A robust, spatially explicit model for identifying oyster restoration sites: case studies on the Atlantic and Gulf Coasts.
KEY WORDS: eastern oyster, Crassostrea virginica, habitat suitability modeling, spatially explicit, geographic information systems, habitat suitability index
Eastern oyster (Crassostrea virginica) reefs are essential components of estuarine ecosystems along the Atlantic and Gulf Coasts of North America, and they provide numerous ecosystem services, including water quality improvements (Newell et al. 2002, Kellogg et al. 2013), landscape diversity (Eggleston 1999), storm surge protection (Meyer et al. 1997, Piazza et al. 2005), and habitat for reef-dwelling and benthic communities (Coen et al. 1999, Posey et al. 1999, Tolley & Volety 2005), among others (Powers et al. 2009, Harding et al. 2010). Reef abundance is currently at its nadir, estimated at 15% of historic levels worldwide (Beck et al. 2011). These declines have been attributed to overfishing, disease and predation, and large-scale human-mediated events (e.g., freshwater diversions). Given the importance of oysters in the estuarine community, significant resources have been dedicated to restoring oyster reefs.
Successful reef restoration depends on choosing sites that sustain reefs over long periods of time (Pollack et al. 2012). Restoration sites should be chosen so they optimize survival (i.e., mitigate mortality factors), which requires an understanding of the complex interactions between oysters and their environment. Often, ecosystem restoration projects are scheduled for locations that have not been well studied and have limited data available, yet time and budget constraints prevent long-term field studies and analysis. Therefore, modeling is often used to determine the best locations for restoration activities. Simplified modeling approaches such as habitat suitability index (HSI) models can provide natural resource managers with a standardized approach for habitat mapping and restoration planning, and have been used extensively by resource agencies for planning and impact assessments for wildlife management, and water resource and ecosystem restoration projects (Brooks 1997, Roloff & Kernohan 1999, Duberstein et al. 2007, U.S. Fish and Wildlife Service 1981). One such example is an HSI developed for the eastern oyster habitat as detailed in Louisiana's Comprehensive Master Plan for a Sustainable Coast (Coastal Protection & Restoration Authority 2012). Although the methods in the plan were designed specifically to assess the impacts of coastal protection and restoration projects on oyster habitat, the overall model approach is considered useful for a broad range of oyster-related restoration efforts, and was adapted for use in this study (Soniat 2012). Briefly, HSI models consist of a priori hypotheses that represent the critical relationships between a species and the environmental parameters that affect species mortalities and distributions (Tirpak et al. 2009). These hypotheses are translated into relative assessments of habitat suitability (scaled from 0-1, representing worst to best habitat, respectively) for a particular species based on its species-specific habitat associations. Suitability scores are then combined into a composite score, also scaled from 0-1, that represents the overall quality of a location for particular species and, therefore, for restoration efforts (U.S. Fish and Wildlife Service 1981). Habitat suitability index models were developed initially to assess habitat quality based on field measurements of habitat attributes extrapolated across large areas (e.g., a forest stand, a management unit), and advances in geographic information systems (GIS) and remote sensing have allowed the application of HSI models at a variety of spatial scales and extents to meet specific management objectives. HSI models can be incorporated into a GIS in a spatially explicit framework that can reduce uncertainty associated with trial-and-error approaches and can provide standardized, broadly applicable methods (Curnutt et al. 2000, Store & Kangas 2001).
There are several benefits to using an HSI approach. These models can be constructed rapidly and can be developed with a variety of data types, including scientific literature, field studies, modeling results, monitoring data, and/or expert opinion, giving resource managers flexibility when time and budget constraints prevent long-term field data collection. The pliancy of data inputs allows data of different types to be used in HSI models; however, applying a model parameterized with lower fidelity data limits the extent to which the model can be considered reliable. That is, if there is a lot of uncertainty associated with a particular component of the model, then that uncertainty can affect model results and limit its applicability. These models are also designed for portability and can be used among many different sites rather than be restricted to specific locations, as is often the case with complex ecological models (Soniat & Brody 1988). Conversely, HSI approaches have been criticized for their lack of scientific rigor and reliability (Cole & Smith 1983, Roloff & Kemohan 1999). Recent improvements in data quality as well as more rigorous evaluation criteria have improved the reliability of these approaches (Brooks 1997).
The objectives of this study were (1) to develop a spatially explicit, flexible HSI model that can be used to determine locations appropriate for restoration of Crassostrea virginica reefs throughout the western Atlantic and Gulf coasts and (2) to apply the model in 2 study areas--1 in the Chesapeake Bay, a data rich environment, and the other in the western Mississippi Sound (northern Gulf of Mexico), a data poor environment--and discuss the implications of using data of varying quality when applying the model for restoration.
MATERIALS AND METHODS
The Oyster Habitat Suitability Index Model (OHSIM) is designed as a spatially explicit, grid-based model that uses a series of linear equations to calculate habitat suitability for restoration of Crassostrea virginica. The model presented here is a modification of that of Soniat's (2012) and it follows the methodology established by Cake (1983) and Soniat and Brody (1988). The terminology and model evaluation techniques were adapted from Pollack et al. (2012). The model is composed of 4 variables, with each being assigned a dimensionless oyster suitability index (OSI) value that represents the relationship between an environmental variable and a stage of the oyster's life history. Each OSI is represented quantitatively as a series of linear suitability curves, with a minimum value of 0 for unsuitable to 1.0 for optimal habitats. Suitability curves are formulated as step-functions with linear approximations between each step. A restoration suitability index (RSI) is calculated as the geometric mean of the OSI values and represents the overall suitability of a particular location for restoration (Pollack et al. 2012). Data and equations are imported into a GIS and applied to specific geo-referenced locations.
The overarching assumption of the OHSIM is that substrate and salinity can describe quantitatively suitable oyster habitat for restoration. We adapted the model designed for Louisiana's Comprehensive Master Plan for a Sustainable Coast (Coastal Protection & Restoration Authority 2012) with the following modifications: (1) differences in data type, origin, spatial resolution, and content; (2) update to GIS methods, including the interpolation techniques; and (3) changes to 1 variable, such that we did not consider land building or conversion, and thus analyzed aquatic areas only. Suitable substrate (i.e., cultch) is described as the percentage of the bottom covered with hard substrate (e.g., oyster shell or other suitable bottom). Salinity is resolved into the following 3 variables that address different relationships between salinity and the oyster's life history: (1) mean salinity during the spawning season (MSSS), in which spawning and spat set have a greater optimal salinity than for survival of adults; (2) annual mean salinity (AS), which is the expected range over which adult oysters are viable; and (3) minimum annual salinity (MAS), which defines the impacts of high-mortality events resulting from lower salinities resulting from freshwater influxes (Soniat 2012).
