Spatiotemporal variability of suitable habitat for American lobster (Homarus Americanus) in Long Island Sound.
KEY WORDS: lobster, habitat, suitability index, habitat suitability index model, Homarus americanus, Long Island Sound
The American lobster (Homarus americanus) is a benthic crustacean distributed throughout coastal Northwest Atlantic waters, most commonly from Newfoundland, Canada through North Carolina (Thunberg 2007). The species can be found in waters ranging from the intertidal zone to depths of up to 700 m, but tend to be most abundant in coastal waters shallower than 50 m (Lawton & Lavalli 1995, Van der Meeren et al. 2010). Coarse rocky substrates (cobbles and boulders) are the most common habitat, but lobsters can also be found on several other substrates including mud, sand base with rock, shell, eelgrass, or algae (Lawton & Lavalli 1995).
The American lobster is an ectothermic species with a specific preferred thermal range for optimum physiological functionality (Reynolds & Casterlin 1979). Water temperature has a significant impact on the physiology of juvenile and adult lobsters, especially in nonoptimal dissolved oxygen and salinity conditions (Mercaldo-Allen & Kuropat 1994). Lobsters have been found in waters of temperatures ranging from 0 to 25[degrees]C and with salinities ranging from 15-32 ppt, but lobsters prefer a thermal range between 12 and 18[degrees]C (Crossin et al. 1998) and salinities of 20-32 ppt (Harding 1992, Jury et al. 1994, ASMFC 2009). Lobsters use more energy for respiration in warmer water conditions leaving less energy for feeding, growth, immune response, and reproduction (Qadri et al. 2007). As water temperatures rise above 20[degrees]C, lobster show various physiological stress responses such as increased respiration rates and decrease in immunocompetence (Dove et al. 2005, Fogarty et al. 2007).
The spatial distribution of lobster is known to vary by sex, size, and season (Chang et al. 2010). American lobsters typically remain within a home range of about 5-15 [km.sup.2] (Lawton & Lavalli 1995). Large adults are more likely to be found in deeper, cooler waters, but migrate to shallow coastal waters seasonally to reproduce. Large mature lobsters in the Gulf of Maine (GOM) move inshore and into estuaries in spring (Watson et al. 1999) and often remain close to shore in the summer, then move back offshore in late fall to escape winter turbulence (Chen et al. 2006a). Small juvenile lobsters are more likely to be found inshore at depths of less than 10 m and do not make seasonal migrations offshore in winter (Cooper et al. 1975). These differences
in spatial distribution suggest size-specific responses to environmental variables such as bottom temperature and salinity (Jury et al. 1994, Factor 1995).
The American lobster fishery in the northeastern United States has experienced significant expansion in both effort and landings over the last 60 years (Berger 2014). Lobster landings were generally around 25 million pounds until the early 1950s, increasing to roughly 150 million pounds in 2012 (Berger 2014). Approximately 94% of total U.S. landings came from the GOM (Berger 2014). The 2009 benchmark stock assessment showed record high stock abundance and recruitment in the GOM and Georges Bank (GBK); however, the Southern New England (SNE) lobster stock was found to be in poor condition because of prolonged low abundance and persistently poor recruitment (ASMFC 2009). Changes in water temperature, salinity, and dissolved oxygen coupled with continued high fishing mortality have been identified as principal causes of low recruitment and poor stock condition (ASMFC 2009).
The lobster population in Long Island Sound (LIS) is a part of the Southern New England lobster stock. With ex-vessel values over $40 million, the LIS lobster fishery accounted for over 90% of the value of commercial landings in the region and remained the third largest lobster fishery in the United States until 1998 (Shields 2013). The LIS lobster stock has experienced a substantial decrease in abundance over the last 20 years because of deteriorating habitat and heavy exploitation (ASMFC 2009). Triggered by the major mortality event, possibly caused by stress from warm temperatures, pollutants and decreasing oxygen concentrations, landings declined by 89% in 1999 with cumulative landings from 1999 to 2010 only reaching 0.42 million pounds (CTDEP 2010). Epizootic shell disease has also become an increasing threat to the stability of the LIS lobster fishery (Bell et al. 2012, Castro et al. 2012). As climate change continues to alter Northwest Atlantic coastal ecosystems (Mills et al. 2013), the poorly adapted LIS lobster stock is under increasing stress caused by changes in suitable habitat availability. To illustrate the extent that habitat deterioration has influenced the American lobster's decline in LIS, it is necessary to quantify changes in suitable lobster habitat over time.
Habitat suitability index (HSI) models are widely used in wildlife management to describe the relations between species abundance and ecological variables (Franklin 2010, Chang et al. 2012, Morrison et al. 2012). An empirical HSI model is derived from observations of the species in the field, and reflects the impacts of multiple habitat variables given the input data (e.g., abundance index or relative biomass) (Dettki et al. 2003, Ahmadi-Nedushan et al. 2006, Tian et al. 2009, Chen et al. 2010). HSI models standardize habitat suitability a target species on a scale from 0 to 1, representing "least suitable" to "most suitable" habitat qualities respectively (Schamberger et al. 1982). Oftentimes evaluation of habitat suitability is based on a limited number of habitat variables that influence organism abundance and distribution. Therefore, HSI implies relative habitat quality rather than actual population levels (Jian et al. 2013). Habitat suitability index modeling results in combination with GIS provide an effective means of evaluating spatiotemporal variability in habitat conditions of a target species and produce habitat maps that can be used by managers and policymakers to make informed decisions (Terrell 1984, Bovee & Zuboy 1988). In fisheries management, the HSI model is often used to characterize fish habitat preference, availability, and quality (Morris & Ball 2006). For the lobster fishery, the HSI model can evaluate variability of suitable lobster habitat considering all key environmental variables for different life history stages.
The objective of this study is to develop an HSI model for evaluating the spatiotemporal variability of suitable habitats for LIS juvenile and adult lobsters in spring (April-June) and fall (September-October) from 1978 to 2012. The model is used to describe how the proportion and spatial trend of suitable habitat have changed over time. Finally, HSI model results were used to identify possible periods when lobster habitat conditions were extremely poor in LIS to determine whether habitat conditions have influenced the decline of lobsters in this region.
