Spiny lobster, Jasus edwardsii, recovery in New Zealand marine reserves

Biological Conservation 92 (2000) 359±369 www.elsevier.com/locate/biocon Spiny lobster, Jasus edwardsii, recovery in New Zealand marine reserves S. Kelly a,*, D. Scott b, A.B. MacDiarmid c, R.C. Babcock a a University of Auckland, Leigh Marine Laboratory, PO Box 349, Warkworth, New Zealand b University of Auckland, Tamaki Campus, Private Bag 92019, Auckland, New Zealand c NIWA, PO Box 14-901, Kilbirnie, Wellington, New Zealand Received 11 May 1998; received in revised form 27 May 1999; accepted 25 June 1999 Abstract The abundance, size, biomass and reproductive output of spiny lobsters, Jasus edwardsii, from replicated sites nested within four marine reserves and similar non-reserve locations in north-eastern New Zealand were compared. No time±series data were available from three of the reserves so the ages of the reserves (3±21 years) were used to infer temporal patterns of lobster population recovery. Linear models indicated that the mean density of the total lobster population increased 3.9 and 9.5% in shallow (<10 m depth) and deep sites (>10 m depth), respectively, for each year in which the reserves were established, while the mean size of lobsters was estimated to increase by 1.14 mm per year of protection. As a consequence lobster biomass (kg/500 mÿ2) was conservatively estimated to increase by 5.4% per year of protection in shallow sites and 10.9% per year of protection in the deep sites and egg production (eggs/500 mÿ2) by 4.8 and 9.1% per year of protection for shallow and deep sites respectively. # 2000 Elsevier Science Ltd. All rights reserved. Keywords: Lobster; Marine reserves; Jasus edwardsii; Marine protected areas 1. Introduction Over the last 20 years considerable attention has been directed toward examining the role marine reserves can have in marine conservation and ®sheries management. The potential advantages of providing areas free from extractive exploitation have been extensively reviewed (Bohnsack, 1990; Roberts and Polunin, 1991, 1993; Dugan and Davis, 1993; Jones et al., 1993; Rowley, 1994), but empirical evidence which supports proposed bene®ts, such as allowing populations of exploited species to recover, is limited, especially from temperate locations. Published data on the impact of temperate marine reserves has only been obtained from a few sites in the Mediterranean (Bell, 1983; GarciÂa-Rubies and Zabala, 1990; Dufour et al., 1995; Harmelin et al., 1995), South Africa (Buxton and Smale, 1989, 1991; Bennett and Attwood, 1993), Australia (Edgar and Barrett, 1997), and New Zealand (McCormick and * Corresponding author. E-mail address: casl@xtra.co.nz (S. Kelly). Choat, 1987, Cole et al., 1990, MacDiarmid and Breen, 1993), with most attention focused on determining the impact of protection on the biomass, size, abundance, or species assemblages of ®sh. Results have been mixed, with some species showing clear patterns of recovery while others have apparently not responded to protection or researchers have failed to adequately demonstrate evidence of recovery. The failure to demonstrate the e€ects of protection does not necessarily mean populations have not responded, rather it may re¯ect the limitations of commonly used sampling methodologies and analysis. The ability to generalise about the e€ects of temperate marine reserves has also been hampered by the use of poorly designed surveys with inadequate levels of replication or controls (Jones et al., 1993, Rowley, 1994). Jones et al., (1993) suggested the design of most surveys on marine reserves failed to meet the most basic requirements for any impact study. They highlighted the fact that most studies were confounded because they only compared a single protected location with a non-protected control, leading to the possibility that observed di€erences were due to variability from location to location rather 0006-3207/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved. PII: S0006-3207(99)00109-3 360 S. Kelly et al. / Biological Conservation 92 (2000) 359±369 than protection e€ects. They were also critical of the lack of data from the periods prior to the establishment of most marine reserves which would have enabled before and after comparisons to be made. Some studies did not even attempt to provide comparisons with controls but used changes in abundance estimates within a reserve through time as a measure of protection e€ects (Bennett and Attwood 1991, 1993). Of the 11 known published studies examining the impact of temperate marine reserves, all concluded that protected areas allowed the recovery of at least some species; however, nine only looked at a single marine reserve, one compared a single site inside the reserve with two control sites outside, six only compared one site within the protected area with one control site outside and two had no control sites at all (Table 1). Inadequacies in sampling design clearly mean that conclusions drawn from previous studies of temperate reserves must be treated with a degree of caution. We know of only one published study on temperate marine reserves which has provided data from more than one protected area in order to increase the degree of generality. In a study incorporating both temporal and spatial data MacDiarmid and Breen (1993) compared the abundance of spiny lobsters, Jasus edwardsii, at two marine reserves with three control locations in northeast New Zealand. Their sampling design included replicated samples taken at a number of sites nested within each location. However, they obtained con¯icting results with lobsters occurring in highest densities within one reserve and lowest densities within the other. The lack of shallow-water boulder habitat at the low density reserve