The western rock lobster, Panulirus cygnus, is endemic to Western Australia and supports substantial commercial and recreational fisheries. Due to and its wide distribution and the commercial and recreational importance of the species a key component of managing western rock lobster is understanding the ecological processes and interactions that may influence lobster abundance and distribution. Using terrain analyses and distribution models of substrate and benthic biota, we assess the physical drivers that influence the distribution of lobsters at a key fishery site.
Methods and Findings
Using data collected from hydroacoustic and towed video surveys, 20 variables (including geophysical, substrate and biota variables) were developed to predict the distributions of substrate type (three classes of reef, rhodoliths and sand) and dominant biota (kelp, sessile invertebrates and macroalgae) within a 40 km2 area about 30 km off the west Australian coast. Lobster presence/absence data were collected within this area using georeferenced pots. These datasets were used to develop a classification tree model for predicting the distribution of the western rock lobster. Interestingly, kelp and reef were not selected as predictors. Instead, the model selected geophysical and geomorphic scalar variables, which emphasise a mix of terrain within limited distances. The model of lobster presence had an adjusted D2 of 64 and an 80% correct classification.
Species distribution models indicate that juxtaposition in fine scale terrain is most important to the western rock lobster. While key features like kelp and reef may be important to lobster distribution at a broad scale, it is the fine scale features in terrain that are likely to define its ecological niche. Determining the most appropriate landscape configuration and scale will be essential to refining niche habitats and will aid in selecting appropriate sites for protecting critical lobster habitats.
Citation: Hovey RK, Van Niel KP, Bellchambers LM, Pember MB (2012) Modelling Deep Water Habitats to Develop a Spatially Explicit, Fine Scale Understanding of the Distribution of the Western Rock Lobster, Panulirus cygnus. PLoS ONE 7(4): e34476. https://doi.org/10.1371/journal.pone.0034476
Editor: Simon Thrush, National Institute of Water & Atmospheric Research, New Zealand
Received: November 3, 2011; Accepted: March 5, 2012; Published: April 10, 2012
Copyright: © 2012 Hovey et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was undertaken as part of Fisheries Research and Development Coporation (FRDC) Project No. 2008/013. The Western Australian Marine Science Institution (WAMSI) provided financial support to MBP. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Marine ecosystems worldwide are under increasing pressure due to anthropogenic impacts such as coastal development, pollution and overfishing –. Previous studies have suggested that the impacts of these pressures are being reflected in a decline in biodiversity, outbreaks of marine pests and declining habitat quality –. In recent years ecosystem-based fisheries management (EBFM) has been implemented in an attempt to manage exploited marine resources in a more holistic manner –. One component of EMFB is marine spatial management such as multiple use areas or reserves . However, the implementation of any type of spatial management of exploited species requires an understanding the interaction of the organism with their physical habitat , . Establishing key environmental characteristics that influence the distribution of fish and crustacean populations is thus a critical aspect of maintaining fisheries resources –.
The western rock lobster Panulirus cygnus George is endemic to the west coast of Australia, with a range from the North West Cape south to Cape Leeuwin , where it is the dominant benthic consumer. It also supports Australia's most valuable single-species fishery and significant recreational fisheries . A dramatic decline in puerulus numbers in 2008/09  resulted in management actions to reduce catch to ensure sustainability raised questions about the influence of ecological interactions and processes in determining lobster abundance and distribution. The need to understand interactions and processes that may potentially drive the abundance and spatial distribution of rock lobster resulted in the establishment of a closed area in the deep water (>40 m) portion for the fishery at Leeman in 2011.
The life history of the western rock lobster is spatially segregated; juveniles dominate inshore reefs and majority of sexually mature adults are found on offshore reefs , . Most western rock lobsters are targeted when they migrate offshore as they approach sexual maturity and reach legal minimum size . Despite the importance of offshore reefs and deepwater habitats to the sustainability of the species, little is known about what factors of the environment influence their distribution, beyond the very generic requirements of reef and mixed assemblages dominated by the kelp Ecklonia sp. .