MATERIALS AND METHODS
The LIS is an estuary 181.9 km long and 33.8 km wide at its widest and covers approximately 3419 [km.sup.2] area (Fig. 1). The depth of LIS varies from 4.6 to 60.4 m, averaging 22.6 m. Salinity ranges from 23 ppt at the western end to 35 ppt at the eastern end (Gottschall 2013).
Fisheries-independent data tend to provide a better representation of species distribution and abundance than fisheries-dependent data as they are normally collected using standardized gear and sampling methods with a clearly defined spatiotemporal scale (Tian et al. 2009). Bottom trawl survey data collected from LIS by the Connecticut Department of Energy and Environmental Protection (CTDEP) from 1984 to 2012 were used to develop the HSI models in this study (Fig. 2).
The LIS trawl survey is a semiannual fishery-independent survey operated by the CTDEP. The survey encompasses an area from longitude 72[degrees] 03' (New London, CT) to longitude 73[degrees] 39' (Greenwich, CT), and includes both Connecticut and New York state waters from 5 to 46 m in depth over mud, sand, and transitional (mud/sand) substrate types. The survey is based on a stratified-random sampling design, and the survey area is divided into 1.85 x 3.7 [km.sup.2] sites assigned to 12 strata classified by depth (0-9 m, 9.1-18.2 m, 18.3-27.3 m, and 27.4 + m) and bottom substrate type (Gottschall & Pacileo 2013).
The survey was conducted in the spring, from April through June, and during the fall, from September through October, with 40 sites sampled monthly for a total of 200 sites annually. It was done during daylight hours with a 14 m otter trawl with a 51 mm codend sampling gear to reduce the sampling bias associated with diurnal variability in catchability (Sissenwine & Bowman 1978). Target tow duration was 30 min at 3.5 knots to cover a mean distance of 3241 m at each site (CTDEP 2013). At each site, tow date, tow location (latitude and longitude), tow duration, environmental variables (e.g., bottom temperature, bottom salinity, and depth), and biological information of the catch [e.g., carapace length (CL), weight, cull condition, and shell disease presence] were recorded (Gottschall 2013).
No information with regard to measure of area swept was available before 2012 (CTDEP 2012, 2013). The size specifications for the trawl net and associated gear remain unchanged as far as since 1992 (Reid et al. 1999). The standardized survey design allows for temporal comparisons of lobster catch and distribution. In this study, a total of 5353 tows that sampled 156,202 lobsters between fall 1984 and spring 2012 were analyzed. Lobster CL ranged between 16.1 and 112 mm and between 16 and 117 mm for the spring and fall surveys, respectively.
The Finite-Volume Community Ocean Model (FVCOM) was used to produce bottom temperature and bottom salinity estimates by depth, time, and location in LIS from 1978 to 2012. This regional coastal ocean circulation model was developed by University of Massachusetts Dartmouth-Woods Hole Oceanographic Institution joint efforts and is suited for forecasting and hindcasting the ecosystem dynamics for areas characterized by complex coastlines and intertidal zones (Chen et al. 2006b). In addition, data on distribution of surficial substrate (resolution: 0.00001 decimal degrees or 1.11 m) throughout LIS was obtained from the U.S. Geological Survey (Poppe & Seekins 2000). Bottom substrate types in LIS include gravel (pebbles defined as 2.00-64.00 mm, cobbles defined as 64-256 mm, boulder defined as above 256 mm), gravel-sand (0.62-2.00 mm), sand-clay (0.001-0.004 mm), silt (0.004-0.062 mm)/sand, sand-clay/silt, sand-silt/clay, and sand/silt/clay (Poppe et al. 2000). Bathymetry data were obtained from the U.S. Coastal Relief Model--Northeast Atlantic by the NGDC-NOAA (National Geophysical Data Center 1999).
Habitat Suitability Index Mode!
HSI is a numerical index based on suitability indices (SI) that can quantify the habitat conditions from 0 (least suitable habitat) to 1 (most suitable habitat) for key habitat variables. The SI can be calibrated from presence/absence data, presence only data, or using expert knowledge (Franklin 2010). Development of HSI model requires: (1) selection of habitat variables to include in the model, (2) development of SI for each habitat variable, and (3) combination of those SI via a mathematical equation to produce a composite HSI (Schamberger et al. 1982). Based on the literature on American lobster ecology and behavior (ASMFC 2009, Chang et al. 2010), the following four environmental variables were chosen for their potential influence on American lobster habitat: bottom temperature ([degrees]C), bottom salinity (ppt), depth (m), and bottom substrate type.
Data Analysis and Processing
Bottom trawl survey data for American lobster in LIS from 1982 to 2012 were used in this study. To depict behavioral difference throughout lobster life stage, the dataset was divided into two size classes, juveniles ([less than or equal to] 0 mm CL) and adults (>60 mm CL), as 60 mm represents the minimum size at maturity defined by Atlantic States Marine Fisheries Commission (ASMFC 2009). The spring and fall survey data were analyzed separately. This approach resulted in four groups of lobster (2 size classes x 2 seasons). Each lobster group was modeled independently.
The abundance index derived from LIS bottom trawl survey was considered a good indicator of lobster abundance in developing SI and HSI models in this study (Chang et al. 2010). The nominal abundance index, calculated as a survey catch per unit of sampling effort (CPUE) at sampling station i, in season j, and year y, was calculated as
[CPUE.sub.ijy] = ([Count.sub.ijy] / [Tow duration.sub.ijy]) * 20 (1)
where count is the total number of either adult or juvenile lobsters caught. Tow duration is towing time duration measured in minutes, which usually varied from 20 to 30 min but was standardized to 20 min at each sampling station.
