Dredging Operations and Environmental Research Program: Building Tools for Objective Determination of Environmental Windows Douglas Clarke1 , Jerald Ault 2 , Deborah French3 , and Billy Johnson 4 Abstract: Environmental windows have become a prevalent dredging project management practice. Frequently decisions regarding windows are made in the absence of conclusive evidence that dredging does or does not pose a significant threat to specific resources. Windows are often based on a risk-averse approach, which in many cases may be over-restrictive. Given the host of factors that influence dredging- related risk, resource and dredging project managers have been handicapped by a lack of tools and methods for evaluation of the need for protective measures. Recent efforts within the Dredging Operations and Environmental Research (DOER) Program have sought to fill this gap along several fronts: development of models designed to simulate dredging processes; development of enhanced technologies for characterization of dredging- induced perturbations; and development of databases that provide access to relevant data. Among the modeling tools are SSFATE, SSDOSE, and FISHFATE, models that respectively simulate sediment resuspension and transport/dispersion, calculate exposures of aquatic organisms to sediment plumes, and estimate the population dynamics consequences of hydraulic entrainment. The ultimate goal of these development efforts is provision of a complete “toolbox” for use by biological resource and dredging project managers, whereby environmental windows can be evaluated in terms of need and effectiveness. These tools can also be used proactively to predict the likely effects of various dredging project scenarios, such that if windows are justified, their dates can be determined to provide optimal resource protection. SSFATE is a flexible model built on a GIS platform that simulates the fate of sediment resuspended by hopper, cutterhead, and bucket dredges. The user customizes the dredging scenario to accommodate best available knowledge of the dredging mode of operation, in situ sediment characteristics, and local bathymetry and flow fields. Output includes particle tracking plume animation in tidal and non-tidal situations and time series plots of suspended sediment concentration at any location in the model domain. In SSDOSE, the user can calculate the hypothetical exposure of various organisms (e.g., sessile bottom invertebrates, passively drifting plankton, and adult fishes with directed movement) as they encounter plumes generated in SSFATE. With FISHFATE, the user can place estimated rates of mortality due to hydraulic entrainment into context with other sources of mortality acting upon a given stock, and predict the short and long-term consequences of multiple dredging project scenarios. Each model is capable of examining alternative dredging practices and providing insights into risk minimization via optimal windows. ________________________ 1 U.S. Army Engineer Research and Development Center, 3909 Halls Ferry Road, Vicksburg, MS 39180, USA.
[email protected] 2 The Ault Group, Inc., Key Biscayne, FL 3 Applied Science Associates, Narragansett, RI 4 Computational Hydraulics and Transport, Edwards, MS
INTRODUCTION Environmental windows, temporal restrictions placed on the conduct of dredging operations, represent a striking example of the difficulty inherent in achieving a balance between biological resource protection and cost-effective navigation dredging. Environmental windows (periods when dredging and dredged material disposal is allowed), also known as seasonal restrictions (when dredging is curtailed) or time-of-year constraints, are applied to a large majority of Federal navigation dredging projects in the United States (Reine et al. 1998). Frequently, decisions regarding windows are made in the absence of conclusive evidence that dredging does or does not pose a significant threat to specific resources. Hence, windows are often based on a risk-averse approach, which in many cases may be over-restrictive. Because many sets of dates for starting and ending dredging windows are fixed, i.e. predetermined and retained over the course of many years and multiple dredging cycles, they may not accommodate inter-annual variation in factors that govern biological processes. For example, in a given waterway, dates that water temperatures rise above or decline below certain thresholds due to mild or severe winters may shift by weeks or months from year to year. If a critical temperaturedriven process such as fish spawning occurs during a non-typical year, the intended protection of a fixed window, unless it has been established using very broad criteria, may be lost. Likewise, the timeline for planning, coordination, contracting, mobilization, and execution of a dredging project is inherently inflexible. In most cases, short-term adjustments to the timing of a dredging project based on known conditions at the project site would be impractical and cost-prohibitive. It is not surprising then, that environmental windows have emerged as a controversial topic. Given the host of factors that influence dredging-related risk (e.g., type and size of dredge plant, waterway geomorphology, hydrodynamic conditions, and life history stage and behavior of organisms of concern), determining the actual need for an environmental window is an exceedingly