With very little known about deepwater lobster habitats, species distribution models offer (1) the potential to develop detailed baseline information of the spatial ecology of this particularly cryptic species, and (2) the chance to better understand how environmental characteristics affect species distribution patterns. Both outcomes have the potential to provide useful spatially-explicit information on critical habitats that can be incorporated into fisheries management. Originally applied to terrestrial ecology , species distribution modelling is now widely used in marine research –. Using full coverage bathymetry and tow video data, species distribution models can be used to map terrain, substrate and seafloor biota, providing detailed information on the variation, composition and configuration of seafloor habitats. Predictive species modelling can then proceed by developing statistical or mathematical models that link the target species to observed or measured habitat attributes, leading to predictions of contemporary species distributions. For example, Holmes et al.  conducted broad scale marine mapping of sessile benthos at Point Addis Marine Park, Australia, using terrain analysis and species distributions models of biota (macroalgae, rhodoliths, sponges, ascidians, and soft corals). Chatfield et al.  used species distribution models and data on fish and habitats from baited underwater video systems (BRUVS) to assess the level of influence of substrate type, macroalgal type and sessile biota on the distribution of fishes. Work on crustaceans has included the development of habitat maps and catch data for the spiny lobster (Panulirus argus) to improve stock assessment . Also, Galparsoro et al.  used ecological niche factor analysis for developing species distribution models, and found the distribution of the European lobster (Homarus gammarus) was mainly influenced by distance to rock substrates, benthic position index, wave flux at the seafloor, and bathymetry.
In this study, species distribution modelling was based on lobster pot data from the closed area at Leeman. Classification trees (CTs) were used to define species distribution, because they are well suited to exploring and modelling complex ecological data , , . They also provide greater deviance explained compared to general additive models (GAMs), while maintaining adequate predictive performance . A necessary component to species distribution is high resolution, spatially explicit, continuous data such as detailed habitat maps and bathymetry, a very rare commodity in marine deep water environments . Thus, habitat maps of the terrain, substrates, and biota at the study area were created first, derived from hydroacoustic surveys and towed video. These three data sets were then used to develop CT models for predicting the distribution of the western rock lobster.
The study site was located approximately 30 km offshore from Leeman (29°56.94 S, 114°58.74E), Western Australia (Fig. 1). A portion of the study site was closed to lobster fishing in March 2011 to provide a research area in the absence of lobster exploitation.
Lobster sampling and data collection
Lobsters were sampled in water depths 45–80 m using standard baited commercial pots, with closed escape gaps, deployed from chartered commercial western rock lobster vessels as part of the western rock lobster independent breeding stock survey (IBSS). This annual standardised ten day fishery independent survey has been conducted in mid to late October at a number of sites along the Western Australian coast since the early 1990s . The area off Leeman area was added to the existing survey in 2008, providing three years of pre-closure information on the distribution, size and abundance of lobsters. Three hundred and thirty (330) standard commercial pots were deployed, 400 m apart, in 11 lines of 30 pots running north to south parallel with the coast. Each pot has its own GPS coordinate so the pot locations are consistently sampled between years. The data collected include the number of lobsters per pot, carapace length in millimetres, sex, reproductive state and general condition of lobsters caught in each pot. All lobsters were released following data collection. For this study, we focus only on the presence or absence of lobsters in pots.
Bathymetry and biological data collection
To complete full coverage information on terrain, substrates and benthic biota, detailed information of the seafloor was captured. A full coverage dataset for the study area consisted of bathymetry from hydroacoustic surveys collected in 2010, using a SEABAT 8101 Reson Multibeam. Bathymetry spot heights were averaged using the moving window technique to produce a digital bathymetry model at 3×3 m resolution. A detrended bathymetry was created to remove depth gradient effects. A trend was calculated from the bathymetry data points (at the same resolution as the bathymetric data set: 3 m) using a linear polynomial. The trend was then subtracted from the bathymetry to create the detrended surface. This enabled the development of important variables such as seafloor rugosity, topographic position, reef relief and aspect, which would otherwise be affected by the strong depth gradient trend in the bathymetric data. Secondary datasets assessing geomorphic terrain features were developed from the bathymetry and detrended bathymetry (Table 1).