The relationship between lobster CPUE and habitat variables from 1984 to 2012 was identified. For each habitat variable, a SI based on species abundance (CPUE) was first developed. The SI were estimated using a common approach known as the histogram method (Vinagre et al. 2006, Chen et al. 2010). The three continuous habitat variables (bottom temperature, bottom salinity, and depth) were delineated into 10 classes using Fisher's natural breaks classification method (Bivand 2013), whereas the categorical habitat variable bottom substrate was classified into seven substrate types (Poppe et al. 2000). For class k of habitat variable i in each lobster group, the average CPUE over all the sampling stations falling within the class was calculated as [CPUE.sub.i,k]. The SI value of class k for habitat variable i, [SI.sub.i,k], was then calculated on a scale of 0.0-1.0 using the following formula (Chang et al. 2012):
[SI.sub.i,k] = [CPUE.sub.i,k] - [CPUE.sub.i,min] / [CPUE.sub.i,max] - [CPUE.sub.i,min] (2)
where [CPUE.sub.i,min] and [CPUE.sub.i,max] are the minimum and maximum values of the average CPUEs of all the classes for habitat variable d. Thus, the SI for the most suitable class should have a value of 1, whereas the SI for the least suitable class should have a value of 0. An SI value was assigned to every class of the habitat variables in the form of a linear transfer function to qualitatively analyze the relationships between the habitat variable and lobster abundance. As a result, a total of 16 SI were calculated (i.e., four SI corresponding to the four environmental variables for four lobster groups including two seasons and two stages).
For the purpose of sensitivity analysis, the SI curves were first drawn by mean function, and then redrawn by trimmed mean function to remove any missing values and 5% of the highest and lowest scores (Tukey 1977, Crawley 2007). The suitable ranges were identified as area under both SI curves. The SI values derived from each habitat variable were then combined to form composite HSIs also scaled 0-1 and proportional to habitat quality. The following two empirical HSI models were developed in this study (Cooperrider et al. 1986) (Fig. 3):
1. Arithmetic mean model (AMM):
His = [n.summation over (i=1)] [SI.sub.i] / n (3)
2. Geometric mean model (GMM):
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4)
where [SI.sub.i] is a value of SI associated with the ith habitat variable and n is the number of habitat variables included in the HSI model.
A cross-validation approach was applied for evaluating the predictive performance of the HSI models. Models were developed independently for each lobster group (e.g., spring-adult, spring-juvenile, fall-adult, and fall-juvenile) using a randomly selected subset of data representing 80% of all the data, referred to as training data. The remaining 20% of the data, referred to as testing data, were set aside for the cross-validation to assess the predicting ability of models developed from training data (Smith 1994, Zuur et al. 2007). The predicted HSI values were compared against the observed HSI values. Linear regression analysis was performed on predicted versus observed HSI values, and the regression intercept, slope, [R.sup.2] value, and the Akaike information criterion (AIC) score were used to evaluate the predictive performance of the HSI model. An unbiased prediction should have an intercept parameter not significantly different from 0, a slope not significantly different from 1, and a high [R.sup.2]. In all, 100 rounds of cross validation were conducted using random selection in each round to obtain 100 sets of regression parameters. This validation process was conducted for both AMM and GMM HSI models to determine which model performed better.
Mapping HSI Values
The predicted HSI values were assigned to every FVCOM grid in LIS, and this procedure was conducted for every year that was available in the FVCOM dataset between 1978 and 2012. The ordinary Kriging method using the exponential semivariogram function was applied to create continuous HSI maps. The area with the highest class of HSI (e.g., 0.6-1.0) was designated as good habitat and correspondingly the area with the lowest HSI (e.g., 0-0.2) as poor habitat. The spatial distribution of median HSI values for a total of 35 years was mapped to observe the overall spatial trend in suitable habitat distribution for each group of lobsters. The median HSI maps were then compared with spatial trends in CPUE from the survey to verify the model. Finally, a median HSI value for each year was calculated with a fitted linear regression model to analyze whether there was any statistically significant trend in suitable habitat. The following R packages were used to implement this analysis; sp (Pebesma et al. 2014), maptools (Lewin-Koh et al. 2014), rgdal (Bivand et al. 2014), gstat (Pebesma & Graeler 2014), maps (Becker et al. 2014), and fields (Nychka et al. 2014).
The highest SI for bottom temperature differed by season and lobster size. The suitable thermal range (i.e., bottom temperature with SI > 0.8) for spring-adult lobsters was found to be 11.1-12.4[degrees]C (Fig. 4A), whereas the suitable thermal range for spring-juvenile lobsters was 8.45-9.55[degrees]C (Fig. 4B). In spring, the suitable depth range for adult and juvenile lobsters was similar at 31.9-37.2 m (Fig. 4C) and 31.9-37.3 m (Fig. 4D), respectively. The suitable salinity range for spring-adult lobsters was 21-23.7 ppt (Fig. 4E), whereas the spring-juvenile lobsters had a suitable salinity range of 21.9-24.4 ppt (Fig. 4F).
The suitable thermal range for fall-adults was 17.9-19.2[degrees]C (Fig. 5A), and 15.6-16.6[degrees]C as well as 17.4-18.4[degrees]C for fall-juveniles (Fig. 5B). The suitable depth range for fall-adult lobsters was between 14.8-17.9 m and 31.9-37.3 m (Fig. 5C), whereas the suitable depth range for fall-juvenile lobsters was 15-17.9 m (Fig. 5D). Higher suitable salinity ranges were observed in fall for both size groups: 27.6-28.1 ppt for adults (Fig. 5E) and 26.6-27.4 ppt and 29.4-30.4 ppt for juveniles (Fig. 5F).
Sand/silt/clay was found to be the substrate type with the highest SI values for adult lobsters in both seasons (Figs. 4G and 5G). For the spring-juvenile group, sand-silt/clay showed the highest SI values, whereas gravel appears to be the most suitable substrate type for fall-juveniles (Figs. 4H and 5H).
Model Validation and Selection
The GMM model showed intercepts ([alpha]) closer to 0, whereas the AMM model showed slopes ([beta]) closer to 1 in the regression between predicted and observed HSI values in cross validation; however, the AMM-HSI models showed higher [R.sup.2] values in all four groups. When compared with an ideal model without prediction bias (i.e., [alpha] = 0, [beta] = 1, and [R.sup.2] = 1), predictive performance for the spring-adult lobster was found to be the best ([alpha] = 0.145, [beta] = 0.732, [R.sup.2] = 0.611), and predictive performance for the spring-juvenile lobster was the poorest ([alpha] = 0.212, [beta] = 0.595, median [R.sup.2] = 0.456). The AMM model also predicted HSI values better, because AIC values were smaller in all modeling groups (Table 1). Finally, because of the nature of geometric mean algorithm, GMM-HSI model yielded a "0" HSI value when the model included an SI value of 0. Thus, the AMM can better distinguish subtle differences in areas of low SI values, and was determined to be more appropriate than the GMM for estimating lobster HSI in LIS.
