli et al., 2013). Different models are currently used to forecast the planar and the vertical extension of the plume dynamics close (nearfield models) and far from to the suspension sources (far field models) (e.g. Shankar et al., 1997; Kim and Je, 2006; Bilgili et al., 2005; Bell and Reeve, 2010). Modelling results are usually presented for “extreme scenarios”, mainly covering only one or few tidal cycles, high-energy or extreme events (e.g. storm or low-frequency flood event) and “seasonal scenarios” (Jiang and Fissel, 2011; Liu et al., 2002; IMDC, 2012). Under common guidelines (e.g. GBRMPA, 2012), results are rarely reported to cover a full year (Deltares, 2009), but are rather referred to much shorter periods. As stated by Johnson et al. (2000), to be truly effective as a dredging project management tool with respect to environmental protection, models should be capable of running multiple simulations in a relatively short time so that a number of alternative dredging scenarios can be evaluated to determine those with the least probabilities of detrimental impacts. A different approach, presented by SKM (2013), is focused on long-term migration of sediments and related effects on water quality and ecosystem condition, modelling the movement of dredged material both during dredging disposal operations and over 12 months. Increases of SSC and DEP parameters at a distance from the dredging source are mainly used to evaluate the extension of the area interested by dispersion and deposition of suspended sediments. The SSC can be expressed as the depth-averaged value or as Total Suspended Solids Concentration (TSSC) (e.g. Bell and Reeve, 2010; Fitzpatrick et al., 2009; Jiang and Fissel, 2011; Je et al., 2007). Maximum excess of SSC is usually expressed in relation to certain thresholds (IMDC, 2012; Bell and Reeve, 2010; Deltares, 2009). GBRMPA (2012) recommends that model results should include, as minimum requirement, maps showing the predicted maximum and mean SSC at mid-depth and near the seafloor, and the predicted deposition rate (g/m2) as well as time-series predictions of these three parameters at key sites over the duration of the project. Descriptive statistics of the SSC, such as the median, 95th percentile and the maximum (Hadfield, 2014) are sometimes reported. Total sedimentation and bottom thickness maps are usually presented for single time-step, at the end of the specified

dredged material placement scenario, or at a certain time after placement (SKM, 2013). The duration of environmental effects is usually reported in term of exceedance probability, which is calculated as the percentage of time during which an SSC threshold is exceeded throughout the dredging and dumping operation (IMDC, 2012; Savioli et al., 2013). The percentage of time that the SSC is above the “critical” threshold for more than 12 h within a 24 h period, is also reported (Fitzpatrick et al., 2009). In few cases, magnitude is related to duration and frequency of resuspension (Schoellhamer, 2002). The CCME WQI is proposed by Canadian Council of Ministers of the Environment (CCME, 2001) as an index for simplifying the reporting of water quality data. The CCME WQI is based on three individual factors, relating the extent of water quality guideline non-compliance over the time period of interest (factor 1: scope), the percentage of individual tests (“failed tests”) that do not meet objectives (factor 2: frequency), and the amount of failed test with values that do not meet objectives (factor 3: amplitude). This index gives a measure of water quality referred to the length of a vector calculated by combining the three factors and scaled to range between zero and 100. The risk level and the severity of environmental impacts depend on the closeness of environmentally sensitive areas (i.e. Sites of Community Importance with specific habitats, etc.) to the dredging zone and local hydrodynamics. Thus, to address the severity of environmental effects related to dredging, it is appropriate to estimate the magnitude, duration and frequency of the exposure to SSC (or higher sedimentation; Clarke and Wilber, 2000). It also relates to pre-existing habitat stress, which may affect the tolerance of species to elevated turbidity and sedimentation. In particular, frequent short-term or chronic long-term exposure to high SSC or sedimentation events may result in mortality for some species, while moderate levels of increased SSC and sedimentation persisting for particularly long time may cause changes in diversity for more sensitive species that are then gradually replaced by more tolerant ones. Thus, the environmental management of the dredging works requires the quantification of these different aspects through the determination of temporal and spatial variability of SSC in the water column (PIANC, 2010; Clarke and Wilber, 2000). At present, there is a lack of tools that synthesize results of validated numerical models and make them usable for decision support and environmental management (SKM, 2013). To address this need, this paper describes an integrated, flexible and replicable methodological approach for synthesizing parameters related to water quality variations that arise from dredging activities. This approach is designed for different dredging techniques in coastal areas, with a main focus on estuarine and semi-enclosed basin. The main objective of the approach is to capture the spatial (vertical and horizontal) and temporal variability of SSC and DEP levels using simple and intuitive parameters. The model considers a full year and multiple scenarios to account for seasonal variations. It results in a realistic understanding of outcomes related to the dredging plume development. Events of exceedance of SSC thresholds are spatially described in term of magnitude, duration and frequency, and through the definition of a new integrated index (SSCnum). The proposed approach is applied to Augusta Harbour (Eastern coast of SicilyeItaly) case study, for a hypothetical dredging project and for a SSC threshold arbitrarily defined. 2. Material and methods 2.1. Integrated methodological approach The proposed methodological approach, Dr-EAM, has been developed to support environmental studies (Environmental