and its location in relation to the continental shelf and major currents were cited as possible explanations for this result, highlighting the diculty of making generalisations with limited levels of replication. Some of the shortcomings evident in prior studies were due to the lack of data from before the reserves were established and the low number of reserves available for study, however, the number of marine reserves has increased to the point where it is now possible to improve on previous sampling regimes. Our aims were therefore to (1) increase the number of marine reserves examined to improve the generality of the study and (2) to provide an estimate of the rate of lobster population recovery within marine reserves. To do this we compared the size and abundance of lobsters at sites nested within four marine reserves and four similar non-reserve locations in northeast New Zealand. Time series data were not available for three of the marine reserves so temporal changes at individual reserves could not be examined. Therefore a di€erent approach was adopted to examine temporal patterns of lobster recovery. The age of the reserves ranged from 3 to 21 years. Di€erences in lobster abundance, size, biomass and egg production were compared between the reserves and non-reserve controls with the expectation that these parameters would increase in magnitude as the age of reserve increased. This approach allowed tentative estimates of the rates of increase in abundance, mean size, biomass and egg production to be made. 2. Methods Surveys were conducted at sites within and outside four marine protected areas (hereafter called marine reserves or locations) on the eastern coast of North Island, New Zealand (Fig. 1). No ®shing of any kind is allowed in any of the protected areas which varied in age from 3 years (Cathedral Cove Marine Reserve, and Tuhua Marine Reserve) to 14 years (Tawharanui Marine Park) and 21 years (Leigh Marine Reserve). Sampling was strati®ed and limited to broken boulder areas and/or fractured reef, on which Jasus edwardsii commonly occur (MacDiarmid, 1991, 1994, pers obs). All sites were dominated by a mixture of laminarian or fucalean kelp forests and urchin zones (Evechinus chloroticus) at shallow to intermediate depths (<15 m). In deeper areas (>15 m) kelp and urchin density decreased Table 1 Breakdown of the survey designs for temperate marine reserves showing the level of replication for reserve and control sites Authors Year # Reserve locations # Control locations Nested reserve sites Nested control sites Habitat types Temporal data Bell McCormick and Choat Buxton and Smale GarcõÂa-Rubies and Zabala Bennet and Attwood Cole et al. Bennet and Attwood MacDiarmid and Breen Dufour et al. Harmelin et al. Edgar and Barrett 1983 1987 1989 1990 1991 1991 1993 1993 1995 1995 1997 1 1 1 1 1 1 1 2 1 1 4 1 1 1 2 0 1 0 4 1 1 4 1 1 1 1 2 5 2 5 1 1 1±6 1 1 1 1 0 3 0 5 1 1 2±10 2 7±10 1 3 1 1±3 1 1 2 1 1 No No No No Yes Yes Yes Yes Yes Yes Yes 361 S. Kelly et al. / Biological Conservation 92 (2000) 359±369 and sponges generally became larger and more abundant. Tuhua Marine Reserve di€ered from the other locations in that it is an o€shore island. At Tuhua water clarity was substantially higher, kelp cover extended to greater depths, and there was a more diverse range of ®sh and algal species (Jones and Garrick, 1991). Two reef sites were selected within each marine reserve and two from nearby areas of coast. Within the Leigh Marine Reserve, where there is a history of lobster research (MacDiarmid, 1987, 1991, 1994; Cole et al., 1990, 1991; MacDiarmid, et al., 1991; MacDiarmid and Breen, 1993, 1994), sample sites were chosen with a random number generator from a selection of ®ve potential reef sites scattered throughout the reserve. At the other locations, where there was no prior knowledge of lobster abundance, formal randomisation procedures were not carried out. Rather, reef sites were haphazardly selected from the surface prior to diving. Two depths (<10 and 10±25 m) were sampled at each site to allow for seasonal changes in the depth distribution of lobsters (MacDiarmid, 1991). At each location the size of every lobster within ®ve, 50  10 m transects was recorded. The starting position and direction of each transect was determined by the censor swimming for a predetermined time in a predetermined direction. The choice of the 50  10 m transect size was based on a pilot study conducted by MacDiarmid (1991). He compared the precision of three di€erent sized transects: 10  10 m, 25  10 m and 50  10 m, in estimating lobster abundance within the Leigh Marine Reserve and found it to be similar in all cases. The 50  10 m transect was therefore selected to permit at least one transect per dive to be completed in areas of high lobster abundance, and limit the number of zero counts in areas of low lobster abundance. The visual size estimation method was used (MacDiarmid, 1991), with divers estimating carapace length (C.L.) to within 5 mm without capturing or handling individual lobsters. This level of accuracy was achieved through a series of calibration dives where the size of individual lobsters were ®rst estimated, after which each lobster was caught by hand and measured with vernier callipers to obtain a true length measurement. Sampling at Tuhua, Cathedral Cove and Leigh was done by the same censor, whereas he and two other censors sampled Tawharanui Marine Park. An analysis of covariance (ANCOVA) was used to test for di€erences between the Tawharanui Marine Park censors. No signi®cant di€erence was found between the slopes of the relationship between estimated and actual carapace length for each censor (p=0.3844). However, a signi®cant di€erence was found between the intercepts (p=0.0001). Least squares means were then calculated and pairwise comparisons using Bonferroni corrections carried out (Table 2). The range of the least squares means (2.8 mm) was less than the 5 mm accuracy margin required by each censor and pairwise comparisons were unable to detect any signi®cant di€erences among the censors. Therefore the di€erences detected by the ANCOVA were considered to be trivial. Where possible the sex of every lobster was also determined visually at all sites except those within the Leigh Marine Reserve where these data were available from a separate study conducted in the same month 1 year earlier (Kelly, unpublished data). 