Spatial and Temporal Variability in HSI Values
Using the AMM-HSI model, the spatial distribution of estimated HSI values in LIS was mapped for each lobster group. A visual examination of HSI maps revealed that the suitable habitats (i.e., HSI > 0.6) are concentrated in western-central LIS in spring for both size groups of lobster, but showed clustering patterns throughout LIS in fall (Fig. 6). The season- and size-specific distribution of suitable habitats predicted by the AMM-HSI model generally coincides with high and low lobster catches on the bottom trawl survey (Fig. 6). Temporal variation in proportion of suitable habitat was observed in all four groups over the 35-year time series (Fig. 7). There were no statistically significant temporal trends in suitable habitat change for adult lobsters in spring (P = 0.317) and fall (P = 0.609). For juvenile lobsters, a significant declining trend in proportion of suitable habitat was found in spring ([beta] = -0.003, P = 0.016), and a significant increasing trend was found in fall (P = 0. 002, P = 0.015) (Fig. 8).
In the HSI model, depth and substrate type are static components, whereas bottom temperature and bottom salinity are dynamic components; however, no relationships were observed between temporal trends in temperature, salinity, and availability of suitable habitat (Fig. 9). Possible periods of extremely poor habitat conditions (such as a year when both seasons had an HSI value above 0.6 in less than 10% of the studied area) were identified. For adult lobsters, poor habitat conditions were observed in 1979 to 1980, 1983, 1985, 1988 to 1989, 1994 to 1999, 2004 to 2005, 2007, and 2009. For juvenile lobsters, poor habitat conditions occurred in 1980, 1983, 1985, 1997, 1999, 2002, 2004 to 2005, and 201l (Fig. 10).
This study developed a modeling approach to analyze the spatiotemporal variability of suitable habitat as a function of bottom temperature, bottom salinity, depth, and substrate for lobster in LIS.
The SI results for bottom temperature, bottom salinity, and depth were consistent with past observations of preferential lobster habitat. The SI for bottom temperature identified different suitable thermal ranges between spring and fall for both size classes of lobsters. Adult lobsters showed a slightly higher and broader suitable thermal range compared with juveniles. In spring, the suitable thermal range for adults appeared to be considerably warmer than that of juveniles, whereas the difference in thermal range between the two size classes was less in fall. Overall, the suitable thermal range for lobsters in fall appears to be greater than that in spring. Water temperatures above the thermal threshold were avoided in both seasons (Crossin et al. 1998). Finally, two separate suitable thermal ranges were identified for juveniles in fall. These distinctive suitable thermal ranges between two seasons and two life history stages may be because of differences in behavioral thermoregulation as lobsters mature. It is generally accepted that lobster behavior is strongly regulated by water temperature (Crossin et al. 1998), and that the relationship between lobster density and bottom temperature is dome shaped with a peak somewhere between 14 and 19[degrees]C (Chang et al. 2010). In this study, the suitable thermal range (SI > 0.8) varied from 8.45 to 18.4[degrees]C. This is consistent with a previous study in the GOM, where lobster concentrations observed in areas with water temperature greater than 5[degrees]C in spring and 8[degrees]C in fall (Chang et al. 2010).
The SI for depth showed differences in suitable depth ranges across all groups. In fall, the suitable depth range for adult lobsters was between 14.8-17.9 m and 31.9-37.3 m, which might reflect a skewed spatial distribution by sex caused by migrations of ovigerous lobsters and sex-specific responses to different salinity ranges (ASMFC 2009, Chang et al. 2010). Contrary to previous findings suggesting that small juveniles are more likely to remain inshore at depths of less than 10 m (Cooper et al. 1975), both adult and juvenile lobsters in spring showed a deeper suitable depth range when compared with depths in fall. These findings agree with the in situ observations of seasonal lobster movements in Bonavista Bay, Newfoundland (Ennis 1984) and suggest possible seasonal shift in suitable depth ranges for both life stages of lobster.
Model results indicate a suitable salinity range of 21-30.4 ppt, which is consistent with lobster salinity tolerance observed in past studies of 20-32 ppt with occasional tolerance as low as 15 ppt (Harding 1992, Jury et al. 1994, ASMFC 2009). The bottom salinity SI identified relatively constant suitable salinity ranges between adults and juveniles; however, higher suitable salinity ranges were observed in fall for both age groups, which may indicate different salinity tolerance of lobsters under different thermal regimes (Mercaldo-Allen & Kuropat 1994, Ennis 1995). There were two separate suitable salinity ranges identified for fall juveniles. This may indicate differential distribution of male and female juveniles in the fall. Because females are more sensitive to low salinities, males are generally more dominant in inshore waters and females dominant in offshore waters (Jury et al. 1994, Chang et al. 2010).
This study identified sand/silt/clay as the most suitable bottom substrate type for adult lobsters in both spring and fall. Sand-silt/clay was the substrate with the highest SI for juveniles in spring, but was the least suitable substrate for juveniles in fall. Gravel yielded the highest SI for fall juveniles. These findings are relatively inconsistent with the previously documented postsettled lobster habitat preferences of shelter-providing rocky and boulder substrates (Barshaw & Bryant-Rich 1988, Wahle & Steneck 1991). Several factors have been identified as the potential source of this inconsistency. First, the U.S.G.S. substrate data used in the HSI models did not differentiate boulder or cobble substrate, because the category of gravel includes grain size greater than 2 mm in diameter (Poppe et al. 2000). This overgeneralization of bottom substrate type may have affected SI values derived from the histogram method and may have resulted in underrepresentation of bottom substrate as a habitat variable in the HSI model. Second, the CTDEP bottom trawl survey may have shown biased lobster density as rocky substrate could sometimes interrupt a tow (CTDEP 2013) and boulder, and rocky substrates are generally associated with lower trawl capture efficiency (Steneck & Wilson 2001). Whereas the lack of trawl survey data with certain key substrates cannot be quantified or ignored, these data-driven biases can be potentially corrected by the use of expert knowledge as quantitative assessment criteria (Store & Kangas 2001, Vincenzi et al. 2006, Vincenzi et al. 2007). Furthermore, although shelter-providing rocky/cobble/boulder substrates are generally considered to be preferred habitat by both newly settled and older lobsters (Cooper & Uzmann 1980, Steneck 2006), preference for specific substrates diminishes as lobsters grow out of the early benthic phase (Hudon 1987, Wahle & Steneck 1991). This inconsistency in substrate preference between lobster life stages was also observed in the GOM, where substrate type affected the probability of juveniles, but not adult, presence (Chang et al. 2010). Similarly, mud base (particle size < 0.06 mm) with burrows is known to be a preferred substrate by adult lobsters in inshore and in estuaries where lobsters can create shelters by excavating soft substrate (Lawton & Lavalli 1995). This literature supports the result that sand/silt/clay is the most suitable substrate type for adults in both spring and fall. Overall, despite insufficient resolution of the substrate data and trawl survey bias, agreement of the seasonal size-specific suitable lobster habitat predictions by the HSI model and patterns in survey catch suggest the robustness of modeling results.