difficult challenge. Windows have been used as dredging project management practices since the early 1970s, following passage of the National Environmental Policy Act in 1969. In the interim, relatively little progress has been made toward integrating scientific and engineering knowledge such that decisions regarding the need for windows for individual dredging projects could be made with confidence. During the coordination process, these decisions are generally based on professional judgment at best or gut reaction at worst. Resource and dredging project managers have consistently been handicapped by a lack of tools and methods for evaluation of the relative need for protective measures. Decisions related to optimal timing of a dredging project to provide target resources with acceptable levels of protection are necessarily “up- front” in nature, i.e. the decisions must be made without benefit of real time monitoring data prior to initiation of dredging. The absolute number and frequency of Federal navigation dredging projects in most regions of the United States precludes the collection of comprehensive resource, physical, and dredging process characterization data for each project. Hence, the application of modeling tools represents one of few options
2
available to managers. Although numerous models have been developed to address dredging process-related needs (e.g., sediment transport models), until recently none had been developed specifically to support environmental windows decisions. A task to develop modeling tools designed for windows-related issues was initiated under the U.S. Army Corps of Engineers’ Dredging Operations and Environmental Research (DOER) Program in 1998. Based on the frequencies of particular concerns that are used to justify requests for windows, a prioritization of needed modeling tools was undertaken. Two persistent concerns that appeared to be amenable to model evaluations were identified: effects of suspended sediments on a variety of resources (Clarke and Wilber 2000; Wilber and Clarke 2001), and entrainment of aquatic organisms by hydraulic dredges (Reine and Clarke 1998). Resuspension of sediments is among the most widely cited concerns that lead to restricted dredging schedules. To assess the need for protective windows due to potential effects of suspended sediments, one would need to know the temporal and spatial scales of plumes associated with the dredging operation, either at the dredging or placement sites. One would also need to know the nature of the interaction between the target organis m and the plume. The degree of exposure of individuals or populations to the plume would then determine the probability of detrimental effect. Two models have been developed to address these tasks: Suspended Sediment FATE (SSFATE) to simulate plume dynamics, and Suspended Sediment DOSE (SSDOSE) to simulate interactions between aquatic organisms and plumes generated by SSFATE. Hydraulic entrainment is another frequently cited concern used to justify environmental windows. Questions related to entrainment by hydraulic cutterhead or hopper dredges have arisen concerning early life history stages of oysters and striped bass, Dungeness and blue crabs, juvenile sturgeon, and sea turtles, among other taxa (Reine and Clarke 1998). Various modeling approaches have previously been devised or employed (Carricker et al. 1986; Armstrong 1987; Burton et al. 1992; Wainright et al. 1992), but none of these were designed specifically to be capable of treating all aquatic organisms and a broad spectrum of dredging scena rios. Also, none were adapted to assist in determination of optimal windows for those scenarios. To satisfy this need FISHFATE has been developed. Each model is described in greater detail below. SSFATE Suspended Sediment FATE was developed jointly by the U.S. Army Engineer Research and Development Center (ERDC) and Applied Science Associates (ASA). SSFATE is a versatile computer modeling system with a number of features designed to assist in evaluation of the probabilities of detrimental impacts due to dredginginduced plumes (Swanson et al. 2000). The model operates on a GIS platform such that the model domain can be visualized as layers containing pertinent maps of topographic features or sensitive biological resources. Ambient currents can either be
3
imported from a numerical hydrodynamic model or drawn graphically using interpolation of limited field data. SSFATE uses Lagrangian particles (LPs) to track the movement of suspended sediment in the model domain. Each LP is characterized by its location, mass, and sediment size class. Particle advection is based on the simple relationship that a particle moves linearly with a local velocity, obtained from the hydrodynamic input, for each model time step. Particle diffusion is assumed to follow a simple random- walk process. A diffusion distance, defined as the square of the product of an input diffusion coefficient and the time step, is decomposed into x and y displacements via a random direction function. Particle settling is computed at the end of each time step, after the model has computed the concentration of each sediment size class and the total concentration on a concentration numerical grid. These computations result in the settling of the different grain size classes being interactive. The size of all concentration grid cells is the same, with the total number