© 2006 International Council for the Exploration of the Sea
Habitat suitability modelling of economically important fish species with commercial fisheries data
Primary Industries Research Victoria – Marine and Freshwater Systems PO Box 114, Queenscliff, Victoria 3225, Australia
*Correspondence to L. Morris: tel: +61 3 5258 0111; fax: +61 3 5258 0270. e-mail: liz.morris{at}dpi.vic.gov.au.
In this study we used catch and effort data from a commercial fishery to generate habitat suitability models for Port Phillip Bay, Victoria, Australia. Species modelled were King George whiting (Sillaginodes punctata), greenback flounder (Rhombosolea tapirina), Australian salmon (Arripis trutta and A. truttaceus), and snapper (Pagrus auratus). Locations of commercial catches were reported through a grid system of fishing blocks. Spatial analyses in a Geographic Information System (GIS) were applied to describe each fishing block by its habitat area. A multivariate approach was adopted to group each fishing block by its dominant habitats. Standardized catch per unit effort values were overlaid on these groups to identify those that returned high or low catches for each species. A simple set of rules was then devised to predict the habitat suitability for each habitat combination in a fishing block. The spatial distribution of these habitats was presented in a GIS. These habitat suitability models were consistent with existing anecdotal information and expert opinion. While the models require testing, we have shown that in the absence of adequate fishery-independent data, commercial catch and effort data can be used to produce habitat suitability models at a bay-wide scale.
Keywords: catch and effort, fisheries habitat, Geographic Information Systems, multivariate analysis, Port Phillip Bay
Received 25 May 2005; accepted 26 June 2006.
 |
Introduction |
|---|
 Top Introduction MethodsResultsDiscussionConclusionsReferences |
|---|
The protection of fishery habitat is a vital part of ecosystem-based approaches to fisheries management, and acknowledges that fish populations should not be considered independently of their environment (Sharp, 1997; Parsons and Harrison, 2000). Despite this, we often lack detailed information about the importance of different habitat types to fish species, so may be failing to provide adequate protection for important habitats. Where fish-habitat association data exist, it is possible to combine them with habitat data in a Geographic Information System (GIS) to provide a spatially explicit model of habitat suitability (Gallaway et al., 1999; Rubec et al., 1999, 2001, 2003; Brown et al., 2000; Guisan and Zimmermann, 2000; Stoner et al., 2001).
Commercial fisheries' catch and effort data from vessel monitoring systems and logbooks are routinely collected for stock assessment and fishery management purposes in Australia and internationally. Such data have intrinsic problems when used as a surrogate for fishery-independent data, relating to the scale of information, sampling bias, reporting issues, and confidentiality (Mace, 1997; Starr and Fox, 1997; Zheng et al., 2001; Gallaway et al., 2003a, b). Most commercial fisheries data do not have precise geographic coordinates to define spatial location; more often they use a system of coarse-scale grids for fishers to record catch locations. Further, associated environmental data are not typically recorded with the catch information (Rubec, 1996). Despite these problems, fisheries scientists and managers are increasingly interested in accessing the large amount of information that exists within the fishing community (Bowen, 1997; Maurstad, 2002).
In areas where characteristics of the fishery are well known, one way of sourcing fisher knowledge is to use commercial catch and effort data. An implicit assumption in using these data is that the regions in which fishers are operating have the highest densities of the targeted species. Several recent studies have used logbook data and vessel monitoring systems to investigate the spatial distributions of fish, and have attempted to link this information with environmental data (Denis and Robin, 2001; Zheng et al., 2001; Denis et al., 2002; Kemp and Meaden, 2002; Marrs et al., 2002; Reynolds, 2003). In this study, we extend this approach to predict the distribution of suitable habitats, and by extension, fish distributions, based on commercial catch and effort data in Port Phillip Bay, Victoria, Australia (Figure 1).
|
Port Phillip Bay is a large and relatively shallow marine embayment. It is characterized by predominantly bare sand, silt, and clay sediments, with extensive shallow seagrass beds bordering the southern and western shores (Figure 2). The bay is linked to the oceanic waters of Bass Strait through a narrow entrance called The Rip, and the tidal range in most of the bay is <1 m. Its fishery is managed through a licencing system with restrictions on fishing gear types, minimum allowable fish sizes, fishing seasons, and spatial closures (e.g. marine national parks). There are 59 licenced commercial fishers operating in the Bay, who primarily fish from small vessels with seine-nets (haul, purse, and beach), mesh-nets (also known as gillnets), and longlines (Anon., 2001).
|
In this study we analysed catch and effort data for King George whiting (Sillaginodes punctata), greenback flounder (Rhombosolea tapirina), Australian salmon (Arripis spp.), and snapper (Pagrus auratus) to produce habitat suitability models. These species are all major components of the Port Phillip Bay fishery and include demersal species (King George whiting, greenback flounder, snapper) and a pelagic predator (Australian salmon). We assumed that models for the demersal species would be more reliable because of the closer association of these species with the types of habitat parameters investigated in the study.
The fishery for King George whiting and Australian salmon targets sub-adults, with minimum catch sizes of 27 and 21 cm, respectively. The snapper fishery can be divided into a longline fishery that targets primarily adult fish, and a haul-seine/mesh-net fishery that primarily targets sub-adults, colloquially known as "pinkies". Sub-adult snapper have a minimum legal catch size of 27 cm, whereas adults may reach lengths >80 cm. The greenback flounder fishery targets adults with a minimum catch size of 23 cm.