The model is designed to be flexible with regard to data input and spatial scales and can take input from hydrodynamic models, monitoring stations, scientific literature, and expert opinion. Cell size and spatial extent can vary, but the spatial extent must be large enough to include both suitable and unsuitable habitats for the model to be verifiable (Brooks 1997). One limitation for input data is that a value must be available for each cell within the spatial domain. The model has a wide variety of potential application to any engineering or restoration activity that modifies salinity or substrate, including changes in freshwater inflow (e.g., freshwater diversions or any hydrological modifications that alter salinity), reef creation, land building that replaces oyster bottoms with other habitat, and sediment additions that cover suitable cultch.
Percent cultch is the percent of bottom covered with hard substrate. Oyster larvae require a hard substrate, such as existing oyster reefs (cultch) or other hard surfaces (e.g., limestone, concrete, granite, and so forth), on which to settle and metamorphose. Cake (1983) considered a high-quality bottom to be one in which 50% or more of the area is hard substrate, although no indication was given of the specific spatial scale over which the variable is to be applied. Although the relationship between percent cultch and its OSI is understood at the extremes (i.e., no substrate is unsuitable and 100% coverage is ideal), there is considerable uncertainty in the intermediate range. Cake (1983) considered the relationship between percent cultch and OSI to be linear, from 0%-50% cultch, and ideal (OSI = 1.0) when percent cultch was greater than 50%. We modified Cake's formulation by using the most parsimonious solution and assumed that oyster habitat suitability increases linearly from 0%-100% cultch cover (Eq (1); Fig. 1A).
[OSI.sub.%Cultch] = 0.01 x (% Cultch) (1)
The model was applied to 2 study areas with different types of available benthic habitat data (e.g., Chesapeake Bay and Mississippi Sound, northern Gulf of Mexico). This choice also provides a comparison between Atlantic and Gulf Coast habitats.
Mean salinity during the spawning season represents the mean monthly salinity from May through September, which is the spawning season for Crassostrea virginica. Mean salinity during the spawning season was calculated by averaging daily values of salinity from May 1 through September 30. Mean salinity during the spawning season reflects the greater optimal salinities required for spawning and larval stages (Butler 1953, Cake 1983). The relationship between MSSS and its OSI is formulated as a linear step-function (Fig. 1B). Breakpoints in the step-functions were determined by field validation of Cake (1983) by Soniat and Brody (1988). Values between the steps were interpolated linearly, and OSI values for MSSS were calculated as follows:
MSSS [less than or equal to] 5 or MSSS > 40 [OSI.sub.MSSS] = 0 (2)
5 < MSSS [less than or equal to] 10 [OSI.sub.MSSS] = -0.3 + (0.06 x MSSS) (3)
10 < MSSS [less than or equal to] 15 [OSI.sub.MSSS] = -0.4 + (0.07 x MSSS) (4)
15 < MSSS < 18 [OSI.sub.MSSS] = -1.1 + (0.1167 x MSSS) (5)
18 [less than or equal to] MSSS < 18 [OSI.sub.MSSS] = 1 (6)
22 < MSSS [less than or equal to] 30 [OSI.sub.MSSS] = 2.925 - (0.0875 x MSSS) (7)
30 < MSSS [less than or equal to] 35 [OSI.sub.MSSS] = 1.5 - (0.04 x MSSS) (8)
35 < MSSS [less than or equal to] 40 [OSI.sub.MSSS] = 0.8 - (0.02 x MSSS) (9)
Minimum annual salinity is the minimum value of the 12 monthly mean salinities. This variable is essential to describe freshwater impacts (e.g., freshets, high rainfall years, or freshwater diversions) on oysters and is analogous to the frequency of the killing floods variable used by Cake (1983). Low salinity has a greater negative impact in the summer than in the winter; however, the model does not include a temperature parameter. This could be included easily if month was to serve as a surrogate for salinity, which would require 2 relationships to describe the effect of minimal salinity (1 for the summer months and 1 for the winter months). The relationship between MAS and OSI does not represent any potential positive benefits of increased freshwater, such as reducing predators and disease (Butler 1953, Gunter 1979, LaPeyre et al. 2009). The relationship between MAS and its OS1 is formulated as a linear step-function (Fig. 1C). Breakpoints in the step-functions were determined by the field validation of Cake (1983) by Soniat and Brody (1988). Values between the steps were interpolated linearly and OSI values for MAS were calculated as follows
MAS [less than or equal to] 2 [OSI.sub.MAS] = 0 (10)
2 < MAS [less than or equal to] 4 [OSI.sub.MAS] = -0.05+ (0.025 x MAS) (11)
4 < MAS [less than or equal to] 6 [OSI.sub.MAS] = -0.85+ (0.225 x MAS) (12)
6 < MAS [less than or equal to] 8 [OSI.sub.MAs] = -1 + (0.25 x MAS) (13)
8 < MAS [OSI.sub.MAS] = 1 (14)
Annual mean salinity represents the range of salinities over which adult oysters are viable (Gunter 1955, Calabrese & Davis 1970, Castagna & Chanley 1973, Cake 1983, Chatry et al. 1983). Annual mean salinity is an annual representation of Cake's (1983) historical mean salinity, and was calculated by averaging mean monthly salinity values. The relationship between AS and its OSI follows that of Soniat and Brody (1988), with the exception that the optimum AS in the current model is a range (10-15) and not a discrete point (12.5). The relationship between AS and its OSI is formulated as a linear step-function (Fig. 1D). Breakpoints in the step-functions were determined by field validation of Cake (1983) by Soniat and Brody (1988). Values between the steps were interpolated linearly. OSI values for AS were calculated as follows:
AS [less than or equal to] 5 or AS > 40 [OSI.sub.AS] = 0 (15)
5 < AS [less than or equal to] 10 [OSI.sub.AS] = -1 + (0.2 x AS) (16)
10 < AS [less than or equal to] 15 [OSI.sub.AS] = 1 (17)
15 < AS [less than or equal to] 20 [OSI.sub.As] = -2.2 - (0.08 x AS) (18)
20 < AS [less than or equal to] 25 [OSI.sub.As] = 2 - (0.07 x AS) (19)
25 < AS [less than or equal to] 30 [OSI.sub.AS] = 1 - (0.03 x AS) (20)
30 < AS [less than or equal to] 40 [OSI.sub.AS] = 0.4 - (0.01 x AS) (21)
The RSI is determined as the geometric mean of the OSI values for the 4 component variables (Pollack et al. 2012). If any component OSI is 0 (unsuitable). RSI is 0 (poor-quality habitat). The RSI is calculated as
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (22)
where [OSI.sub.i] represents the OSI value per cell for each environmental variable i, and n represents the number of variables included in the model. Restoration suitability index results were categorized as 0-0.25 (low), 0.25-0.55 (low/medium), 0.55-0.85 (medium/high), and 0.85-1 (high), similar to the categories described by Soniat and Brody (1988) and Brooks (1997).