of cells increasing as the suspended sediment plume moves away from the source. Although the resolution of the spatial grid used by SSFATE can be adjusted, the model was designed to treat far- field processes in which the transport and turbulence associated with ambient currents dominate. After the settling velocity of each particle size class is computed, along with a depositional probability based on shear stress, the deposition of sediment from each size class in each bottom cell is computed and the calculation cycle is repeated. Thus the model computes transport, dispersion, and settling of suspended sediment released to the water column. Sediment source strength and vertical distribution from pipeline cutterhead, hopper, or bucket dredges can be computed internally or specified by the user. The fate of multiple sediment grain size fractions can be simulated simultaneously. Output consists of concentration contours in both horizontal and vertical planes (Figure 1). Time series plots of suspended sediment concentration can easily be generated for any cell within the model domain (Figure 2), as can the distribution of sediments deposited on the substrate. Animations of sediment particle movement and concentration evolution can readily be produced, a very useful capability for sharing results with resource agencies. SSDOSE Suspended Sediment DOSE was also developed by ASA in collaboration with ERDC. Once the spatial and temporal scales of suspended sediment plumes have been determined with an acceptable degree of confidence, the response of biota exposed to the plumes is a primary concern. In very few instances will monitoring data yield this type of information. Calculating hypothetical exposures would therefore provide insights into plume-organism interactions. Newcombe and Jensen (1996) and Wilber and Clarke (2001) have emphasized that effects are related to both suspended sediment concentration and duration of exposure. A suspended sediment dose can be defined as the sum over time of suspended sediment concentration times duration of that exposure (expressed in units of mg/l- hr).
4
Dose is related to the location of an organism relative to the suspended sediment plume. Thus, the pathway of the organism’s movement, if any, relative to the spatial dynamics of dredge- induced plumes needs to be simulated and the exposure along that pathway summed to quantify dose. This modeling problem has previously been addressed by ASA for analogous exposures of aquatic organisms to oil and chemical spills. In these applications (using ASA’s SIMAP model), toxic exposures are simulated by tracking the movements of organisms through the water and over time. Calculations of dose and mortality are then based on toxicity of the concentration and duration of exposure. Mortality is estimated as a percent of organisms in a specified area. The percent loss can then be converted to estimates of loss in terms of either numbers or biomass. This basic approach was employed to quantify dose of suspended sediment exposure. SSFATE creates output files that include scenario definitions and suspended sediment concentrations as a function of three-dimensional space and time. SSDOSE reads these files to calculate dose. At the end of each time step, the SSFATE plume of Lagrangian particles is contour mapped into a three dimensional grid sized to encompass the plume at that time. The grid is divided into 50 X 50 horizontal X 5 vertical cells of equal volume. The summed mass of particles falling in each cell, divided by the cell’s volume, yields concentration in the cell at that time. In SSDOSE the LPs are used to represent groups of organisms that move together within the model domain. Replicate LPs are used to represent different behavior patterns, i.e. where they occur and how they move (or remain stationary). Exposure of each LP is tracked over time to calculate dose, which is the sum over all time steps of suspended sediment concentration (mg/l) times the time interval (mg/l-hr). Within SSDOSE a variety of movement behaviors can be simulated. Categories include: passively drifting plankton (e.g., buoyant fish eggs), stationary on the bottom (e.g., demersal eggs), stationary in the bottom (e.g., benthos or plants, with exposure calculated as sediment loading or mass per unit area), pelagic non-directed movement (e.g., slow random movement of small fishes or rapid random movement of large fishes), non-directed movement just above the bottom (e.g., demersal fishes), pelagic directed movement (e.g., at a speed and path specified by the user, as could be used for anadromous fishes), and demersal directed movement (e.g., at a speed and path just above the bottom specified by the user, as could be used for certain crustaceans). Because habitat maps can be structured within the SIMAP habitat grid, movements of taxa categories can be restricted within specified portions of the habitat grid. This feature allows simulation of site tenacious species. Additional categories could be customized to reflect other movement behaviors. SSDOSE needs as input a model grid of the domain that defines the land-water boundary, habitat type, and water depth. The grid can be drawn and edited by the user, updated from GIS layers (e.g., referenced seagrass beds, oyster reefs), or created from ArcView GIS data in grid format. These gridding tools are imbedded in the SSFATE/ SSDOSE user interface.