Haul-seining is restricted to the shallower areas of the bay. Mesh-nets are used throughout the bay, but most of the effort of this gear type is in the same fishing blocks as the haul-seines. Seine-nets exceeding 460 m in length are not permitted, and the most frequently used mesh sizes for haul-seines range from 30 to 100 mm. Seasonal restrictions apply to the size and length of mesh-nets permitted in the bay, with mesh sizes mostly ranging from 60 to 130 mm.
Longlines consist of a monofilament main line weighted at each end with a maximum of 200 hooks typically attached at 10-m intervals by 1-m snoods (Coutin, 2000). Fishers are only permitted to deploy one longline at a time, and these are typically used in the deeper areas of the bay away from possible seabed snags, interference from recreational fishers and boating, and where the bycatch of low value species is likely to be minimized.
|
Methods |
|---|
|
TopIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
Habitat data
GIS polygon layers for depth, sediment type, and substratum type/biota provided the habitat information to characterize each fishing block (Figure 2). The depth layer was produced by digitizing depth contours from 1:25 000 hydrographic charts sourced from the Port of Melbourne Corporation. The sediment type layer was digitized from a 1:100 000 seabed sediment map presented in a study of grain-size distribution throughout the bay (PMA, 1987). A substratum type/biota polygon layer at a scale of 1:25 000 was available from mapping of seagrass distribution, through interpretation of high-resolution colour aerial photography combined with extensive ground-truthing (Blake and Ball, 2001). Because Port Phillip Bay is predominantly a marine system, salinity and water temperature were not considered to be significant influences on the distribution of the fish species investigated in the study.
All spatial analyses and development of habitat suitability maps were undertaken with the GIS software ARCINFO and ArcView. To determine the habitat characteristics of each fishing block, we combined all habitat layers into a single layer in the GIS. The Identity command in ARCINFO was used to overlay the fishing block layer with substratum type/biota, depth, and sediment polygon layers, and to calculate the geometric intersection of each layer (Figure 3). Two layers were intersected at a time with the Identity command, and the output of the process formed one of the input layers to intersect with the next layer (i.e. a geometric intersection was calculated on the fishing block and substratum type/biota layers first, then the output from this was intersected with the depth layer, and so on until all layers had been intersected). The final output of this process was a single combined layer that retained the spatial features and attributes for each of the input layers (Figure 3). A composite habitat code for each feature in the output layer was then calculated by combining the habitat codes from each input layer. The attributes of the GIS habitat layers used in this analysis are summarized in Table 1.
|
|
The attribute table for the combined fishing block/habitat layer from the Identity process (Figure 3) included all the attribute values from the input layers as well as the area in m2 of each combined spatial feature. We added a further column to the table, consisting of a "composite" habitat code generated by combining the respective codes for substratum type/biota, depth, and sediment into a single code (Table 1). This table was exported to Excel and a pivot table created to summarize each fishing block by the total area of every possible combination of the habitat parameters. In all, 135 habitat combinations present in Port Phillip Bay were identified in the analysis (Table 1).
Fishery catch and effort data
Commercial fishers in Port Phillip Bay are required to submit catch and effort data each month in the form of logbooks that include daily records of the time spent fishing, gear types used, species and weight of the catch, and the catch location. Fishers record the location of their catch through a system of 41 fishing blocks, based on a 5-min grid (approximately 9 km x 9 km; Figure 1). Fishers are required to record the block(s) where the majority of their catch was taken. The current system of fishing blocks was introduced in 1998. Prior to 1998, catch and effort returns for Port Phillip Bay were based on only seven catch blocks, which did not provide an adequate spatial resolution for the analyses in the present study. As a result, the analyses presented here only used catch and effort data for the three years between autumn 1998 and summer 2001 (Table 2).
|
The Port Phillip Bay catch and effort data are stored in a relational database by Primary Industries Research Victoria (PIRVic) on behalf of the state fishery management agency, Fisheries Victoria. To assist querying and displaying this data, PIRVic Marine and Freshwater Systems developed a customized ArcView GIS application known as Catch and Effort Info (Ball and Coots, 2001). This system was used to extract the data required for this study.
We used catch per unit effort (cpue) values, where effort was measured by metre-lifts for mesh-nets, number of shots for haul-seines, and number of hook-lifts for longlines. Catches were recorded in tonnes. One problem with this type of fishery-dependent data is that cpue data tend to be at different scales across different gear types owing to the various units of measurements and the differing gear efficiencies (Hilborn and Walters, 1992). A recommended approach to this problem is to standardize cpue data to provide a consistent index of a species' abundance (Hilborn and Walters, 1992). We chose a simple method to standardize cpue values across the different gear types in which we assumed that the average cpue of each gear type represented a similar density of fish. We divided the cpue within each fishing block for a specific gear type by the average cpue for that gear type over the whole bay. Once the cpue values were standardized so that data from all gear types were effectively unit-less and at the same scale, we combined these relative values by calculating the mean relative cpue for each fishing block.
Statistical analyses
In the present study we were primarily interested in spatial rather than temporal patterns, so we used interaction plots (Quinn and Keough, 2002) to determine whether the pattern of relative cpue was consistent across fishing blocks between seasons. There was no evidence of an interaction between fishing block and season, except for adult snapper, of which there were low to non-existent catches in winter. As we were primarily interested in spatial patterns in fish distributions and their relation to benthic habitats, and not in differences among years or seasons, data were pooled across all seasons and years, so that each fishing block had one value for each species. Because of low catches, we excluded winter data for adult snapper from the analysis.