To determine how sensitive RSI values were to the inclusion of model parameters, a sensitivity analysis was conducted. The sensitivity analysis shows the percent change in RSI value from a 3-parameter model scenario to the inclusive OHSIM 4-parameter model. More specifically, RSI values were calculated for each possible combination of 3-OSI values (e.g.. 1 OSI value removed), and then percent change was calculated from the 3-RSI value to the inclusive OHSIM, 4-RSI value, which reflects the relative importance of each parameter to the model structure. This method is similar to that developed by Pollack et al. (2012), but it considers how the model responds when variables are added to the model rather than removed; the overall interpretation remains the same.
Spatial Data and OHSIM Application
The equations discussed in the previous section were applied in a GIS to a subset of spatial data variables (percent cultch and salinity) to compute an overall RSI. We selected 2 areas for case studies to illustrate the application of the OHSIM: a data-rich area (Chesapeake Bay) and a data-poor area (western Mississippi Sound, northern Gulf of Mexico). By evaluation of both cases, the goal is to illustrate how the OHSIM can be used regardless of origin, condition, or type of input data available. Only the preprocessing of the geospatial data differs in the 2 case scenarios; the application of the equations remains the same. Although the level of granularity differs in the results of the 2 case scenarios (as a direct result of the input data types), the approach is consistent for both areas, yielding examples of the range of results that can be achieved. The following sections describe the application of the OHSIM in 2 case scenarios. All data were processed in ESRI's ArcGIS 10.0/10.1 software (ArcInfo).
The Chesapeake Bay is a well-studied system that is rich in digital data and oyster resources; therefore, it represents an ideal study area to conduct habitat suitability analysis using high-fidelity, oyster-related geospatial data. For the purposes of this study, we chose an 871-[km.sup.2] area along the Lower Rappahannock River (Fig. 2), because this area is among a handful of project sites in the Bay in which detailed seafloor conditions were mapped to produce detailed benthic habitat maps, has had several high-resolution hydrodynamic models applied to it, and has a well-studied oyster fishery. For this case study, we used the National Oceanic and Atmospheric Administration's (NOAA's) integrated benthic characterization database to calculate percent cultch and results from hydrodynamic simulations using the curvilinear-grid hydrodynamics in three-dimensions (CH3D [Kim 2013]) model to derive MSSS, MAS, and AS. We applied OF1SIM using 3 y of CH3D data to determine temporal variability in oyster habitat as it relates to changes in salinity.
For the percent cultch variable, we used data from NOAA's integrated benthic characterization database (NOAA 2013), which consists of detailed side-scan sonar, acoustic surveys, sediment grab samples, and historical data sets, including mainstem sediment polygons, Maryland Bay Bottom Survey polygons, and Virginia Oyster Ground Survey polygons (i.e., the Baylor survey grounds) (Fig. 3). All data were clipped to the study area in Figure 2 and reprojected to UTM Zone 18 North NAD 1983.
To prepare the seabed classification data for the OHSIM, only faunal and man-made reef hard bottoms (mollusc class in Fig. 3) were selected from the integrated data set because mud, sand, and other soft bottoms are not suitable for oyster growth. To generate values for the percent cultch variable, the hard bottom layer was combined with the CFI3D grid cell layer. First, area was computed for the CH3D grid cell layer (grid cells are not uniform in size and shape, but the total area was 871.4 [km.sup.2]) in the attribute table, and then the hard bottom and CFI3D layers were unioned. A new field of percent coverage was created, illustrating the percent hard bottom coverage in each grid cell. Last, a new attribute field was added in which Eq (1) was applied to each cell within the spatial and temporal domains, resulting in 870 OS1 values for percent cultch.
Salinity variables were extracted from hydrodynamic simulations of the Lower Rappahannock River, using the CH3D model, which is a 3-dimensional, finite-difference hydrodynamic model that uses a horizontal curvilinear grid and a vertical z-grid to calculate temporally varying water levels and 3-dimensional velocity, temperature, and salinity (Kim 2013). Annual model runs were archived and calibrated for 8 y, between 1993 and 2000, and bottom salinities were extracted from 870 grid cells for OHSIM application. To evaluate how the OHSIM performed under different environmental conditions, 3 y (1997 to 1999) were selected from the data set--representing average, wet, and dry rainfall conditions, respectively--and providing the opportunity to evaluate a broad range of conditions and their potential influence on oyster suitability.
Bottom salinity values were processed so that they corresponded to the OHSIM salinity variables (AS, MSSS, and MAS). The values were included in the simulation result polygon layer attribute table--3 values per year, resulting in a total of 9 salinity values per grid cell (example of AS bottom values for 1997 are shown in Fig. 4). Therefore, 9 new attribute fields were added to calculate and apply the series of salinity suitability equations (Eqs 2-21). Using the field calculator, the appropriate equation was applied to each grid cell for each year (e.g., Eq (3) was applied to all cells with MSSS values between 5 and 10), resulting in a total of 7,830 salinity-based OSI values for the entire spatial and temporal domains of the Chesapeake Bay case study.
When all the OSI values were computed in the 2 polygon layer attribute tables, the 2 layers were combined using a union function for the application of the RSI equation (Eq (22)). The unioned layer file combined all the attributes, and a new RSI field (1 for each year. 1997 to 1999) was added. The field calculator was used to populate RSI values for each grid cell for each year. The RSI equation was also applied to a salinity-only-based model (i.e., percent cultch removed) to determine the sensitivity of the model results and to illustrate more fully the change and influence of the broad range of salinity conditions over the 3 y.