A. Feola et al. / Journal of Environmental Management 166 (2016) 357e373

Assessment Method) related to dredging activities (Dr) in coastal areas, with a main focus on estuarine and semi-enclosed basin. Two main modules, hereinafter called Hydrodynamic and Transport Module (HTM) and Environmental Assessment Module (EAM), are implemented in series (Fig. 1). Here, classical deterministic models are used for evaluating variations of SSC and pollutants (if any) in the water column and variations of DEP at the seabed, induced by different dredging plumes under the action of different climatic (ordinary and extreme) conditions. Numerical results are analysed and combined to derive suitable statistical parameters and indexes from predicted SSC (mg/l) and DEP (deposition, g/m2) time-series. Maps showing the spatial distribution of magnitude, duration and frequency of high SSC events are produced representing statistical parameters over the duration of the activity. The methodological approach, here used to evaluate the temporal and spatial variation of the SSC, can be replicated to assess the dredging-induced temporal and spatial variation on other water quality parameters and biological features (e.g. pollutants, nutrient, dissolved oxygen, particulate organic matter). This approach is  mez designed to support, with its outputs, Risk Assessment (Go et al., 2014) and Decision Support Systems (DSS). 2.1.1. Hydrodynamic and Transport Module (HTM) HTM is used, as part of Dr-EAM, to evaluate how hydrodynamic parameters affect sediment transport and deposition rate for different dredging techniques and environmental conditions within the intervention area. Three-dimensional hydrodynamic and transport models are highly recommended to forecast plumes arising from dredging operations, providing a complete temporal and spatial picture of advection/dispersion processes in stratified water bodies (e.g. Davies et al., 2002; DHI, 2012). The entire domain should be discretized in horizontal and vertical directions with a resolution appropriate to reproduce the complexity of current flow and physical processes that relate to plume dynamics (advection, diffusion and settling processes of SSC) and how it moves from the dredging site to near-field and farfield zones (Bridges et al., 2008). The resolution and scale of the computational domain should be suitable for environmental studies (PIANC, 2010). An offshore area should be included in the domain in order to avoid boundaries effects on numerical results within the area of interest. A hybrid vertical discretization system can be used to properly simulate the stratification effects and both