5
In addition to dose, the model is capable of calculating a percent mortality if a concentration lethal to 50% of test individuals (LC50) is specified. The model can predict reductions in fish and shellfish catch in present and future years though linkage to conventional fisheries models. A linkage to FISHFATE may serve this purpose in the future. SSDOSE outputs include tabulations of dose for each movement behavior category and percent mortality by each behavior category. Also produced are animated maps of the locations of exposed organisms over time and locations of mortality over time (Figure 3). Both types of maps depict color-coded symbols representing magnitudes of dose and mortality. FISHFATE FISHFATE is an integrated suite of models developed by The Ault Group in collaboration with ERDC. Details of the model’s conceptual design and populationmodeling framework are given in Ault et al. (1998). In brief, FISHFATE uses estimates of natural mortality rates in tandem with fishing and entrainment- induced mortality rates acting on appropriate life history stages of target species to assess potential short and long-term consequences for that population. The model is built on a Unix platform in the C++ language to take advantage of its object-oriented capabilities. An underlying model called STOCAST (Spatial and Temporal Objectoriented Cohort-Structured) (Meester et al. 2001) provides the spatial structure for FISHFATE, which is also compatible with SSFATE. Ultimately, hydrodynamic grids and dredging scenarios built in SSFATE could be linked to FISHFATE. In FISHFATE, the population is simulated through a cohort structure comprised of a group of identical individuals (a cohort) where each cohort is uniquely identified by its time step of recruitment, spatial location at recruitment, and sex. The model contains fundamental population-dynamic processes of growth, mortality, and recruitment specific for the regional population of interest. Female and male cohorts are treated separately to allow for the easy calculation of egg production and to accommodate differential growth, mortality, spawning, and temporal sex changes when known. FISHFATE can be run to simulate any number of finite time steps, from minutes to years, and is capable of simulating fish, shellfish, or crustacean populations. Parameterization of FISHFATE is described in greater detail in Ault et al. (2000). To assess the effects of alternative dredging scenarios within a given year, the model uses differential mortalities due to entrainment during different environmental windows. The model is run using a daily time step for 20 years in order to allow the population to reach an equilibrium state, and the results of alternative windows compared in year 20. Natural mortality acts upon each cohort in each time step, either at a rate calculated by the model (see Ault et al. 2000) or specified by the user. Stochastic or deterministic growth effects on natural mortality can easily be incorporated. Fishing mortality can be integrated into FISHFATE in a spatially
6
explicit manner. Estimates of fishing effort, derived from a catchability coefficient and the total number of fishing vessels, is equally distrib uted among the stock. Dredging mortality estimates are based on measures of exposure to hydraulic entrainment, a factor of the time an organism will be in a dredge “search area.” Although almost all dredging scenarios will assume 100% mortality of organisms entrained, the model can simulate specified levels of survival if deemed appropriate. Multiple dredges and dredge types can be modeled, as well as moving dredges. At present the entrainment rate of a given species-dredge type must be estimated based on past studies. However, entrainment rates can be modified as additional knowledge of the influences of flows fields surrounding dragheads or cutterheads and dredge production rates becomes available. Based on specified input parameters, FISHFATE infers effects at the population level by calculating a probability that the perturbed population will fall below certain thresholds of population size or reproductive effort. This fundamental approach entails separation of natural variability from fishing and dredging-related effects, statistical analysis of the resultant data, determination of appropriate model structure, parameter estimation, modeling, and risk analysis. DISCUSSION In tandem, the models described above represent a greatly enhanced capability to examine dredging projects in as objective