A management strategy evaluation approach for broadbill swordfish, Xiphias gladius, in the south-west Pacific Ocean

Transcription

1 SCTB16 Working Paper BBRG-4 A management strategy evaluation approach for broadbill swordfish, Xiphias gladius, in the south-west Pacific Ocean Robert Campbell and Natalie Dowiing CSIRO, Division of Marine Research Hobart, Tasmania Australia July 03

2 A Management Strategy Evaluation approach for broadbill swordfish, Xiphias gladius, in the south-west Pacific Ocean Robert Campbell and Natalie Bowling CSIRO Marine Research 1. Introduction The longline sector of the Australian fishery which operates off the east coast of Australia, known as the Eastern Tuna and Billfish Fishery (ETBF), underwent considerable expansion during the second half of the 1990s, with the number of hooks deployed increasing from less than 3 million in 1994 to over 11 million hooks in 01. During this period the fishery also increased its targeting of both broadbill swordfish and bigeye tuna. As a result, the catch of swordfish increased from less than 50 t in 1994 to over 3000 t in 1999, while the catch of bigeye increased from 1 t to around 1300 t in 01, While the catch of yellowfin tuna, which until 1995 had been the main target species, also increased over this period (from 1300 t in 1994 to 2700 t in 01), the relative size of the increase was not as large. The recent changes in the ETBF, together with the large spatial range and the highly migratory nature of the targeted stocks and uncertainties in the information available, pose problems regarding the ability to provide specific scientific advice on resource status, sustainable catch levels and the trade off between risk and catch to the managers of this fishery, For example, bigeye tuna, which are presently considered to represent a single stock across the entire Pacific Ocean, are targeted and caught by a large number of fisheries across this region. Furthermore, the catch of bigeye by a number of these fisheries greatly exceeds that taken within the ETBF. As a consequence, the catch of bigeye in the ETBF represents around half-of-one percent of the total catch (estimated to be around 0,000t in 00) taken from the entire Pacific stock of bigeye. Hence, unless it can be shown that the bigeye tuna that occur within the ETBF are a largely self-recruiting sub-population of the Pacific-wide stock, management of the Australian fishery alone will not be able to ensure the sustainability of this fishery. This is because the status of the stock will be, to a large extent, determined by the.size of catches taken outside the ETBF. Of the three principal target species in the longline sector of the ETBF, broadbill swordfish is the species most likely to possess a local or regional stock structure within the south-west Pacific (Reeb 00, Bremer et al 01). Furthermore, recent data indicates that Australian longliners take the largest catch from this regional stock (Campbell and Taylor 00). Consequently, swordfish has become the species most amenable to local management. This may be considered somewhat fortuitous, as coincidentally swordfish has in recent years comprised the largest component of the total longline catches taken in the ETBF. However, as no recent stock assessments have been undertaken on broadbill swordfish in the Pacific Ocean, it remains unknown whether current catches of this species are sustainable, nor is there an identified harvest strategy to deal with the any future developments within the fishery. To date, the Eastern Tuna and Billfish Fishery (ETBF) has had no formal harvest strategies for the principal catch species and work has only recently been undertaken

3 to determine a suite of performance indicators and measures for this fishery. However, in 01 the Australian Fisheries Management Authority (AFMA) agreed to the introduction of a new Management Plan for the ETBF. This Plan will include the introduction of input controls in the form of a total allowable effort (TAE), manifested as "hook days". The management plan is to take effect from mid-04. With the introduction of the management plan there is a need to determine an appropriate initial TAE. Furthermore, concomitant with the setting of this TAE will be the need to identify a range of performance indicators for the fishery and decision rules by which the TAE may be altered. In this paper we outline a method used to help evaluate a range of initial TAEs and decision rules appropriate to the harvesting of swordfish in the ETBF. 2. Evaluation of Harvest Strategies - The MSE Approach A harvest strategy is a set of rules that is used to determine a management action. The set of rules should define the data to be collected from the fishery, how those data are to be analysed, and how the results of the data analyses are to be used to determine actions. Harvest strategies may be very simple (eg. a constant catch/effort strategy) or extremely complicated (such as the determination of annual TACs based on the outcomes of a stock assessment and a set of performance based decision rules). There is currently no harvest strategy been used in the ETBF, but future management action is to be based on the setting of a Total Allowable Effort (TAE). Before any harvest strategy is adopted for the ETBF it should be evaluated against how well it is able to satisfy the management objectives for the fishery. An approach that has been developed to do this is known as Management Strategy Evaluation (MSE; Smith 1994, Punt et al 01). The primary goal of the MSE approach is to identify, in an objective and quantifiable manner, the trade-offs among the management objectives across a range of management actions. Note, it is never possible to perfectly satisfy all of the management objectives, and all harvest strategies consequently achieve some balance among them (eg. high catch, high risk; low catch, low risk). This information on the trade-offs among the management objectives is needed by the decision makers to make an informed decision about management actions, given the importance they assign to each of the objectives. The key steps in the MSE approach are outlined in Figure 1 and involves five basic steps: 1) Identification of the management objectives and representation of these using a set of quantitative performance measures. 2) Identification of alternative harvest strategies. 3) The development and parameterization of a set of alternative operating models that are used to represent the alternative realities in the calculations. 4) Simulation of the future using each harvest strategy to manage the system (as represented by each operating model), 5) The development of summary measures to quantify the performance of each harvest strategy relative to the management objectives of the fishery. An operating model is a mathematical or statistical model of the population dynamics of the fishery being studied. Each operating model reflects an alternative (yet 2