359

the surface and bottom layers, with normalized thicknesses (scoordinate system) in shallow water and horizontal layer thicknesses (z-coordinate system) for the deeper part of offshore area. The simulation of plume evolution requires the definition of the dredging source and the definition of the sediment loss rate in the immediate vicinity of the dredge (“near-field” area). In fact, in the starting phase of the plume development, the suspended rate can greatly vary for different dredging equipment (hydraulic and mechanical dredges) and operational techniques (dredge-head movements, dredge cut depth, production rate, precautions taken in advance or during the dredging operations, dredging cycles, distribution of spill at different depth along the column, etc.; Bridges et al., 2008; Palermo et al., 2008; DHI, 2012). Using conventional mechanical dredges, a high level of sediment release can occur when the grab (or the bucket) hits the seabed and raising dredged material through the water (spillage). In this case, the sediment loss rate is generally assumed constant through the water column. On the contrary, using conventional hydraulic dredges, operating on an almost continuous dredging cycle, the sediment release is mainly due to fractions of the dislodged sediments that escape to the suction pipe during drag-head disturbance at the bottom (Hayes et al., 2000). Thus, in this case, the source is expected to be confined around the drag-head. The properties of the sediment to be dredged (i.e. volume, grain size distribution, watercontent, contamination level, degree to which it aggregates when disturbed by dredging, settling properties, etc.) also greatly affect the amount and size of the sediment release. When attempting to implement the models, key components of the dredging program, such as the dredge methodology, dredge schedule, sediment spill sources, and climatic conditions encountered during dredging activity, may not be well defined. This will typically be reflected in the accuracy of the predictions and in the efficiency of mitigation measures (Savioli et al., 2013). Initial assumptions must be based on the best available information during the planning phase, but they could require later adjustments. As the released sediment moved out of the dredging zone, coarse particles progressively settle in the near field zone, while the finer fractions move away to constitute the far field (or passive) plume. The far field transition generally occurs within few hundred meters far from the dredging zone (John et al., 2000; Bridges et al., 2008), but it is strongly related to advection and diffusion phenomena as much as sediment settling processes, driven by the sitespecific conditions.

Fig. 1. Schematic flow-chart of Dr-EAM.

360

A. Feola et al. / Journal of Environmental Management 166 (2016) 357e373

Basically, at the dredging source, the dredge rate, which is the average suspension rate representative of the mass discharge of suspended sediment (kg/s, eventually defined as discharge, m3/s, and suspended sediment concentration, kg/m3), must be specified. A spill percentage describes the fraction of suspended sediment that is finally dispersed in the water column, taking into account only the fine fractions of particles smaller than 74 mm (e.g. Hayes et al., 2000; Hayes and Wu, 2001). The dredging source location must be specified as it varies in time during work progression. The source can be located only near the bottom, on top layers, or it can be uniformly distributed in the water column. To evaluate SSC and DEP gradients for the most representative scenarios (high frequency and high-energy or extreme events, lowfrequency flood events) modelling should be undertaken using meteomarine conditions corresponding to a full reference year. This helps the model to capture seasonal variations in order to support the identification of best timing for hypothetical dredging periods. The reference year should be defined based on a multi-year statistical analysis on oceanographic and meteomarine datasets. Moreover, numerical modelling should be performed for a time window long enough to establish the suspension source due to the seabed erosion and initial settling that occurs after dumping. In fact, the dredging plume must be considered as the result of the combination of the source spilled from the dredge head action on the seabed and during the dredging cycle, and the seabed erosion caused by current friction that suspends settled material as a residual layer within and outside the dredging area (Palermo and Averett, 2003; Palermo et al., 2008). The proper implementation of HTM requires input data describing the dredging source and other factors affecting plume dispersion. Field data collection used to force the model (free surface oscillation, wind, waves, oceanographic features) and to calibrate and validate numerical results (current speed and direction, SSC, etc.; Jiang, 2014) are highly recommended to forecast advection/dispersion processes. In the literature, a wide range of productivity values for different dredge types is reported (e.g. John et al., 2000; Hayes and Wu, 2001; Anchor Environmental C.A. L.P., 2003). Generally production rate ranges from 50 to 10,000 m3/hr for hydraulic dredges, while values up to 500 m3/hr can be achieved with largest mechanical dredges. Although the sediment mass rate can be derived from previous SSC data collected for a variety of dredging operations, the available data cover only a relatively narrow set of operational (production rate, dredge-head dimension, dredge-head penetration, velocity of dredging cycle, etc.) and environmental (range of sediment, water depth, climatic and oceanographic conditions) factors to serve as a predictive base for distinctly different dredging operations. Then, the adequate definition of mass rate of suspended sediments (kg/s) for a proper implementation of the method requires further collection of SSC data during dredging operation. 2.1.2. Environmental Assessment Module (EAM) At present there is a lack of useful tools to synthesize numerical results and to make these results usable for decision support and purposes of management. EAM is proposed to support the assessment of environmental effects from dredging plume. The quantification of magnitude, duration and frequency of SSC events in the short and long-period is presented. EAM includes automated tools specifically implemented to analyse HTM outputs parameters (i.e. current field, SSC, DEP), resulting in integrated parameters or indexes useful for the quantitative characterization of the dredging plume dynamic in time and space and any related environmental effects. Data analysis requires the comparison of numerical results with

