Burkhart And Slooten 2003

Burkhart & Slooten–Population of Hector’s dolphin2003, Vol. 37: 553–566 New Zealand Journal of Marine viability and Freshwater Research, 0028–8330/03/3702–0553 $7.00 © The Royal Society of New Zealand 2003

553

Population viability analysis for Hector’s dolphin (Cephalorhynchus hectori ): a stochastic population model for local populations STEPHANIE M. BURKHART* ELISABETH SLOOTEN Environmental Science University of Otago P.O. Box 56, Dunedin New Zealand email: [email protected] *Present

address: Coast Guard Pacific Area, Coast Guard Island, Bldg 51-5, Alameda, CA 945015100, United States. email: [email protected]

Abstract Mortality of Hector’s dolphin (Cephalorhynchus hectori) in gill-net fisheries is a threat to local populations throughout its range. This population viability analysis extends previous work by exploring a wider range of fishing levels and population growth rates, by incorporating year-toyear and environmental variability and by reporting results for smaller population units. Ten of the 16 populations are likely to continue to decline, five are indefinite, and one is likely to increase. All populations subjected to high fishing effort are declining. The only population predicted to increase is partly protected by a marine mammal sanctuary (created in 1988) which reduces the amount of gill-net fishing. Conservation measures are most urgently needed for the highly threatened North Island population, in particular the dolphins at the northern and southern end of this range. Keywords Hector’s dolphin; population viability analysis; PVA; extinction risk; gill-net entanglement; bycatch; environmental impacts of fishing

M02041; Published 5 August 2003 Received 7 June 2002; accepted 30 January 2003

INTRODUCTION Hector’s dolphin (Cephalorhynchus hectori, van Beneden, 1881) is endemic to New Zealand waters and is experiencing mortality from entanglement in gill nets throughout its range. The IUCN (International Union for Conservation of Nature and Natural Resources) lists the species as “endangered” and the North Island population as “critically endangered” (Hilton-Taylor 2000). Quantitative estimates for the number of dolphins caught in gill nets are available only for the Canterbury area (Fig. 1). In this paper we use this estimate of the bycatch rate (per dolphin, per year, per metre of gill net, and per km2 of area) to assess the impact of bycatch on Hector’s dolphins around New Zealand. A recent deterministic population viability model (Martien et al. 1999) was the first to take the entire population of Hector’s dolphins into consideration. Martien et al. (1999) used a density-dependent deterministic model to predict the future abundance and geographic distribution of Hector’s dolphin under different scenarios of fisheries management. They examined the sensitivity of the model’s conclusions to uncertainty in the input parameters, and considered the resulting management advice. Uncertainties about biological data and human effects often delay decisions on management of endangered species. This has been a feature of the management of Hector’s dolphin bycatch. Some decision makers argue that uncertainty about the risk posed to a species should lead to precautionary decisions. Others tend to argue for delaying protective measures until there is overwhelming evidence that a human activity is having a serious effect on the species. We incorporate a range of uncertainties into our risk assessment, with the aim of breaking through this impasse and reducing delays in management action. Here, we extend Martien et al.’s (1999) approach by increasing the range of parameter uncertainty considered, adding year-to-year variability in fishing effort, adding environmental stochasticity in the population growth rate, and by reporting results for smaller population units. This stochastic model generates a range of potential outcomes for each

554

New Zealand Journal of Marine and Freshwater Research, 2003, Vol. 37 Fig. 1 Map of New Zealand, showing the potential outcome for each Hector’s dolphin (Cephalorhynchus hectori) population.

Hector’s dolphin population, expressed as pseudoextinction curves. Population viability analysis (PVA) is commonly used to estimate the probability that a population will decline or become extinct under specified conditions within a certain amount of time (Soulé 1987; Shaffer 1990; Boyce 1992; Burgman et al. 1993). Here, we express extinction risk as pseudo-extinction curves (e.g., Ginzburg et al. 1982; Possingham et al. 1993) that indicate the probability of falling below a range of population sizes within a specified time frame. This analysis highlights factors critical to the species’ persistence, identifies high-risk areas in Hector’s dolphin range, and suggests areas for further research to improve the models and conservation strategies.