4 plausible) representation of the status and productivity of the resource and the fishing dynamics of the fleets. The operating model is used to generate observations in the form of pseudo catch, effort and catch-at-length data sets which are then used in the management procedure. Several operating models are considered because the true situation for any fishery is never well known, and a broad range of input parameter values thus needs to be examined to ensure the full range of possible resource and fleet dynamics are covered. In reality, we can only implement one strategy at a time, often with unknown results. The operating model component of the MSE framework enables a range of possible harvest strategies (e.g. TAEs) to be evaluated in terms of how well they can satisfy management objectives, prior to their actual implementation. A key feature of the MSE approach is that it can explicitly take into account a wide range of uncertainties. In particular, it can be used to identify robust harvest strategies in light of the uncertainties in the information available for managing fish resources. This is achieved by incorporating into the operational models not only the uncertainty in the underlying dynamics of the resource in response to management actions, but also the uncertainty in the methods and data used to assess the status of the resource, and uncertainty in the ability to implement management actions. This uncertainty is modelled by constructing a range of operating models each based on the given set of parameter values. As such, the approach is based on recognition that it is the combination of the uncertainties about the dynamics of the system being managed, plus the ability to measure relevant information about the system, that determines the performance and robustness of a management decision-making framework. 3. An Operating Model for Swordfish in the SW-Pacific Within the MSE framework, the broadbill swordfish, Xiphias gladius, fishery in the south-west Pacific is simulated using a range of operating models. As stated in the introductory section, stock structure is a critical factor influencing management of the ETBR Based on the understanding gained from studies previously cited, for the purposes of this study we assumed a single stock of swordfish within the SW Pacific. Adoption of this hypothesis is also warranted from a precautionary perspective, as any mixing of the swordfish population in the SW Pacific with other populations in the broader Western and Central Pacific Ocean (WCPO) will likely dilute the impact of the fishing mortality in this region. The operating model used for characterising the dynamics of the swordfish resource targeted by the ETBF is therefore based on a single stock and explicitly considers the age-structure and sex-structure of the population. Individual variability in growth (and hence the length-structure of the population) is accounted for by dividing each cohort into several groups, each of which grows according to a different growth curve. The time-step used is a quarter of the year. The model assumes that natural and fishing mortality occur continuously through a year whereby values are updated every quarter, and that movement (which is assumed to be a function of size) occurs at the end of each quarter. Spawning and recruitment are assumed occur in only two of the quarters. The model is fleet-specific for the Japanese, Australian and New Zealand longline fleets, and fishing and movement occurs across five regions (see Figure 2). Data for

5 the Japanese fleet was available since 1971 while data for the Australian and New Zealand fleets was available since 1990 and 1991 respectively. Effort for the Japanese fleet was standardised in order to obtain a more appropriate measure of effort targeted at swordfish. The Japanese fleet catch and effort data was also scaled to account for catches taken by other fleets operating in the region. Selectivity of the fishing gear was assumed to be the same for all fleets, though temporal changes in fishing efficiencies (catchabilities) for each fleet were modelled separately. Allowances for the loss of catch due to discarding and / or predation were also incorporated in the model. A full technical description of the operating model framework can be found in Campbell and Dowling (02). Where possible, the biological inputs (e.g. growth rates, age at maturity) were based on the results of published research. However, in developing an operating model for broadbill swordfish, it must be acknowledged that little is known of the biology and behaviour of broadbill swordfish in the south west Pacific. While reproductive studies have been undertaken by Young and Drake (02), little other biological work has been directed at broadbill swordfish in this area. Given the uncertainty surrounding the population dynamics, it was important to examine the sensitivity of the model outputs to a range of plausible values. Thus a range of operating models, each with a different suite of assumed biological inputs (including crosses where two inputs have been simultaneously changed), has been considered (Campbell and Dowling 03). For interest, a concise summary of results from a subset of these alternative biological scenarios are presented in this report. Specifically, results are presented from alternative scenarios where i) steepness in the stock-recruitment relation has been changed to 0.4 (reference value 0.9), ii) the natural mortality-at-age vector has been multiplied by 1.5, and iii) no movement between areas is permitted. The operating model is conditioned on historical data (catch, effort and size data from the three fisheries) for the years 1971 to 01. However, due to the lack of a good stock assessment for swordfish, knowledge concerning the status of the stock at the end of 01 presently remains unknown. Thus, for scenario testing, we considered a range of assumed levels of depletion. This subset is chosen to include upper and lower extremes as well as the assumed "reference" value. We thus considered the following three depletion levels: B(01) = 85%Bo B(0l) = 70% Bo B(01) = 50% Bo These range of values may be thought of as reflecting high, medium or low stock productivity scenarios. 4. Projections using fixed effort scenarios Having 'conditioned' the model on the historical data, the model is then 'projected' for years (80 quarters) under a given effort scenario. We use the 'reality' of the operational model to represent and evaluate the impact of any given effort scenario on the future condition of the fishery and the stock. 4

6 In consultation with industry representatives and managers for the ETBF, a range of future effort scenarios were developed for consideration, These scenarios also had to include future scenarios for the two foreign fleets included in the operational model. After taking a number of factors into consideration, the following future effort scenarios were agreed upon for evaluation: Domestic Effort: 1. Status quo: effort stays at 01 level (11.2 million hooks) (with or without annua! effort creep of 2%; see below) 2. Increase in effort over five years to 1.5 x 01 level (16.8 million hooks) + annual effort creep of 2 % 3. Increase in effort over five years to 2.0 x 01 level (22,4 million hooks) + annual effort creep of 2 % 4. Increase in effort over five years to 2.5 x 01 level (28.0 million hooks) + annual effort creep of 2 % Foreign Effort 1. Status quo: effort stays at 01 level 2, Increase in effort over five years to 2.0 x 01 level + effort creep of 2% per annum for first five years only Each domestic effort scenario was crossed with each foreign effort scenario, giving a total of 8 different effort scenarios. (N.B. when both domestic and foreign effort was held at the status quo, no effort creep was applied to either. If the domestic nominal effort was at the status quo and the foreign effort increased, then effort creep was applied to both fleets). Time series plots of annual domestic and total effort under each of these scenarios are shown in Figure 3. These effort scenarios cover a diverse range of possibilities and will hopefully bracket future changes in domestic and foreign effort. Domestic effort scenario 4 crossed with foreign effort scenario 2 is the most extreme scenario. Holding foreign effort at the status quo while domestic effort increases its effort and efficiency approximates the situation where the domestic fleet "out-competes" the foreign fleets. Under any future effort scenario, the operational model allows predictions to be made of the corresponding annual catches for each fleet and the response of the swordfish population to these catches. One hundred Monte Carlo simulations were run for each effort scenario, under each depletion regime (B(01) = 85%, 70% or 50% Bo), The model predictions were summarised in the following manner across the 0 iterations: i) time-series plot, and 95 th confidence limits, of the mean spawning biomass in Area 2 (expressed as a percentage of the original biomass), ii) time-series plot of the mean and upper 95 percentiles of average weight of fish caught by the Australian fleet in Area 2, iii) time-series plots of the mean percentage of fish in Area 2 (by number) within each of three size classes (< 25kg, 25 < processed weight < 50kg, > 50kg), iv) histogram of the average annual Australian catch over the projection years, v) histogram of the final total spawning biomass relative to initial total spawning biomass, vi) histogram of the probability that the total spawning biomass drops below 30% of its initial level,