values in “reference areas” potentially not affected by the dredging works. The choice of these reference areas requires site-specific evaluations of natural variability of SSC (i.e. induced by the local circulation, river discharge, etc.) and requires a careful and integrated analysis of different selected parameters. The GBRMPA hydrodynamic modelling guidelines for dredging projects (GBRMPA, 2012) encourage the application of numerical models to predict the spatial extent and the severity of the impact. Three different zones are identified: high impact, moderate impact and influence zones. To model zones of impact, quantitative impact thresholds for SSC and DEP must be established. The knowledge of the potentially affected receptors may not be sufficient to establish impact threshold criteria prior to modelling. It will then be necessary to calculate the general spatial distribution of varying levels of SSC and DEP, and use the results to identify potential receptors. The next step is to establish threshold criteria for zones of impact and influence (SKM, 2013). Thus, it is useful to define a discrete number of check-points for extracting time series of output parameters throughout the whole period of the HTM simulation. Check-points could be regularly distributed in the domain with a spatial scale chosen as a function of the spatial variability of the numerical results. Spatial density of check-points can be different from the grid resolution. For each check-point, a SSC time series can be extracted at different depth (i.e. bottom layer, surface layer, etc.) and a DEP time series can be extracted at seabed. Mean and maximum intensity of SSC parameters (respectively SSCmean and SSCmax) and DEP amount are then calculated for each time series (Fig. 2). Quantifiable tolerance limits for SSC (and similarly, light attenuation, sedimentation, and contaminant dispersion), can be identified with respect to regulatory limits and obligations. In the absence of defined ranges of variability, meaningful criteria to limit the extent and intensity of dredging plumes and their effects will always require site-specific evaluations, taking into account both the magnitude and the natural variability of the local background turbidity (Erftemeijer and Lewis, 2006). This identification is complex, but it allows to provide a control parameter for supporting environmental studies and the subsequent indirect monitoring (and modelling) in a management context. In particular, conservative tolerance limits should be preliminary stated in the environmental impact assessment stage and should be refined as part of the subsequent monitoring and management programme

Fig. 2. Time series for SSC. Evaluation of mean and max SSC and of exceedance events in term of intensity (SSCmean_th,j) and duration (tj) for an arbitrarily defined threshold (5 mg/l). SSCnum is defined as sum of products of intensity and duration of single events (eq. (1)).

A. Feola et al. / Journal of Environmental Management 166 (2016) 357e373

during the course of the project (PIANC, 2010). For a given threshold of SSC (here arbitrarily assumed equal to 5 mg/l), a series of exceedance events can be identified. For each exceedance event, the most important parameters are the duration (tj) and mean intensity (SSCmean_th,j, Fig. 2). The total duration of exceedance events (ttot) is given by the sum of durations of all single events, while a frequency parameter is derived by the number of exceedance events (Mi) for each simulation. Neither the duration nor the total suspension alone can fully account for the severity of SSC increase events: an integration of the two parameters is needed (Rapaglia et al., 2011; Erm and Soomere, 2006). In Erm and Soomere (2006), the total impact of wakes was described by a “SSC number” (mg s/l), defined as the integral area below the resuspension peaks. Here SSCnum,i, for the i-th simulation period, is calculated as the sum of products of mean intensity above threshold (SCCmean_th,j) and duration (tj) of exceedance events (with j ¼ 1,.., Mi exceedance events for i-th simulation):