Whereas cross-validation of the AMM-HSI model suggested reasonable predictive performance, the Sis derived from the spline smooth regression method can be used to deal with possible nonlinear relationships between covariate and response variables in a semiparametric manner for further analysis (Maunder & Punt 2004, Chang et al. 2012).
The type and number of habitat variables to be included in the calibration of an HSI model is critical to the successful identification of suitable habitats (Tian et al. 2009). Distribution and abundance of lobster can be influenced by many other environmental variables such as availability of prey, presence of predators, thermal fronts, latitude and longitude, time of the day, light levels, and dissolved oxygen concentration (Wahle & Steneck 1992, Mercaldo-Allen & Kuropat 1994, Crossin et al. 1998, Chang et al. 2010). Consequently, more habitat variables may need to be incorporated and evaluated in future analyses. Whereas these variables are likely to be correlated, application of dimension reduction technique such as principal component analysis can be incorporated to develop more comprehensive HSI model (Daskalov 1999).
Furthermore, although equal weight was assigned to each habitat variable for the empirical HSI model in this study (Vayghan et al. 2013), the relative importance of different habitat variables in regulating lobster spatiotemporal distribution is likely to be variable, which could significantly influence the predictive performance of HSI models (Gong et al. 2012). For the existing HSI models to better predict spatiotemporal distribution of suitable lobster habitat, the impact of differential weighing of habitat variables should be carefully analyzed based on relative contribution to the spatial distribution of lobsters (Chang et al. 2010). The selection and weighting of habitat variables in an empirical HSI model should be further studied to improve the model's hindcasting or forecasting ability. This will be important in promoting the use of HSI models in fishery management and could be particularly useful when considering shifts in the marine environment because of climate change.
Most stock assessments neglect to incorporate habitat information into the assessment models, but habitat data are important to many aspects of the stock assessment process. The SI and HSI modeling results for juvenile lobsters have implications for lobster recruitment, whereas results for adult lobsters have implications for spawning stock biomass. For example, periods of low habitat suitability, such as years when both seasons had an HSI value above 0.6 in less than 10% of the studied area (Fig. 10), partially overlap with periods of estimated low recruitment abundance (2003-2007) and low spawning stock abundance (2004-2007) from the 2009 benchmark assessment (ASMFC 2009). Further analysis may (1) reveal statistically significant correlations between habitat suitability and recruitment or spawning stock abundance, and (2) link availability of suitable habitat to carrying capacity of LIS for American lobsters. Also, change in habitat availability could potentially be related to the recent collapse of LIS lobster stock.
Application of HSI models can improve lobster stock assessment by allowing us to (1) hindcast and forecast periods of distinct lobster productivity and recruitment dynamics in LIS, and (2) define and compare different modeling time periods with respect to these processes. Traditional stock assessment models focus on the context of commercial fishing, where natural mortality is relegated to a single, typically time-invariant parameter that is often not related to lobster ecology. The recent management shift toward ecosystem-based fisheries management requires scientists and managers to develop useful, quantitative measures to illustrate the history of stock fluctuations in an ecological context. Incorporating habitat availability modeling into stock assessments will aid in effective implementation of ecosystem-based management.
We thank Chongliang Zhang, Samuel Truesdell, Bai Li, Jie Cao, Katherine Thompson, and Max Ritchie of University of Maine for their generous support, ideas, and perspectives. We would also like to thank Kurt Gottschal from the Connecticut Department of Energy and Environmental Protection for providing lobster survey data. Financial support for this study was provided by NSF IGERT program, NSF Coastal SEES program, and Maine Sea Grant College Program. This work uses the FVCOM model-predicted database created by Dr. Chen's research team at the Marine Ecosystem Dynamics Modeling Laboratory, University of Massachusetts Dartmouth.
Ahmadi-Nedushan, B., A. St-Hilaire, M. Berube, E. Robichaud, N. Thiemonge & B. Bobee. 2006. A review of statistical methods for the evaluation of aquatic habitat suitability for instream flow assessment. River Res. Appl. 22:503-523.
Atlantic States Marine Fisheries Commission (ASMFC). 2009. Stock Assessment Report No. 09-01 (Supplement) of the Atlantic States Marine Fisheries Commission: American Lobster Stock Assessment Report for Peer Review. Boston, MA: Atlantic States Marine Fisheries Commission. 298 pp.
Barshaw, D. E. & D. R. Bryant-Rich. 1988. A long-term study on the behavior and survival of early juvenile American lobster, Homarus americanus, in three naturalistic substrates: eelgrass, mud and rocks. Fish Bull. 86:789-796.
Becker, R. A., A. R. Wilks, R. Brownrigg & T. P. Minka. 2014. R Package "maps": display of maps (Ver 2.3-9). R-Project. Accessed October 18,2014. Available at: http://cran.r-project.org/web/packages/ maps/index.html.
Bell, S. L., B. Allam, A. McElroy, A. Dove & G. T. Taylor. 2012. Investigation of epizootic shell disease in American lobsters (Homarus americanus) from Long Island Sound: I. Characterization of Associated Microbial Communities. J. Shellfish Res. 31:473-484.
Berger, T. L. (Ed). 2014. Annual Report 2013. Arlington, VA: Atlantic States Marine Fisheries Commission. 52 pp.
Bivand, R. 2013. R Package "classInt": choose univariate class interval (Ver 0.1-21). R-Project. Accessed October 23, 2014. Available at: http://cran.r-project.org/web/packages/classInt/index.html.