Substrates and benthic biota were observed using video footage from an underwater camera towed behind a boat travelling 1–2 knots per hour. The camera was held at approximately 1 m above the seafloor and the position georeferenced using an Ultra Short Base Line (USBL) acoustic positioning system linked to a GPS with satellite differential correction. Video footage was archived with digitally superimposed GPS time stamps at 1 second intervals. Sampling was designed to ensure a broad geographic coverage of the study area with sufficient numbers of georeferenced data for modelling and mapping. Seafloor information from the hydroacoustic survey enabled a design that captured benthos across the full range of habitats (Fig. 1). A total of 55 km of seafloor was sampled with the video over a 52 km2, giving a ratio of video footage to area of 1.06. Transects designed to run perpendicular and parallel to the coastline to cover ecological gradients and areas where high and low variability were expected. Video classification involved identifying primary and secondary substrates and benthic biota, as well as biota density. Primary substrate was classified as sand, rhodoliths (hard structures of coralline algae typically at low points on sandy substrates), obscured reef (hard substrate covered with sand veneer), flat reef, or low, medium, or high profile reef. Biota categories were Ecklonia spp. (kelp), other macroalgae and sessile invertebrates.
Predictive modelling of habitats
Bathymetry and derived terrain datasets were used as input to species distribution models for predicting substrate type and biota at unsampled locations. Classification trees (CTs) were developed in S Plus ® 8.2 for Windows (TIBCO Software Inc, Palo Alto, California, USA) and used to predict (in order) substrates, benthic biota, and finally lobster distributions.CTs explain the variation of a response variable by one or more predictor variables and are constructed by recursively partitioning data and splitting into mutually exclusive groups. The objective is to partition the response into homogeneous groups while keeping the tree size small. Splitting continues until the stopping criterion (e.g. minimum deviance) is reached then the tree is pruned back to an optimal size using cross-validation. The benthos datasets from video observations were randomly split using the set seed function in S Plus before modelling, with 75% of the data used to develop models while 25% was held back as a validation dataset to test the models predictive ability. The set seed function puts the random number generator, which is based upon a single uniform random number generator, in a reproducible state so that identical results from multiple runs can be obtained. After full models were developed, a 10-fold cross-validation was used to identify key predictors and optimal tree size based on the misclassification rate threshold of 15% . The amount of variance in substrate and biotic distribution explained by the models was calculated using adjusted deviance (Adj D2). This measure was chosen as it takes into account the differences in the number of observations and parameters used in the development of each model .
Separate models were developed for each type of substrate. Video observations of medium and high relief reef were merged together with low relief reef creating a comprehensive reef class. Classes modelled to create a full coverage substrate map were sand, rhodoliths, obscured reef, flat reef and reef. Full models for biota categories were developed using bathymetry, terrain datasets and substrate types.
Predictive modelling of lobsters
Lobster presence and absence data were derived via catch data from three hundred and thirty (330) standard commercial-sized, baited pots (described above). It is common for species distribution models (SDMs) to use presence/absence data to predict the distribution of a species as a first step to understanding species presence and absence in a landscape (e.g. in this case, seafloor habitats). In addition, frequency histograms of lobsters per pot show a Poisson distribution, with most pots having either one or no lobsters (Figure S1). For this study, catch data were combined across years and any pots with lobsters were treated as present. The Classification tree (CTs) for the lobster model was developed using 75% (chosen using the set seed function in S Plus) of the pot catch data over three years as the dependent variable with predictor variables from bathymetry, terrain, substrate and benthic biota classes. Like substrate and biota models, the 10-fold cross-validation method was used to determine key predictors and optimal tree size. The remaining 25% of the data were used to validate model.
Model accuracy was assessed using receiver-operator characteristics (ROC) analysis (ROC AUC software package, Schröder 2004) which evaluated CT performance from the validation dataset (25% of original dataset). ROC analysis is graphical plot of the sensitivity (true positive rate) and specificity (false positive rate) of a model output, commonly referred to as ROC curves. The larger the ‘area-under-the-curve’ (AUC), the more effective the model at prediction. Pfair was chosen as the threshold to convert predicted probabilities of occurrence to presence/absences values, by balancing the number of false positives and false negative predictions. It also provides a measure of how well the model predicts both presences and absences.
Full coverage maps of the Leeman site were created once the final models were applied to the predictor datasets. The first map produced for each category modelled showed the probability of presences at 3 m pixel resolution. Binary presence/absence maps were then constructed using the P-fair probability thresholds from ROC analysis.