SSCnum;i ¼

Mi X

SSCmean

th;j $tj

(1)

j¼1

In order to have a quantitative description of the spatial and temporal variability of effects related to plume dynamics as a function of distance from the dredging zone, geostatistical parameters of SSCnum can be derived from analysis of time-series extracted at each checkpoint. In particular, the variogram 2g(h) (Eq. (2)) provides a description of how a spatial random field, or stochastic process Z(s), correlates with distance. The semivariogram function was originally defined by Matheron (1963) as half of the averaged squared difference between points separated by a vector h.

 i h  ! 1 ! ! ! g h ¼ Var Z s þ h  Zð s Þ 2

(2)

The appropriate variogram model is chosen by matching the shape of the curve of the empirical variogram to the shape of the curve of the mathematical function (i.e. linear, exponential, spherical, etc.) that best describes spatial relationships (Fig. 3). Most variograms are defined through several parameters (Fig. 3): the Nugget (microscale variation or measurement error, g(0)), the Sill (variance of the far field), and the Range (the distance above which data are no longer correlated). The range parameter gives a

Fig. 3. Empirical (dots) and modelled (line) semivariogram (g(h)) with characteristic parameters (Nugget, Sill, Range).

361

direct measure of the spatial scale of the dispersion process and can here be used to directly compare plumes produced by different dredge systems. All statistical parameters can be evaluated, for each position, for single simulation or as a mean value of seasonal and annual conditions. In order to quantify differences in term of dispersion of sediments for different technical operating systems and different weather conditions, spatially distributed parameters (SSCmean, SSCmax, and, for defined threshold, M, Ttot and SSCnum) are represented through GIS application. In particular, the domain is represented through a series of squares centred on each discrete checkpoint, with each side equal to the mutual distance. A map can be easily produced for each parameter derived from time-series analysis for specific period of time (i.e. seasonal and annual mean) and specific vertical level trough the water column. Coherently with GBRMPA (2012), when applying Dr-EAM to different case studies, the thresholds used to model ecological response must be clearly indicated and supported by peerreviewed scientific papers and/or field campaigns (Sofonia and Unsworth, 2010) and compared against model outputs. Moreover, the output maps obtained from the model should be overlaid upon maps of sensitive habitats and ecological receptors in order to relate the sediment plume dynamic with different targets. In the following, Dr-EAM is applied to a hypothetical dredging project within the Augusta Harbour case study. 2.2. Case study: Augusta Harbour Augusta Harbour, located in the eastern coast of Sicily (Italy), was formed by the closure of most of the natural Augusta Bay (early 1950s) through the construction of three breakwaters (Fig. 4). This harbour may be considered one of the most important Italian harbours for bunker operations, ship repairs, ship maintenance, and goods loading and unloading. It is also one of the most important petrochemical pole, with several wharves devoted mainly to loading of oil refineries. Also a commercial harbour is present. The sheltered area of this harbour is roughly 24 km2, extends ~7.0 km alongshore and 3.5 km cross-shore and has an average depth of 15.0 m. Bathymetry, reported in Fig. 4, was obtained by multi-beam and side-scan sonar technique acquired in 2005 (ISPRA, 2008). The harbour is connected to the open sea by two entrances: the eastern inlet (about 450 m wide and 40 m deep) and the southern inlet (about 300 m wide and 20 m deep). In addition, it connects with the open sea through the Ponte Rivellino channel, with extremely limited water depth that does not exceed 1.0 m. The northern and central sections of this water body are influenced by the runoff of several small rivers (i.e. Mulinello, Marcellino and Cantera rivers), which are characterised by seasonal and discontinuous freshwater discharges and which drain mainly MesozoiceQuaternary carbonatic rocks. Most of the harbour area was extensively dredged in 1979 while only very limited sectors were dredged in the 1980 and in 1990 (Romano et al., 2013). The strong industrial activity has had heavy consequences on the terrestrial and marine environment. During the 1960s this site underwent a fast industrial development, and several oil-refineries and petrochemical industries are still active. Particularly, marine sediments have been highly polluted by heavy metals (especially Hg), PAHs and PCBs mainly in the southern area (Romano et al., 2013). In addition, untreated urban and agricultural wastes seem responsible for eutrophication due to diminished water exchange in the enclosed environment of the bay (Romano et al., 2009). The most recent chemical and physical characterization of sediments (ISPRA, 2008) confirmed a widespread contamination, mainly in

