A Modeling and Field Approach to Identify Essential Fish Habitat for Juvenile Bay Whiff (Citharichthys spilopterus) and Southern Flounder (Paralichthys lethostigma) Within the Aransas Bay Complex, TX
- First Online:
- Cite this article as:
- Froeschke, B.F., Stunz, G.W., Robillard, M.M.R. et al. Estuaries and Coasts (2013) 36: 881. doi:10.1007/s12237-013-9600-9
- 377 Views
The goal of this study was to use an ecosystem-based approach to consider the effect of environmental conditions on the distribution and abundance of juvenile bay whiff and southern flounder within the Aransas Bay Complex, TX, USA. Species habitat models for both species were developed using boosted regression trees. Juvenile bay whiff were associated with low temperatures (<15 °C, 20–23 °C), moderate percent dry weight of sediments (25–60 %), salinity >10, and moderate to high dissolved oxygen (6–9 mg O2/l, 10–14 mg/l). Juvenile southern flounder were associated with low temperatures (<15 °C), low percent dry weight of sediment (<25 %), seagrass habitat, shallow depths (<1.2 m), and high dissolved oxygen (>8 mg O2/l). Our results indicate that conservation measures should focus along the eastern side of Aransas Bay and the north corner of Copano Bay to protect essential fish habitat. These findings provide a valuable new tool for fisheries managers to aid in the sustainable management of bay whiff and southern flounder and provide crucial information needed to prioritize areas for habitat conservation.
KeywordsManagementParalichthys lethostigmaCitharichthys spilopterusNursery habitatBoosted regression treesEssential fish habitat
Habitat loss due to human impacts is a primary cause of population depletion in fishes (Ruckelshaus et al. 2002; Dulvy et al. 2003; Pyke 2004; Levin and Stunz 2005; Lotze et al. 2006). Declining fish stocks and loss of habitat threaten the health of marine ecosystems (Jackson et al. 2001; Pauly et al. 2002; Hilborn et al. 2003; Pyke 2004; Hughes et al. 2005; Lotze et al. 2006; Crowder et al 2008; Halpern et al. 2008; NMFS 2008; Worm and Lotze 2009; Zhou et al. 2010), and it has been hypothesized that the overfished populations and ecosystems that they inhabit are more susceptible to other anthropogenic impacts (Jackson et al. 2001; Halpern et al. 2008). In the Gulf of Mexico, declining populations of important fish stocks such as southern flounder (Paralichthys lethostigma; Froeschke et al. 2011) accentuate the importance of defining critical habitats as well as the processes that contribute to habitat quality (Houde and Rutherford 1993; Allen and Baltz 1997). Southern flounder support an important fishery in the Gulf of Mexico, yet essential fish habitat (EFH) has not been described distribution-wide for this species (VanderKooy 2000). An improved understanding of the relationship between abiotic (e.g., temperature, hydrodynamics, oxygen, salinity) and biotic factors (e.g., organic content, habitat) with respect to life history and habitat requirements is essential for robust management of this fishery.
Along the Texas coast, flounder have historically supported a multi-million dollar commercial and recreational fishery (Matlock 1991; VanderKooy 2000). Southern flounder represent over 95 % of harvested flounder and is one of the top three fish species targeted by recreational anglers (Riechers 2008). Despite increased commercial and recreational fishing regulation in Texas, the southern flounder population is declining at an alarming rate (Froeschke et al. 2011). A fisheries management plan for the Gulf of Mexico flounder fishery was developed and determined that identification of EFH for the flounder fishery is crucial for effective management (VanderKooy 2000). Initial studies on EFH for young-of-the-year southern flounder in Aransas Bay and Copano Bay, TX, USA showed that they occur in vegetated habitats (seagrass and marsh edge) near tidal inlets in Aransas Bay (Nañez-James et al. 2009). However, their abundance and distribution in conjunction with abiotic factors were not evaluated. Texas estuaries are physically dynamic, and the distribution of fishes is strongly affected by environmental conditions (Froeschke et al. 2010; Froeschke and Froeschke 2011).
Flatfish as a group are important components of coastal ecosystems. For example, bay whiff are among the most common flatfishes in Gulf of Mexico estuaries (Allen and Baltz 1997; Castillo-Rivera et al. 2000) and North Carolina (Walsh and Peters 1999). Bay whiff are not a recreational or commercially targeted species and little is known about their habitat use along the Texas coast. However, they exhibit similar temporal recruitment patterns to southern flounder. Moreover, it has been hypothesized that bay whiff are habitat generalists (Allen and Baltz 1997; Walsh and Peters 1999). For example, in North Carolina, the abundance of bay whiff was not significantly different among 21 stations sampled, which included marsh, seagrass, and non-vegetated habitats, implying that bay whiff are associated with all estuarine habitats (Walsh and Peters 1999).