2.1. Data analysis 2.1.1. Abundance Two features complicated the analysis of abundance data. Firstly, the count data was not normally distributed and was skewed. Secondly, the sites within Table 2 Pairwise comparisons and least square means obtained from carapace length estimates for each censora p Values of pairwise comparisons Censor 1 Censor 2 Censor 3 Fig. 1. Map of north-eastern New Zealand showing the locations of the four marine protected areas surveyed in the study. a Censor 1 ± 0.4911 0.0414 LS Mean (S.E.) Censor 2 ± ± 0.1499 123.9 (0.8718) 124.4 (0.8695) 126.7 (1.0417) Note the signi®cant p value after Bonferroni correction is 0.0167. 362 S. Kelly et al. / Biological Conservation 92 (2000) 359±369 locations and the transects within sites were not considered to be independent. The count data were therefore transformed [ln(count +1)] and analysed using linear mixed models which allowed for both ®xed and random e€ects, and for correlations between observations. The particular software used was the function lme developed by Pinheiro, Bates, and Lindstrom (Pinheiro and Bates, 1995) and incorporated into the software package S-Plus (Statistical Sciences, 1993). An accessible introduction to the software is given in Venables and Ripley (1997, Section 10.3), and a comprehensive discussion of the models and the modelling processes used is given in Longford (1993). Two sets of counts were analysed, total lobster counts, and counts of lobsters above the legal size limit (>100 mm C.L.). Linear mixed models were ®tted and the signi®cance of terms in the models was tested using likelihood ratio tests. In addition to the tests, linear model diagnostics were used to check the model assumptions. For each model the residuals were plotted against ®tted values, data against ®tted values, and normal quantile±quantile or qqplots of residuals obtained. These plots were used to check the assumption of normally distributed errors, to check for outliers, heteroscedasticity, or other problems. The value of the Akaike Information Criterion (AIC) was also considered to determine the most appropriate model. The model which had the minimum AIC was considered optimal and represented a compromise between model ®t and complexity (Akaike 1973, 1974). were skewed, and that a logarithmic transformation was required. This caused problems with transects in which no lobsters were observed, giving an estimated zero biomass. The solution to this problem was to omit those transects with no lobsters. This was not entirely satisfactory because it arti®cially in¯ated the mean biomass of sites containing samples with no lobsters, thereby introducing bias into the ®tted models. The results of this section of the analysis should therefore only be regarded as approximate. However, the bias was conservative and underestimated the relative e€ect of protection on lobster biomass, as there were fewer zero counts in reserve than non-reserve transects (7 cf.19). The fecundity of individual females was estimated according to the formula given by MacDiarmid (1989) where fecundity =0.169 CL3.0091. Using this formula the total number of eggs produced per transect was calculated and log transformed prior to analysis. Within the Leigh Marine Reserve MacDiarmid (1989) found that the size at onset of maturity for female J. edwardsii was 87.5 mm C.L. Females above 90 mm C.L. were therefore assumed to be mature. As with biomass, where the sex of an individual lobster was not known it was randomly assigned for each location based on the size-speci®c probability of being male or female. 2.1.2. Size, biomass and egg production To examine the response of lobster size, biomass and egg production to protection, mixed linear models that allowed for the lack of independence between observations were again ®tted using S-Plus. Signi®cance testing was carried out using likelihood ratio tests, and the AIC was considered in deciding on the appropriate model. Diagnostic plots were used to assess the validity of the model assumptions. Separate length±weight relationships were established for male and female lobsters (MacDiarmid, unpublished data) and biomass was determined by converting length to weight after log-log transformation. Lobsters whose sex could not be determined visually were randomly assigned a sex based on the probability of being male or female in a particular 10 mm C.L. size class at each location. This method was deemed preferable to simply assigning equal probabilities to males and females as the size frequency distributions for the sexes varied markedly. As no sex data was collected within the Leigh Marine Reserve, data from a survey conducted one year earlier was used to determine the size class probabilities and assign sexes accordingly. In attempting to model the e€ect of status, depth and age of reserve on biomass, it was found that distributions The abundance of individual sexes was not examined separately, rather the sex data were pooled and total lobster counts analysed. Boxplots of the pooled data for the total number of lobsters against status (reserve or non-reserve), and depth (deep or