7 vii) histogram of the probability that the total spawning biomass drops below 50% of its initial level. The results were also summarised using various performance indicators and measures to evaluate each future effort scenario across each of the assumed levels of depletion. These performance measures fall into two broad classes - economic and conservation measures. For example, the average size of the annual catch is an economic measure of the outcome of the harvest strategy whilst the final biomass is a proxy conservation measure. For each performance indicator or measure, the mean (or median) was taken across each of the 0 simulations. The performance indicators / measures which were considered were as follows: Economic 1. Average annual Australian catch (tonnes), 2. Median percentage change in the Australian catch between years, 3. Average weight of individual fish caught by the Australian fleet in Area 2, 4. Average percent of fish in the Australian catch from Area 2 that are greater than 50kg, 5. Average annual value of the Australian catch (millions of dollars) (calculated assuming values of $4.50/kg for 0-25kg fish, $6.50/kg for 25-50kg fish, and $ll/kg for >50kg fish). Conservation 1. Average final spawning biomass relative to the initial ("virgin") spawning biomass, 2. Average probability that the spawning biomass drops below 30% of its initial value (as measured by the average proportion of quarterly time steps in which the spawning biomass is less than 30% of its initial value), 3. Average probability that the spawning biomass drops below 50% of its initial value (as measured by the average proportion of quarterly time steps in which the spawning biomass is less than 50% of its initial value). 5. Results i) Time series and histograms Examples of time series and histogram results for the 8 effort scenarios listed above are presented in Figures 4-. These are based on the reference conditioning scenario, which incorporates the "best guess" of the biology (on the basis of published research) for the swordfish population together with an assumed depletion at the end of 01 of 30% (i.e. total spawning biomass at end of 01 is equal to 70% of the initial total spawning biomass). The impact of the increased catches of swordfish by Australia and New Zealand after the mid-1990s is seen in the steeper decrease in the Area 2 biomass trajectories after the 25 l year. The tightening of the 95 lh percentile ranges after the conditioning years is due to the level of spawning biomass depletion being constrained at 30% within the mode! fitting. Indeed, plotting total spawning biomass across all areas shows bottlenecking about year 32, the first projection year. The standard deviations for the total spawning biomass ranged from 6% to 8% across all effort scenarios. 6

8 Under the status quo scenario, the spawning biomass in area 2 was predicted, on average, to decline to around 49% of its initial level by the end of the years of projections (Figure 4). The 95 th confidence interval for this estimate was between 34% and 64% of the initial spawning biomass. The introduction of 2% annual effort creep and a doubling in foreign effort (scenario 2) further reduced the predicted spawning biomass to 32% of its initial level. In the most extreme scenario, where Australian effort is increased by 2.5 times to 28 million hooks over 5 years, with the same effort creep and rease, the final spawning biomass was estimated to be depleted to around 17% of its initial value. In general, the spawning biomass, both in area 2 and overall, decreased with increasing effort (Figures 4 & 8). Relative to the scenarios with effort creep and a doubling of foreign effort, those where the foreign effort was held at status quo resulted in 29% to 32% higher mean final spawning biomasses, for corresponding levels of increase (with effort creep) in Australian effort (Figure 8). Time-series plots of the mean weight of fish caught by the Australian fleet in Area 2 under each of the 8 effort scenarios are shown in Figure 5. The drop at time step 89 was due to a zero recorded catch. Declines in the mean weight of fish caught were seen for all scenarios, though the decline for the status quo scenario was relatively minimal, As with the spawning biomass, the mean weight of fish decreased with increasing effort. The mean weight was relatively insensitive to the effort scenarios, with declines more pronounced for the upper 95 th percentile. A similar result was also noted by Punt et al., (01). Under the most extreme scenario, where Australian effort is increased by 2.5 times to 28 million hooks over 5 years, with the same effort creep and rease, the upper 95 th weight percentile declined from approximately 115kg to approximately 65kg. Finally, time-series plots of the percentage, by weight, of fish caught by the Australian fleet in Area 2 that are within three size classes are shown in Figure 6. Under all scenarios there was a switch in the dominant size class of fish in the catch, At the start of the projection years, fish greater than 50kg comprised % of the catch, but by the th year medium (25-50kg) and small fish (<25kg) dominated, with these categories comprising between 32% and 41% of the catch, with a greater relative increase in the small size category. Note, the proportion of medium fish was 33% at the start of the projection years, and the proportion of small fish 27%. The exception was the status quo scenario, where the medium and large-sized fish dominated after the th projection year. The transition in size dominance in the catch from large to small and medium fish occurred between the 4 lh and 7 lh projection years for the scenarios where domestic effort was doubled or increased by 2,5 times (to 28 million hooks) over 5 years, and between the 6 th and 13 th projection years for those scenarios where domestic effort increased by 1.5 times (to 16.8 million hooks), or only effort creep was applied. The decrease in the proportion of large fish in the catch became more substantial the larger the future effort. Histograms of the average annual Australian catch over all projection years, together with the overall mean and standard deviation, are shown in Figure 7. As would be expected, the annual size of the Australian catch over the projection years was generally greater with greater effort. Additionally, the spread of the catch about the mean across the 0 simulations increased with increasing effort and effort creep. Mean catches ranged from 2878 tonnes under the status quo scenario, to 5764 tonnes