Species distribution data often display spatial dependence where locations close together exhibit more similar values than those further apart. This is due to dependence of a response variable on explanatory variables (e.g. physical structure) that are themselves spatially structured , . One of the key assumptions of statistical analyses is that sample data are independent and violation of this assumption may bias parameter estimates and inflate accuracy . Therefore, it is necessary to understand the effect spatial dependence has on species distribution models. Patterns of spatial dependence of substrate and biota categories were analysed using indicator semivariograms (WinGsLib 1.5.6, Statios Software and Services). In the presence of spatial autocorrelation, the amount of deviance it explained in each model was calculated. Then, the effect on model accuracy was estimated by testing models against blind validation data that was spatially independent from the model data.
A total of 3122 georeferenced video frames were described (Table 2). The ratio between hard and soft substrates observed was approximately 1∶1, with less than 10% of the substrates described being over 1 m in height (medium and high profile reef). The most prevalent primary substrate observed was Rhodolith beds (31%). Macroalgal dominant communities (not including Ecklonia) accounted for 29% of frames observed and included mostly red algae. Sessile invertebrate (mostly undifferentiated sponges) dominant habitats accounted for 20% of frames observed while Ecklonia dominant communities accounted for 16%. A total of 2087 lobsters (P. cygnus) were caught in pots over the three year survey.
Model fit and variable contribution
Substrate models explained 53% to 87% of the total deviance (AdjD2) (Table 3) indicating solid associations with geophysical environmental variables. CT models developed for hard substrates had the greatest associations, with reef models explaining 73% (low reef) to 86% (flat reef) of deviance. Rhodoliths also showed strong associations with environmental variables, with the model explaining 81% of deviance. In contrast, models for soft substrates including sand and obscured reef (hard substrate with sandy veneer) explained less than 70% of the total deviance (Table 3). Bathymetry and detrended bathymetry were the most influential variable for most substrate models, accounting for 18 to 83% of explained deviance (see Table 4). Focal variables (standard deviation or mean surface area around a focal cell) were also influential, with standard deviation of 7 m kernel radius from focal cell (F7s) retained in four of the six substrate models (Table 4). Dropping this term reduced deviance explained by up to 48% for reef. Dropping the focal variable that calculates the mean surface area within 21 m from the focal cell (F21M) from obscured reef, rhodoliths and sand reduced deviance explained by 79%, 17% and 6% respectively. Other geophysical variables generally accounted for less than 10% of explained deviance.
For biota categories, sessile invertebrates had the highest deviance explained by the model with an AdjD2 of 87%, followed by other macroalgae (70%) and Ecklonia(48%). Ecklonia and sessile invertebrates showed strong associations with depth contributing 57% and 45% of explained deviance respectively, followed by hard substrate types which contributed to 17% and 41% of explained deviance, respectively (Table 5). In contrast, the contribution of variables for other macroalgae was spread more evenly among predictors, with range (62.5 m radius) contributing the most explained deviance at 24%.
The CT model for lobsters explained 64% of total deviance. The model retained only geophysical variables, including three based on local neighbourhood measures; hypsometric index with 12.5 m kernel radius (HYP 5), range with 62.5 m kernel radius (RNG 25) and focal analysis using standard deviation statistic (F7s), as well as depth (Fig. 2). Depth and focal analysis contributed to 80% of explained deviance while the other variables each contributed 10% of explained deviance.
Focal analysis was calculated based on standard deviation of surface area over a 7 m kernel radius. Hypsometric Index and Range were calculated over a 12.5 m and 62.5 m kernal radius, respectively. Bathymetry (detrend) was the most influential variable, contributing to 47% of the variation explained, followed by focal analysis (33%), range (10%) and hypsometric index (10%).
Of the six substrate models, five had acceptable predictive power (AUC range between 0.72 and 0.77, Table 3). Flat reef had poor predictive power with an AUC value of 0.61, despite the model explaining the highest total deviance. The ability of the model to correctly predict presence of flat reef was low, as suggested by low sensitivity (34%). Sensitivity values of other substrates ranged from 48 to 76% (Table 3). For all substrates except sand, specificity values (correct absences) where higher than sensitivity values and ranged from 65 to 94% (Table 3). The total number correct predictions (both presence and absence) ranged from 66% for sand to 88% for obscured reef (Table 3).
Evaluation of Ecklonia and sessile invertebrate models showed good predictive power (AUC=0.94 and 0.80, respectively) and correct classification rates above 80%. Ecklonia had the best predictive performance with both high sensitivity (90%) and specificity (84%). Other macroalgae model performance was poorer (AUC=0.70), with low sensitivity and specificity values, and correct classification rate (Table 3). For the lobster model, high sensitivity increased overall performance with an AUC value of 0.88 and a correct classification rate of 80%.