To link fish distributions with habitat parameters, we assumed that habitats with higher cpue would also have greater habitat suitability for that particular species. Because of problems associated with working with commercial catch data, and in particular the differences in spatial scale between catch and habitat data, a multivariate approach was used to link benthic habitat to fish distributions. The first step was to create a data matrix of habitat combinations that was independent of the differences in the spatial size of fishing blocks (Figure 1). We took the summary table of fishing blocks vs. total area of habitat combinations from the spatial analyses described previously (Figure 3), and calculated the proportion (percentage) by area of each habitat combination in a fishing block vs. the total area of that fishing block. The resulting data matrix consisted of an array of rows (habitat combinations) and columns (fishing blocks). Then we created a similarity matrix in which the similarities between each fishing block were calculated using the Bray–Curtis coefficient (Bray and Curtis, 1957).
To test the a priori hypothesis that fishing blocks where a species was caught differed in their habitat parameters from those where no fish were caught, an analysis of similarities (ANOSIM) was carried out. This is a non-parametric randomization procedure that provides an R-statistic that ranges between –1 and +1, and a probability of getting this R-statistic if the null hypothesis is true (Clarke, 1993). Where there was a significant difference between the habitat parameters of fishing blocks with fish and blocks without (p < 0.05), the relationship between the habitat parameters of those fishing blocks was further explored with ordination and cluster analyses. For the ordination, non-metric multidimensional scaling (nMDS) was used in an attempt to place the fishing blocks on a "map" in such a way that the rank order between the fishing blocks represented the rank order of the similarities in the similarity matrix (Clarke and Warwick, 2001). The cluster analysis progressively links the samples based on the calculated similarities among hierarchical groups, and the analysis is represented in the form of a dendrogram (Clarke and Warwick, 2001). Primer Version 5 (© PRIMER-E 2000) was used for all multivariate analyses.
Fishing blocks were grouped arbitrarily according to the cluster analysis, so fishing blocks that had at least 40% similarity were considered to have similar habitat combinations. These groupings were then overlaid on the ordination, with the relative cpue values of each species positively related to the size of a bubble plot. The groups of fishing blocks determined from the cluster analysis were designated as either high or low density groups according to the size of the relative cpue values within each group for each species. This step was based on the assumption that the relative cpue values were positively related to the actual densities of a species. A series of simple rulings were then used to determine whether each habitat combination was of "high", "medium", or "low" suitability for the species in question (Table 3). These rulings were based on the assumption that the consistent presence of a habitat parameter in a cluster group would be important in determining the density of that species, based on its mean cpue. Primer Version 5 was used to produce the bubble plots and to extract the information on the presence of habitat parameters in the cluster groups.
|
Once each habitat combination was defined as high, medium, low, or undefined, the polygons in the combined habitat layer (Figure 3) were reclassified in the GIS to their corresponding suitability value to create a predictive model in the form of a map of habitat suitability. Ideally the habitat suitability models would have been validated with fishery-independent data, but suitable data were not available. However, numerous studies have investigated different aspects of the fishery in Port Phillip Bay over the years. Port Phillip Bay also has a large recreational fishery, and there is considerable anecdotal information and expert opinion detailing fish distributions within the bay. This information was used to provide a qualitative test of model predictions to assess whether use of commercial catch data for modelling habitat suitability was a valid approach.
|
Results |
|---|
|
TopIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
There were significant differences (ANOSIM) in habitat characteristics between fishing blocks where fish had and had not been caught, for all species examined (Table 4). Ten cluster groups were defined from the 40% similarity level (Figures 4a and 5), and there was a good correspondence between cluster groupings and the 2-dimensional nMDS ordination (Figure 4b). Blocks A6 and F4 clustered out singly (Figure 4a) and could not be used in the analysis described above because more than one block was required per cluster group. Fishing block E4 (Swan Bay) was excluded from the cluster analysis because no fishing is permitted in the area. Fishing block E9 clustered out in group 1 (Figure 4a), but this area was excluded from cluster group 1 because it appeared to be a large outlier in several cases; this is discussed further below.
|
|
|
King George whiting (Sillaginodes punctata)
The cluster groups designated as high and low density for King George whiting are shown in Figure 6a. The relative cpue values for King George whiting were greatest in block G6 (Figure 6a), at the southern end of the bay (Figure 1). The correspondence between cluster groupings and the relative cpue values was reasonably good across all groups. The only exceptions were the moderate relative cpue values in areas B7, B9, and D6, all of which were in cluster groups assigned low density, and the comparatively low cpue in area C5, which was in a high density cluster group (Figure 6a).
|
Reclassifying the habitat composites with suitability categories following our simple ruling system (Table 3) resulted in the predictive model of habitat suitability for King George whiting (>27 cm) shown in Figure 6b. The most notable feature of this model was the high suitability of all habitats that include seagrass or seagrass-edge, which are primarily in the southern and western areas of the bay. The shallow bare areas on fine sediment in the northern part of the bay as well as the reef areas along the northeastern shores of the bay were also predicted to provide suitable habitat for King George whiting. The areas classed as low suitability habitats were mainly the deeper bare substratum in the centre of the bay, and the coarse sand habitat on the eastern side of the bay.
Greenback flounder (Rhombosolea tapirina)
The catch of greenback flounder came from exactly the same cells as King George whiting, and may well be a bycatch of the more highly valued King George whiting. As a result, the habitat suitability model was identical to that of King George whiting (Figure 6b).