9 when Australian effort was increased to 28 million hooks, with effort creep, and foreign effort was held at status quo. Note that, for the same effort, Australian catches were 12% to 13% higher when foreign effort was held at status quo (scenarios 6-8), as compared to when foreign effort doubled, This suggests that increases in both the domestic and foreign fleets resulted in fishing down of the biomass to levels such that catches began to decline. As a collorary to the previous result, with increased effort and catches there was an increased probability that the biomass would be fished down to levels below 30% and 50% of its initial level (Figures 9 & ). However, within each set of effort creep / foreign effort scenarios, the increase in the average probability that the spawning biomass dropped below 50% as domestic effort increased was only slight (maximum range of 7.4%). For the status quo scenario, spawning biomass never dropped below 30% (Figure 9), and for 55 of the 0 simulations, the spawning biomass never dropped below 50% of its initial level, and on average dropped below 50% in only 8 percent of ail time steps (Figure ). Otherwise, on average, the spawning biomass dropped below 50% of its initial level between 24.7% (effort creep and foreign effort doubled) and 30.4% (increase to 28 million domestic hooks in 5 years, with effort creep) of all time steps (Figure ). For the three scenarios when domestic effort was increased and foreign effort doubled, with effort creep, the biomass dropped below 30% its initial level in % (increase to 16.8 million domestic hooks in 5 years), 16% (increase to 22.4 million domestic hooks in 5 years) and 21% (increase to 28 million domestic hooks in 5 years) of all quarters on average. Under the corresponding scenarios when foreign effort was held at the status quo, these proportions of quarters in which the spawning biomass dropped below 30% were reduced to, on average, 2%, 7% and 15% (Figure 9). These results indicated that large increases in the combined effort of both the Australian and foreign fleets would place pressure on the swordfish population. For the most extreme fishing scenario (increase to 28 million domestic hooks in 5 years, doubling of foreign effort, effort creep), the probability that the population drops below 30% of the initial biomass was estimated to be 98%. Indeed, the combined effect of increased domestic effort, doubling of foreign effort and effort creep yielded the worst-case scenario for the fishery, in terms of spawning biomass, weight of fish in the Australian catch, and the average Australian catch. ii) Performance indicators and measures: comparison across effort scenarios and assumed depletions Table 1 gives the results for each of the performance measures under each of the effort scenarios and across the range of assumed depletions (note that each fixed effort scenario is assigned a reference number). These results are displayed graphically in Figure 11 for six of the performance indicators. The median annual change in Australian catch was relatively insensitive to both the effort scenario and assumed depletion, ranging only between.15% and 13.12%. It decreased slightly with increased depletion, and increased with increasing Australian effort. Trends in average annual value of the Australian catch tracked the changes in the average annual Australian catch, and so are not plotted. 8

10 The performance of the fishery improved (greatest catches, lowest depletion of the swordfish biomass) as the assumed depletion was reduced. This follows since this assumption implies that the stock has been more resilient to the historical fishing effort. Within each assumed depletion scenario, the relative trends across the 8 future effort scenarios were generally consistent. The status quo scenario gave the best fishery performance in terms of the conservation indicators (spawning biomass and average weight of fish in the Australian catch), but the worst in terms of average catch, while the scenario with 28 million hooks, effort creep, and foreign effort doubled resulted in the worst performance in terms of the conservation indicators, while it yielded the second-highest domestic catches within the 15% and 30% depletion regimes, and the 4 th highest within the 50% depletion regime. However, average Australian catches were consistently higher for corresponding levels of effort when foreign effort was held at status quo, such that an increase to 28 million domestic hooks over 5 years, with effort creep, but foreign effort at status quo gave the highest average catches within each depletion regime. For the 50% depletion regime, higher Australian catches were achieved for each effort level when foreign effort was held at status quo, than those achieved at the highest effort level in scenarios where foreign effort was doubled. Indeed, the average Australian catch showed stronger relative sensitivity to the assumed depletion for a subset of effort scenarios. Average catch for scenario 6 (increase to 16.8 million domestic hooks over 5 years, effort creep, foreign status quo) showed a lower relative decrease with increasing assumed depletion levels. When 15% and 30% depletion was assumed, a lower average catch was achieved for scenario 6 than for scenarios 4 and 5 (increase to 22.4 and 28 million domestic hooks over 5 years, foreign effort doubled, effort creep). However, for 50% depletion, the average catch for scenario 6 was higher than that for scenarios 4 and 5. Additionally, average catch for scenarios 5 (increase to 28 million domestic hooks over 5 years, effort creep, foreign effort doubled) and 2 (foreign effort doubled, effort creep applied to all fleets) showed a greater relative decrease with increasing assumed depletion levels. When 15% depletion was assumed, scenario 5 yielded higher average catches than scenarios 6 and 7 (increase to 16.8 million and 22.4 million domestic hooks over 5 years, effort creep, foreign status quo), but lower average catches for both scenarios 6 and 7 when 50% depletion was assumed. Scenario 2 resulted in higher average Australian catches than the status quo scenario when 15% or 30% depletion was assumed, but lower catches than the status quo scenario when 50% depletion was assumed. Correspondingly, it was noted that the final spawning biomasses for scenarios 4 and 5 dropped to less than 7% of the initial "virgin" value when 50% depletion was assumed, while that for scenario 6 was at 15%. That is, when 50% depletion was assumed, greater domestic catches were achieved for lesser domestic effort when the total effort was reduced. This shows that the population is less resilient to increases in total effort when the assumed depletion is high. Lowered resilience with increasing total effort was evident to a lesser degree across all depletion regimes, in that greater domestic catches were taken for the same level of effort when the foreign effort was held at status quo.