In most instances, the effect of spatial dependence on deviance explained by the model or model accuracy was minimal (less than 5%) (Table 3). Deviance explained by the Ecklonia and lobster models was reduced by 9% and 10% respectively. However, model accuracy was good according to accepted statistical tests, with AUC values above 0.79 and correct classifications above 73%.
Final maps provide a detailed representation of dominant substrates and biota in the study area (Fig. 3, Fig. 4 and Fig. 5). The maps document a clear gradient of Ecklonia dominance on the ridge and lee of the reef that aligns with lobster distribution. Sessile invertebrates dominate the flat reef structure in deeper waters (>65 m) and sandy substrates in shallow waters. Rhodoliths tend to dominate on the shallow side of the reef and extend into the sandy substrates to the east. Some mapping artefacts are visible in areas of soft substrates, where boat motion likely introduced striping effects.
The development of spatially explicit, detailed habitat maps in this study has allowed us to accurately represent the benthic environment in an area of known importance for the western rock lobster. The models can be used to explore complex geomorphic characteristics and the major drivers in benthos distributions in deep water lobster habitats. For three of the four biota categories (including lobster), more than 60% of the variation in distribution could be explained by depth and geophysical variables. This confirms that depth and geomorphology are the principle drivers of biota distribution at this location. Moreover, these variables are ecologically relevant as they reflect important physiological (e.g., light requirements), environmental or ecological (e.g., hard substrates to attach holdfasts) limitations , .
Though time consuming and expensive to produce, habitat maps are critical for assessing the spatial relationship of important marine resources to their environment. For example, habitat suitability maps for European lobsters showed that they are more abundant on the lee side of peaks or ridges , where it has been suggested that they feed most actively. Similarly, the western rock lobster also appears to have a preference for the leeward side of the reef ridge (Fig. 3 and Fig. 5). Based on our current understanding of lobster ecology, it was expected that substrate types such as boulders , coral reef  or rocky reefs with mixed kelp assemblages , , could be significant variables in defining lobster distribution. However, our model for lobsters did not incorporate any of the substrate or biota classes developed from towed video surveys.
Integrating geo-referenced pots, hydroacoustic surveys and detailed habitat maps has been demonstrated as an effective and efficient method in a number of lobster studies –. In this study, the combined benefits of pots and seafloor mapping have revealed significant and somewhat unexpected effects of habitat on lobster distribution. We found that geophysical variables based on local neighbourhood analysis were key predictors of lobster distribution, suggesting that a fine scale mix of terrain is important to western rock lobster. The selection of geophysical variables indicates a preference for geomorphologic complexity. Studies on different lobster species around the world have also shown a strong association between sea floor complexity and lobster distribution , –. Habitat complexity is mainly used as a proxy for shelter quality, where highly complex habitats provide greater protection from predators.
Complex habitats may also support high abundances of organisms that serve as food for lobsters, as shown in seagrass meadows . On the other hand, food resources are sometimes spatially segregated from refuge habitats ,  and is supported by the strong association between lobster abundance and proximity to edges of habitats that differ in shelter quality and resource availability , . Thus, the geophysical variables used in the model in this study (hypsometric index, focal analysis, and range) may typify a landscape with a juxtaposition of habitats suitable for both sheltering and foraging for western rock lobster. Lobster behaviours such as feeding, foraging, social interaction, movement and migration are regulated by a combination of chemosensors, hormones and magnetoreceptors. The inclusion of environmental variables that relate to the sensory biology, such as oceanographic conditions (e.g. wave energy, water currents) will further help to understand the complex relationship of lobsters to the range of habitats they utilise.
An earlier study on the relationship between the western rock lobster and habitat established a significant (though moderate) association, with mixed assemblages dominated by Ecklonia sp. accounting for 28% of variation in abundance of western rock lobster .While Ecklonia and reef are abundant at the study site, we found that a fine scale mix of terrain is the strongest driver of lobster distributions. The approach used in this study provides a more holistic view of the ecological and spatial structure of the western rock lobster and their preferred habitats. Crucially, the relationship between fine scale features in the landscape and lobster distribution will have strong implications for the spatial management of the western rock lobster fishery. The fundamental first step will be assessing how much preferred habitat is available to the western rock lobster within its geographical range.