Gulf of Mexico
In contrast to the Chesapeake Bay, many areas do not have ideal geospatial data resources, such as archived, high-resolution hydrodynamic model simulations and detailed seabed classifications, and thus it is important to address how the OHSIM can be applied under such conditions. The Gulf of Mexico, although rich in oyster resources, does not have detailed seabed or salinity data; therefore, it represents a good example of how to make use of different data types that are more coarse in spatial resolution. The OHSIM was applied to a 942-knr area in the western Mississippi Sound.
Data from the Oyster Reef Mapping Project, collected in 2005, were used to assess the condition of oyster reefs after hurricane Katrina and were generated by the Mississippi Department of Marine Resources and NOAA's National Coastal Data Development Center (NCDDC). Briefly, seafloor samples were collected following predetermined transects and were recorded as a range of different bottom types (e.g., soft mud, shell, and so on). The data were provided directly from NCDDC as a GIS point file. The following designations were considered suitable for the percent cultch variable: live oysters, scattered live oysters, and shell or hash (Fig. 5). The extracted point data were interpolated to a grid surface to illustrate continuous coverage of suitable bottom conditions in a gridded system. The output grid cell size selected was 100 m (the default grid cell resolution was 90 m, which was rounded up). The resultant grid was converted to a polygon layer for integration with the salinity variables for RSI calculations.
For the salinity variables, data from NOAA's National Oceanographic Data Center (NODC) were obtained online (http:// www.node.noaa.gov/OC5/regional_climate/GOMclimatology/). More specifically, the Gulf of Mexico Regional Climatology data includes a set of mean fields at 1[degrees], 0.25[degrees], and 0.10[degrees] resolutions for temperature, salinity, oxygen, phosphate, silicate, and nitrate. Statistical mean values for surface salinity were downloaded in GIS point file format for 0.10[degrees] and 0.25[degrees] for the winter, spring, summer, and autumn seasons (surface values were used in this case study, because numerical values were consistently missing at other depth levels). Statistical mean values are defined as the average of all unflagged interpolated values at each standard depth level for each variable in a given resolution grid cell (i.e., 0.10[degrees]) containing at least 1 measurement for a particular variable (refer to http://www.nodc.noaa.gov/OC5/regional_climate/ GOMclimatology/ for more detailed information). In the western Mississippi Sound, 15 points from both the 0.10[degrees] and 0.25[degrees] data sets were used to interpolate the 3 salinity parameters. These points also helped determine the extent of the study area from which salinity could be interpolated, including the lower part of Bay St. Louis (Fig. 6). The data from the 0.10[degrees] data set were used primarily in the analysis because they had the greatest spatial resolution. However, in some cases, missing values were obtained from nearby points in the 0.25[degrees] data set. Because neither data set contained monthly values, the lowest value of the seasonal minima was selected for MAS. For MSSS, the mean was calculated from the spring and summer values, whereas for AS, the mean was calculated from all 4 seasonal values. All 3 point data sets were interpolated to a gridded surface using the salinity values and an output cell size of 100 m, matching the grid cell size of the percent cultch layer. Figure 6 shows the spatial extent of the study area and AS interpolation results. The results were converted to polygon layers and unioned into 1 overall salinity variable layer. Eqs 2-21 were applied using the field calculator, resulting in 3 salinity OSI fields.
To compare the cultch and salinity layers, the study area extent file was edited so that only water grid cells were analyzed. This was accomplished by creating a land/water mask that was digitized using current aerial photography. Water areas were delineated and then a 100-m-grid cell fishnet was overlaid to place grid cells in the study area (aligned with the salinity layer grid cells). The gridded study area layer was used to mask the salinity variable layer to ensure the same extent and matching cell size for all data and reprojected to UTM Zone 16 North, NAD 1983. In addition, each grid cell was assigned a percent cultch value--not present in a grid cell (0%) or covering the entire grid cell (100%)--as a result of the interpolation of suitable bottom points to a grid layer. Then, the percent cultch OSI field was created using the field calculator to apply Eq (1) and, last, the RSI equation was applied combining the 4 OSI values in each grid cell.
In the Chesapeake Bay, salinity conditions varied during the 3 y period. In 1997, MSSS ranged from 3.6-20.7, MAS ranged from 0.4-16.9, and AS ranged from 2.6-20.1. In 1998, MSSS ranged from 2.5-19.2, MAS ranged from 0-16.2, and AS ranged from 3.2-19.5. In 1999, MSSS ranged from 7.65-22.1, MAS ranged from 0.4-19.8, and AS ranged from 5.6-22.0. Although the salinities varied during the 3-y period, the general trend shows greater salinities in the eastern part of the study area, nearer to the central part of the Bay, whereas lower salinities and/or more variable salinity conditions were observed in the western part of the study area, making up part of the Lower Rappahannock River, and thus were more influenced by freshwater pulses (Fig. 4). Suitable cultch conditions were estimated for 50.9 [km.sup.2] of 871 [km.sup.2], or approximately 6% of the study area. These conditions were found primarily from the middle to lower reach of the Rappahannock River within the study area boundary (Fig. 3). Restoration suitability index values were calculated for 3 y of data (1997 to 1999), illustrating conditions for average, wet, and dry years, respectively, in the inclusive OHSIM as well as a salinity-only-based model (e.g., percent cultch value removed), as shown in Figure 7. Restoration suitability index values ranged from 0-1, which is necessary to distinguish suitability among sites (Brooks 1997). Tables 1 and 2 summarize RSI statistics for both the inclusive OHSIM (4-RSI) and the salinity-only-based model (3-RSI).