Australian salmon (Arripis trutta and A. truttaceus)
The relative cpue values of Australian salmon were dominated by the large value for fishing block E9 (Figure 7a), but we treated this block as an outlier (see below). There was good agreement with the cpue values and the classification of the cluster groups to high or low density. The only exceptions were fishing blocks C5 and F5, which had low relative cpue values and were in cluster groups assigned to high density. The resultant suitability codes and habitat suitability model (Figure 7b) were similar to that of King George whiting, in that the majority of seagrass-associated habitat and shallow fine sediment were classified as high suitability, whereas the low suitability habitat was again the deeper central region of the bay. The main difference in predicted habitat suitability from that of King George whiting was the shallow strip of coarse sand along the eastern edge of the bay, which was classed as medium suitability, and the shallow strip of sandy sediment along the western shore, which was classed as high suitability.
|
Sub-adult snapper (Pagrus auratus)
The majority of the catch of sub-adult snapper came from cluster groups 4, 5, 8, and 9 (Figure 8a), with fishing block A7 in group 5 dominating the relative cpue values. There was a reasonably good agreement between the assignment of cluster groups to high or low density groupings and to the relative cpue values. The main exceptions were blocks B9 and F5, both of which were in cluster groups assigned low density ratings even though they had moderate relative cpue values, and C5, which had a low relative cpue and was in a high density cluster group. The predictive map of habitat suitability for sub-adult snapper was similar to those of the other species, but it had more high suitability habitat in the north of the bay and at the entrance to the Geelong Arm, and more low suitability habitat around the southern and eastern fringes of the bay (Figure 8b).
|
Adult snapper (Pagrus auratus)
The catch of adult snapper was mainly in different fishing blocks from those of the other species (Figure 9a), and the assignment of cluster groups to low or high density groups was reasonably consistent with the relative cpue values. Cluster group 5 was assigned a high density rating, although blocks A8, A7, and C3 had comparatively low relative cpue values; this group could arguably have been assigned a low density rating, but all blocks within the group did return at least some adult snapper. The habitat suitability map (Figure 9b) for adult snapper differs from those for the other species. Suitable habitat for adult snapper was predicted to be in the deeper areas of the bay, while shallow seagrass habitat and the coarser sediment on the eastern side of the bay were predicted to be of low habitat suitability for this life history stage (Figure 9b).
|
|
Discussion |
|---|
|
TopIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
The habitat suitability models for King George whiting, greenback flounder, Australian salmon, and sub-adult snapper emphasized the importance of shallow habitat, but highlighted subtle differences between species. Validating and testing the habitat suitability models presented here was hindered by a lack of fishery-independent data. Although data exist for some species, the unequal spatial distribution and the concentration of sampling over only some of the habitat types (mostly bare sediment in depths >7 m) restricted data utility in model validation. As a consequence, we provide a qualitative assessment of the overall patterns of suitable habitat distribution.
The habitat suitability model for sub-adult King George whiting (Figure 6b) was consistent with the existing information on habitat affinities and distributions for the species. King George whiting recruit into shallow, seagrass-dominated areas, and move into reef and bare shallow areas as they become older (Fowler and McGarvey, 1995; Smith and MacDonald, 1997; Jenkins and Wheatley, 1998). Our model predicted that the sheltered, shallow, and seagrass habitats were of high suitability, including Swan Bay (fishing block E4) and the seagrass areas in fishing block G6, and both are important nursery areas for the species (Jenkins et al., 1993; Jenkins and Hamer, 2001). The strip of low suitability habitat along the eastern side of the bay corresponds to results from fishery surveys (Parry et al., 1995) and recreational angling returns (Coutin et al., 1995; Conron and Coutin, 1998), and appears consistent with a lack of structural habitat and a coarse sandy sediment (Figure 2). While independent validation of the model will be necessary, the consistency of the model for King George whiting with other sources of information suggests reasonable confidence in the model predictions.
Greenback flounder are probably less restricted to the shallow areas than the model suggests (Figure 6b; Kuiter, 1993; Gomon et al., 1994). Data from trawl surveys in Port Phillip Bay report flounder in the deeper, more central areas of the bay (Parry et al., 1995). However, most flounder are caught with haul-seines, and fishing effort using this gear type is concentrated in shallower water (<10 m). Mesh-nets deployed in slightly deeper areas do not target flounder particularly well, creating a bias in the data towards the shallower areas for this species. Flounder are also associated with bare organic-rich substratum and have been recorded in large numbers in the bare areas interspersed between patches of seagrass in Swan Bay (Jenkins et al., 1993). The high suitability area predicted in Corio Bay is compatible with the large areas of seagrass and organic-rich clays, and similarly the high suitability classification around the mouth of the Yarra River at the northern end of the bay also corresponds with an organic-rich clay sediment.
Juvenile Australian salmon recruit to a wide range of coastal habitats, from medium energy sandy areas to sheltered mangrove-lined tidal creeks (Jones, 1999). As they grow older, schooling behaviour becomes more apparent. In Port Phillip Bay they have been described as transient and gregarious, and have been recorded from shallow water over mosaics of seagrass and rocky reef interspersed with patches of unvegetated sand (Hindell et al., 2000a, b). Dietary studies from Port Phillip Bay also showed that Australian salmon consume juveniles of seagrass-associated fish (Hindell et al., 2000b). The habitat suitability model (Figure 7b) for Australian salmon in Port Phillip Bay is fairly consistent with this information, with a wide range of shallow habitats classed as of high suitability.
Fishing block E9 (Figure 1) had only a small amount of habitat classed as highly suitable for Australian salmon, despite the very large catch of the species there. Fishers utilizing the area tend to be based locally and be very experienced, and they target transient schools of Australian salmon. The schooling behaviour of salmon combined with the targeted effort may make it possible to obtain very large catches from a small area of highly suitable habitat, or alternatively from a reasonably large area of medium suitability habitat. The comparatively small area of low suitability habitat for this species within the bay overall is also consistent with the fact that Australian salmon are pelagic predators and therefore less likely to be strongly tied to a particular benthic habitat. Rather, they are more likely to move relatively large distances through the water column in search of suitable prey (Hoedt and Dimmlich, 1994).