shallow) showed evidence of more lobsters in the reserves compared to nonreserves, although there was one obvious outlier in the non-reserves and some possible outliers in the reserves (Fig. 2). A boxplot of total number of lobsters against the age of the reserve also gave some indication of increasing numbers with age (Fig. 3a.) and counts of lobsters above the legal size limit (i.e. >100 mm C.L.) showed a similar trend (Fig. 3b). The initial model used for analysis of both the total counts and the counts of lobsters over 100 mm incorporated ®xed e€ects for reserve (Cathedral Cove, Tawharanui, Tuhua or Leigh), status (in or outside the reserve), depth (deep or shallow) and all possible interactions between these three factors (including the threeway interaction), plus random e€ects for reserve and depth and the interaction between reserve and depth. Fitting both random and ®xed e€ects for particular factors is equivalent to treating the e€ect of a factor as being random with a normal distribution, but di€erent levels of the factor have distributions with di€erent 3. Results 3.1. Abundance S. Kelly et al. / Biological Conservation 92 (2000) 359±369 363 Fig. 2. Boxplots of the raw count data comparing (a) transects from reserve and non-reserve areas, and (b) deep and shallow sites. The central boxes enclose data falling between the 25th and 75th percentiles, while the stems indicate the extent of data falling between the 10th and 90th percentiles. Data points falling outside these ranges are plotted individually. Fig. 3. Boxplots of (a) raw count data, and (b) counts of lobsters over 100 mm carapace length, from reserves of di€erent ages. Data from nonreserve areas has been are pooled at age zero, and data from Cathedral Cove and Tuhua marine reserves pooled at age 3. The central boxes enclose data falling between the 25th and 75th percentiles, while the stems indicate the extent of data falling between the 10th and 90th percentiles. Data points falling outside these ranges are plotted individually. means, which are given by the ®xed e€ects. Considering reserves for example, the number of lobsters was expected to vary between reserves simply because of di€erences in the age of reserves. In addition to an e€ect due to a reserve being a particular age, that is the ®xed e€ect, the reserves may also be considered as a sample from a population of possible reserves of the same age. The additional variability introduced by this was modelled by including the random e€ect. The correlation between transects at the various sites was taken into account by treating transects within a given site as a cluster, where each cluster represented observations that were expected to be related (Pinheiro and Bates, 1995). The ®rst step in the modelling process was to see which random e€ects were needed. The results of this procedure were di€erent for the total counts compared to the lobsters over the legal size limit. For the total counts no random e€ects were necessary, while for the counts of larger lobster, only the interaction between reserve and depth could be eliminated. However, further 364 S. Kelly et al. / Biological Conservation 92 (2000) 359±369 consideration of the design suggested that random e€ects for reserve and for depth were appropriate and should be included in both cases. A random e€ect for reserve allowed for correlation between sites within a reserve and also allowed the generalisation of results by treating the reserves surveyed as a sample of possible reserves. Similar arguments supported the inclusion of a random e€ect for depth. Therefore for consistency, and because of these design aspects, random e€ects for reserve and depth were included in subsequent models of both total lobster counts and counts of legal sized lobsters. In the initial phase of the analysis possible outliers were also identi®ed. For the total count data, it was decided to eliminate a shallow non-reserve transect at Cathedral Cove where a very large number of small lobsters was observed. A shallow reserve transect at Leigh was also omitted from the analysis of larger lobsters. In this transect no lobsters were observed, while all of the other transects at that site contained substantial numbers of lobsters. These observations were not errors in the data, but simply re¯ected the degree of variation present, and it was considered that their inclusion might have lead to the model assumptions being violated. After deciding on outliers and random e€ects attention was turned to ®xed e€ects. A major concern of the study was to determine the e€ect of age of reserve on lobster numbers. With this aim in mind, models using the variable ``age of reserve'' were tried, where age was taken to be zero for the unprotected sites and equal to the number of years the reserve had been established for sites within each reserve. With this de®nition, knowledge of the status of a site, and the reserve in which the site is situated, gives knowledge of the value of ``age of reserve''. Thus models containing the variable age of reserve were submodels of models containing the variables status and reserve. The model selected by this process for both the total counts and the counts of larger lobsters included ®xed e€ects for reserve, age of reserve, and depth, plus interactions between reserve and depth, and age of reserve and depth. For total counts in the deep sites the age of reserve term was highly signi®cant (p <0.0001), and gave a coecient of 0.0906, which represented the change in ln(count +1) for each year of protection. The exponent of this value gives the proportional increase in the expected lobster count +1, per year of protection. When the counts are large this is approximately equal to the ratio of counts between consecutive years, and hence gives the percentage increase. For the deep transects the expected annual rate of increase was e(0.0906)=1.0948, or approximately 9.5% per year of protection. For shallow transects the age of reserve term was signi®cant and gave an approximate rate of increase of 3.9% per annum (p=0.025). The model without the interaction between age of reserve and depth provided a less satisfactory ®t, but gave an approximate rate of increase in total counts of 7.4% per annum (p<0.0001). The analysis of the counts of lobsters over 100 mm in carapace length revealed approximate rates of increase of 6.2% per annum (p<0.0001) for deep transects and 12.2% per annum (p<0.0001) for shallow transects. The rate of growth in the model without interaction was 7.4% per annum (p<0.0001). 