Bivand, R., T. Keitt, B. Rowlingson, E. Pebesma, M. Summer, R. Hijmans & E. Rouault. 2014. R Package "rgdal": bindings for the Geospatial Data Abstraction Library (Ver 0.9-1). R-Project.
Accessed October 15, 2014. Available at: http://cran.r-project.org/web/packages/rgdal/index.html.
Bovee, K. & J. R. Zuboy. 1988. Proceedings of a workshop on the development and evaluation of habitat suitability criteria (No. FWS-88 (11)): a compilation of papers and discussions presented at Colorado State University. Fort Collins, CO: National Ecology Research Center, U.S. Department of the Interior, Fish and Wildlife Service. Accessed July 12, 2014. Available at: http://www.dtic.mil/cgi-bin/GetTRDoc?AD=ADA322764.
Castro, K. M., J. S. Cobb, M. Gomez-Chiarri & M. Tlusty. 2012. Epizootic shell disease in American lobsters Homarus americanus in southern New England: past, present and future. Dis. Aquat. Organ. 100:149-158.
Chang, J.-H., Y. Chen, D. Holland & J. Grabowski. 2010. Estimating spatial distribution of American lobster Homarus americanus using habitat variables. Mar. Ecol. Prog. Ser. 420:145-156.
Chang, Y.-J., C.-L. Sun, Y. Chen, S.-Z. Yeh & G. Dinardo. 2012. Habitat suitability analysis and identification of potential fishing grounds for swordfish, Xiphias gladius, in the South Atlantic Ocean. Remote Sens. 33:7523-7541.
Chen, C., R. C. Beardsley & G. W. Cowles. 2006b. An unstructured-grid, finite-volume coastal ocean model (FVCOM) system. Oceanography (Wash. D.C.) 19:78-89.
Chen, Y., S. Sherman, C. Wilson, J. Sowles & K. Minoru. 2006a. A comparison of two fishery-independent survey programs used to define the population structure of American lobster (Homarus americanus) in the Gulf of Maine. Fish Bull. 104:247-255.
Chen, X., S. Tian, Y. Chen & B. Liu. 2010. A modeling approach to identify optimal habitat and suitable fishing grounds for neon flying squid (Ommastrephes bartramii) in the Northwest Pacific Ocean. Fish Bull. 108:1-14.
Connecticut Department of Energy and Environmental Protection (CTDEP). 2010. Lobster landings. Long Island Sound Study. Accessed August 17, 2014. Available at: http://longislandsoundstudy.net/2010/07/lobster-landings/.
Connecticut Department of Energy and Environmental Protection (CTDEP). 2012. A Study of Marine Recreational Fisheries in Connecticut. Hartford, CT. 140 pp.
Connecticut Department of Energy and Environmental Protection (CTDEP). 2013. A Study of Marine Recreational Fisheries in Connecticut. Hartford, CT. 145 pp.
Cooper, R. A., R. A. Clifford & C. D. Newell. 1975. Seasonal abundance of the American Lobster, Homarus americanus, in the Boothbay Region of Maine. Trans. Am. Fish. Soc. 104:669-674.
Cooper, R. A. & J. R. Uzmann. 1980. Ecology of juvenile and adult Homarus. In: J. S. Cobb, B. F. Phillips, editors. The biology and management of lobsters. New York, NY: Academic Press, Inc. pp. 97-142.
Cooperrider, A. Y., R. J. Boyd & H. R. Stuart. 1986. Inventory and monitoring of wildlife habitat. Denver, CO: Service Center, Bureau of Land Management, U.S. Department of Interior; 858 pp. Available from the Superintendent of Documents, U.S. Government Printing Office, Washington, DC 20402; GPO # 024-011-00170-1.
Crawley, M. J. 2007. The R Book. Chichester, UK: John Wiley & Sons, Ltd. 980 pp.
Crossin, G., S. Al-Ayoub, S. Jury & W. Howell. 1998. Behavioral thermoregulation in the American lobster Homarus americanus. J. Exp. Biol. 201:365-374.
Daskalov, G. 1999. Relating fish recruitment to stock biomass and physical environment in the Black Sea using generalized additive models. Fish. Res. 41:1-23.
Dettki, H., R. Lofstrand & L. Edenius. 2003. Modeling habitat suitability for moose in coastal northern Sweden: empirical vs process-oriented approaches. Ambio 32:549-556.
Dove, A. D. M., B. Allam, J. J. Powers & M. S. Sokolowski. 2005. A prolonged thermal stress experiment on the American lobster, Homarus americanus. J. Shellfish Res. 24:761-765.
Ennis, G. P. 1984. Small-scale seasonal movements of the American lobster Homarus americanus. Trans. Am. Fish. Soc. 113:336-338.
Ennis, G. P. 1995. Larval and postlarval ecology. In: Factor JR, editor. Biology of the lobster Homarus americanus. San Diego, CA: Academic Press, pp. 23-88.
Factor, J. R. (Ed). 1995. Biology of the lobster: Homarus americanus. San Diego, CA: Academic Press. 528 pp.
Fogarty, M., L. Incze, R. A. Wahle, D. Mountain, A. Robinson, K. Hayhoe, A. Richards & J. Manning. 2007. Potential climate change impacts on marine resources of the Northeastern United States. Northeast Climate Impacts Assessment Technical Series Accessed April 28, 2014. Available at: http://www.northeastclimateimpacts.org/pdf/miti/fogarty_et_al.pdf.
Franklin, J. 2010. Mapping species distributions: spatial inference and prediction. Cambridge, United Kingdom: Cambridge University Press. 340 pp.
Gong, C., X. Chen, F. Gao & Y. Chen. 2012. Importance of weighting for multi-variable habitat suitability index model: a case study of winter-spring cohort of Ommastrephes bartramii in the Northwestern Pacific Ocean. J. Ocean Univ. China 11:241-248.
Gottschall, K. 2013. Long Island Sound trawl survey, 1984 to 2013 [Data set]. Hartford, CT: Connecticut Department of Energy and Environmental Protection.
Gottschall, K. & D. Pacileo. 2013. Marine finfish survey, part 1: Long Island Sound trawl survey. In: A study of marine recreational fisheries in Connecticut. Hartford, CT: Department of Environmental Protection, Marine Fisheries Division. US Department of Interior, Fish and Wildlife Service Federal Aid in Sport Fish Restoration Grant F-54-R-32, Annual performance report. 145 pp.