METHODS The population units used in modelling were the 16 fisheries management units for which fishing effort

is available (Fig. 1). We used the basic modelling approach outlined in Martien et al. (1999). However, unlike Martien et al. (1999), we report on the results for each of the 16 management units. We did not include movement of individuals from one population unit to another, because studies of Hector’s dolphin movement show very limited home ranges (Bräger et al. 2002) and because we wanted to more explicitly explore the effects of fishing on local populations. This means that if bycatch in one area was unsustainable it would not be compensated for by movement of individuals from adjacent areas. Likewise, this analysis does not consider potential population declines caused by bycatch immediately outside the area under consideration. One of the population units (unit 12) includes the Banks Peninsula Marine Mammal Sanctuary. In this area, commercial gill netting has been banned since 1988 and recreational gill netting is restricted (DOC & MFish 1994). We made the optimistic assumption that the Hector’s dolphins found off this part of the

Burkhart & Slooten–Population viability of Hector’s dolphin coastline have been completely protected from gillnet bycatch since 1988. The model projected the population size of each unit forward 100 years from the 1985 population estimates using a Schaefer (1954) surplus production model: Nt+1 = Nt [1 + (lmax.– 1)(1 – Nt/K)] – Nt ct

(1)

where Nt = population size at time t, lmax. = maximum population growth rate, K = carrying capacity, and ct = the proportion of the Hector’s dolphin population killed by entanglement in gill nets at time t. The proportion of the dolphin population killed each year (ct in Equation 1) was estimated using fishing effort, the entanglement rate and the size of the area: ct = M Et

(2)

where Et = fishing effort at time t (metres of gill net per km2) and M = dolphin entanglement rate, estimated using data on bycatch, fishing effort, and population size from the Canterbury area (see Martien et al. 1999). The New Zealand Ministry of Fisheries (MFish) provided data on commercial fishing effort in the form of Catch Effort and Landing Return (CELR) summaries. Fishing effort was first reported in 1975 (Maria Struzak, MFish pers. comm.). From 1975 to 1982, fishers recorded the time of year and the total weight of fish caught, but were not required to specify the location or the amount of effort. Beginning in 1983, a new system for reporting fishing effort was implemented (Tony Laidler, MFish pers. comm.). MFish divided the waters surrounding New Zealand into statistical fishing units and fishing effort data were summarised by unit. There are a total of 52 statistical fishing units. Hector’s dolphins occur in 16 of these (Fig. 1), and gill-net fisheries are found in all 16. For each unit, we divided the fishing effort data by the area (km2) of the unit to provide an estimate of fishing effort per km2 for each population unit. Each unit extended out to the 100 m contour line because Hector’s dolphins and most, if not all, gill netting occurs within this depth (Colin Sutton, Sea Food Industry Council pers. comm.). The area of each unit was measured with a PLANIX 7 (Tamaya Technics Incorporated 1996) digital planimeter. Since we were only interested in gill nets that were likely to encounter dolphins, we requested assistance from the Sea Food Industry Council in identifying and excluding fishing effort that did not