3.2. Size Of the 2100 lobsters counted, the sizes of 2074 were estimated. When a lobster was not in full view a size estimate could not be made. We judged that lobsters whose carapace length could not be estimated did not di€er in any systematic way from to those whose sizes were determined, so they were therefore ignored. Plots of the size data indicated that lobsters were larger inside the reserves than outside, and that there was a relationship between the age of the reserve and lobster size, with the largest lobsters being found in the older reserves (Fig. 4). However, there was no obvious di€erence in the mean size of lobsters in the two older reserves, nor between mean sizes for the group of more recently established reserves. Modelling of the size data was more complicated than the other analyses because there was an additional level of sampling. For abundance, biomass and egg production, the response was at the transect level. For the size data, the response was the size of individual lobsters within transects. Transects were therefore treated as clusters and sites as random e€ects to model the possible correlation between transects at the same site. A random e€ect for depth was also included in the initial model as well, but reserve and status random e€ects were not included since these were confounded with site because of the design used. Plots of the data and of residuals from the various models examined indicated that transformation was not necessary so the raw data was used in this analysis. Testing of random e€ects indicated that these were also unnecessary. The ®xed e€ects were then examined; starting with a model incorporating ®xed e€ects for reserve, status and depth, plus all interactions. The model with the smallest AIC had e€ects for reserve, age of reserve, depth, and the interaction between depth and reserve. The age of reserve term was highly signi®cant (p<0.0001) and indicated an increase in average size of 1.14 mm per year. 3.3. Biomass Initial descriptive analysis indicated that biomass appeared to increase in proportion to the number of years the reserve had been established (Fig. 5a), and S. Kelly et al. / Biological Conservation 92 (2000) 359±369 365 Fig. 4. Size distributions of lobsters from protected and unprotected sites in and around Cathedral Cove Marine Reserve, Tuhua Marine Reserve, Tawharanui Marine Park and Leigh Marine Reserve. Data from all of the deep and shallow sites at each location were pooled. that greater biomass was supported in shallow water than in deep water. Modelling was carried out using the same procedure as for the count data with the response being the natural log of biomass. Testing revealed no need to incorporate random e€ects for reserve and depth, and these were omitted from the model. The selected model included ®xed e€ects for reserve, age of reserve, depth and the interaction between age of reserve and depth. The coecient for age of reserve for deep transects was 0.103 (p<0.0001) giving an estimated 366 S. Kelly et al. / Biological Conservation 92 (2000) 359±369 factor, and that the number of eggs produced within a given area increased with the age of the reserve (Fig. 5b). The drop apparent in the boxplot of the 3 year old reserves was due to lower egg production in Tuhua Marine Reserve (Fig. 5b). Random e€ects for reserve and depth were not needed, and the model selected had the same ®xed e€ects as in the model for biomass. The estimated rate of increase in egg production was determined to be 9.1% in deep transects (p<0.0001), and 4.8% in shallow transects (p=0.0009). The model without an interaction between age of reserve and depth was acceptable and gave an overall rate of increase of 6.7% per year (p<0.0001). 4. Discussion Fig. 5. Boxplot of (a) biomass, and (b) egg production comparing reserves of di€erent ages. Data from non-reserve areas has been pooled at age zero, and data from Cathedral Cove and Tuhua marine reserves pooled at age 3. The central boxes enclose data falling between the 25th and 75th percentiles, while the stems indicate the extent of data falling between the 10th and 90th percentiles. Data points falling outside these ranges are plotted individually. increase in biomass of 10.9% per year. For shallow transects, the rate of growth was estimated to be 5.4% per year (p=0.003). The model without an interaction between age of reserve and depth did not ®t particularly well but gave a rate of increase of 8.1% per year (p=0.0002). 