362

A. Feola et al. / Journal of Environmental Management 166 (2016) 357e373

Fig. 4. a) Location of Augusta Harbour (Eastern coast of Sicily, Italy) and b) location of measurement stations for water levels (Catania station), wind (Belvedere station) and for vertical temperature and salinity profiles. c) Model domain, hypothetical dredging area and discrete check points for result extractions.

the southern part of the bay and with a thickness of sediments of about one meter. A physical characterization of seabed determined mainly sandy-mud and muddy sediments, except some areas close to the western coastal zone and breakwaters, characterized by the outcropping of bedrock and coarser sediments. 2.2.1. Sediment transport modelling in relation to dredging activities The complex topography, the presence of the wharves, and the average depth suggest that a 3D numerical code is required to reproduce the hydrodynamics inside the Augusta Harbour (De Marchis et al., 2014). A three-dimensional finite volume MIKE 3 Flow Model (DHI, 2012) has been used to model the stratified flow field and the cohesive sediment plume dynamics arising from the bottom during hypothetical dredging activities. A test dredging area was chosen inside the harbour (Fig. 4). A Mud Transport module (MIKE 3 MT) was added to the hydrodynamic module simulation to evaluate erosion, transport and deposition of sand/mud mixtures under action of currents and waves (DHI, 2012). Hydrodynamic and Mud Transport models were run decoupled, meaning that two simulations were executed separately in the hypothesis that any major erosion or deposition event do not alter the hydrodynamic behaviour of the site (Lumborg, 2005). The MIKE 3 Flow Model uses the three-dimensional, Reynoldsaveraged, NaviereStokes equations for solving the full non-linear equations of continuity and conservation of momentum (Rasmussen et al., 1999). The hydrostatic assumption and Boussinesq approximation were used in the formulation of the dynamic equations to yield the free-surface elevation and the three dimensional water velocity. Based on the given boundary conditions and the concepts of advection and dispersion, the model also calculates the water temperature and salinity (T, S) and thus stratification effects are included. The turbulence is parametrized through the estimation of the eddy viscosity, introduced using the

Smagornisky formulation for the horizontal eddy viscosity and the k-ε scheme for the vertical eddy viscosity. The Mud Transport module simulates, through an eulerian formulation, erosion, transport and deposition processes for fine grained sediments under currents and waves action calculated by the Hydrodynamic Module. The sediment can move from the water column to the bed (deposition) and from the bed to the water column (erosion) depending on hydrodynamic conditions, characteristics of grains (affecting the settling velocity) and bed (the degree of consolidation affects the tendency of the sediment to be eroded). The Mud Transport module applies a stochastic model for flow and sediment interaction and uses empirical relations. In particular, deposition is calculated according to the Krone formulation (1962) as a function of settling velocity (ws, m/s), near bed sediment concentration (cb, kg/m3) and the probability of deposition, function of the relationship between bed shear stress (tb, N/m2) and critical bed shear stress for deposition (tcd, N/m2). The erosion and the resuspension of deposited sediments in dense and consolidated beds are formulated by Partheniades (1965) as a function of a coefficient of erodibility (E, kg/m2/s) and of the relationship between bed shear stress (tb, N/m2) and critical bed shear stress for erosion (tce, N/m2; e.g. Lumborg, 2005). In the present work, for the model set-up, ws was set equal to 0.2 mm/s, tcd was set equal to 0.04 N/m2, while tce was set equal to 0.05 N/m2, according to tce values for dredged material applied in other modelling studies (see e.g. Parchure and Mehta, 1985; Milburn and Krishnappan, 2003). The computational domain covers the whole harbour area and an offshore area, which is needed to avoid boundaries effects on numerical results within the area of interest (Fig. 4). To ensure an appropriate resolution of the study area, a sensitivity analysis was performed, discretizing the physical domain in horizontal and vertical directions based on the bathymetry (Fig. 4) and the main forces acting on the harbour. In particular, a finite-volume flexible