The objective of this study was to develop species habitat models for two important flatfish species, southern flounder and bay whiff. Specifically, the relationship between abiotic (temperature, salinity, turbidity, dissolved oxygen, and pH) and biotic factors (habitat, depth, and organic content) with the frequency of occurrence of bay whiff and southern flounder was investigated within the Aransas Bay Complex (Mission-Aransas National Estuarine Research Reserve—MANERR), TX, USA. To examine this relationship, we developed spatially explicit distribution patterns of juvenile bay whiff and southern flounder. We used boosted regression trees (BRT) (De’ath 2007; Elith et al. 2008), a powerful yet relatively new approach to modeling species–environment relationships. Boosted regression trees is an ensemble method that combines statistical and machine learning techniques and has shown to be an effective method to identify relationships between fish distribution patterns and environmental predictors (Leathwick et al. 2006; 2008; Froeschke et al. 2010a; Froeschke and Froeschke 2011). The species habitat models of southern flounder and bay whiff will provide natural resource managers crucial information needed to conserve habitats for various developmental stages of flatfish within the Aransas Bay Complex, TX, USA.
Materials and Methods
A stratified, randomized experimental design was used to identify EFH for juvenile bay whiff and southern flounder within the Aransas Bay Complex from February to May 2010 during peak flatfish recruitment season (Nañez-James et al. 2009; Froeschke et al. 2011). Sites were selected by converting the study area into 100-m2 grid cells. Habitat type for each cell was determined using existing habitat maps (http://www.csc.noaa.gov/digitalcoast/data/benthiccover/download.html). Using this grid, 40 100-m2 sites were sampled each month in three habitat types: seagrass (n = 10), oyster (n = 10), and non-vegetated bottom habitats (n = 20). Sampling effort per habitat type was determined based on the proportion of each habitat within the Aransas Bay Complex. Sample sites were selected without replacement using a randomized selection of sites from the sampling grid.
Samples with a low percent of dry weight were considered to have a higher percentage of organic content than samples with a higher percent of dry weight. Thus, low percentage of dry weight is correlated with higher quality of sediments.
Juvenile bay whiff and southern flounder were collected using a 2-m-wide beam trawl with 6-mm-stretch mesh liner towed for 50 m (total area 100 m2) at a constant speed (5 knots). Trawl samples were rough-sorted in the field to remove excessive algae, seagrass, and debris, preserved in 10 % formalin, and returned to the laboratory for further processing. All flatfishes were identified, enumerated, and measured to the nearest millimeter standard length (SL).
Saltwater and larval exchange occurs via the Aransas tidal inlet, and flatfish use the tidal inlet to migrate offshore for spawning as adults and as an ingress pathway during the larval stage. Therefore, to examine a potential relationship between juvenile bay whiff and southern flounder with the connection to the Gulf of Mexico, the distance from the Aransas tidal inlet to each sampling location was calculated using the cost distance function in the spatial analyst extension in ArcGIS (ESRI, Redlands, CA, USA), using the shoreline as a buffer (Whaley et al. 2007). The cost distance function is used to calculate the shortest distance between two points that are constrained within a geographic boundary to provide more accurate relative distance estimates than Euclidian methods (Froeschke et al. 2010).
Boosted Regression Trees
Relationships between both juvenile bay whiff and southern flounder density and biological, physical, spatial, and temporal variables were determined using a forward fit, stage-wise, binomial boosted regression tree model (De'ath 2007). Boosted regression trees (1) accept different types of predictor variables, (2) accommodate missing values through the use of surrogates, (3) resist the effects of outliers, and (4) automatically fit interactions between predictors (Elith et al. 2006; Leathwick et al. 2006; Elith et al. 2008; Leathwick et al. 2008). Unlike traditional regression techniques, BRT combines the strength of two algorithms, regression trees and boosting, to combine large numbers of relatively simple tree models instead of a single “best” model (Elith et al. 2006; Leathwick et al. 2006; Elith et al. 2008; Leathwick et al. 2008). Each individual model consists of a simple regression tree assembled by a rule-based classifier that partitions observations into groups having similar values for the response variable based on a series of binary splits constructed from predictor variables (Friedman 2001; Leathwick et al. 2006; Elith et al. 2008). The BRT often has a higher predictive performance than single tree methods due to the inherent strengths of regression trees and the robustness of model averaging that improves predictive performance. Overfitting is minimized by incorporating tenfold cross-validation into the model fitting process (Elith et al. 2006; Leathwick et al. 2006; Elith et al. 2008; Leathwick et al. 2008). The fitting of a BRT model is a stochastic process. To examine within model variability, the BRT model was refit using (n = 1,000) randomized variations of the original dataset. Mean predicted probability of occurrence and 95 % confidence limits were determined in the study area.