The habitat suitability models for sub-adult and adult snapper (Figures 8b and 9b) are considerably different. Sub-adult snapper were predicted to occur in the shallower areas in the northern and western parts of the bay; findings which correspond with studies of the Port Phillip Bay recreational fishery that primarily targets sub-adult snapper (Conron and Coutin, 1998). Although most of the shallow areas along the eastern edge of the bay were predicted to be of low suitability for sub-adult snapper, there are small areas of reef along this strip that were predicted to be of high suitability (Figure 2). This is consistent with information provided by recreational fishing guide books (Wilson, 1986; Classon and Wilson, 2002) and anecdotal evidence about this species.
Although the shallow reefs along the northeastern shores were identified as highly suitable for sub-adult snapper, the adjacent band of coarse sand up to depths of about 15 m (Figure 2) was predicted to be of low suitability (Figure 8b). However, the same area also features reefs, rubble, lace coral, and cunji beds (Pyura stolonifera), which are recognized as good fishing sites for sub-adult snapper by recreational fishers (Wilson, 1986; Classon and Wilson, 2002). Two studies (GHD, 1997; Hamer et al., 1997) also report the presence of P. stolonifera beds in the area, and Hamer et al. (1997) suggested that juvenile snapper may be associated with the presence of sessile organisms such as P. stolonifera. These habitats are not currently represented in the GIS substratum type/biota layer owing to the limitations of mapping to these depths in Port Phillip Bay from aerial photography, and the area is primarily defined as bare coarse sand at present (Figure 2). As a result, the presence of these habitats was not accounted for in either the analysis of fishing blocks vs. habitat variables, or in the production of the predictive habitat suitability models. These habitats are probably also avoided by commercial fishers because of the potential for snagging their nets.
Interestingly, Swan Bay (block E4) was predicted to be of primarily high suitability for sub-adult snapper owing to the intertidal seagrass coverage on muddy substratum, but anecdotal information and previous studies (e.g. Jenkins et al., 1993) suggest that they are not particularly abundant there. Good sub-adult catches are recorded in deeper areas immediately adjacent to shallow seagrass beds in other parts of the bay, so at the scale of modelling undertaken here, this may be influencing the classification of this habitat type as highly suitable.
Snapper move into deeper water with age (Gomon et al., 1994; Coutin, 2000), and the habitat suitability model for adult fish (Figure 9b) predicted that the deeper areas of Port Phillip Bay had the most suitable habitat for this life stage of snapper. Most of Geelong Arm/Corio Bay were classified as low suitability for adult snapper (Figure 9b), with the balance of the area being classified as medium suitability. Recreational fishers catch good numbers of adult snapper in the western Geelong Arm, and the area is recognized as having a good winter recreational fishery for the species. Commercial fishers do not longline in this relatively shallow and enclosed area, but instead use haul-seines that target sub-adult snapper. As a result, the presence of any adult snapper in the area would be underestimated by this method of modelling.
If the two habitat maps for snapper were to be combined, the majority of Port Phillip Bay would be predicted to be of high or medium suitability. This is consistent with available information about snapper and, in particular, dietary data. Snapper are demersal predators, but appear to eat a wide range of prey species from a variety of habitats (Parry et al., 1995; Coutin, 2000). They are also a highly aggregative species that can move considerable distances and so are likely to utilize a range of habitats, and the predictive maps of habitat suitability agree with this information.
The modelling approach
Although the available information suggests that, with the exception of greenback flounder, we can have confidence in the broad patterns predicted by the habitat suitability models, there were still some anomalies. A fishing block that had similar habitat parameters to blocks where catches were high could have a consistently low catch and effort for all species (e.g. C5). There may be a number of reasons for this type of anomaly, and they may affect our original assumption that fishers will target areas with the highest densities of fish. For example, there is a considerable amount of drifting macroalgae in block C5 (Blake and Ball, 2001), which may make fishing difficult or less cost effective. Alternatively, fishers may not consider it cost effective to travel large distances from port, or if they do will have less time available for fishing. As a result, fishing effort may not be equally distributed among highly suitable habitats, or certain gear types might be excluded from highly suitable areas.
There may also be sources of variation in fishing effort that we have not measured and that have the potential to affect fishing efficiency and, in turn, cpue. These may include intangibles such as the experience and skill of the particular fishers that work an area, as well as differences in technology used by fishers targeting different areas. Conversely, where there may be suitable habitat for a species, other environmental factors that have not been measured may also be important in determining distributions of a species; examples are hydrodynamics, effects of pollution, and the distribution of prey items or introduced marine pests. The reverse situation is one where a fishing block had a high relative cpue for several species, but had habitat more similar to blocks with low relative cpue values (e.g. E9). In that case, fishers might apply their experience and concentrate their efforts within a small area and, as previously discussed, if schooling fish are successfully targeted, relative cpue values can be very high.
An advantage of the approach we used in this study is that these anomalies in relative cpue did not affect the predictions we made about suitable habitat. Fishing blocks were classified into high or low density groups depending on the standardized cpue values for all blocks in a cluster group, so effectively we were averaging across all blocks that had similar habitat. At the same time, obvious outliers (such as block E9) were excluded from the analysis, although habitats within fishing blocks that were excluded from the analysis because no fishing took place in that block (e.g. E4) were still included in the predictive model of habitat suitability. This was as a result of all habitat combinations being assigned a suitability rating, so that predictions could be transferred to areas with no fisheries data as long as the same habitat combinations were present.