555

affect dolphins. For example, the majority of gill netting in unit 11 was deep water netting for groper, terakihi, and school sharks (depths >100 m) that occurred outside Hector’s dolphin range (Colin Sutton, Sea Food Industry Council pers. comm.). For our analysis, we therefore considered only 10% of the total fishing effort taking place in unit 11. It was necessary to estimate fishing effort for the years 1970–82 as well as future fishing effort. In mid 1986, the Quota Management System (QMS) was introduced, changing both the magnitude and distribution of fishing effort in New Zealand (Clark et al. 1988; DOC & MFish 1994). The level of fishing effort for the years 1970–82 was estimated by taking the average of the effort reported between 1983 and 1985, before implementation of the QMS. MFish staff felt that fishing effort data from before 1983, for this fishery, were not sufficiently reliable to be used in these analyses. For future fishing effort we used the average level of fishing from 1987 onwards, after the QMS was introduced. We examined the effects of five different levels of fishing effort, to explore our uncertainty about the potential level of bycatch in future years. All five levels were also subjected to year-to-year variability, to represent the variability in fishing effort caused by varying market prices for the target fish stocks and other reasons for changes in the behaviour of fishers between years. The first level was estimated by taking the mean of the reported effort after 1987, when the QMS was in place. The other four levels of future fishing effort were determined by increasing and decreasing the mean (first level of fishing effort) by 10% and 25% successively. Yearto-year variability was incorporated into the fishing effort levels by establishing a normal distribution with a mean corresponding to one of the five fishing effort levels and a standard deviation (SD) equal to the SD of the reported effort since 1987 (post-QMS fishing effort). For each year during each simulation, a value was randomly selected from this distribution to represent the level of fishing effort for that year. The method of back-calculation described in Martien et al. (1999) and Smith & Polachek (1979) was used to estimate population size in 1970 when populations were assumed to be at carrying capacity. Commercial gill netting rapidly expanded in New Zealand waters in the early 1970s with the introduction of nets made of monofilament plastic (Massey & Francis 1989; Dawson 1991). We used the same entanglement rate as Martien et al. (1999). To assess the sensitivity of the model to this parameter, scenarios were run in which the

556

New Zealand Journal of Marine and Freshwater Research, 2003, Vol. 37

Fig. 2 Risk curves for final and minimum population size for the unit 4 population, with the highest maximum population growth rate (1.049), an entanglement rate decreased by 20%, and five different levels of fishing effort (FE): , FE; , +10% FE; - - -, +25% FE; , –10% FE; , –25% FE. Threshold population sizes are given as a proportion of initial population size.

entanglement rate was increased and decreased by 20%. These back-calculations were carried out for each of the different scenarios (e.g., different values for lambda max, different levels of mean fishing effort). We considered four levels of maximum population growth rate for Hector’s dolphin, again to account for uncertainty in the estimation of this parameter: 1.018, 1.023, 1.044, and 1.049 (see Slooten & Lad 1991). This is the population growth rate expected to be achieved in very small populations, with little competition for food, space, and other resources. To represent environmental stochasticity, each of the four levels of maximum population growth rate was expressed as a normal distribution with a mean corresponding to one of the four maximum population growth rates and a SD of 6%, based on Slooten et al. (2000). They estimated the SD between years for adult survival at 6%, and found that population growth closely tracks survival rate (a 1% change in adult survival translates into a 1% change in population growth). For each year during each simulation, a maximum population growth rate was randomly selected from this distribution to give a slightly different maximum population growth rate each year. All calculations were carried out in Microsoft Excel, using SOLVER (Frontline Systems Incorporated 1994) for the backcalculations, and using macros for repetitive calculations. The viability of Hector’s dolphins is expressed as the risk of the minimum and final populations falling below certain thresholds. The median time to extinction was calculated for those populations that became extinct during the 100-year projection period.

The deterministic analogue of the stochastic model was run for each of the 60 scenarios for comparative purposes. The deterministic version had the same parameter levels as the stochastic model. The only difference was that year-to-year variability was removed from the maximum population growth rate and fishing effort. Instead of selecting a value from a distribution for each year, a mean maximum population growth rate (1.018, 1.023, 1.044, or 1.049) and a mean fishing effort (FE) level (or +10%, –10%, +25%, or –25% FE) were used in the respective scenarios as constant values for the parameters for the 100-year projection.