Habitat Suitability Models
We used ordinary kriging with a spherical semivariogram of predicted probability of occurrence of bay whiff and southern flounder (Froeschke et al. 2010) to develop spatially explicit predictions. Kriging is a spatial interpolation algorithm that was used to predict values at unsampled sites in the study area (Saveliev et al. 2007). This routine was carried out for each iteration of the fitted BRT model (n = 1,000) to determine the mean (and 95 % confidence limits) of predicted probability of occurrence across the study area. Kriging was carried out using the automap (Hiemstra et. al. 2008) and raster (Hijmans and van Etten 2012) libraries in R (version 2.15, R Development Core Team).
Abiotic and Biotic Parameters
Mean (± standard error) parameter ranges by habitat from 160 sites (seagrass n = 40, oyster reef n = 40, and non-vegetated n = 80) sampled from February to May 2010 within the Aransas Bay Complex
21.55 ± 2.41
21.97 ± 3.47
22.99 ± 3.64
14.74 ± 1.65
13.13 ± 2.08
18.93 ± 2.99
81.12 ± 9.07
73.10 ± 11.56
56 ± 8.85
3.59 ± 0.40
2.78 ± 0.44
2.15 ± 0.34
Dissolved Oxygen (mg O2/l)
7.26 ± 0.81
7.89 ± 1.25
9.03 ± 1.43
8.14 ± 0.91
8.22 ± 1.30
8.44 ± 1.33
Dry Weight (%)
47.83 ± 5.49
29.06 ± 4.59
Habitat Model for Bay Whiff and Southern Flounder
Predictive performance of boosted regression trees (BRT) models for juvenile bay whiff and southern flounder. tc = tree complexity, lr = learning rate, and bf = bag fraction
Percentage Deviance Explained
Mean ROC Cross-Validation
Mean ROC Cross-Validation SE
This study demonstrates the importance of incorporating biological, physical, and spatial variables in species habitat models to identify the frequency of occurrence patterns of estuarine organisms. The occurrence of juvenile bay whiff and southern flounder demonstrated strong relationships with biological (habitat type, dry weight of sediments), physical (depth, dissolved oxygen, temperature, turbidity, and pH), and spatial (distance to inlet) variables. The occurrence of bay whiff was most strongly influenced by % dry weight of sediments, distance to inlet, water temperature, salinity, and dissolved oxygen. The occurrence of southern flounder was driven by water temperature, dry weight of sediments, habitat type, month of collection, depth, and dissolved oxygen. Others have shown biological variables such as prey abundance, predators, habitat structure, water depth, and physical factors (temperature, salinity, oxygen, and hydrodynamics) to be major factors affecting the growth, survival, and recruitment of flatfishes (Gibson 1994; Allen and Baltz 1997; Stoner et al. 2001; Glass et al. 2008).
Due to a paucity of information about bay whiff in the Gulf of Mexico, this study is valuable in beginning to understand environmental constraints for this highly abundant species. Habitat type was not detected as an important variable in predicting the occurrence of bay whiff. The probability of occurrence for juvenile bay whiff was instead associated with low temperatures, moderate percent dry weight of sediments, low salinities, and high dissolved oxygen levels. Results of the BRT model indicated that environmental conditions were more influential than habitat type (e.g., seagrass). Our results suggest that bay whiff are habitat generalists, which is consistent with previous findings (Allen and Baltz; Walsh and Peters 1999). These results suggest that management of bay whiff should focus more on habitat quality rather than structured habitat type and should consider the effect of environmental conditions on fish habitat quality.
Similar to other studies, juvenile southern flounder were relatively rare in our samples, particularly compared to bay whiff (Hoese and Moore 1998; Walsh and Peters 1999; McEachran and Fechhelm 2006; Nañez-James et al. 2009). Results indicate that juvenile southern flounder are most likely to occur in areas with low temperatures, low percent dry weight of sediment, shallow depth, seagrass habitat, and high dissolved oxygen content. The highest occurrence rates of juvenile southern flounder temperatures below 15 °C is consistent with previous studies in Texas (captured between 14.5 and 21.6 °C; Günter 1945). However, previous work has shown that the optimum recruitment temperature of southern flounder is 16–16.2 °C (Stokes 1977). Given the importance of temperature on occurrence patterns, projected sea temperature increases are of potential concern for this species. Seawater temperature is projected to increase by 4 °C in the twenty-first century (Thuiller 2007). Both AppleBaum et al. (2005) and Fodrie et al. (2010) reported rising sea temperatures within the Gulf of Mexico. These predicted increases in temperature could have substantial effects on the temporal and spatial recruitment patterns and, ultimately, population size of southern flounder.