A. Feola et al. / Journal of Environmental Management 166 (2016) 357e373

mesh of about 11,250 elements was constructed in the horizontal direction from digitized geometry, allowing finer discretization where necessary. Since no discernible differences in the current flow were found by changing the mesh resolution, a mean grid dimension of 75 m was applied to the inner domain located between the harbour entrance and solid boundaries, and a mean grid dimension of 250 m was applied in the domain outside the harbour. In the vertical direction, a non-uniform grid was used with a refinement near the bottom and the free surface. In particular, a hybrid coordinate system (S- and Z-coordinates) was used in the vertical direction in order to efficiently simulate, at the same time, near bed velocities in shallow water and stratification effects. The S-coordinates layers (eight sigma-layer, with thickness related to local bathymetry) were used in the shallow water inside the harbour, while Z-coordinates layers (two Z-layers) were used in the deeper part of the vertical offshore area. This resulted in roughly 90,000 elements in total. In order to evaluate SSC spatial and temporal evolution as a function of hypothetical dredging period and flow field seasonal variability, an entire reference year of climatic conditions was needed. The stratified current field was reproduced using both wind (assumed spatially uniform) and tidal forcing. Wind stress has an important influence on surface currents and circulation patterns in a semi-enclosed basin such as the Augusta Harbour (Lisi et al., 2009). Wind data, including wind speed (m/s) and wind direction (degree) components, were available from the Belvedere station, located 10 km south from Augusta Harbour (Fig. 4). Wind data were used to represent a “typical mean year”. The statistics of the wind data (temporal interval 1998/10/ 01e2008/09/30) identified year 2003 as similar to the long-term values in term of mean values of wind speed and direction, with differences smaller than 5%. Moreover, wind data analysis showed that prevailing surface wind events come from the Northeast (NE), with a mean annual occurrence of 35%, and from the Northwest (NW), with a mean annual occurrence of 33%. Higher values of wind speed are often of about 12 m/s and rarely exceed 16 m/s, with maximum (between 19 and 29 m/s) registered only from NE and E-NE. The chosen year is characterised by the presence of peaks of wind intensity for each of the main angular sectors. Water level measurements, collected during 2003 at Catania Station, located 35 km north from Augusta Harbour (Fig. 4) and belonging to the Italian Mareographic Network (http://www. mareografico.it), were applied at the open ocean boundary. The analysis of level data showed that the mean astronomic tide excursion is of about ±0.15 m, with a maximum amplitude of about ±0.30 m. Wind wave forcing, mainly from NNE and SSE, is not considered here because preliminary numerical simulations showed that breakwaters screen wave penetration and disrupt relevant wave breaking phenomena. Wave-induced circulation within the harbour is not an important phenomena as compared to tidal and wind induced circulation (Lisi et al., 2009). Salinity (S) and temperature (T) series collected from three oceanographic stations (http://www.sidimar.tutelamare.it, Fig. 4) were used to define the effects of water column vertical stratification on circulation and transport/mixing processes. In particular, starting from 6 years of measurements (time-series 2001e2007), typical series of monthly S and T vertical profiles (assumed spatially uniform) at sea open boundary were obtained by averaging single months from different years. Although the hydrodynamic circulation can be influenced by density gradient, the system does not include freshwater inputs from rivers because they flow into Augusta Harbour for only a short time of the year and with highly localized hydrodynamic effects (Lisi et al., 2009). Moreover, no evaporation or precipitation was included.