There are several important assumptions that underlie the method of determining suitable habitat presented in this paper. The first is that the areas targeted by fishers corresponded to high densities of the species under investigation. For species that are not the primary target, this may not be the case. For example, greenback flounder are mainly caught with haul-seines in the same areas as King George whiting (the primary target species). As haul-seines are not used in the deeper, unvegetated, soft-sediment habitats of Port Phillip Bay, that are known to provide suitable habitat for greenback flounder (Parry et al., 1995), the assumption that fishers only target areas of high density is unlikely to be true for this species. We also assumed that demersal species would provide more reliable habitat suitability models than pelagic species, but it appears that the degree to which a species is actively targeted is at least, if not more, important than the life history characteristics of that species. Similarly, the operational limitations of a gear type may also exert an influence on where species are caught. Longlines are not typically used in Corio Bay or most of the Geelong Arm because of the relatively shallow and enclosed nature of the area. Consequently, species targeted by these gear types and potentially present in the area, such as adult snapper, do not appear in the catch and effort data.
The second major assumption of the method relates to our prediction that the consistent presence of a habitat type in areas that have similar environmental attributes will be important in determining the yield of a fish species. In fact, we do not know over what scale habitat is important, and this type of scale-dependent habitat information would improve models of habitat suitability.
Modelling of habitat suitability was also influenced by the accuracy of the spatial data for substratum type/biota and sediment type. The substratum type/biota mapping used in this study (Figure 2) was based on interpretation of aerial photography and ground-truthing, which only allowed accurate definition of habitats to depths of about 7 m (Blake and Ball, 2001). While 7 m is about the depth limit of seagrass in the bay, other habitats such as rocky reef and Pyura beds that are known to provide important habitat for sub-adult snapper at depths >7 m were not defined in the substratum type/biota layer. Consequently, the composite habitat suitability models could not identify the presence of highly suitable habitat at these locations. Some substratum type/biota categories, such as seagrass and algal beds, may also display seasonal and annual variations in their distribution. The seabed sediment layer also represented a broad-scale pattern of sediment distribution in the bay, but did not represent localized variations in sediment type.
The models we produced are likely to be "over-protective", owing to the methods used to classify habitat suitability. Emphasis was placed mainly on the fishing blocks where fish were caught, rather than the blocks where they were not caught. This meant that the models probably had more high and medium quality habitats than was actually the case. A model that over-predicts these upper levels of suitable habitat would, though, seem preferable if it is likely to be used in initially identifying areas of important habitat.
|
Conclusions |
|---|
|
TopIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
The spatial habitat suitability models produced in this study present a simplified picture of habitat suitability and do not account for many complex relationships and interactions both between species and between species and environmental variables. However, in the absence of a more complete knowledge of the nature of these relationships and the spatial scales at which they operate, the habitat suitability modelling approach presents a relatively effective method for conducting a first-pass identification of likely distributions of important fishery habitats.
In the absence of suitable fishery-independent data, catch and effort data from a commercial fishery can be used to create habitat suitability models. The method developed allows prediction of suitable habitat and, by extension, fish distribution, at a smaller scale than the actual catch returns. Within this method, the best models will relate to species that are much targeted by commercial fishers because of the implicit assumption that areas targeted by commercial fishers have the highest densities of the species being modelled. Although there is evidence that the habitat suitability models we produced provide good predictive information on fish habitat and fish distribution for some species, the models and the hypotheses generated from the modelling process require further testing. In the absence of suitable fishery-independent monitoring data, the approach described here provides a valuable step in developing spatial models to define important fishery habitats at a bay-wide scale.
|
Acknowledgements |
|---|
This project was funded by the Fisheries Research and Development Corporation and Fisheries Victoria. We thank PIRVic personnel Allister Coots for data extraction, and Greg Jenkins, Anne Gason, and Jeremy Hindell for comments on the manuscript. We also gratefully acknowledge the constructive comments of two anonymous reviewers for improving the manuscript.
|
References |
|---|
|
TopIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
-
Anon. (2001) Fisheries Victoria commercial fish production information bulletin(Marine and Freshwater Resources Institute, Queenscliff, Victoria, Australia) 28 pp.
Ball D. and Coots A. G. (2001) Catch and effort info: an ArcView GIS application for querying and displaying commercial fishery catch data. AURISA 2001, 29th Annual Conference of the Australasian Urban and Regional Information Systems Association19–23 NovemberMelbourne, Victoria, Australia.
Blake S. and Ball D. (2001) Seagrass Mapping of Port Phillip Bay. MAFRI Report, 39(Marine and Freshwater Resources Institute, Queenscliff, Victoria, Australia) 79 pp.
Bowen B.K. (1997) Developing and Sustaining World Fisheries Resources. The State of Science and Management. Proceedings of the 2nd World Fisheries Congress(CSIRO PublishingIn Hancock D.A., Smith D.C., Grant A., Beumer J.P. (Eds.). , Melbourne) pp. pp. 169–176.
Bray J. and Curtis J. (1957) An ordination of the upland forest communities of southern Wisconsin. Ecological Monographs 27: 325–349.
Brown S.K., Buja K.R., Jury S.H., Monaco M.E., Banner A. (2000) Habitat suitability index models for eight fish and invertebrate species in Casco and Sheepscot Bays, Maine. North-American Journal of Fisheries Management 20:408–435.[CrossRef]
Clarke K.R. (1993) Non-parametric multivariate analyses of changes in community structure. Australian Journal of Ecology 18:117–143.[CrossRef][Web of Science]
Clarke K.R. and Warwick R. (2001) Change in Marine Communities: an Approach to Statistical Analysis and Interpretation 2nd edn (PRIMER-E, Plymouth) 169 pp.