The difference in model variables in the two western rock lobster studies likely relates to the scale of characterisation of habitats. Bellchambers et al.  used a broad-scale presence-absence approach, while our study used bathymetry data (at 3×3 m pixel resolution) to characterise terrain down to 12.5 m kernels (Table 1). The role of scale in habitat characteristics on abundance and distribution has been investigated in the California spiny lobster (P. interruptus)  and discussed with respect to the American lobster (Homarus americanus) . These studies suggest that lobsters respond to habitat characteristics at local (e.g. shelter) and landscape scales (e.g. kelp dominated habitat), with microhabitat influencing the distribution of lobsters within patches of broadly suitable habitat. The results of the present study, in conjunction with Bellchambers et al. , suggest similar broad and fine scale habitat characteristics influence the distribution of the western rock lobster.
The development of spatially-explicit, detailed distribution maps in this study allowed an area of known importance to the western rock lobster to be accurately characterised, despite the depth (60–90 m). We were able to demonstrate for the first time that kelp and reef are not the intrinsic drivers of lobster distribution off the coast of Western Australia, as first thought. Clearly, kelp and reefs do play a role in the ecological structure of western rock lobsters, as they align closely with lobster distribution. The use of species distribution models in this study, however, shows that it is the fine scale features within the geophysical landscape that strongly influence patterns in lobster distribution. Our results imply that the juxtaposition of habitat, terrain and scale are most likely to define the ecological niche of western rock lobster. In essence, the preferred habitat is not limited by the presence of reef or kelp, rather a geomorphic complexity that is likely related to the presence of quality shelter and food resources. This study provides new and critical information to the ecological and spatial structure of western rock lobster which will be an integral part of the way ecosystem –based fisheries management (EBFM) is implemented for the sustainability of the western rock lobster fishery. For example, we can start to predict the carrying capacity of the WRL, based on how much preferred habitat is available within its geographical range. Congruent with this is the chance to assess the potential impacts of fishing without the confounding effects of habitat quality . More importantly, understanding the spatial structure of habitats as perceived by the organism in question is crucial to gaining insight into the ecological processes necessary for population persistence and maintenance, which are fundamental to managing a sustainable fishery. The next step is to determine the most appropriate landscape configuration, composition and scale to refine the ecological niche for western rock lobster. However, understanding the interaction of the organism with their physical habitat must extend past the just sexually mature lobsters (brood stock) and move towards conserving an intact life history and all the habitats associated with different life stages.
The authors would like to thank Scott Evans and Nick Konzewitsch and the crew of Kool Change and Si-Bon III for help with lobster sampling; Fugro Survey Pty. Ltd. for hydroacoustic data collection and processing; Kris Waddington and Centre for Marine Futures for towed video analysis; Ben Radford and Karen Holmes for their help with script development for distribution models. Finally, we thank the reviewers whose constructive feedback resulted in an improved manuscript.
Conceived and designed the experiments: LMB MBP. Performed the experiments: LMB MBP. Analyzed the data: RKH KVN. Contributed reagents/materials/analysis tools: RKH KVN. Wrote the paper: RKH KVN.
- 1. Jackson JBC, Kirby MX, Berger WH, Bjorndal KA, Botsford LW, et al. (2001) Historical Overfishing and the Recent Collapse of Coastal Ecosystems. Science 293: 629–637.
- 2. Lotze HK, Lenihan HS, Bourque BJ, Bradbury RH, Cooke RG, et al. (2006) Depletion, Degradation, and Recovery Potential of Estuaries and Coastal Seas. Science 312: 1806–1809.
- 3. Pandolfi JM, Bradbury RH, Sala E, Hughes TP, Bjorndal KA, et al. (2003) Global Trajectories of the Long-Term Decline of Coral Reef Ecosystems. Science 301: 955–958.
- 4. Sala E, Knowlton N (2006) Global Marine Biodiversity Trends. Annual Review of Environment and Resources 31: 93–122.
- 5. Worm B, Barbier EB, Beaumont N, Duffy JE, Folke C, et al. (2006) Impacts of Biodiversity Loss on Ocean Ecosystem Services. Science 314: 787–790.