The greatest RSI values generally occurred in the western part of the study area, in the middle to lower reach of the Rappahannock River, corresponding to areas with the most suitable cultch conditions (Fig. 7). The lowest RSI values occurred in the eastern part of the study area, nearer to the central part of the Bay. Although salinity conditions were often suitable in the eastern part of the study area, it ranked lower as a result of unsuitable cultch conditions that were not found in deeper waters. Average salinity conditions in 1997 resulted in 2.5 [km.sup.2] of high suitability, 110 [km.sup.2] of medium to high suitability, 165 [km.sup.2] of low to medium suitability, and 595 [km.sup.2] of low suitability in the inclusive OHSIM results, or 4-RSI (Table 1). The year 1998 was considered a wet year in terms of rainfall conditions, and thus salinity conditions were less ideal and RSI values tended toward lower suitability than in 1997. As a result, no area was found to be highly suitable in 1998 (Table 1), and the mean 3-RSI (salinity-only model) value decreased from 0.80 to 0.73 (Table 2), illustrating that rainfall and/or freshwater influences resulted in less area with suitable salinity. Therefore, lower suitability classes increased in area in 1998, with 4-RSI values increasing to 604 [km.sup.2] of low suitability and 182 [km.sup.2] of low to medium suitability. In contrast, 1999 proved to be a dry year in terms of rainfall, resulting in more area with suitable salinities compared with 1997. For example, the mean 3-RSI value increased from 0.80 in 1997 to 0.84 in 1999 (Table 2), and resulted in an increase in high-suitability area from 2.5 [km.sup.2] in 1997 to 9.4 [km.sup.2] in 1999 (Table 1).
The sensitivity analysis indicated that percent cultch has the most influence on model results (Fig. 8), which was expected given the linear relationship between percent cultch and OSI. Furthermore, not only did most RSI values decrease when the percent cultch OSI value was added, but they decreased by 90% or more, illustrating that salinity conditions were highly suitable in a given grid cell in the salinity-only-based model; without suitable cultch, the value decreased significantly. In contrast, the model was less sensitive to a particular salinity parameter, with much of the habitat in the eastern part of the study area (closer to the central part of the Bay) having a slight decrease or little/no change in RSI value (blue cells) when any salinity OSI value was added to the model (Fig. 8C, D).
Gulf of Mexico
In the western Mississippi Sound, Gulf of Mexico, salinity conditions were as follows: MSSS ranged from 6.0-29.0, MAS ranged from 4.8-26.0, and AS ranged from 6.9-31.2. In general, lower salinities and/or higher salinity variability were observed along the shoreline and in the lower part of Bay St. Louis, which is closer to inlets and other freshwater sources (Fig. 6). Salinities increased moving away from the shoreline, with the highest salinities occurring in the southeastern and eastern parts of the study area toward the central part of the Gulf of Mexico. Suitable cultch conditions were estimated for 83.7 [km.sup.2] of 942 [km.sup.2], or approximately 9% of the study area. In general, these conditions extended from the lower part of Bay St. Louis to a concentrated area south of Pass Christian as well as a few small, scattered areas in other parts of the study area (Fig. 5). Restoration suitability index values ranged from 0-1, which is necessary to distinguish suitability among sites (Brooks 1997). Restoration suitability index values were calculated for the inclusive OHSIM (4-RSI) and the salinity-only-based model (3-RSI, percent cultch value removed), as shown in Figure 9A and B, respectively. Tables 3 and 4 summarize RSI statistics for both the inclusive OHSIM (4-RSI) and the salinity-only--based model (3-RSI).
The greatest RSI values generally occurred in an area extending from the lower part of Bay St. Louis to a concentrated area south of Pass Christian, and corresponded to areas with the most suitable cultch conditions (Fig. 9). Low RSI values occurred throughout the study area, especially in the east. Although salinity conditions were often suitable throughout the study area, many areas ranked low as a result of unsuitable cultch conditions that were either not found in deeper waters or not found along some parts of the shoreline. Most of the study area had low suitability (more than 90%; Table 3). In addition, less than 1 % had low to medium suitability, 4.1 % had medium to high suitability, and almost 5% had high suitability in the inclusive OHSIM (4-RSI) results. Table 3 also reports areas for the salinity-only-based model (3-RSI), in which less than 1 % had low suitability, 26.4% had low to medium suitability, 71.2% had medium to high suitability, and 2.1 % had high suitability. Minimum and maximum RSI values were similar for the 2 models; however, given the large number of low-suitability RSI values in the OHSIM 4-RSI result, the mean value was only 0.1, whereas in the salinity-only-based model 3-RSI result, it was 0.63 (Table 4). The majority of highly suitable conditions for both cultch and salinity was concentrated in an area offshore and south of Pass Christian (Figs. 5, 6, and 9) and corresponded to the location of known commercial oyster reefs (Fig. 10).
As with the Chesapeake Bay, a sensitivity analysis was conducted for the western Mississippi Sound case study. The same approach was used, whereby the analysis shows the percent change in RSI value from a 3-parameter model scenario to the inclusive OHSIM, 4-parameter model, illustrating the sensitivity of the model to each parameter (Fig. 11). Similar to the Chesapeake Bay case study, much of the area had favorable salinity conditions; however, with the addition of the percent cultch OSI value, most RSI values decreased (Fig. 11A). This is especially true in the areas close to the shoreline, where salinity-only-based model 3-RSI values ranked high (Fig. 9B); but, because of the lack of suitable cultch, the RSI values decreased by more than 70% (Fig. 11 A). The influence of percent cultch is also illustrated in Table 3, in which more than 70% of the study area had medium to high or high suitability in the salinity-only-based model 3-RSI results, decreasing to less than 10% area with medium to high or high suitability in the inclusive OHSIM 4-RSI results. The exception to the decreasing trend was in areas where suitable cultch conditions existed (Fig. 11A), and some values increased by as much as 30%.
Much like the sensitivity analysis for the Chesapeake Bay, RSI values in the Gulf of Mexico were not as sensitive to the addition of the salinity parameters (Fig. 11C, D). For example, when MSSS or AS were added to the model, the majority of the cells showed minimal percent change (-3 to 2%; Fig. 11C, D). When MAS was added to the model, some RSI values increased in areas with suitable cultch (2%-10%). This is also illustrated in a few areas near the mouth of Bay St. Louis when AS was added to the model (Fig. 11D), although areas with suitable cultch farther offshore experienced a decrease in RSI value (by as much as 35%). The reverse trend was shown when MSSS was added to the model, whereby areas with suitable cultch near the mouth of Bay St. Louis decreased in RSI value (by as much as 54%) and areas farther offshore increased (2%-10%).