RESULTS The stochastic model predictions were used to construct two types of quasiextinction risk graphs (e.g., Fig. 2). For each population, we graphed the proportion of simulation runs for which the final population after 100 years fell below certain threshold population sizes. A second set of graphs shows the risk to the population of dropping down to minimum threshold population sizes any time during the same 100-year period (1984– 2084). Each of these figures shows the results for 1000 runs of the model. Ten of the 16 populations were likely to decline, five were indefinite, and one was likely to increase. All populations subjected to high fishing effort were declining (see Fig. 1, Table 1). Most of the populations that had a high probability of decreasing were subjected to either moderate or high fishing effort. Unit 12 was the only population likely to increase. Although it has a moderate level of fishing effort,

Burkhart & Slooten–Population viability of Hector’s dolphin most of the dolphins in this population are found within the boundaries of the Banks Peninsula Marine Mammal Sanctuary and we have assumed that these have not been exposed to gill nets since 1988 when the sanctuary was created. Serious risk of extinction The simulation results indicate the unit 4 population is in serious danger of extinction. In most scenarios, there was a high probability of unit 4 falling below 10% of the initial population size within 100 years. Even under the most optimistic scenario (maximum population growth 1.049, entanglement rate –20%, fishing effort –25%) there was still a 50% chance that the final population size was below 10% of initial size and the minimum population level was 7% of initial population size (Fig. 2). Under the most reasonable scenarios (maximum population growth 1.023 or 1.044, mean entanglement rate and mean fishing effort) the population had a high extinction risk (see Table 2). We did not model decreases in fishing effort by more than 25%, but this would obviously improve the prognosis for all populations and unit 4 in particular. High risk of population decline Population units 1, 3, 5, 7, 8, 10, 11, 13, and 15 (e.g., Fig. 3) show declining populations and a high

557

probability that the final size would fall below 50% of the starting population size in most scenarios. Units 3 and 11 show the highest risk of population decline (Fig. 3). Even under the most optimistic scenario (maximum population growth 1.049, entanglement rate –20%, and fishing effort –25%), there was still a 50% chance that the final populations in both units would fall below 63% and 75% of the starting size for units 3 and 11, respectively. In general, the risk curves for units 1 and 15 were relatively steep for scenarios run at the higher maximum population growth rates, suggesting a sharp increase in the probability of population decline over a short range of threshold population sizes (Fig. 3). Moderate risk of population decline at average-to-high fishing effort levels Population units 2 and 9 showed a high risk of declining in scenarios with the lower maximum population growth rates. However, most scenarios with the upper maximum population growth rates (1.044 and 1.049) and reduced levels of fishing effort (–10% and –25% FE) had a better than 50% chance of increasing (Fig. 4). Populations in both units were especially sensitive to changes in the fishing effort level. In scenarios with higher maximum population growth rates, the effects of the varying fishing effort were more pronounced and increased levels of

Table 1 Summary of characteristics for each population unit. Population unit 1 2 3 4 5 6 7 8 9 10 11 12‡ 13 14 15 16 *Estimated

Population size in 1970, or K* 6 475 36 60 15 995 614 810 216 236 1149 2895 225 62 3 604

Population size in 1985† 6 122 12 6 14 423 409 476 65 113 237 1024 76 56 3 341

Area (km2) 2521.8 3947.5 6244.2 4169.8 1168.8 2322.4 3416.6 2085.3 4502.8 2684.1 1373.6 20055.6 3897.2 3820.4 4895.6 3662.1

Mean fishing effort (m gill net/km2) 84 133 205 382 109 69 179 85 106 174 227 98 174 12 52 48

by back-calculation, as explained under Methods, value given here is for lambda max. 1.023 and mean fishing effort presented in this table. †From population survey carried out during 1984–85 (see Dawson & Slooten 1988). ‡Unit 12 contains a marine mammal sanctuary.