a manner as possible. Each model offers unique features that can be customized to simulate diverse dredging project scenarios in levels of detail heretofore unattainable. The models can potentia lly be linked to investigate multiple aspects of risk posed by dredging operations. To date SSFATE has been used to simulate dredging scenarios for the Providence River Dredging Project. Plumes generated by bucket dredges in several reaches of the project channel were simulated to assess potential impacts on several organisms in the Providence River and Narragansett Bay, Rhode Island. Results of the simulations formed the basis for predictions of impacts on early life history stages of winter flounder (Pleuronectes americanus) and other target species, and were incorporated in the Environmental Impact Statement for the project. This application was built using a “high end” hydrodynamic model to drive transport and dispersion of simulated sediment particles. Although a source of some debate during the interagency coordination process, the model’s output was an effective means for visualization of plume dynamics, and did contribute to rational discussions of alternative means to conduct the dredging wit hout resort to overly restrictive environmental windows. SSFATE has also been used by ERDC to simulate dredging projects in riverine and estuarine waterways where high-end hydrodynamic models were unavailable. In these cases the model serves as a screening tool. Using limited data on current structure, SSFATE can be effectively used to build simplistic flow fields to drive sediment particle transport and dispersion. For example, one set of scenarios was
7
constructed to generate first order examinations of plume movement in the vicinity of seagrass beds. The GIS platform of the model is well suited to display plumes, suspended sediment concentration gradients, and depositional patterns such that likely “hot spots” were identified. These types of data can be used not only to establish the need for protective measures, including windows, but also to assist in the design of monitoring plans. As is appropriate for all modeling tools, SSFATE requires actual field measurements of plume dynamics in order to be validated. At present ERDC is pursuing opportunities to gather validation data at actual dredging projects. One important validation exercise will be conducted in conjunction with the Providence River Dredging Project, scheduled to begin in the fall of 2002. SSDOSE is newly developed and has been applied only to hypothetical dredging scenarios. Current plans are to use the Providence River SSFATE scenarios to generate plumes, and then to expose hypothetical organisms to these plumes using SSDOSE. To the extent possible, these data will be validated by supplemental monitoring efforts. Clearly, some degree of confidence derived from validation of SSFATE is a prerequisite for full implementation of SSDOSE. Likewise, FISHFATE has been applied to hypothetical dredging scenarios. Applications are being pursued that will test the utility of the model. In addition to validation exercises, further refinements will be made to each model. For example, currently SSFATE computes and displays suspended sediment concentrations attributable to dredge- induced resuspension only, and does not display background or ambient conditions. A capability to superimpose dredge-related increments on a static or dynamic (e.g., tidally driven) set of ambient concentrations would enhance interpretation of output with respect to biological responses to total sediment exposures. This capability would obviously benefit SSDOSE computations as well. Interpretation of model output should always be done with full recognition of model assumptions and limitations. However, with due acknowledgement of inherent assumptions, model results can provide numerous insights into the likely scales of dredging- induced perturbations where otherwise decisions would necessarily be based on speculation and anecdote. Future refinements of the modeling tools described herein should improve their overall utility and ability to shed light on complex dredging issues. REFERENCES Armstrong, D. 1987. Model of Dredging Impact on Dungeness Crab in Grays Harbor, Washington. Fisheries Research Institute Report FRI-UW-8702, School of Fisheries, University of Washington, Seattle, WA. Ault, J. S., Lindeman, K. C., and Clarke, D. G. 1998. FISHFATE: Population Dynamics Models to Assess Risks of Hydraulic Entrainment by Dredges.