11 The average individual weight of fish caught by the domestic fleet in Area 2 was the least sensitive/most robust performance indicator to the range of assumed depletions and to the effort scenarios. There was a 4.4kg range within the 15% depletion regime, which equated to a maximum of 9.3% variation between the effort scenarios. The variability increased with assumed depletion, with a maximum 24% variation for the 50% depletion regime. Consistent with the time series plots of mean weights (Figure 5), the upper weight percentiles were more sensitive to the effort scenarios, and to the assumed depletion. Under the status quo scenario (that with the least variation across depletion regimes), the average percentage of fish taken from Area 2 and weighing more than 50kg was 41% assuming 15% depletion, (maximum variation of 14% between effort scenarios within this regime), and 25% assuming 50% depletion (maximum variation of 58% between effort scenarios within this regime). Final spawning biomass was the most sensitive performance indicator between effort scenarios and across depletion regimes. The relative final spawning biomass ranged from 38% for scenario 5 (increase to 28 million domestic hooks over 5 years, doubling of foreign effort, effort creep) to 71% for the status quo scenario, assuming 15% depletion. For the 50% depletion regime, final spawning biomass ranged from 30% of its initial value for the status quo scenario, to 4.7% under scenario 5. Indeed, for the 50% depletion regime, all but the status quo scenario resulted in final spawning biomasses less than 16% of the initial level, while for the 30% depletion regime, all but two effort scenarios (1 and 6) gave final spawning biomasses less than 35% of the initial level. The probability that, or proportion of quarterly time steps in which, the spawning biomass dropped below 30% of its initial value, was highly sensitive to the assumed depletion. When 50% depletion was assumed, this probability was 26-32% for all but the status quo effort scenario. That is, at 50% depletion, the probability of spawning biomass dropping to below 30% Bo was relatively insensitive to the amount by which effort was increased. At an assumed depletion of 30%, these probabilities were more sensitive to the effort scenario. Those with a 2.5 times increase in Australian effort (to 28 million hooks), with effort creep, and scenario 4 (2.0 times increase in Australian and foreign effort, with effort creep), all yielded probabilities greater than 14%, while scenarios 2 (effort creep and foreign doubling) and 6 (increase to 16.8 million domestic hooks over 5 years, effort creep, foreign status quo), had less than 5% probability of spawning biomass declining to less than 30% Bo. For an assumed. depletion of 15%, only scenario 5 (increase to 28 million domestic hooks over 5 years, doubling of foreign effort, effort creep) had a greater than 2% probability of spawning biomass declining to less than 30% Bo, For an assumed depletion of 50%, the probability that the spawning biomass dropped below 50% of its initial values was insensitive to the assumed effort scenario. The level of depletion was such that at the start of the projection years, the spawning biomass dropped below 50% of its initial level and did not recover, such that 39.4% of all time steps recorded a spawning biomass less than 50% of the initial value. At lower assumed depletions, these probabilities were more sensitive to the effort scenarios. For 30% depletion, scenarios where Australian effort was at least doubled resulted in the spawning biomass dropping below 50% in over 25% of the time steps (with the exception of when Australian effort was increased to 28 million hooks over

12 15 years with no effort creep). For the 15% depletion regime, the proportion of time steps in which the spawning biomass dropped below 50% of its initial value ranged from 2.6% for scenario 6 (increase to 16.8 million domestic hooks in 5 years, effort creep, foreign status quo) to 16% for scenario 5 (increase to 28 million domestic hooks in 5 years, foreign effort doubled, effort creep). Hi) Alternative operating models: sensitivity to biological inputs A concise summary of the results from the three alternative operating models, each with an altered assumed biological input, and from the baseline scenario, is presented in Figure 12. It should be noted that the reference level of depletion (30%) was assumed for all scenarios. The effect of changing these biological inputs on the average annual Australian catch and mean final spawning biomass proportion is presented for two future fixed effort scenarios: the status quo (no increase in effort for Australian or foreign fleets; no effort creep) and the most extreme effort scenario of increasing domestic effort to 28 million hooks over 5 years, and a doubling of foreign effort, with effort creep. None of the alternative biological scenarios impacted the Australian catch or the final spawning biomass to the same extent as when the assumed depletions were changed (Figure 11 versus Figure 12). The relative changes in catch and biomass between the biological scenarios were generally more pronounced for the higher effort scenario, suggesting interaction between the biological variables and the effort level (Figure 12). For the lower steepness, the discrepancy between the recruitment at lower versus higher steepness was more pronounced at lower biomasses, which were more likely to result when effort was higher. Decreasing steepness resulted in both lower domestic catches and a decrease in final spawning biomass, relative to the reference scenario, since a less productive stock ultimately results in a smaller biomass. Increasing natural mortality decreased the domestic catch, but, as a result of compensatory increased productivity, the spawning biomass increased relative to the reference scenario. Eliminating fish movement between the 5 regions of the model had the strongest relative impact on the performance measures, reducing the mean Australian catch but increasing the total spawning biomass. This occurred because the more heavily fished areas were not replenished and hence catches declined, and the remaining biomass was conserved in less heavily-fished areas, rather than moving into "sink" areas where it would have been removed. At higher effort levels, the relative magnitude of this conservation effect increased. 6. Discussion The results for the fixed effort scenarios presented here are taken from an operating model that assumes the baseline biological parameters; that is, the "best guess" based on the knowledge to date. The outcomes were sensitive to assumptions made about the natural mortality, recruitment and extent of movement within the population, However, the model outputs were equally or more sensitive to the assumed depletion level at the end of the historical time-period used to condition the model. Clearly, the results for the various fixed effort strategies are highly sensitive to the assumed level of depletion. This is to be expected, given that as the assumed depletion increases, the inference is that the swordfish population has been less resilient to the historical levels of fishing effort. Thus, as future effort is increased, the stock it

13 performs more poorly if depletion levels are higher. When the assumed depletion is set at 50%, future increases in actual or effective effect are likely to be at a high risk to the population, with spawning biomass being depleted to less than 16% of its initial "virgin" value. If 30% depletion is assumed, then 2.5-fold increases in Australian effort are also likely to be a high risk (spawning biomass reduced to less than 29% of its initial level), especially foreign effort also doubles. For the 30% depletion regime, the three lowest effort scenarios (i.e. 1) status quo, 2} domestic status quo with effort creep and rease, and 3) domestic increase of 1.5 times over 5 years with effort creep and foreign status quo), were the only ones that kept the final spawning biomass at over 30% of its initial level on average. Under an assumed depletion level of only 15%, all future effort scenarios had more favourable outcomes, in terms of both economic and conservation performance indicators, with only scenario 5 (increase to 28 million domestic hooks in 5 years, effort creep, foreign effort doubled) resulting in greater than 60% decline in the mean final spawning biomass relative to its initial level. However, it may be a higher risk strategy to assume that the impact of the historical fishing has been as small as this. The question remains as to which is the most realistic depletion assumption. The only data available to address this is the New Zealand length-frequency and the weightfrequency data from Australian observers on Japanese vessels. The goodness of fit of the model to this data may give some indication as to the most appropriate level of depletion. The weight-frequency data is available by quarter for areas 1-3, while the length frequency data is available for quarters 2 and 3 within area 5. The total residual sums of squares across these 14 distributions was x ' 2 for the 30% depletion regime x " 2 for the 15% depletion regime, and x ' 2 for the 50% depletion regime. Thus there was very little difference in goodness of fit of the model to the data. However, the total sums of squares for the 30% depletion regime was 2.1% lower than that for the 15% depletion regime, and 1.4% lower than that for the 50% depletion regime. This may suggest that the true depletion lies closer to 30% than to 15% or 50%. In the absence of any additional evidence to support the 30% depletion assumption, it is still probably the most sensible one on which to base management decisions, given that 15% is not precautionary and 50% is likely to be extreme, thus invoking unnecessary or premature limitations on recommended future effort levels. The spawning biomass, and the performance measure pertaining to spawning biomass (the probability of it dropping below 30% of its initial value) are the most sensitive indicators in response to the effort scenarios and assumed depletions. However, at present, information on the status of the stock biomass is unavailable. As many assessments are based on catch trends, which are less sensitive to the changes in underlying biomass, extra caution should be exercised in interpreting the outcomes of these assessments. The above analyses have shown that the upper percentiles of the mean weight of individual fish in the catch are a more sensitive fishery dependent indicator of the population's response to changes in effort or assumed depletion. Indeed, for the 30% depletion regime, all of the effort scenarios resulted in a switch in the dominant size class from large to small and medium fish. As such, these indicators should also be considered in addition to the outcomes of assessments based on catch and effort data. 12