558

New Zealand Journal of Marine and Freshwater Research, 2003, Vol. 37

Fig. 3 Risk curves for final and minimum population size for units 3, 11, and 15, with mean fishing effort, an entanglement rate of M, and four different maximum population growth rates: , 1.049; , 1.044; , 1.023; , 1.018. Threshold population sizes are given as a proportion of initial population size.

fishing effort resulted in lower final and minimum population sizes. Likely to increase with high maximum population growth rates and decrease with low maximum population growth rates Population units 6, 14, and 16 generally had a better than 50% chance of increasing with the higher maximum population growth rates (1.044 and 1.049). For scenarios with the lower maximum population growth rates (1.018 and 1.023) these populations consistently decreased (Fig. 5).

Likely to increase In most scenarios, unit 12 had a better than 50% chance of increasing (e.g., Fig. 6). For the worst-case scenario (maximum population growth 1.018, M + 20% and FE + 25%), there was a 50% risk that the final population would drop below 79% of initial population size. For scenarios with the upper maximum population growth rates (1.044 and 1.049), the minimum population size dropped to 90% of the starting population size in most simulations (Fig. 6). For scenarios with the lower maximum population growth rates (1.018 and 1.023), most populations dropped to a minimum size of 75%

Burkhart & Slooten–Population viability of Hector’s dolphin

559

Fig 4 Risk curves for the final population size of units 2 and 9, for two maximum population growth rates (1.049 and 1.018), with an entanglement rate of M, and five fishing effort levels: , FE; , +10% FE; - - -, +25% FE; , –10% FE; , –25% FE. Threshold population sizes are given as a proportion of initial population size.

Table 2 Median time to extinction (in years) for the unit 4 population under various scenarios of risk (M = entanglement rate which was increased and decreased by 20%; lmax. = maximum population growth rate; and FE = fishing effort which was increased and decreased by 10% and 25%).

Parameters lmax. = 1.049, FE lmax. = 1.049, +10% FE lmax. = 1.049, –10% FE lmax. = 1.049, +25% FE lmax. = 1.049, –25% FE lmax. = 1.044, FE lmax. = 1.044, +10% FE lmax. = 1.044, –10% FE lmax. = 1.044, +25% FE lmax. = 1.044, –25% FE lmax. = 1.023, FE lmax. = 1.023, +10% FE lmax. = 1.023, –10% FE lmax. = 1.023, +25% FE lmax. = 1.023, –25% FE lmax. = 1.018, FE lmax. = 1.018, +10% FE lmax. = 1.018, –10% FE lmax. = 1.018, +25% FE lmax. = 1.018, –25% FE

M

Entanglement rates +20% M

–20% M

28±0.27 26±0.24 31±0.34 24±0.17 36±0.52 26±0.26 24±0.21 28±0.30 23±0.16 32±0.44 20±0.15 19±0.13 21±0.18 18±0.11 22±0.25 19±0.14 18±0.12 19±0.17 18±0.10 21±0.20

20±0.18 20±0.16 22±0.21 19±0.12 24±0.31 20±0.17 19±0.14 20±0.19 18±0.11 22±0.26 16±0.11 15±0.09 16±0.12 15±0.08 16±0.14 15±0.10 15±0.09 15±0.11 15±0.07 15±0.13

42±0.46 38±0.36 50±0.60 33±0.28 67±0.76 38±0.41 34±0.33 43±0.49 30±0.25 57±0.69 27±0.23 25±0.19 28±0.25 24±0.15 32±0.35 25±0.20 24±0.16 26±0.23 22±0.14 29±0.30

560

New Zealand Journal of Marine and Freshwater Research, 2003, Vol. 37

Fig. 5 Risk curves for final and minimum population size for units 6, 14, and 16, with mean fishing effort, an entanglement rate of M, and four different maximum population growth rates: , 1.049; , 1.044; , 1.023; , 1.018. Threshold population sizes are given as a proportion of initial population size.