In Classon B. and Wilson G. (Eds.). Fishing Victoria's Coastline (2002) (Australian Fishing Network, South Croydon, Victoria, Australia) 112 pp.
Conron S. and Coutin P. (1998) The Recreational Snapper Catch from Port Phillip Bay: a Pilot Survey of the Boat Based Fishery 1994/95. MAFRI Internal Report(Marine and Freshwater Resources Institute, Queenscliff, Victoria, Australia) 11: 22 pp.
Coutin P. (2000) Snapper – 1998. Fisheries Victoria Assessment Report(Department of Natural Resources and Environment, Marine and Freshwater Resources Institute, Queenscliff, Victoria, Australia) 19: 43 pp.
Coutin P., Conron S., MacDonald M. (1995) The Daytime Recreational Fishery in Port Phillip Bay, 1989–94(Department of Conservation and Natural Resources, Victorian Fisheries Research Institute, Queenscliff, Victoria, Australia) 43 pp.
Denis V., Lejeune J., Robin J-P. (2002) Spatio-temporal analysis of commercial trawler data using general additive models: patterns of loliginid squid abundance in the north-east Atlantic. ICES Journal of Marine Science 59:633–648.
Denis V. and Robin J-P. (2001) Present status of the French Atlantic fishery for cuttlefish (Sepia officinalis). Fisheries Research 52:11–22.[CrossRef][Web of Science]
Fowler A.J. and McGarvey R. (1995) Development of an Integrated Fisheries Management Model for King George Whiting (Sillaginodes punctata) in South Australia(South Australian Research and Development Institute, Henley Beach) 231 pp.
Gallaway B.J., Cole J.G., Martin L.R., Nance J.M., Longnecker M. (2003) Description of a simple electronic logbook designed to measure effort in the Gulf of Mexico shrimp fishery. North-American Journal of Fisheries Management 23:581–589.[CrossRef]
Gallaway B.J., Cole J.G., Martin L.R., Nance J.M., Longnecker M. (2003) An evaluation of an electronic logbook (ELB) as a more accurate method of estimating spatial patterns of trawling effort and bycatch in the Gulf of Mexico shrimp fishery. North-American Journal of Fisheries Management 23:787–809.[CrossRef]
Gallaway B.J., Cole J.G., Meyer R., Roscigno P. (1999) Delineation of essential fish habitat for juvenile red snapper in the northwestern Gulf of Mexico. Transactions of the American Fisheries Society 128:713–726.[CrossRef]
GHD (Gutteridge, Haskins, and Davey). (1997) Port Phillip Bay coastal values assessment. Physical characteristics and environmental values. Part A. Report to Parks Victoria. Gutteridge, Haskins and Davey Pty Ltd, Melbourne. 45 pp.
Gomon M.F., Glover J.C.M., Kuiter R.H. (1994) The Fishes of Australia's South Coast(State Print, Adelaide) 992 pp.
Guisan A. and Zimmermann N.E. (2000) Predictive habitat distribution models in ecology. Ecological Modelling 135:147–186.[CrossRef][Web of Science]
Hamer P., Jenkins G., Welsford D. (1997) Sampling of Newly Settled Snapper, Pagrus auratus, and Identification of Preferred Habitats in Port Phillip Bay – a Pilot Study(Marine and Freshwater Resources Institute, Queenscliff, Victoria, Australia) 53 pp.
Hilborn R. and Walters C. (1992) Quantitative Fisheries Stock Assessment. Choice, Dynamics and Uncertainty(Chapman and Hall, London) 570 pp.
Hindell J.S., Jenkins G.P., Keough M.J. (2000) Evaluating the impact of predation by fish on the assemblage structure of fishes associated with seagrass (Heterozostera tasmanica) (Martens ex Ascherson) den Hartog, and unvegetated sand habitats. Journal of Experimental Marine Biology and Ecology 255:153–174.[CrossRef][Web of Science][Medline]
Hindell J.S., Jenkins G.P., Keough M.J. (2000) Variability in abundance of fishes associated with seagrass habitats in relation to diets of predatory fishes. Marine Biology 136:725–737.[CrossRef]
Hoedt F.E. and Dimmlich W.F. (1994) Diet of subadult Australian salmon, Arripis truttaceus, in Western Port, Victoria. Australian Journal of Marine and Freshwater Research 45:617–623.[CrossRef]
Jenkins G.P. and Hamer P.A. (2001) Spatial variation in the use of seagrass and unvegetated habitats by post-settlement King George whiting (Percoidei: Sillaginidae) in relation to meiofaunal distribution and macrophyte structure. Marine Ecology Progress Series 224:219–229.[Web of Science]
Jenkins G. P., Watson G. F., Hammond L. S. (1993) Patterns of utilisation of seagrass (Heterozostera) dominated habitats as nursery areas by commercially important fish. Technical Report, 19. Victorian Institute of Marine Sciences, East Melbourne, Victoria, Australia. 100 pp.
Jenkins G.P. and Wheatley M.J. (1998) The influence of habitat structure on nearshore fish assemblages in a southern Australian embayment: comparison of shallow seagrass, reef-algal and unvegetated sand habitats, with emphasis on their importance to recruitment. Journal of Experimental Marine Biology and Ecology 221:147–172.[CrossRef][Web of Science]
Jones K.J. (1999) Australian Salmon (Arripis truttacea)(South Australian Research and Development Institute, Aquatic Sciences, Henley Beach) 13 pp.