Harding, G. C. 1992. American lobster (Homarus Americanus Milne Edwards): a discussion paper on their environmental requirements and the known anthropogenic effects on their populations, No. 1887. Fisheries and Oceans Canada. Accessed September 30, 2014. Available at: http://publications.gc.ca/collections/collection_2014/mpo-dfo/Fs97-6-1887-eng.pdf.
Hudon, C. 1987. Ecology and growth of post-larval and juvenile lobster, Homarus americanus, off lies de la Madeleine (Quebec). Can. J. Fish Aquat. Sci. 44:1855-1869.
Jian, J., H. Jiang, Z. Jiang, G. Zhou, S. Yu, S. Peng, S. Liu, S. Liu & J. Wang. 2013. Predicting giant panda habitat with climate data and calculated habitat suitability index (HSI) map. Meleorol. Appl. 21:210-217.
Jury, S. H., M. T. Kinnison, W. Huntting Howell & W. H. Watson III. 1994. The behavior of lobsters in response to reduced salinity. J. Exp. Mar. Biol. Ecol. 180:23-37.
Lewin-Koh, N. J., E. Pebesma, E. Archer, A. Baddeley, S. Dray, D. Forrest, M. Friendly, P. Giraudoux, D. Golicher, V. Gomez, P. Hausmann, K. Ove, T. Jagger, S. Luque, D. MacQueen, M. Stokely & R. Turner. 2014. R Package "maptools": tools for reading and handling spatial objects. Ver 0.8-30. R-Project. Accessed September 15, 2014. Available at: http://cran.r-project.org/web/packages/maptools/index.html.
Lawton, P. & K. L. Lavalli. 1995. Postlarval, juvenile, adolescent, and adult ecology. In: J. R. Factor, editor. Biology of the lobster Homarus americanus. San Diego, CA: Academic Press, pp. 47-88.
Maunder, M. N. & A. E. Punt. 2004. Standardizing catch and effort data: a review of recent approaches. Fish. Res. 70:141-159.
Mercaldo-Allen, R. & C. A. Kuropat. 1994. Review of American lobster (Homarus americanus) habitat requirements and responses to contaminant exposures. NOAA Technical Memorandum NMFS-NE-105, NOAA, NMFS, Department of Commerce. 52 pp.
Mills, K. E., A. J. Pershing, C. J. Brown, Y. Chen, F.-S. Chiang, D. S. Holland, S. Lehuta, J. A. Nye, J. C. Sun, A. C. Thomas & R. A. Wahle. 2013. Fisheries management in a changing climate: lessons from the 2012 ocean heat wave in the Northwest Atlantic. Oceanography (Wash. D.C.) 26:191-195.
Morris. L. & D. Ball. 2006. Habitat suitability modelling of economically important fish species with commercial fisheries data. ICES J. Mar. Sci. 63:1590-1603.
Morrison, M. L., B. Marcot & W. Mannan. 2012. Wildlife-habitat relationships: concepts and applications. Washington, DC: Island Press. 520 pp.
National Geophysical Data Center. 1999. U.S. Coastal Relief Model: Northeast Atlantic. National Geophysical Data Center, NOAA. doi:10.7289/V5MS3QNZ. Accessed July 28, 2014. Available at: http://www.ngdc.noaa.gov/docucomp/page?xml=NOAA/NESDIS/ NGDC/MGG/DEM/iso/xml/713.xml&view=getDataView&header=none.
Nychka, D., R. Furrer & S. Sain. 2014. R Package "fields": tools for spatial data (Ver 7.1). R-Project. Accessed June 8, 2014. Available at: http://cran.r-project.org/web/packages/fields/index.html.
Pebesma, E., R. Bivand, B. Rowlingson & V. Gomez-Rubio. 2014. R Package "sp": classes and methods for spatial data (Ver 1.0-15). R-Project. Accessed August 9, 2014. Available at: http://cran.r-project. org/web/packages/sp/index.html.
Pebesma, E. & B. Graeler. 2014. R Package "gstat": spatial and spatiotemporal geostatistical modelling, prediction and simulation (Ver 1.0-19). R-Project. Accessed August 2, 2014. Available at: http://cran.r-project.org/web/packages/gstat/index.html.
Poppe, L. J. & B. A. Seekins. 2000. Distribution of surficial sediments in Long Island Sound. In: Paskevich V, Poppe L, editors. Georeferenced sea-floor mapping and bottom photography in Long Island Sound. Woods Hole, MA: U.S. Geological Survey. Accessed February 15, 2015. Available at: http://pubs.usgs.gov/of/2000/of00-304/lisound/data/chap04/listex.htm.
Poppe, L. J.. H. J. Knebel, B. A. Seekins & M. E. Hastings. 2000. Map showing the distribution of surficial sediments in Long Island Sound. In: Paskevich V.F., Poppe L.J., editors. U.S. Geological Survey Open-File Report 00-304: Georeferenced Sea-Floor Mapping and Bottom Photography in Long Island Sound. Woods Hole, MA: U. S. Geological Survey. 245 pp.
Qadri, S. A., J. Camacho, H. Wang, J. R. Taylor, M. Grosell & M. K. Worden. 2007. Temperature and acid-base balance in the American lobster Homarus americanus. J. Exp. Biol. 210:1245-1254.
Reid, R. N., F. P. Almeida & C. A. Zetlin. 1999. Essential fish habitat source document: fishery-independent surveys, data sources, and methods. Woods Hole, MA: NOAA Tech Memo NMFS NE 122. 39 pp.
Reynolds, W. W. & M. E. Casterlin. 1979. Behavioral thermoregulation and activity in Homarus americanus. Comp. Biochem. Physiol. 64:25-28.
Schamberger, M., A. H. Farmer & J. W. Terrell. 1982. Habitat Suitability Index Models: U.S.D. I. Fish and Wildlife Service. Fort Collins, CO: FWS/OBS-82/10. 2 pp.
Shields, J. D. 2013. Complex etiologies of emerging diseases in lobsters (Homarus americanus) from Long Island Sound. Can. J. Fish. Aquat. Sci. 70:1576-1587.