For agencies faced with the task of restoring oyster populations, choosing sites that sustain reefs under dynamic environmental conditions is essential. Often, natural resource managers are not afforded the luxury of long-term field studies that can reduce myriad uncertainties associated with site selection. The application of integrated HSI-GIS approaches provides a standardized, flexible, and rapid approach that managers can use to reduce the uncertainty associated with the trial-and-error of site selection (Pollack et al. 2012). In this study, we developed a generalized OSI model that determined suitable habitat for oyster restoration based on 3 salinity variables and suitable substrate. The OHSIM is a simplified version of the one developed by Soniat (2012). Our goal was to create a model that could be developed rapidly using available data and then be applied throughout the Atlantic and Gulf coasts. We considered salinity and substrate only, because these parameters capture the critical relationships among environmental factors and the oyster's life history. Model results showed that the OHSIM captured general trends in oyster habitat suitability. During wet years, oysters are impacted negatively by being exposed to lower salinities (Hofmann et al. 1994, Dekshenieks et al. 2000), which is reflected in the 1998 results from the Chesapeake Bay case study (Fig. 7E). In contrast, moderate years (1997 and 1999) were more suitable. One trend not captured is the effect of extreme salinities. The available data never experienced those extremes, but the phenomenon is represented in the equations for MSSS and AS, and would be reflected in RSI values under those conditions.
Salinity is a recognized driver for oyster dynamics (Gunter 1955, Kennedy et al. 1996) and our parameterization captured the critical aspects of that relationship, with the optimal range of salinities for each OSIsaimity being in mesohaline conditions, which facilitates oyster growth in disease-prone waters (Carnegie & Burreson 2011, Levinton et al. 2011). Restoration suitability index values were not extremely sensitive to changes in any salinity variable compared with percent cultch. There were, however, differences in how the model responded between Chesapeake Bay and the Gulf of Mexico. Chesapeake Bay was more sensitive to MAS, indicating that available habitat in the Bay was more dependent on freshwater dynamics. The Gulf of Mexico was more sensitive to MSSS, indicating that available habitat was more dependent on summertime salinities.
The presence of hard substrate (represented as percent cultch in this study) has been included in some oyster HSI models (Cake 1983, Soniat & Brody 1988, Soniat 2012, the current study), but not others (Barnes et al. 2007, Pollack et al. 2012). Our results indicated that the OHSIM is highly sensitive to percent cultch. When it was added, the overall amount of suitable habitat was reduced (i.e., RSI values decreased; Figs. 8A and 11 A), because of the number of cells that did not have any hard substrate. This is a direct result of the equation that was used to parameterize percent cultch, which stated that no hard substrate resulted in an OSI value of 0. Without suitable substrate, oyster larvae cannot settle and, therefore, our parameterization seems reasonable. By including percent cultch as a variable, areas that did not have hard substrate, but were otherwise suitable, received an RSI score of 0, effectively removing these areas from consideration for restoration. For projects that plan on restoring oyster reefs in areas where they do not currently exist, simulated, geo-referenced reef polygons would need to be added to the model to determine RSI values accurately for those locations. The type of data used to parameterize the percent cultch data layer impacts inferences made from this model. For example, if percent cultch is parameterized with bottom layer data consisting only of existing shell beds/oyster reefs, then oyster suitability is determined by where oysters already exists (with other areas receiving RSI scores of 0). Conversely, if the percent cultch data layer consists of other types of hard substrate where oysters do not currently exist, or if polygons are created to represent where hard substrate could be installed, then the RSI scores for those locations would be more reflective of that location's potential for successful restoration. Given the confounding nature of this variable, it is important to quantify its impact by exploring the parameter space thoroughly through sensitivity analysis as well as by running a version of the model without percent cultch included. In cases when benthic habitat characterization data are available and can be incorporated easily into the HSI framework, it is reasonable to include this variable to examine inclusive RSI values. Future research should explore the functional form of the OSI-percent cultch relationship as well as considering weighing [OSI.sub.percent cultch] differently in the RSI calculation.
To determine how robust the OHSIM was to data input, we applied it to 2 regions that had different data resources. Chesapeake Bay is a well-studied system, and salinity values from high-resolution hydrodynamic codes (CH3D) and percent cultch values from detailed seabed classifications were used in the OHSIM. One of the limitations of using hydrodynamic modeling results is that model runs may exist only for historical time periods. For example, in this study, model runs existed from 1993 to 2000, but for demonstration purposes only data from 1997, 1998, and 1999 were used in the OHSIM. Other years were not available and it was cost prohibitive to run the model for current years; however, the 3 selected years illustrated the effects of wet (1998), dry (1999), and average (1997) conditions on the overall RSI.
The Gulf of Mexico did not have high-resolution hydrodynamic model data or detailed seabed classifications available, so surface salinity values were interpolated from mean salinities (points) from NOAA's NODC, whereas percent cultch was interpolated from seafloor sample (points) from the Oyster Reef Mapping Project. Bottom salinity values are generally more appropriate for quantifying oyster suitability, because oysters are found on the seafloor. However, we were trying to determine how robust the OHSIM was when nonideal data were available. For the Gulf of Mexico case study, we were able to obtain a pseudo-independent data set (the commercial reef data) that provided a metric for model validation (Fig. 10) and allowed comparison of model results to existing oyster abundance (as described in Tirpak et al. ). Results from this case study indicated that surface salinity and percent cultch interpolations provided a good indicator of oyster suitability based on our evaluation. Nonetheless, it is important to note that validating an oyster HSI that is parameterized with a percent cultch variable will often result in positive validation because oysters were likely already present in locations with hard substrates and suitable salinities.
The OHSIM represents a generalized model for determining locations suitable for oyster restoration throughout the Atlantic and Gulf coasts, and it provides a scientifically based support tool for natural resource managers and project planners. It was designed intentionally to include only the minimum factors required for oyster suitability--namely, substrate and salinity. Given the complexities of restoring reefs and sustaining them over long periods of time, other local conditions may influence reef sustainability, which should likewise be considered for determining restoration potential. For example, Pollack et al. (2012) determined temperature and turbidity were important in the Mission Aransas estuary in the Texas Gulf, and Barnes et al. (2007) determined that the number of high flow days (>4,000 cfs) per month were important (although our AS variable could serve as a surrogate for that variable). Other potential factors that might impact restoration include substrate firmness and stability, slope of shorelines for intertidal reef restoration, and disease prevalence and intensity. The OHSIM is flexible enough that other variables can be integrated easily into the framework, and local conditions should be considered before using the OHSIM exclusively. The sensitivity analysis also illustrated the importance of evaluating quantitatively the relationship between model inputs, equations, and results for all HSI models. Future work should include additional exploration and refinement of the quantitative relationship between percent cultch and overall RSI values.