3.4. Egg production Egg production was modelled in a similar fashion to biomass. Plots of the natural logarithm of egg production indicated that depth might not be an important This is the ®rst published study to demonstrate population level responses to protection at more than one temperate marine reserve. Inconclusive results and the examination of single marine reserves has meant that previous studies were unable to generalise to other areas (Bell, 1983; McCormick and Choat, 1987; Buxton and Smale, 1989; GarciÂa-Rubies and Zabala, 1990; Cole et al., 1990; Bennett and Attwood, 1991, 1993; MacDiarmid and Breen, 1993; Dufour et al., 1995; Harmelin et al., 1995). In our study the use of several marine reserves and the inclusion of a temporal component strengthens the level of con®dence with which we can make large scale inferences about the response of lobsters to protection from ®shing. Lobster biomass was higher within the protected zones than in unprotected areas, with the di€erence estimated to increase by 5.4 to 10.9% per year depending on depth. As transects with no lobsters were omitted from the analysis, and the unprotected sites had more transects with zero counts, this estimate is considered to be conservative. The increase in biomass was due to an increase in both the size and abundance of lobsters within the protected zones. These results demonstrate the potential of marine reserves to allow populations to recover, but they also illustrate the heavy impact ®shing has on lobster populations. The inclusion of reserves of di€erent ages allowed an indirect measure of the rate of population recovery. The data indicate that protected lobster populations could increase in abundance by 3.9% per annum in shallow sites and 9.5% per annum in deeper sites. However these estimates were based on data from only four reserves. The relatively low level of replication does not provide a good measure of inter-site variability; so model parameters derived from it must be treated cautiously. For example, marked di€erences were found in the recovery patterns of the two 3 year old reserves. Many factors may in¯uence recovery rates at individual sites and contribute to variability. For instance, recruit S. Kelly et al. / Biological Conservation 92 (2000) 359±369 supply and juvenile survivorship may vary from place to place and over time (Booth, 1994; Briones-Fourzan, 1994; Butler et al., 1997; Gosselin and Qian, 1997; Hunt and Scheibling, 1997; Wahle and Incze, 1997). Ultimately, population growth may be limited by the ecological suitability of a particular site or the carrying capacity of an area (Bologna and Steneck, 1993; Parrish and Polovina, 1994). However, comparison with the results presented by MacDiarmid and Breen (1993) suggest that lobster populations can respond rapidly to protection and that our estimates of the rate of population recovery may be conservative. MacDiarmid and Breen (1993) examined the abundance of lobsters in shallow sites of less than 10 m within the Leigh Marine Reserve and found that overall, mean densities ranged from 10.07 lobsters per 100 m2 in March/April 1985 to 6.08 lobsters per 100 m2 in July/August 1985. Converting our density data to the same unit area gave values for the Leigh Marine Reserve of 5.89 lobsters per 100 m2 which is below either of their estimates. This suggests that lobster densities within the Leigh Marine Reserve reached very high levels within 8 years of its establishment and have since declined. A number of factors limit the validity of making such direct comparisons with our study. For instance MacDiarmid and Breen (1993) sampled from di€erent sites within the reserve, conducted their survey at di€erent times of year, used di€erent sampling unit sizes, and limited their sampling to a single depth strata. Nevertheless, the fact that the abundances they reported were substantially higher than any found in our study supports their assertion that recovery within the Leigh Marine Reserve occurred rapidly. Given that marine reserves allow J. edwardsii populations to recover it seems prudent to examine the role they could play in the management of this species. There are a number of possible bene®ts in providing areas free from ®shing pressure. Roberts and Polunin (1993) proposed that adults within marine reserves may help restock unprotected areas through the export of larvae and recruits. As egg production increases exponentially with body length (Roberts and Polunin 1991, 1993), and there are more, larger, egg producing adults within marine reserves, the contribution of protected areas to overall egg production should exceed the ratio of protected area to ®shing grounds. Our comparisons of egg production within the reserves and non-reserve areas support this assertion. However, it is more dicult to determine what, if any impact increased egg production would have on recruitment levels, as stock±recruitment relationships are notoriously dicult to demonstrate (Caddy, 1986; Caputi, 1993). In fact there is even some debate as to whether recruitment is related to stock size at all, except at extremely low levels of egg production (Koslow, 1992; Myers and Barrowman, 1996; Francis, 1997; Gilbert, 1997; Hillborn, 1997; Myers, 1997). This 367 is particularly so in species like J. edwardsii whose free swimming larval life span may range from 12 to 24 months (Booth and Phillips, 1994). Because of the poor relationship between spawning stock and recruitment, marine reserves should only play a signi®cant role in maintaining or enhancing recruitment in unprotected areas when stock sizes in those areas are severely depleted (Roberts and Polunin, 1991). Although potentially serious reductions in J. edwardsii stocks have occurred in some New Zealand management areas, current management strategies are considered sucient to allow them to rebuild (Annala and Sullivan, 1996; Breen and Kendrick, 1997). However, problems with conventional ®sheries management techniques (e.g. Hofmann and Powell, 1998) have led directly to the promotion of marine reserves as insurance against poor management, stock depletion and ultimately recruitment failure. A system of marine reserves may therefore be a prudent management strategy to ensure that adequate spawner biomass is maintained in J. edwardsii