14 The results indicate that many of the future effort scenarios considered may result in biomass levels being fished down below desired reference levels. However, the achievement of management objectives for any individual fleet will be dependent on the future effort levels of the other fleets. For example, when foreign effort remains at the status quo, as opposed to being doubled, the same level of Australian effort had a reduced probability of driving the spawning biomass to below 30% of its initial value, and Australia achieved higher catches for identical or lower levels of effort, Additionally, when 50% depletion was assumed, the final spawning biomasses were less than 11% of the initial "virgin" value for scenarios with effort creep, doubling of foreign effort, and doubling or more of Australian effort. As such, the impact of a set level of domestic effort also depends largely on the changes in effort of the foreign fleets, and on their efficiency. This reinforces the need for multi-lateral management arrangements for widely distributed and highly migratory stocks such as swordfish. The investigation of future projections under set levels of effort has been valuable in examining how hard the swordfish population can be fished, and in determining an appropriate initial TAE for the domestic fleet. It has enabled the effects of alternative levels of domestic and foreign effort, and that of effort creep, to be assessed. However, in practise, it is unlikely that any management plan would permit a longterm fixed effort level that disregards changes in performance indicators and results from ongoing assessments. Ideally, harvest strategies should incorporate decision rules whereby the status of the fishery is regularly assessed, whether empirically or using a model-based stock assessment, and alternative management strategies are applied depending on the results. Without management feedback loops, high levels of combined effort will continue to drive down the biomass, as has been illustrated in this report. However, if effort levels can be adjusted according to empirical or model based stock assessment outcome, there is less conservation and economic risk associated with increases in effort. In the attached Appendix, two empirical approaches for updating a TAE are outlined. In summary, the MSB framework is a valuable tool for ongoing analysis of alternative harvest strategies, and the results presented here are but a small subset of the management strategies that may potentially be evaluated using this technique. While this report focuses on the use of the MSE approach in the context of evaluating fixed effort strategies to assist with establishing an initial TAE for the Australian domestic fleet, the operating model may as readily, and indeed should, incorporate formal stock assessments for broadbill swordfish when they become available. Furthermore, with some adaptations to the operating model to accommodate the population dynamics of the tunas, it could also be used to evaluate management strategies for yellowfin or bigeye tuna. Indeed, the definitive MULTIFAN-CL stock assessments that have been developed by the SPC for yellowfin, bigeye, skipjack and albacore in the WCPO could be readily incorporated into an MSE framework for these species in the WCPO. This would enable a comprehensive evaluation of proposed harvest strategies within the "reality" described by the operating model(s). Acknowledgements Thanks are extended to Andre Punt, Tony Smith, Tom Polacheck and Marinelle Basson for guidance and suggestions in the completion of this work. Paavo

15 Jumppanen and Dale Kolody are also thanked for assistance with the use of AD Model Builder. The authors also gratefully acknowledge the support received from the Fisheries Research and Development Corporation and AFMA towards this project.

16 References Bremer, J. A., Hinton, M. and Greig, T. 01. Genetic analyses of nuclear and mitochrondrial DNA data indicate heterogeneity within the Pacific Ocean. Handbook and Abstracts, Third International Billfish Symposium, held August, Cairns, Australia. Campbell, R.A. and Dowling, N.A. (03). Harvest strategy evaluation: examination of alternative biological scenarios. FRDC 99/7 Draft chapter for final report. Campbell, R.A. and Dowling, N.A. (02). Technical description of the operating model to be used for evaluation of harvest strategies for swordfish in the eastern tuna and billfish fishery. FRDC 99/7 Milestone Report - April 02. Campbell, R.A. and Taylor, N.A. (00). Data and biological parameter specifications for a spatially structured operating model for broadbill swordfish and bigeye tuna in the south-west Pacific. FRDC 99/7 Milestone Report - August 00. Punt, A.E., and Smith, A.D.M.(1999). Harvest strategy evaluation for the eastern stock of gemfish (Rexea solandri). ICES J. Mar. Sci. 56: Punt, A. E., Cui, G, and Smith, A. D. M. (01). Defining robust harvest strategies, performance indicators and monitoring strategies for the SEE Fisheries Research and Development Corporation Report 98/2, Australia Reeb, C.A., Arcangeli, L., and Block, B.A. (00) Structure and miogration corridors in Pacific populations of the swordfish, Xiphias gladius, as inferred through analysis of mitochondrial DNA. Working paper BBRG-13 presented at the I3,h meeting of the Standing Committee on Tuna and Billfish, held 5-12 July 00, Noumea, New Caledonia. Smith, A.D.M. (1994). Management Strategy Evaluation - The Light on the Hill. pp In: D.A. Hancock [Ed.] Population dynamics for fisheries management. Australian Society for Fish Biology Workshop Proceedings, Perth August 1993, Australian Society for Fish Biology, Perth. Young, J., and Drake, A. (02). Reproductive dynamics of broadbill swordfish {Xiphias gladius) in the domestic longline fishery off eastern Australia. FRDC 1999/8 Final Report. 15