We thank C. Cerco. M. Kjelland, T. W. Clumpkin, and the anonymous reviewers for providing thoughtful, constructive comments.
Barnes, T. K., A. K. Volety, K. Chartier, F. J. Mazzotti & L. Pearlstine. 2007. A habitat suitability index model for the eastern oyster (Crassostrea virginica), a tool for restoration of the Caloosahatchee Estuary, Florida. J. Shellfish Res. 26:949-959.
Beck, M. W., R. D. Brumbaugh, L. Airoldi, A. Carranza, L. D. Coen, C. Crawford, O. Defeo, G. J. Edgar, B. Hancock, M. C. Kay. H. S. Lenihan, M. W. Luckenbach, C. L. Toropova, G. F. Zhang & X. M. Guo. 2011. Oyster reefs at risk and recommendations for conservation, restoration, and management. Bioscience 61:107-116.
Brooks, R. P. 1997. Improving habitat suitability index models. Wildl. Soc. Bull. 25:163-167.
Butler, P. A. 1953. The southern oyster drill. Proc. Natl. Shellfish. Assoc. 44:67-75.
Cake, E. W. 1983. Habitat suitability index models: Gulf of Mexico American Oyster. Ocean Springs, MS: U.S. Fish and Wildlife Service. 37 pp.
Calabrese, A. & H. S. Davis. 1970. Tolerances and requirements of bivalve molluscs. Helgoland. Wissen. Meeresunt. 20:553-564.
Carnegie, R. B. & E. M. Burreson. 2011. Declining impact of an introduced pathogen: Haplosporidium nelsoni in the oyster Crassostrea virginica in Chesapeake Bay. Mar. Ecol. Prog. Ser. 432:1-15.
Castagna, M. & P. Chanley. 1973. Salinity tolerance of some marine bivalves from inshore and estuarine environments in Virginia on the western mid-Atlantic coast. Malacologia 12:47-96.
Chatry, M., R. J. Dugas & K. A. Easley. 1983. Optimum salinity regime for oyster production on Louisiana's state seed grounds. Contrib. Mar. Sci. 26:81-94.
Coastal Protection & Restoration Authority. 2012. Louisiana's Comprehensive Master Plan for a Sustainable Coast. Baton Rouge, LA: Coastal Protection and Restoration Authority. Available at: http:// coastalmasterplan.la.gov.
Coen, L. D., M. W. Luckenbach & D. L. Breitburg. 1999. The role of oyster reefs as essential fish habitat: a review of current knowledge and some new perspectives. Proc. Amer. Fish. Soc. Symp. 22:438-454.
Cole, C. A. & R. L. Smith. 1983. Habitat suitability indices for monitoring wildlife populations: an evaluation. Trans. N. Amer. Wildl. Nat. Res. 48:367-375.
Curnutt, J. L., J. Comiskey, M. P. Nott & L. J. Gross. 2000. Landscape-based spatially explicit species index models for Everglades restoration. Eco!. Appl. 10:1849-1860.
Dekshenieks, M., E. Hofmann, J. Klinck & E. Powell. 2000. Quantifying the effects of environmental change on an oyster population: a modeling study. Estuaries Coasts 23:593-610.
Duberstein, C. A., M. A. Simmons, M. R. Sackschewsky & J. M. Becker. 2007. Development of a habitat suitability index model for the sage sparrow on the Hanford site. Richland, WA: Pacific Northwest National Laboratory. 25 pp.
Eggleston, D. B. 1999. Application of landscape ecological principles to oyster reef habitat restoration. In: M. W. Luckenbach, R. Mann & J. A. Weston, editors. Oyster reef habitat restoration: a synopsis and synthesis of approaches. Gloucester Point, VA: Virginia Institute of Marine Science Press, pp. 213-227.
Gunter, G. 1955. Mortality of oysters and abundance of certain associates as related to salinity. Ecology 36:601-605.
Gunter, G. 1979. Studies of the southern oyster borer, Thais haemostoma. Gulf Res. Rep. 6:249-260.
Harding, J. M., R. Mann, M. J. Southworth & J. A. Wesson. 2010. Management of the Piankatank River, Virginia, in support of oyster (Crassostrea virginica, Gmelin 1791) fishery repletion. J. Shellfish Res. 29:867-888.
Hofmann, E. E., J. M. Klinck, E. N. Powell, S. Boyles & M. S. Ellis. 1994. Modeling oyster populations II: adult size and reproductive effort. J. Shellfish Res. 13:165-182.
Kellogg, M. L., J. C. Cornwell, M. S. Owens & K. T. Paynter. 2013. Denitrification and nutrient assimilation on a restored oyster reef. Mar. Ecol. Prog. Ser. 480:1-19.
Kennedy, V. S., R. I. E. Newell & A. F. Eble, editors. 1996. The eastern oyster Crassostrea virginica. College Park, MD: Maryland Sea Grant. 734 pp.
Kim, S.- C. 2013. Evaluation of a three-dimensional hydrodynamic model applied to Chesapeake Bay through long-term simulation of transport processes. J. Am. Water Resour. Assoc. 49:1078-1090.
La Peyre, M. K., B. Gossman & J. F. La Peyre. 2009. Defining optimal freshwater flow for oyster production: effects of freshet rate and magnitude of change and duration on eastern oysters and Perkinsus marinus infection. Estuaries Coasts 32:522-534.
Levinton, J., M. Doall, D. Ralston, A. Starke & B. Allam. 2011. Climate change, precipitation and impacts on an estuarine refuge from disease. PLoS One 6:e18849.
Meyer. D. L., E. C. Townsend & G. W. Thayer. 1997. Stabilization and erosion control value of oyster cultch for intertidal marsh. Restor. Ecol. 5:93-99.
Newell, R. I. E., .1. C. Cornwell & M. S. Owens. 2002. Influence of simulated bivalve biodeposition and microphyrobenthos on sediment nitrogen dynamics: a laboratory study. Limnol. Oceanogr. 47:1367-1379.
NOAA. 2013. Lower Rappahannock River, VA, benthic habitat characterization. Available at: http://chesapeakebay.noaa.gov/ acoustic-seafloor-mapping/lower-rappahannock-river-va-benthichabitat-characterization.
Piazza, B. P., P. D. Banks & M. K. La Peyre. 2005. The potential for created oyster shell reefs as a sustainable shoreline protection strategy in Louisiana. Restor. Ecol. 13:499-506.