Kemp Z. and Meaden G. (2002) Visualization for fisheries management from a spatiotemporal perspective. ICES Journal of Marine Science 59:190–202.
Kuiter R.H. (1993) Coastal Fishes of South-Eastern Australia(Crawford House Press, Bathurst, New South Wales, Australia) 437 pp.
Mace P.M. (1997) Developing and Sustaining World Fisheries Resources. The State of Science and Management. Proceedings of the 2nd World Fisheries Congress(CSIRO PublishingIn Hancock D.A., Smith D.C., Grant A., Beumer J.P. (Eds.). , Melbourne) pp. pp. 1–22.
Marrs S.J., Tuck I.D., Atkinson R.J.A., Stevenson T.D.I., Hall C. (2002) Position data loggers and logbooks as tools in fisheries research: results of a pilot study and some recommendations. Fisheries Research 58:109–117.[CrossRef][Web of Science]
Maurstad A. (2002) Fishing in murky waters – ethics and politics of research on fisher knowledge. Marine Policy 26:159–166.[CrossRef][Web of Science]
Parry G.D., Hobday D.K., Currie D.R., Officer R.A., Gason A.S. (1995) The distribution, abundance and diets of demersal fish in Port Phillip Bay. CSIRO Port Phillip Bay Environmental Study Technical Report(Victorian Fisheries Research Institute, Queenscliff, Victoria, Australia) 21: 107 pp.
Parsons T.R. and Harrison P.J. (2000) Introduction. In Harrison P.J. and Parsons T.R. (Eds.). Fisheries Oceanography. An Integrative Approach to Fisheries Ecology and Management.(Blackwell Science, Oxford) 347 pp.
PMA (Port of Melbourne Authority). (1987) Sediments of Port Phillip Bay. Report No. 87-8-15, Marine Model Laboratory, Port of Melbourne Authority, Melbourne, Victoria, Australia. 48 pp.
Quinn G.P. and Keough M.J. (2002) Experimental Design and Data Analysis for Biologists(Cambridge University Press, Cambridge, UK) 537 pp.
Reynolds J.A. (2003) Quantifying habitat associations in marine fisheries: a generalization of the Kolmogorov–Smirnov statistic using commercial logbook records linked to archived environmental data. Canadian Journal of Fisheries and Aquatic Sciences 60:370–378.
Rubec P. J. (1996) GIS applications for fisheries: for data base management, data sharing sampling, analysis, and visualization in support of ecosystem management. GIS Applications for Fisheries and Coastal Resources Management, Gulf States Marine Fisheries Commission, Biloxi, Mississippi pp. 157–192.
Rubec P.J., Bexley J.C.W., Norris H., Coyne M.S., Monaco M.E., Smith S.G., Ault J.S. (1999) Suitability modeling to delineate habitat essential to sustainable fisheries. American Fisheries Society Symposium 22:108–113.
Rubec P.J., Smith S.G., Coyne M.S., White M., Sullivan A., Wilder D., MacDonald T., McMichael R.H., Monaco M.E., Ault J.S. (2001) Spatial modeling of fish habitat suitability in Florida estuaries. In Kruse G.H., Bez N., Booth A., Dorn M.W., Hills S., Lipcus R.N., Pelletier D., Roy C., Smith S.J., Witherell D. (Eds.). Spatial Processes and Management of Marine Populations(Alaska Sea Grant (College Program), Fairbanks, Alaska) pp. 1–18 AG-SG-01-02.
Rubec P. J., Whaley S. D., Henderson G. E., Lewis J., White M., Sullivan A. M., Vadas R. L., Ruiz-Carus R., Wilder D. T., Westergren C., McMichael R. H., MacDonald T. (2003) Development and Evaluation of Methods for Habitat Suitability Modeling in Florida Estuaries. Report submitted to U.S. Fish and Wildlife Service, Atlanta, Georgia, associated with Sport Fish Restoration Grant No. F-96b, Florida Marine Research Institute. 69 pp.
Sharp G.D. (1997) Its about time: rethinking fisheries management. Proceedings of the 2nd World Fisheries CongressDeveloping and Sustaining World Fisheries Resources. The State of Science and Management (CSIRO PublishingIn Hancock D.A., Smith D.C., Grant A., Beumer J.P. (Eds.). , Melbourne) pp. 731–736.
Smith D.C. and MacDonald C.M. (1997) King George Whiting 1996. Compiled by the Bay and Inlet Fisheries Stock Assessment Group. Fisheries Victoria Assessment Report(Fisheries Victoria, East Melbourne, Victoria, Australia) 15: 18 pp.
Starr R.M. and Fox D.S. (1997) Can fishery catch data supplement research cruise data? A geographical comparison of research and commercial catch data. In Hancock D.A., Smith D.C., Grant A., Beumer J.P. (Eds.). Developing and Sustaining World Fisheries Resources. The State of Science and Management(CSIRO Publishing, Melbourne) pp. 190–197 Proceedings of the 2nd World Fisheries Congress.
Stoner A.W., Manderson J.P., Pessutti J.P. (2001) Spatially explicit analysis of estuarine habitat for juvenile winter flounder: combining generalized additive models and geographic information systems. Marine Ecology Progress Series 213:253–271.[Web of Science]
Wilson G. (1986) Snapper. When, where and how to catch them. A Guide to Port Phillip and Westernport Bay(Geoff Wilson, Geelong, Victoria, Australia) 104 pp.
Zheng X., Pierce G.J., Reid D.G. (2001) Spatial patterns of whiting abundance in Scottish waters and relationships with environmental variables. Fisheries Research 50:259–270.[CrossRef][Web of Science]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||