17 Table 1 Mean and standard deviations (across 0 model iterations) of all performance measures, for each of the 8 fixed effort scenarios, under three alternative depletions and future effort scenarios. Effort scenario (see befowfor description) Average annua! Australian catch (tonnes) mean St. dev Medan annual change in Australian catch (%) mean St. dev a Average weight of fish in Australian Area 2 catch (kg) mean Z St. dev a Average % fish in Australian Area 2 catch >50kg mean St. dev Average annual value of Australian catch ($million) mean St. dev Final spawning btomass relative to initial spawning biomass (%) mean I sldev , , Average P(SpBio<0.3Bo)(%) mean St. dev Average P(SpBio<0.5Bo)(%) mean st. dev B Status quo 2 Effort creep & rease 3 Domestic increase of 1.5 over 5 years, with effort creep 4 Domestic increase of 2.0 over 5 years, with effort creep 5 Domestic increase of 2.5 over 5 years, with effort creep 6 Domestic increase of 1.5 over 5 years, with effort creep, but foreign status quo 7 Domestic increase of 2.0 over 5 years, with effort creep, but foreign status quo 8 Domestic increase of 2.5 over 5 years, with effort creep, but foreign status quo

18 Table 2. Comparative summary of mean values for the main economic and biological performance indicators, for each fixed effort scenario, grouped by the assumed depletion level. Values for the economic indicators are expressed relative to those obtained for the status quo effort scenario. B(01) = 85%Bo Management Objectives Economic (domestic fleet) Conservation Total Fish Proportion Final Proportion Proportion Catch Weights Large spawning Biomass Biomass Relative to Status Quo Biomass <30% Bo <50% Bo c Status quo status quo & rease + effort creep 1, , , mill 5yrs + effort creep & foreign status quo 7 mil! 5yrs + effort creep & rease , , mill 5yrs + effort creep & foreign status quo 12 mill 5yrs + effort creep & rease mill 5yrs + effort creep & foreign status quo 28 mill 5yrs + effort creep & rease B(01) = 70%Bo Management Objectives Economic (domestic fleet) Conservation Total Fish Proportion Final Proportion Proportion Catch Weights Large Spawning Biomass Biomass Relative to Status Quo Biomass <30% Bo <50% Bo Status quo Status quo & rease + effort creep mill 5yrs + effort creep & foreign status quo 17 mill Syrs + effort creep & rease mill 5yrs + effort creep & foreign status quo 22 mill Syrs + effort creep & rease , mill Syrs + effort creep & foreign status quo 28 mill 5yrs + effort creep & rease B(01) = 50%Bo Management Objectives Economic (domestic fleet) Conservation Total Fish Proportion Final Proportion Proportion Catch Weights Large Spawning Biomass Biomass Relative to Status Quo Biomass <30% Bo <50% Bo Status quo Status quo & rease + effort creep , mill 5yrs + effort creep & foreign status quo 17 mill Syrs + effort creep & rease , ,96 39, mill Syrs + effort creep & foreign status quo 22 mill 5yrs + effort creep & rease , mill 5yrs + effort creep & foreign status quo 28 mill Syrs + effort creep & rease ,

19 Specify Management Objectives I)ey4qpf»«rformance Measures; Develop Operating Model Develop Harvest Strategies Generate Annual Data I ADDIV Harvest Strategy Apply Stock Assessment 1 Update Performance Indicators Update Sworcifish Population Apply Decision Rule Model Output Performance Measures Figure 1. Outline of the MSE approach.

20 Figure 2. Spatial structure used in the operating model of the broadbill swordfish fishery in the SW Pacific.

21 Status quo - " - Status quo - effort creep & =K«=»*16.8 mill 5yrs 22.4 mill 5yrs -28mill5yrs % 1-1- S -l 80 H ' «* Year Status quo -Status quo-effort creep & mill 5yrs foreign s.q mill Syrs foreign Inc mill 5yrs foreign s.q mill 5yrs foreign Inc 2Bmill Syrs 28.0 mill 5yrs foreign tnc Figure 3. Time series of annual (a) domestic effort, and (b) total effort, under each of the scenarios described in the text,

22 i Scenario 1: status quo - lowef 95th percentile -mean upper 95th percentile mean linal biomass = 49% 1» I 5 log 3 1 " Scenarki 2: effcreep &foreignincrease - lower 95th pefcenlite mean upper 95th percentile mean final biomass = 32*& «is 3> as JO *s» 6 5» '5» 30» ««Figure 4. Time^series plot of the expected Area 2 spawning biomass (expressed as a percentage of the original biomass), together with the 95' confidence limits, under the 8 effort scenarios described in the text. 21

23 Scenario l status quo Scenario 2: etlcreep & rease MO : 1 \ 0 I a so S i e so 0 0 time -mean upper 951h percentile S 80 itl E 60 0 J 0 nr 0 lime mean upper 95lii percentile Scenano 3: ettcreep& 16,Srnill 5yrs Scenano 4: eftcreep & 22.4miH Syrs Scenario 5: eticreep & 28mifl 5yrs 1 1 -mean upper 951h percentile 1 1 -mean upper 95!b percentile 1 1 -mean -upper 95th percentile SO 0 0 time «0 E time 0 1 B E GO 0 nr J L 0 time 1 0 I 80 E 60 JO 0 Scenario B: eftcreep 16.8mill Syrs & foreign status quo mean - upper &5th percentile! ' * * *,! v. _. " r-taj.-., 0 0 time 1 I 1 0 S «> m _ E GO Scenano 7: etfcreep 22.4mill 5yre & foreign status quo ^. f^^w,-! V """" '^^ } mean / " ta " _. 1 " upp 6 ' 95th percentile "* ~~"' l *-~V'\s time I 80 ' E GO - Scenano B: etfcreep 28mil Syrs foreign status quo) mean upper 95th percentile - -1 '"%,! 8 l " - 0 SO time Figure 5. Time-series plot of the expected mean weight of fish caught by the Australian fleet in Area 2, together with the upper 95 th limit, under the 8 effort scenarios described in the text. confidence 22

24 Figure 6. Time-series plot of the expected proportion of the total catch by the Australian fleet in Area 2 within various size classes, under the 8 effort scenarios described in the text. 23