- 6. Halpern BS, Lester SE, McLeod KL (2010) Placing marine protected areas onto the ecosystem-based management seascape. Proceedings of the National Academy of Sciences 107: 18312–18317.
- 7. Marasco RJ, Goodman D, Grimes CB, Lawson PW, Punt AE, et al. (2007) Ecosystem-based fisheries management: some practical suggestions. Canadian Journal of Fisheries and Aquatic Sciences 64: 928–939.
- 8. Levin PS, Fogarty MJ, Murawski SA, Fluharty D (2009) Integrated Ecosystem Assessments: Developing the Scientific Basis for Ecosystem-Based Management of the Ocean. PLoS Biol 7: e1000014.
- 9. Smith ADM, Fulton EJ, Hobday AJ, Smith DC, Shoulder P (2007) Scientific tools to support the practical implementation of ecosystem-based fisheries management. ICES Journal of Marine Science: Journal du Conseil 64: 633–639.
- 10. Laidig TE, Watters DL, Yoklavich MM (2009) Demersal fish and habitat associations from visual surveys on the central California shelf. Estuarine, Coastal and Shelf Science 83: 629–637.
- 11. Cardinale M, Bartolino V, Llope M, Maiorano L, Sköld M, et al. (2011) Historical spatial baselines in conservation and management of marine resources. Fish and Fisheries 12: 289–298.
- 12. Galparsoro I, Borja Á, Bald J, Liria P, Chust G (2009) Predicting suitable habitat for the European lobster (Homarus gammarus), on the Basque continental shelf (Bay of Biscay), using Ecological-Niche Factor Analysis. Ecological Modelling 220: 556–567.
- 13. Chatfield BS, Van Niel KP, Kendrick GA, Harvey ES (2010) Combining environmental gradients to explain and predict the structure of demersal fish distributions. Journal of Biogeography 37: 593–605.
- 14. Harper SJM, Bates CR, Guzman HM, Mair JM (2010) Acoustic mapping of fish aggregation areas to improve fisheries management in Las Perlas Archipelago, Pacific Panama. Ocean & Coastal Management 53: 615–623.
- 15. Phillips BF, Melville-Smith R (2005) Sustainability of the Western Rock Lobster Fishery: A Review of Past Progress and Future Challenges. Bulletin of Marine Science 76: 485–500.
- 16. de Lestang S, Thomson A, Rossbach M (2010) West Coast Rock Lobster Fishery status report. Department of Fisheries, Western Australia 39–48.
- 17. Cobb J (1981) Behaviour of the Western Australian spiny lobster, Panulirus cygnus George, in the field and laboratory. Marine and Freshwater Research 32: 399–409.
- 18. Bellchambers LM, Evans SN, Meeuwig JJ (2010) Abundance and size of western rock lobster (Panulirus cygnus) as a function of benthic habitat: implications for ecosystem-based fisheries management. Marine and Freshwater Research 61: 279–287.
- 19. Guisan A, Zimmermann NE (2000) Predictive habitat distribution models in ecology. Ecological Modelling 135: 147–186.
- 20. De'ath G, Fabricius KE (2000) Classification and regressions trees: a powerful yet simple technique for ecological data analysis. Ecology 81: 3178–3192.
- 21. Maravelias CD, Reid DG, Swartzman G (2000) Seabed substrate, water depth and zooplankton as determinants of the prespawning spatial aggregation of North Atlantic herring. Marine Ecology Progress Series 195: 249–259.
- 22. Pittman SJ, Christensen JD, Caldow C, Menza C, Monaco ME (2007) Predictive mapping of fish species richness across shallow-water seascapes in the Caribbean. Ecological Modelling 204: 9–21.
- 23. Bello PJ, Rios LV, Liceaga CMA, Zetina MC, Cervera CK, et al. (2005) Incorporating spatial analysis of habitat into spiny lobster (Panulirus argus) stock assessment at Alacranes reef, Yucatan, México. Fisheries Research 73: 37–47.
- 24. Breiman L, Friedman JH, Olshen RA, Stone CJ (1984) Classification and regression Trees. New York: Chapman and Hall.
- 25. Chambers JM, Hastie TJ (1993) Statistical Models. London: Chapman and Hall.
- 26. Moore CH, Harvey ES, Van Niel KP (2009) Spatial prediction of demersal fish distributions: enhancing our understanding of species–environment relationships. ICES Journal of Marine Science: Journal du Conseil 66: 2068–2075.