363

2.2.1.1. Dredging source. In this study, the sediment source was defined taking into account: (i) the estimation of dredged sediment concentration; (ii) the definition of vertical spill through the water column depending on dredge typology; (iii) the definition of the grain size distribution of released sediment moving out from the dredging zone into the far field; and (iv) the dredge position during dredging sessions. Accordingly, several parameters are relevant for numerically reproducing the three-dimensionality of the sediment plume arising from hydraulic and mechanical dredges and their dispersion throughout the entire water column (Collins, 1995; Kim and Lim, 2009; Bai et al., 2003). Sediment-related parameters were based on available field data while reasonable technical project assumptions were based on literature. When comparing mechanical and hydraulic dredging, the model assumes equal volumes of sediment removed for each scenario. In particular, the same “unit volume to be dredged” was chosen as reference unit for any dredging scenario possibly seen as a multiple of this reference activity. Following literature data (e.g. John et al., 2000; Hayes and Wu, 2001), the imposed production rate was 800 m3/cycle for the mechanical dredge and 3,000 m3/cycle for the hydraulic dredge (Table 1) for three dredging sessions, so the daily production rate became 2,400 m3 and 9,000 m3 respectively for mechanical and hydraulic dredge. Choosing a dredging period of approximately one month for the mechanical dredge and one week for the hydraulic dredge, a unit volume of about 70,000 m3 was derived. A dredging depth of 1 m was assumed based on available soft substrate thickness data. An hypothetical area of about 70,000 m2 (28 consecutive cells of about 2,500 m2 each, see map in Fig. 4) was identified choosing from areas with almost constant bathymetry (15e20 m, Fig. 4) and homogeneous sediment particle size (~70% of fine sediments; ISPRA, 2008). A sediment spill was imposed equal to 5% of the volume to be dredged, both for mechanical and hydraulic dredges. This estimate closely matches estimates from many previous studies (Pennekamp et al., 1996; John et al., 2000). For the mechanical dredge, the sediment release of dredged material was considered constant throughout the whole water column, from the seabed to the surface. For the hydraulic dredge, the sediment source is expected to remain confined around the dredging head (spillage layer) for the starting phases of the plume development (Collins, 1995; Palermo and Averett, 2003; Palermo et al., 2008), and the dredging release is expected to be relatively constant in the case of homogeneous materials to be removed (HR Wallingford Ltd & Dredging Research Ltd, 2003). Hydraulic and mechanical dredges complete the unit work within almost 30 and 8 days respectively (Table 1). An extra-time of 3 days is added to the end to take into account the sedimentation of most of the suspended sediment. Different dredging simulations, with different climatic conditions and subsequent hydrodynamics along the entire year, start with a lap time of one week for a total amount of 48 simulations (Fig. 5). It is worth noting that the dredging activity in this study is considered only in the bed excavation phase. It is known that overflow during the loading phases (into the dredge or into barges), if allowed, can greatly affect the sediment release and need to be considered for assessing environmental effects, especially in the dredging of contaminated mud (HR Wallingford Ltd & Dredging Research Ltd, 2003). In order to compare numerical results, a regular grid (step 250 m) of discrete check points has been identified inside the domain (Fig. 4, 408 check points). For each simulation and for each check point, time series of the current speed, SSC, and DEP output parameters have been extracted for different vertical levels: the bed layer, at ~25% from the bottom, at ~75% from the bottom and

364

A. Feola et al. / Journal of Environmental Management 166 (2016) 357e373

Table 1 Parameters for dredging implementation e Unit dredging activity of 70,000 m3 of sediments. Dredge

Mechanical

Hydraulic

Dredged volume per dredging session (ds: 8 h) [m3/ds] Number of dredging sessions per day Dredged volume per day [m3/d] Dredging time [d] Unit volume to be dredged [m3]

800 3 2,400 ~30 70,000

3,000 3 9,000 ~8

Sediment density [kg/m3] % fine sediment (