8
DOER Technical Notes Collection (TN-DOER-E4), U.S. Army Engineer Research and Development Center, Vicksburg, MS. www.wes. army.mil/el/dots/doer/pdf/doere4/pdf Ault, J. S., Meester, G. A., Lindeman, K. C., and Juo, J. 2000. FISHFATE Users Guide: Spacially Temporally Explicit Population Simulation Model. DOER Technical Notes Collection (TN-DOER-E11), U.S. Army Engineer Research and Development Center, Vicksburg, MS. www.wes.army.mil/el/dots/doer/pdf/doere11/pdf Burton, W., Weisberg, S., and Jacobson, P. 1992. Entrainment Effects of Maintenance Hydraulic Dredging in the Delaware River Estuary on Striped Bass Ichthyoplankton. Report submitted to the Delaware Basin Fish and Wildlife Management Cooperative, Trenton, NJ, by Versar, Inc. Carricker, M., LaSalle, M., Mann, R., and Pritchard, D. 1986. Entrainment of Oyster Larvae by Hydraulic Cutterhead Dredging Operations: Workshop Conclusions and Recommendations. American Malacological Bulletin, Special Edition 3: 71-74. Clarke, D. G., and Wilber, D. H. 2000. Assessment of Potential Impacts of Dredging Operations Due to Sediment Resuspension. DOER Technical Notes Collection (TN-DOER-E9), U.S. Army Engineer Research and Development Center, Vicksburg, MS. www.wes.army.mil/el/dots/doer/pdf/doere9/pdf Johnson, B. H., Andersen, E., Isaji, T., Teeter, A. M., and Clarke, D. G. 2000. Description of the SSFATE Numerical Modeling System. DOER Technical Notes Collection (TN-DOER-E10), U.S. Army Engineer Research and Development Center, Vicksburg, MS. www.wes.army.mil/el/dots/doer/pdf/ doere10/pdf Newcombe, C. P., and Jensen, J. O. T. 1996. Channel Suspended Sediment and Fisheries: A Synthesis for Quantitative Assessment of Risk and Impact. North American Journal of Fisheries Management 16(4): 693-727. Meester, G. A., Ault, J. S., Smith, S. G., and Mehrotra, A. 2001. An Integrated Simulation Modeling and Operations Research Approach to Spatial Management Decision Making. Sarsia, 86: 543-558. Reine, K., and Clarke, D. 1998. Entrainment by Hydraulic Dredges – A Review of Potential Impacts. DOER Technical Notes Collection (TN-DOER-E1), U.S. Army Engineer Research and Development Center, Vicksburg, MS. www.wes.army.mil/el/dots/doer/pdf/doere1/pdf Reine, K. J., Dickerson, D.D., and Clarke, D. G. 1998. Environmental Windows Associated with Dredging Operations. DOER Technical Notes Collection (TNDOER-E2), U. S. Army Engineer Research and Development Center, Vicksburg, MS. www.wes.army.mil/el/dots/doer/pdf/doere2/pdf Swanson, J. C., Isaji, T., Ward, M., Johnson, B. H., Teeter, A., and Clarke, D. G. 2000. Demonstration of the SSFATE Numerical Modeling System. DOER Technical Notes Collection (TN-DOER-E12), U.S. Army Engineer Research and Development Center, Vicksburg, MS. www.wes.army.mil/el/dots/doer/ pdf/doere12/pdf
9
Wainwright, T., Armstrong, D., Dinnel, P., Orensanz, J., and McGraw, K. 1992. Predicting Effects of Dredging on a Crab Population: An Equivalent Adult Loss Approach. Fishery Bulletin, 90:171-182 Wilber, D. H., and Clarke, D. G. 2001. Biological Effects of Suspended Sediments: A Review of Suspended Sediment Impacts on Fish and Shellfish with Relation to Dredging Activities in Estuaries. North American Journal of Fisheries Management, 21(4): 855-875.
10
Figure 1. An example of SSFATE output depicting a simulation of a suspended sediment plume in New York/New Jersey Harbor. Concentrations are indicated in the legend on the right side of the figure. The source (dredge) is indicated by a cross in the central portion of the plume.
11
Figure 2. Illustration of SSFATE time series output (inset) of concentration at a selected location and depth stratum within a simulated plume.
12
Figure 3. Example of SSDOSE output showing estimated exposures of small pelagic fishes to a hypothetical suspended sediment plume in New York/New Jersey Harbor. Color-coded dose concentrations are shown in the legend box.
13