Sissenwine, M. P. & E. W. Bowman. 1978. An analysis of some factors affecting the catchability of fish by bottom trawls. ICNAF. Res. Bull. (Sun Chiwawitthaya Thang Tltale Phuket) 13:81-87.
Smith, P. A. 1994. Autocorrelation in logistic regression modelling of species' distributions. Glob. Ecol. Biogeogr. Lett. 4:47-61.
Steneck, R. S. & C. J. Wilson. 2001. Large-scale and long-term, spatial and temporal patterns in demography and landings of the American lobster, Homarus americanus, in Maine. Mar. Freshw. Res. 52:1303-1319.
Steneck, R. S. 2006. Possible demographic consequences of intraspecific shelter competition among American lobsters. J. Crustac. Biol. 26:628-638.
Store, R. & J. Kangas. 2001. Integrating spatial multi-criteria evaluation and expert knowledge for GIS-based habitat suitability modelling. Landsc. Urban Plan. 55:79-93.
Terrell, J. W. 1984. Proceedings of a workshop on fish habitat suitability index models. Washington, DC: U.S. Fish and Wildlife Service, Biological Report 85. 393 pp.
Thunberg, E. M. 2007. Demographic and economic trends in the northeastern United States lobster (Homarus americanus) fishery, 1970-2005. Woods Hole, MA: US Department of Commerce, Northeast Fishery Science Center Reference Document 07-17. pp. 64.
Tian, S.. X. Chen, Y. Chen, L. Xu & X. Dai. 2009. Evaluating habitat suitability indices derived from CPUE and fishing effort data for Ommatrephes bratramii in the northwestern Pacific Ocean. Fish. Res. 95:181-188.
Tukey, J. W. 1977. Exploratory data analysis. In: Exploratory data analysis. 1st edition. F. Mosteller, consulting ed. Reading, MA: Addison-Wesley. 688 pp.
Van der Meeren, G. I., J. Stottrup, M. Ulmestrand, V. Oresland, J. A. Knutsen & A. L. Agnalt. 2010. Invasive alien species fact sheet: Homarus americanus. Accessed May 4, 2014. Available at: http://www.nobanis.org/files/factsheets/homarus_americanus.pdf.
Vayghan, A. H., H. Poorbagher, H. T. Shahraiyni, H. Fazli & H. N. Saravi. 2013. Suitability indices and habitat suitability index model of Caspian kutum (Rutilus frisii kutum) in the southern Caspian Sea. Aquat. Ecol. 47:441-451.
Vinagre, C., V. Fonseca, H. Cabra & M. J. Costa. 2006. Habitat suitability index models for the juvenile soles, Solea solea and Solea senegalensis, in the Tagus estuary: defining variables for species management. Fish. Res. 82:140-149.
Vincenzi, S., G. Caramori, R. Rossi & G. A. De Leo. 2006. A GIS-based habitat suitability model for commercial yield estimation of Tapes philippinarum in a Mediterranean coastal lagoon (Sacca di Goro, Italy). Ecol. Model!. 193:90-104.
Vincenzi, S., G. Caramori, R. Rossi & G. A. De Leo. 2007. A comparative analysis of three habitat suitability models for commercial yield estimation of Tapes philippinarum in a North Adriatic coastal lagoon (Sacca di Goro, Italy). Mar. Pollut. Bull. 55:579-590.
Wahle, R. A. & R. Steneck. 1991. Recruitment habitats and nursery grounds of the American lobster Homarus Americanus: a demographic bottleneck? Mar. Ecol. Prog. Ser. 69:231-243.
Wahle, R. A. & R. Steneck. 1992. Habitat restrictions in early benthic life: experiments on habitat selection and in situ predation with the American lobster. J. Exp. Mar. Biol. Ecol. 157:91-114.
Watson, W. H., Ill, A. Vetrovs & W. H. Howell. 1999. Lobster movements in an estuary. Mar. Biol. 134:65-75.
Zuur, A. F., E. N. Ieno & G. M. Smith. 2007. Analysing ecological data. Vol. 680. New York, NY: Springer. 672 pp.
KISEI TANAKA * AND YONG CHEN
School of Marine Sciences, University of Maine, 5741 Libby Hall, Orono, ME 04469
* Corresponding author. E-mail: email@example.com
TABLE 1. Summary of regression analyses from 100 runs of cross validations. Intercept ([alpha]) Model Life stage Season Mean Median (95% CI) AMM Adult Spring 0.145 0.143 0.042 0.266 Juvenile Spring 0.212 0.212 0.082 0.368 Adult Fall 0.161 0.168 0.007 0.271 Juvenile Fall 0.179 0.183 0.075 0.291 GMM Adult Spring 0.128 0.124 0.012 0.280 Juvenile Spring 0.136 0.136 0.036 0.284 Adult Fall 0.161 0.166 0.001 0.305 Juvenile Fall 0.152 0.162 0.052 0.247 Slope ([beta]) Model Life stage Mean Median (95% CI) AMM Adult 0.732 0.732 0.529 0.922 Juvenile 0.594 0.595 0.397 0.797 Adult 0.682 0.676 0.493 0.949 Juvenile 0.683 0.681 0.465 0.867 GMM Adult 0.750 0.746 0.484 1.123 Juvenile 0.680 0.669 0.343 1.194 Adult 0.628 0.615 0.341 0.954 Juvenile 0.679 0.673 0.470 0.910 [R.sup.2] AIC Model Life stage Mean Median Mean Median AMM Adult 0.594 0.611 -533 -539 Juvenile 0.458 0.456 -357 -351 Adult 0.495 0.492 -274 -273 Juvenile 0.549 0.550 -199 -194 GMM Adult 0.536 0.541 -306 -298 Juvenile 0.441 0.448 -194 -181 Adult 0.384 0.358 -94 -84 Juvenile 0.507 0.512 -148 -144 The table shows model parameters for the linear regression between the predicted and observed HSI values and AIC for the two HSI models AMM and GMM.
|Printer friendly Cite/link Email Feedback|
|Author:||Tanaka, Kisei; Chen, Yong|
|Publication:||Journal of Shellfish Research|
|Date:||Aug 1, 2015|
|Previous Article:||Comparing the response of the brown shrimp Crangon crangon (Linnaeus, 1758) to prolonged deprivation of food in two seasons.|
|Next Article:||Effect of temperature on the behavior of stage IV American lobster (Homarus americanus) larvae.|