stocks. Marine reserves have also been attributed with having the potential to maintain or enhance the yield of adjacent ®sheries (Alcala and Russ, 1990; Kelly, 1999). As populations build up within marine reserves food or habitat requirements may eventually limit further population expansion. Mobile species can respond by emigrating to unprotected areas were there is less pressure on resources, and in the process become susceptible to capture. Alternatively individuals may leave protected areas through general di€usive or migratory movements (Kelly, 1999). Alcala and Russ (1990) proposed that ®shing yields adjacent to the Sumilon Marine Reserve in the Philippines were enhanced by such processes. They found that when ®shing was re-established within the reserve the total yield around Sumilon Island declined by 54%, despite a larger area of reef being ®shed (i.e. both the reserve and non-reserve areas). However, modelling suggests that the establishment of marine reserves will in most cases lead to a reduction in the yield of ®sh stocks (Polacheck, 1990; DeMartini, 1993; Attwood and Bennett, 1995). Only under conditions of heavy ®shing mortality and moderate rates of transfer were small increases in yield predicted. Nevertheless, intensive trapping for lobsters occurs around the boundaries of at least three of the marine reserves surveyed in this study, and the boundaries of all the reserves are popular dive sites for SCUBA divers hunting for lobsters. The popularity of these areas for catching both lobsters and ®n-®sh suggests that there is a public perception that reserves contribute to the local ®shery. This perception is backed up by the commercial catch rates of lobsters obtained around the boundary of the Leigh Marine Reserve which show strong seasonal variability but are relatively high compared with areas remote from the reserve (Kelly, 1999). 368 S. Kelly et al. / Biological Conservation 92 (2000) 359±369 While direct bene®ts to lobster management remain unclear, areas of zero exploitation can be used to obtain valuable information for ®sheries research and management. The recovery of lobsters within marine reserves has provided ecologists with the opportunity to work on relatively natural populations free from human interference. As a result marine reserves have made a substantial contribution to our understanding of J. edwardsii ecology (MacDiarmid, 1987, 1991, 1994; MacDiarmid et al., 1991,; Kelly, 1999; Kelly et al., 1999) and may permit more e€ective management strategies to be developed. Acknowledgements This research was funded by the Department of Conservation and the Graduate Research in Industry Fellowship Scheme. Additional support was provided by Leigh Fisheries, University of Auckland, Bay of Plenty Polytech, Leigh Marine Laboratory, Keith Gregor, Gilli Adams, Peter Carter, Andy Garrick, Alan Jones and Rawiri Tuawau. We would also like to thank all of the students and research assistants who helped with the ®eld work. References Akaike, H., 1973. Information theory and an extension of the maximum likelihood principle. In: Petrov, B. N., Csaki, F. (Eds.), Proc. 2nd Int. Symp. Information Theory, Akademia Kiado, Budapest, pp. 267±281. Akiake, H., 1974. A new look at statistical model identi®cation. IEEE Trans. Automatic Control 19, 716±723. Alcala, A.C., Russ, G.R., 1990. A direct test of the e€ects of protective management on abundance and yield of tropical marine resources. J. Cons. Int. Explor. Mer 46, 40±47. Annala, J.H., Sullivan, K.J., 1996. Report from the mid-year ®shery assessment plenary, November 1996: stock assessments and yield estimates. NIWA, Wellington. Attwood, C.G., Bennett, B.A., 1995. Modelling the e€ect of marine reserves on the recreational shore-®shery of the South-Western Cape, South Africa. S. Afr. J. Mar. Sci 16, 227±240. Bell, J.D., 1983. E€ects of depth and marine reserve ®shing restrictions on the structure of a rocky reef assemblage in the north-western Mediterranean Sea. J. Appl. Ecol. 20, 357±369. Bennett, B.A., Attwood, C.G., 1991. Evidence for the recovery of a surf-zone ®sh assemblage following the establishment of a marine reserve on the southern coast of South Africa. Mar. Ecol. Prog. Ser 75, 173±181. Bennett, B.A., Attwood, C.G., 1993. Shore-angling catches in the De Hoop Nature Reserve, South Africa, and further evidence for the protective value of marine reserves. S. Afr. J. Mar. Sci 13, 213±222. Bohnsack, J.A., 1990. The potential of marine ®sheries reserves for reef ®sh management in the U.S. Southern Atlantic. Plan Development Team, NOAA Technical Memorandum, NMFS-SEFC-261. Bologna, P.A.X., Steneck, R.S., 1993. Kelp beds as habitat for American lobster Homarus americanus. Mar. Ecol. Prog. Ser 100, 127±134. Booth, J.D., 1994. Jasus edwardsii larval recruitment o€ the east coast of New Zealand. Crustaceana 66, 295±317. Booth, J.D., Phillips, B.F., 1994. Early life history of spiny lobster. Crustaceana 66, 271±294. Breen, P.A., Kendrick, T.H., 1997. A ®sheries management success story: the Gisborne, New Zealand, ®shery for red rock lobsters (Jasus edwardsii). Mar. Freshwater Res 48, 1103±1110. Briones-Fourzan, P., 1994. Variability in postlarval recruitment of the spiny lobster Panulirus argus (Latreille, 1804) to the Mexican Caribbean coast. Crustaceana 66, 326±340. Butler, M.J., Herrnkind, W.F., Hunt, J.H., 1997. Factors a€ecting the recruitment of juvenile Caribbean spiny lobsters dwelling in macroalgae. Bull. Mar. Sci 61, 3±19. Buxton, C.D., Smale, M.J., 1989. Abundance and distribution patterns of three temperate marine reef ®sh (Teleostei: Sparidae) in exploited and unexploited areas o€ the southern Cape coast. J. Appl. Ecol 26, 441±451. Caddy, J.F., 1986. Modelling stock-recruitment processes in crustacea: some practical and