of the initial population size (Fig. 6). Quasiextinction risk curves for minimum population size were relatively insensitive to changes in fishing effort, especially at upper maximum population growth rates (Fig. 7). Sensitivity to input parameters Quasiextinction risk clearly depended on the maximum population growth rate used. In general, as maximum population growth rate increased, the

quasiextinction risk curves shifted to the right and the risk of population decline decreased (e.g., Fig. 8). The quasiextinction risk curves for threshold final and minimum population size were also sensitive to the entanglement rate, M. When M was increased by 20%, the risk curves moved to the left, indicating that the chances of quasiextinction were substantially higher for populations with an entanglement rate increased by 20% than for populations with an entanglement rate of M or –20% M (Fig. 9). In

Burkhart & Slooten–Population viability of Hector’s dolphin

561

Fig. 6 Risk curves for final and minimum population size of unit 12 with an entanglement rate of M, a mean fishing effort, and four different maximum population growth rates: , 1.049; , 1.044; , 1.023; , 1.018. Threshold population sizes are given as a proportion of initial population size.

Fig. 7 Risk curves for minimum population size of unit 12 under two maximum population growth rates (1.049 and 1.018), an entanglement rate of M, and five different fishing effort levels: , FE; , +10% FE; - - -, +25% FE; , –10% FE; , –25% FE. Threshold population sizes are given as a proportion of initial population size.

general, when M was decreased by 20%, the quasiextinction curves moved to the right, predicting a lower risk of population decline (Fig. 9). Likewise, increasing the level of fishing effort tended to shift the quasiextinction risk curves to the left, predicting a higher risk of population decline. Decreasing the level of fishing effort tended to shift the quasiextinction risk curves to the right, predicting less risk to the population (e.g., Fig. 10). Fishing effort affected the risk of population decline in all units. However, some units were much more sensitive to changes in fishing effort than others. For example, in very small populations with relatively high levels of fishing effort, reducing fishing effort by 10–25% did not substantially reduce the risk of decline. For these populations, fishing effort would need to be reduced by much more than 25% to show a significant

reduction in the risk of population decline. Likewise, in large populations with a relatively low level of fishing effort, the risk of decline was relatively insensitive to changes in fishing effort in the range of 10–25%. In general the effect of changing the level of fishing effort (up or down by 10–25%), and of stochastic changes in fishing effort from year to year, was lower in populations with relatively low and/or constant fishing effort. For example, population unit 14 had the lowest and least variable fishing effort. Therefore, the risk curves in Fig. 5 are relatively steep and affected most strongly by the maximum population growth rate. Each of the individual curves in the risk curves plots is for a different scenario, and therefore a different level of K or estimated population size in 1970. For example, in Fig. 2 a comparison of the

562

New Zealand Journal of Marine and Freshwater Research, 2003, Vol. 37

Fig. 8 Quasiextinction risk curves for final and minimum population size for unit 6 with an entanglement rate of M, a mean fishing effort, and four maximum population growth rates: , 1.049; , 1.044; , 1.023; , 1.018. Threshold population sizes are given as a proportion of initial population size.

Fig. 9 Risk curves for final and minimum population size for the unit 2 population with a maximum population growth rate of 1.049, a mean fishing effort level, and three entanglement rates: , +20% M; , M; , –20% M. Threshold population sizes are given as a proportion of initial population size.

Fig. 10 Risk curves for final and minimum population size for unit 9, with a maximum population growth rate of , FE; , +10% FE; 1.049, an entanglement rate of M, and five levels of fishing effort: - - -, +25% FE; , –10% FE; , –25% FE. Threshold population sizes are given as a proportion of initial population size.