Biological variables percent dry weight of sediments, depth, and habitat type were the second through fourth most important variables. Previously, EFH for young-of-the-year southern flounder in Aransas Bay and Copano Bay, TX, USA was identified as vegetated habitats (seagrass and marsh edge) that occur closest to the tidal inlet between Aransas Bay and the Gulf of Mexico and in high-salinity areas (Nañez-James et al. 2009), and our models support those results. However, based on the results of this study, we suggest that when incorporating both habitat type and distance to inlet in predictive models, habitat type contributes more to occurrence rates of juvenile southern flounder than distance to inlet. The relationship between habitat type and distance to inlet implies that there is a correlation with habitat type and the distance to inlet that may be caused by increased habitat quality near the inlets (increased water exchange with the Gulf of Mexico). Clearly, identifying EFH for southern flounder is a component of sound management for this species. Additionally, in Newport River and Back Sound estuaries in North Carolina, no size-specific patterns in habitat utilization were found, but the abundance of southern flounder was significantly higher in the spring in the middle and upper estuary on mud substrates with detritus and in the fall in areas near marsh edges with mud substrates and detritus (Walsh and Peters 1999). Glass et al. (2008) concluded that variation seen in density of southern flounder is more influenced at the bay scale than at the habitat scale. These results underscore the value of considering biotic factors (e.g., seagrass) as well as the suite of environmental characteristics (abiotic factors) and how these factors interact to ultimately determine habitat quality for southern flounder.
Dissolved oxygen, pH, salinity, and turbidity were less important predictors of occurrence. While dissolved oxygen levels can influence the distribution, abundance, and diversity of organisms (Breitburg 2002, Vaquer-Sunyer 2008, Montagna and Froeschke 2009), this primarily occurs at low oxygen levels (i.e., < 2 mg O2/l). In this study, few samples were taken in low DO conditions, but low dissolved oxygen events (e.g., hypoxia) are increasing in frequency and spatial extent in Texas estuaries (Applebaum et al. 2005, Montagna and Froeschke 2009). These data suggest that oxygen levels could influence the distribution and abundance of southern flounder. Southern flounder are euryhaline (Deubler 1960), but survivorship and growth rates increase in lower-salinity waters (Stickney and White 1974, Hickman 1968). This study supports these prior findings as the occurrence of southern flounder was more prevalent in the low-salinity environments. This result illuminates potential ramifications of reduced freshwater inflow into the Aransas Bay Complex as historic inflows are increasingly diverted for human usage.
Abiotic factors were important in predicting the distribution of both bay whiff and southern flounder. Although both dissolved oxygen and % dry weight of sediments were important abiotic variables, their ranges differed between species. We suggest that conservation measures for both flatfish species within the Aransas Bay Complex should prioritize areas that include high probabilities of occurrence for juvenile bay whiff and juvenile southern flounder in the same locations, specifically along the eastern side of Aransas Bay and the north corner of Copano Bay.
Despite the strengths of our modeling approach, there are some inherent limitations. Cross-validated model evaluation indicated good performance of the BRT for both bay whiff and southern flounder. It is possible that other factors affecting their distribution or frequency of occurrence may not have been incorporated into the model, for example, biotic components: spawning location, prey and predator density, using % dry weight as an indicator of organic content. However, we were able to examine several variables simultaneously that were related to habitat suitability, providing timely information for the conservation and management of bay whiff and southern flounder within the Aransas Bay Complex.
This study demonstrated the importance of incorporating environmental and biological variables in species habitat models to identify areas suitable for EFH designation. Habitat is clearly a driving factor for most estuarine-dependent species; however, establishing EFH should also extend beyond the first steps of delineating habitat–density relationships by including interactions among suitable biotic and abiotic constraints within particular areas (Hayes et al. 1996). The complex nature of many marine life history strategies coupled with the lack of research on other ecosystem-level interactions has made progress toward determining EFH problematic (Shutter 1990; and Guisan and Thuiller 2005), and these types of relationships had not been established for flatfish in Texas estuaries. Evidence from this study will lead to more comprehensive management strategies as species habitat models can provide much-needed information to better identify EFH. The modeling approach developed in this study also provides a framework for natural resource managers to identify crucial nursery habitats for various developmental stages of fishery species. Climate changes will certainly alter abiotic factors within all marine environments; therefore, we must understand the importance of these changes to develop a more effective ecosystem-based management system (Chittaro et al. 2009).
We thank the Mission-Aransas National Estuarine Research Reserve Fellowship Program and the Fisheries and Ocean Health Laboratory at the Harte Research Institute for the Gulf of Mexico Studies for funding and support of this work. In addition, we thank Jason Slocum, Laura Payne, Rachel Brewton, and Brittany Blomberg for all of their hard work in the field. We also thank S. Bortone, P. Tissot, B. Sterba-Boatwright, J. Fox, and L. McKinney for their assistance with and comments on the manuscript.