Pollack, J. B., A. Cleveland. T. A. Palmer, A. S. Reisinger & P. A. Montagna. 2012. A restoration suitability index model for the eastern oyster (Crassostrea virginica) in the Mission-Aransas estuary, TX, USA. PLoS One 7:e40839.
Posey, M., T. D. Alphin, C. M. Powell & E. Townsend. 1999. Oyster reefs as habitat for fish and decapods. In: M. W. Luckenbach, R. Mann & J. A. Weston, editors. Oyster reef habitat restoration: a synopsis and synthesis of approaches. Gloucester Point, VA: Virginia Institute of Marine Science Press, pp. 229-238.
Powers, S. P., C. H. Peterson, J. H. Grabowski & H. S. Lenihan. 2009. Evaluating the success of constructed oyster reefs in no-harvest sanctuaries: implications for restoration. Mar. Ecol. Prog. Ser. 389:159-170.
Roloff, G. J. & B. .1. Kernohan. 1999. Evaluating reliability of habitat suitability index models. Wildl. Soc. Bull. 27:973-985.
Soniat, T. M. 2012. Eastern oyster habitat suitability index. In: Louisiana's Comprehensive Master Plan for a Sustainable Coast. Baton Rouge, LA: Coastal Protection and Restoration Authority. Appendix D-13. Available at: http://www.coastalmasterplan. louisiana.gov/2012-master-plan/master-plan-appendices/.
Soniat, T. M. & M. S. Brody. 1988. Field validation of a habitat suitability index model for the American oyster. Estuaries 11:87-95.
Store, R. & J. Kangas. 2001. Integrating spatial and multi-criteria evaluation and expert knowledge for GIS-based habitat suitability modeling. Landsc. Urban Plan. 55:79-93.
Tirpak, J. M., D. Jones Farrand, F. R. Thompson, III, D. J. Twedt, C. K. Baxter, J. A. Fitzgerald & W. B. Uihlein, III. 2009. Assessing ecoregional scale habitat suitability index models for priority landbirds. J. Wildl. Manage. 73:1307-1315.
Tolley, S. G. & A. K. Volety. 2005. The role of oysters in habitat use of oyster reefs by resident fishes and decapod crustaceans. J. Shellfish Res. 24:1007-1012.
U.S. Fish and Wildlife Service. 1981. Standards for the development of habitat suitability index models. Washington, DC: Division of Ecological Services, U.S. Fish and Wildlife Service, Department of the Interior. 103 pp.
TODD M. SWANNACK, (1,2) * MOLLY REIF (1,3) AND THOMAS M. SONIAT (4)
(1) U.S. Army Engineer Research and Development Center, 3909 Halls Ferry Road, Vicksburg MS 39180; (2) Department of Biology, Texas State University, 601 University Drive, San Marcos, TX 78666; (3) U.S. Army Research and Development Center, Joint Airborne Lidar Bathymetry Technical Center of Expertise, 7225 Stennis Airport Road, Suite 100, Kiln, MS 39556; (4) Oyster Research Laboratory, Department of Biological Sciences and Pontchartrain Institute for Environmental Science, University of New Orleans, Lakefront Campus, 2000 Lakeshore Drive, New Orleans, LA 70148
* Corresponding author. E-mail: email@example.com
TABLE 1. RSI area statistics for the inclusive OHSIM (4-RSI) and the salinity- only-based model (3-RSI) results (1997 to 1999) for the Lower Rappahannock River in the Chesapeake Bay. Area ([km.sup.2]) RSI/suitability 4-RS1 1997 4-RS1 1998 4-RSI 1999 0-0.25/low 594.5 603.9 592.9 0.25-0.55/ 164.7 181.8 159.2 low-Medium 0.55-0.85/ 109.7 85.6 109.8 medium-high 0.85-1.0/high 2.5 0 9.4 Total area 871.4 ([km.sup.2]) Area ([km.sup.2]) RSI/suitability 3-RSI 1997 3-RSI 1998 3-RSI 1999 0-0.25/low 34.7 56.8 11.1 0.25-0.55/ 19.2 27.2 11.5 low-Medium 0.55-0.85/ 221.3 644.1 312.4 medium-high 0.85-1.0/high 596.2 143.2 536.4 Total area ([km.sup.2]) TABLE 2. RSI summary statistics for the inclusive OHSIM (4-RSI) and the salinity-only-based model (3-RSI) results (1997 to 1999) for the Lower Rappahannock River in the Chesapeake Bay. Year 4-RSI min 3-RSI min 4-RSI max 3-RSI max 1997 0.00 0.00 0.86 0.92 1998 0.00 0.00 0.79 0.89 1999 0.00 0.00 0.92 0.93 Year 4-RSI mean 3-RSI mean 4-RSI SD 3-RSI SD 1997 0.17 0.80 0.25 0.20 1998 0.15 0.73 0.23 0.23 1999 0.17 0.84 0.27 0.12 TABLE 3. RSI area statistics for the inclusive OHSIM (4-RSI) and the salinity-only-hased model (3-RSI) results for the Gulf of Mexico. Area ([km.sup.2]) 4-RSI 3-RSI RSI/suitability 4-RSI area (%) 3-RSI area (%) 0-0.25/low 858.1 91.11 2.38 0.25 0.25-0.55/low-medium 0.12 0.01 249 26.44 0.55-0.85/medium-high 38.6 4.10 670.7 71.21 0.85 1/high 45 4.78 19.75 2.10 Total area ([km.sup.2]) 941.8 100.00 100.00 TABLE 4. RSI summary statistics for the inclusive OHSIM (4-RSI) and the salinity-only-based model (3-RSI) results for the Gulf of Mexico. 4-RSI min 3-RSI min 4-RSI max 3-RSI max 4-RSI mean 0 0.17 0.89 0.86 0.1 4-RSI min 3-RSI mean 4-RSI SD 3-RSI SD 0 0.63 0.24 0.14
|Printer friendly Cite/link Email Feedback|
|Author:||Swannack, Todd M.; Reif, Molly; Soniat, Thomas M.|
|Publication:||Journal of Shellfish Research|
|Date:||Aug 1, 2014|
|Previous Article:||Estimating sustainable harvests of eastern oysters, Crassostrea virginica.|
|Next Article:||Depuration of bacterial populations in the Indian backwater oyster Crassostrea madrasensis (Preston, 1916): effects on surface and bottom held...|