25 - ' 1 Scenario 1: status quo Scenario 2. effcreep & rease mean - 287Q.41 st.dev B r mean ^ side* = r^p "1 fl,.-i i-i- - -v... CO W CO c\l <M ru n <n tn m u> to O t B N lo w PI n co to in m UJ (D as J Scenario 3: effcreep & 16.6min 5yrs DL mean et stdev = Scenario 4: effcreep & 22.4mill 5yrs ril -- r mean = stdev. = n Q O O O -V (O D CO Scenario 5: effcreep & 28miN 5yrs ^flftfm M n n n 8 S mean = st.dev Scenario 6: effcreep 16Bmi>4 Syrs & foreign status quo Scenario 7: effcreep 22.4mfll 5yrs ft foreign status quo Scenario 8: effcreep 28m?Ji Syrs & foreign status quo 15 5 CO o CM rl -, r-i 1. - mean = st.dev. = o o o o o o o o a o o o o o o o mean = stdev. = J r ^-1 t ; i 1- _ e o o o o o o o o o o o o o o o o o o o o Q ^ C O C U C O Q ^ T C O Q I t O O * * d i o r N c l o ^ - ' T ^ i r j i n c o c a,, III II mean = st.dev. = j P-, r~i r T- _J.. T -hr 3 =H O C S C O Q O O O D O O O O O O O O O O O o ^ - t s c u c D c i ^ c D r ^ c D O ^ n p j m o x f c v o i c M C? n «^ ^ i n t o t o r D C D l - - r > - c a 4 C i n Figure 7. Histograms (over 0 simulations) of the average annual Australian catch (in tonnes) over the projection years, under the 8 effort scenarios described in the text. 24

26 Scenario 1: status quo Scenario 2: eticreep & rease mean= 0.50 st. dev. = mean= 0.33 st. dev. = r~ «M a -Pi tu Scenario 3: effereep & 16.8mill 5yrs Scenario 4: effereep mill 5yrs Scenario 5: effereep S 28mill 5yrs mean= St. dev. = mean= st dev. = D. 0 J 1-4 Scenario 6: eticreep 16.8mill 5yrs & loreign status quo Scenario 7: eticreep 22.4mill 5yrs & (oieign status quo Scenario 6: eflcreep 28mW 5vrs & loreign status quo mean= si. dev. = mean= St. dev. = 0.06S mean= st. dev. = 0.059,. r ' ~U ti~ Figure 8. Histograms (over 0 simulations) of the final spawn.ng biomass relative to initial ("virgin") spawning biomass, under the 8 effort scenarios described in the text. 25

27 0 90 BO Scenario 1: slatus quo mean =0.0 si tfev. = SO 30 - ttu Scenario 2: elfcreep & rease to mean = st. dev = t^! i -! I SO i 0.05 dbite 0.15 Scenario 3: ettcreep & 16.8mill 5yrs mean = st. dew. = T Scenario 4: ehcteep miH 5yrs mean = SI dev = Scenario 5: effcreep & 26m315yrs u mean = 0.7 st. dev S4 Scenario 6: effcreep 16.8mill 5yrs & loretgn status quo Scenario 7: effcreep 22.4mi0 5yrs & foreign stalus quo Scenario B: effcreep 28mil 5yrs K foreign status quo A-f O.05 TU 0.15 ' 0.25 ' mean = st, (lev, = ' i- cy mean = St. dev. = BO O-l T 'I -4^m pi in o r^pt *1 ^ H" 1 1"! r v> tn in m n o o o o o mean = si. dev. = in Kt m m <0 I -1 - OJ tn O O O o Figure 9. Histograms (over 0 simulations) of the probability that the spawning bioniass dropped below 30% of its initial level, under the 8 effort scenarios described in the text. 26

28 Scenario 1: status quo Scenario 2: effcreep & rease mean = 0,079 St. dev mean = 0247 ~rrn ^ a Scenario 3: effcreep S 16.8miU Syrs mean = si. dev. = f-h" o - 1 Scenario 4: effcreep & 22 4milt Syrs St, dev. = mean = st. dev. = in o 0.15 Scenario S: effcreep & 28mll 5yrs mean = st. dev =00164 m (rt m ir> <n m f^ SJ CJ -<t tfi <q r-:? o o o o cj o o 0.95 j i I! i i I Scenario 6: etfcreep 16.8mU Syrs & foreign status quo Scenario 7: effcreep 22.4mSI 5yrs foreign status quo Scenario 8: effcreep 28mili mean = st. dev. = H ' 0.05 r^ r_n 1 m in ir> «o *- CM n ^t o cj o o mean = st. dev. = Figure. Histograms (over 0 simulations) of the probability that the spawning biomass dropped below 50% of its initial level, under the 8 effort scenarios described in the text. 27

29 Average annual Australian catch (t) 15% Depletion 30% Depletion 50% Depletion Status quo H Status quo - effort creep & ai6.8mill5yrs E 16.8 mill 5yrs 22.4 mill 5yrs E mill 5yrs D28.0 mill 5yrs 28.0 mill 5yrs Final spawning biomass as percentage of initial ("virgin") biomass 15% Depletion 30% Depletion 50% Depletion Status quo B Status quo - effort creep & 16.8 mill 5yrs (316.8 mill 5yrs 22.4 mill 5yrs H 22.4 mill 5yrs 28.0 mill 5yrs H 28.0 mill 5yrs Average weight of fish caught (kg) in Area 2 - projection years 50- m - S.304 _o 2 15% Depletion 30% Depletion 50% Depletion Status quo 13 Status quo - effort creep S 16.8 mill 5yrs mill Syrs mill 5yrs [ mill 5yrs 28.0 mill 5yrs 28.0 mill Syrs Average Pr(spawning biomass < 0.3Bo) (%) 15% Depletion 30% Depletion 50% Depletion Status quo Status quo - effort creep & 16.8 mill 5yrs m 16.8 mill 5yrs 22.4 mill 5yrs H 22.4 mill 5yrs 28.0 mill 5yrs D 28.0 mill 5yrs Average percentage ol fish >50kg in Area 2 - projection years 45' ID S' 0 15% Depletion 30% Depletion 50% Depletion 9 Status quo S Status quo - effort creep & 16.8 mill 5yrs mill 5yrs 22.4 mill 5yrs (322.4 mill 5yrs 28.0 mill 5yrs 28.0 mill 5yrs Average Pr(spawning biomass < 0.3Bo) (%) 15% Depletion 30% Depletion 50% Depletion Status quo Status quo - effort creep & 16.8 mill 5yrs mill 5yrs 22.4 mill 5yrs E322.4 mill 5yrs D 28.0 mill 5yrs 28.0 mill 5yrs Figure 11. Pictorial comparison of performance measures for each future effort scenario and depletion regime described in the text. 28