Burkhart & Slooten–Population viability of Hector’s dolphin individual risk curves indicates the uncertainty in the risk assessment as it relates to uncertainty in fishing effort. Fig. 3 shows how final and minimum population sizes depend on the estimate of lambda max, and Fig. 9 indicates how uncertainty in the entanglement rate affects the risk assessment. Comparison with deterministic model The results of the stochastic scenarios and the deterministic scenarios were similar in terms of the relative status of the population units. Population units 1, 3, 4, 5, 7, 8, 10, 11, 13, and 15 were highly likely to decline. Population units 2 and 9 could increase if maximum population growth rate was 1.044 or 1.049 and fishing effort was reduced (–10% and –25% FE). Population units 6, 14, and 16 were likely to increase with a maximum population growth rate of either 1.044 or 1.049 and likely to decrease with a maximum population growth rate of either 1.018 or 1.023. Population unit 12 was likely to increase. However, the stochastic scenarios always predicted a higher risk of population decline. For example, unit 2 had a moderate risk of population decline. The deterministic scenario for unit 2 with a maximum population growth rate of 1.049, entanglement rate of –20% M and mean fishing effort indicated that there was a 100% probability that population size in 100 years would be larger than the initial size. Under those same conditions, the stochastic scenario predicted a 50% chance of increase and a 50% chance of decrease.

DISCUSSION Since a primary goal in conservation is to minimise the risk of population decline (Burgman et al. 1993; Hedrick et al. 1996), our results were expressed as quasiextinction curves. Risk curves are a better indicator of the likely future of a population than are mean population trajectories because they reflect the variation behind the means (Burgman et al. 1993). For each population, we graphed the risk that the final population after 100 years would fall below certain threshold population sizes and the risk of the population dropping down to certain minimum threshold population sizes any time during the 100year period. The results of the stochastic models indicate that 10 of the 16 Hector’s dolphin population units are declining and will continue to do so unless fishing effort is reduced. One of these units could become extinct within a few decades. The status of five of the 16 units is uncertain.

563

Only one population (unit 12), the only one that includes a sanctuary area with restrictions on fishing effort to protect the dolphins, is likely to increase. Martien et al. (1999) predicted that the east coast South Island population (which they defined as population units 9–16), would increase if maximum population growth rates were relatively high (>1.044) and fishing effort did not increase in future. Our analysis indicates that although the prognosis for unit 12 is reasonably good, several of the other population units off the South Island east coast are likely to decline. Therefore, there would likely be some reduction and fragmentation of the geographical range occupied by Hector’s dolphins in this area. Importantly, both analyses assumed that the Banks Peninsula Marine Mammal Sanctuary is 100% effective, and eliminates bycatch mortality for the dolphins on that stretch of coastline (part of unit 12). Therefore, our prognosis for this population may be too optimistic (e.g., see Slooten et al. 2000). Unit 4 on the west coast of the North Island is most in need of conservation intervention to reduce or eliminate gill-net entanglement. This area has a very small dolphin population and very high fishing effort. Even under the most optimistic conditions (maximum population growth rate of 1.049, fishing effort level reduced by 25%, and entanglement rate decreased by 20%), unit 4 has a median time to extinction of 67 years. Under more reasonable scenarios (maximum population growth rate of 1.023–1.044, entanglement rate of M and a mean fishing effort level) the population could become extinct within 20–26 years. Fishing effort would need to be reduced substantially before population growth could occur. Unless the level of fishing effort is reduced, this population unit will face extinction in the very near future. All populations were sensitive to changes in the maximum population growth rate, entanglement rate and the level of fishing effort. The largest maximum population growth rate modelled (1.049) was estimated using human survivorship curves (Slooten & Lad 1991) and is likely to be an absolute maximum population growth rate for Hector’s dolphin. For example, the United States National Marine Fisheries Service uses a default value of 1.04 for maximum population growth rate in their population management models for whales and dolphins (e.g., Wade 1998). This leads to some important research recommendations. More data on reproduction and survival will help narrow the range of likely maximum population growth rates, and will make it possible