- 27. Holmes KW, Van Niel KP, Radford B, Kendrick GA, Grove SL (2008) Modelling distribution of marine benthos from hydroacoustics and underwater video. Continental Shelf Research 28: 1800–1810.
- 28. Melville-Smith R, Chubb CF, Caputi N, Cheng YW, Christianopoulos D (1998) Fishery independent survey of the breeding stock and migration of the western rock lobster (Panulirus cygnus). Canberra: Fisheries Research and Development Corporation.
- 29. Legendre P (1993) Spatial Autocorrelation: Trouble or New Paradigm? Ecology 74: 1659–1673.
- 30. Legendre P, Dale MRT, Fortin M-J, Gurevitch J, Hohn M, et al. (2002) The consequences of spatial structure for the design and analysis of ecological field surveys. Ecography 25: 601–615.
- 31. Anderson MJ, Miller RB (2004) Spatial variation and effects of habitat on temperate reef fish assemblages in northeastern New Zealand. Journal of Experimental Marine Biology and Ecology 305: 191–221.
- 32. Tremblay MJ, Smith SJ, Todd BJ, Clement PM, McKeown DL (2009) Associations of lobsters (Homarus americanus) off southwestern Nova Scotia with bottom type from images and geophysical maps. ICES Journal of Marine Science: Journal du Conseil 66: 2060–2067.
- 33. Acosta CA (1999) Benthic Dispersal of Caribbean Spiny Lobsters among Insular Habitats: Implications for the Conservation of Exploited Marine Species. Conservation Biology 13: 603–612.
- 34. Parnell PE, Dayton PK, Margiotta F (2007) Spatial and Temporal Patterns of Lobster Trap Fishing: A Survey of Fishing Effort and Habitat Structure. Bulletin, Southern California Academy of Sciences 106: 27–37.
- 35. Dunnington MJ, Wahle RA, Bell MC, Geraldi NR (2005) Evaluating local population dynamics of the American lobster, Homarus americanus, with trap-based mark-recapture methods and seabed mapping. New Zealand Journal of Marine and Freshwater Research 39: 1253–1276.
- 36. Geraldi NR, Wahle RA, Dunnington M (2009) Habitat effects on American lobster (Homarus americanus) movement and density: insights from georeferenced trap arrays, seabed mapping, and tagging. Canadian Journal of Fisheries and Aquatic Sciences 66: 460–470.
- 37. O'Donnell KP, Wahle RA, Bell M, Dunnington M (2007) Spatially referenced trap arrays detect sediment disposal impacts on lobsters and crabs in a New England estuary. Marine Ecology Progress Series 348: 249–260.
- 38. Rios-Lara V, Salas S, Javier B-P, Ayora PI (2007) Distribution patterns of spiny lobster (Panulirus argus) at Alacranes reef, Yucatan: Spatial analysis and inference of preferential habitat. Fisheries Research 87: 35–45.
- 39. Mai TT, Hovel KA (2007) Influence of local-scale and landscape-scale habitat characteristics on California spiny lobster (Panulirus interruptus) abundance and survival. Marine and Freshwater Research 58: 419–428.
- 40. Herrnkind WF (1980) Spiny lobsters: Patterns of movement. In: Cobb JS, Phillips BF, editors. The biology and management of lobsters (Volume 1). New York: Academic. pp. 349–407.
- 41. Lawton P, Lavalli KL (1995) Postlarval, juvenile, adolescent and adult ecology. In: Factor JR, editor. Biology of the lobster. San Diego: Academic Press. pp. 47–88.
- 42. Polovina JJ, Haight WR, Moffitt RB, Parrish FA (1995) The Role of Benthic Habitat, Oceanography, and Fishing On the Population Dynamics of the Spiny Lobster, Panulirus Marginatus (Decapoda, Palinuridae), in the Hawaiian Archipelago. Crustaceana 68: 203–212.
- 43. Selgrath JC, Hovel KA, Wahle RA (2007) Effects of habitat edges on American lobster abundance and survival. Journal of Experimental Marine Biology and Ecology 353: 253–264.
- 44. Wynne SP, Côté IM (2007) Effects of habitat quality and fishing on Caribbean spotted spiny lobster populations. Journal of Applied Ecology 44: 488–494.