theoretical perspectives. Can. J. Fish. Aquat. Sci. 43, 2330±2344. Caputi, N., 1993. Aspects of spawner-recruit relationships; with particular reference to crustacean stocks; a review. Aust. J. Mar. Freshwater Res 44, 589±607. Cole, R.G., Ayling, M.A., Creese, R.G., 1990. E€ects of marine reserve protection at Goat Island northern New Zealand. N.Z. J. Mar. Freshwater Res 24, 297. DeMartini, E.E., 1993. Modelling the potential of ®shery reserves for managing Paci®c coral reef ®shes. Fish. Bull 91, 414±427. Dufour, V., Jouvenel, J., Galzin, R., 1995. Study of a Mediterranean reef ®sh assemblage. Comparisons of population distributions between depths in protected and unprotected areas over one decade. Aquat. Living. Resour 8, 17±25. Dugan, J.E., Davis, G.E., 1993. Applications of marine refugia to coastal ®sheries management. Can. J. Fish. Aquat. Sci 50, 2029±2042. Edgar, G.J., Barrett, N.S., 1997. Short term biotic change in Tasmanian marine reserves. J. Exp. Mar. Biol. Ecol. 213, 261±279. Francis, R.I.C.C., 1997. Comment: How should ®sheries scientists and managers react to uncertainty about stock-recruit relationships. Can. J. Fish. Aquat. Sci 52, 982±983. GarcõÂa-Rubies, A., Zabala, M., 1990. E€ects of total ®shing prohibition on the rocky ®sh assemblages of Medes Islands marine reserve (NW Mediterranean). Sci. Mar 54, 317±328. Gilbert, D.J., 1997. Towards a new recruitment paradigm for ®sh stocks. Can. J. Fish. Aquat. Sci 54, 969±977. Gosselin, L.A., Qian, P., 1997. Juvenile mortality in benthic marine invertebrates. Mar. Ecol. Prog. Ser 146, 265±282. Harmelin, J.G., Bachet, F., Garcia, F., 1995. Mediterranean marine reserves: Fish indices as tests of protection eciency. Mar. Ecol. 16, 233±250. Hilborn, R., 1997. Comment: Recruitment paradigms for ®sh stocks. Can. J. ®sh. Aquat. Sci 52, 984±985. Hofmann, E.E., Powell, T.M., 1998. Environmental variability e€ects on marine ®sheries: four case histories. Ecol. Appl 8, 23±32. Hunt, H.L., Scheibling, R.E., 1997. Role of early post-settlement mortality in recruitment of benthic marine invertebrates. Mar. Ecol. Prog. Ser 155, 269±301. Jones, A.J., Garrick, A.S., 1991. Tuhua (Mayor Island) marine habitats. A preliminary intertidal and subtidal survey. Technical Report Series no 4, Department of Conservation, Rotorua, New Zealand. Jones, G.P., Cole, R.C., Battershill, C.N., 1993. Marine reserves: Do they work? In: Battershill, C.N., Schiel, D.R., Jones, G.P., Creese, R.G., MacDiarmid, A.B. (Eds.), Proceedings of the Second International Temperate Reef Symposium. NIWA Marine, Wellington, pp. 29±45. Kelly, S., 1999. Marine Reserves and the Spiny Lobster, Jasus edwardsii. Unpublished Ph.D. thesis, University of Auckland. Kelly, S., MacDiarmid, A.B., Babcock, R.C., 1999. Characteristics of spiny lobster, Jasus edwardsii, aggregations in exposed reef and sandy areas. Mar. Freshwater Res. 50, 409±416. S. Kelly et al. / Biological Conservation 92 (2000) 359±369 Koslow, J.A., 1992. Fecundity and the stock-recruit relationship. Can. J. Fish. Aquat. Sci 49, 210±217. Longford, N., 1993. Random Coecient Models. Clarendon Press, Oxford. MacDiarmid, A.B., 1987. Rock Lobster Ecology. Unpublished Ph.D. Thesis, University of Auckland. MacDiarmid, A.B., 1989. Size at onset of maturity and size-dependent reproductive output of female and male spiny lobsters Jasus edwardsii (Hutton) (Decapoda, Palinuridae) in northern New Zealand. J. Exp. Mar. Biol. Ecol 127, 229±243. MacDiarmid, A.B., 1991. Seasonal changes in depth distribution, sex ratio and size frequency of spiny lobster Jasus edwardsii on a coastal reef in northern New Zealand. Mar. Ecol. Prog. Ser 70, 129±141. MacDiarmid, A.B., 1994. Cohabitation in the spiny lobster Jasus edwardsii (Hutton 1875). Crustaceana 66, 341±355. MacDiarmid, A.B., Breen, P.A., 1993. Spiny lobster population changes in a marine reserve. In: Battershill, C.N., Schiel, D.R., Jones, G.P., Creese, R.G., MacDiarmid, A.B. (Eds.), Proceedings of the Second International Temperate Reef Symposium. NIWA Marine, Wellington, pp. 47±56. MacDiarmid, A.B., Hickey, B., Maller, R.A., 1991. Daily movement patterns of the spiny lobster Jasus edwardsii (Hutton) on a shallow reef in northern New Zealand. J. Exp. Mar. Biol Ecol 147, 185±205. McCormick, M.I., Choat, J.H., 1987. Estimating total abundance of a large temperate-reef ®sh using visual strip-transects. Mar. Biol 96, 469±478. 369 Myers, R.A., 1997. Comment and reanalysis: paradigms for recruitment studies. Can. J. Fish. Aquat. Sci 54, 978±981. Myers, R.A., Barrowman, N.J., 1996. Is ®sh recruitment related to spawner abundance? Fish. Bull 94, 707±724. Parrish, F.A., Polovina, J.J., 1994. Habitat thresholds and bottlenecks in production of the spiny lobster (Panulirus marginatus) in the northwestern Hawaiian islands. Bull. Mar. Sci 54, 151±163. Pinheiro, J.C., Bates, D., 1995. Mixed-e€ects models, methods and classes for S-PLUS, Technical Report No. 89, Department of Biostatistics, University of Wisconsin Ð Madison. Polacheck, T., 1990. Year around closed areas as a management tool. Nat. Res. Model 4, 327±354. Roberts, C.M., Polunin, N.V.C., 1991. Are marine reserves e€ective in the management of reef ®sheries? Rev. Fish. Biol. Fish 1, 65± 91. Roberts, C.M., Polunin, N.V.C., 1993. Marine reserves: simple solutions to managing complex ®sheries? Ambio 22, 363±368. Rowley, R.J., 1994. Marine reserves in ®sheries management. Aquat. Conserv. Mar. Freshwat. Ecosyst 4, 233±254. Statistical Sciences., 1993. S-PLUS User's Manual, Version 3.2. Seattle, WA. Venables, W.N., Ripley, B.D., 1997. Modern Applied Statistics with S-Plus, 2nd Ed. Springer-Verlag, New York. Wahle, R.A., Incze, L.S., 1997. Pre- and post-settlement processes in recruitment of the American lobster. J. Exp. Mar. Biol. Ecol 217, 179±207.