564

New Zealand Journal of Marine and Freshwater Research, 2003, Vol. 37

to draw more specific conclusions regarding extinction risk. A nationwide observer programme, placing independent observers on fishing boats, would provide more reliable data on entanglement rate. At this time, the only possible way to estimate extinction risk for Hector’s dolphin populations around the country is to use the entanglement rate data from the Canterbury area and apply them to fishing effort in other parts of the country. (We used entanglement rates from before 1988 when the Banks Peninsula Marine Mammal Sanctuary was created.) It is important to emphasise that these analyses do not include all forms of fishing known to entangle Hector’s dolphins. We were able to include only commercial gill netting. No information is available on fishing effort by amateur gill netters. New Zealand is unusual in allowing recreational fishers to use gill nets. However, these fishers are not licenced and no quantitative data are available to allow us to include recreational fishing effort into the model. Likewise, trawling is not included in this analysis. From time to time, dolphins are entangled and killed in trawl gear (e.g., Baird & Bradford 2000). However, no quantitative data are available to estimate the catch rate of dolphins in trawling. Therefore, we have not included trawling in our analysis. The strong relationship between fishing effort and extinction risk leads to some obvious research and management recommendations. Reductions in gillnet fishing effort, or a shift to fishing methods that do not catch dolphins, would clearly benefit Hector’s dolphin populations nationwide. It is also important to keep working on improving the process by which fishing effort is recorded, in particular the spatial scale. Commercial gill-net effort is recorded only to the level of the fairly large areas indicated in Fig. 1. More detailed effort data for gill-net fisheries, with information about the precise locations where gill nets are used, would be very helpful in determining the effect of this fishing method on Hector’s dolphins. This would also allow more sophisticated spatial modelling of the population and the impacts on it, including the overlap between fishing effort and dolphin distribution, and would allow modelling of source-sink dynamics. It would be important, however, to keep sight of the areas in which fishing effort is unsustainable (from the point of view of dolphin bycatch). Population declines could occur in adjacent areas as well as the area in which the fishing effort occurs. Therefore, it will be important to distinguish declines caused by local fishing effort

from declines caused by dolphins moving into an adjacent area with high fishing effort. Our results are broadly similar to those of Martien et al. (1999). They grouped the 16 population units into three populations: North Island (units 1–4); South Island west coast (units 5–8); and South Island east coast (units 9–16). They found that two of these populations (North Island and South Island west coast) were predicted to decline, even when the most optimistic parameter estimates were used. The status of the third population (South Island east coast) depended on maximum population growth rate and assumptions about the effectiveness of the Banks Peninsula Marine Mammal Sanctuary. The advantage of splitting the population into smaller units is that it allows a more fine-scale assessment of the effects of fisheries mortality in individual populations. Both genetic data and information on dolphin movements indicate that Hector’s dolphins have a very limited range of movement. The average home range is 31 km of coastline (Bräger et al. 2002) and there are strong genetic differences among local populations (Pichler et al. 1998). In addition, fisheries are strongly localised. Source-sink dynamics may act to “rescue” a population from bycatch by adding individuals from adjoining populations. However, it would be inappropriate to conclude that a local population can therefore support a higher level of bycatch, if this “action at a distance” would act to deplete the adjoining populations over time and eventually the local population itself. Conversely, a population may be declining due to bycatch in an adjoining area. Again, it is important to consider fishing pressure in the area in which it occurs, even if it acts at a distance on adjoining populations. Our analysis provides clear management advice by showing exactly which areas have unsustainable levels of fishing, regardless of whether that fishing activity is acting on adjoining populations as well as the local population (and regardless of whether the local population is “rescued” by immigration from adjoining populations). We modelled a broader range of parameter inputs than Martien et al. (1999), and included year-to-year variability in input parameters. As expected (e.g., Beudels et al. 1992; Milner-Gulland 1994; McCarthy 1996; Marmontel et al. 1997; Manly 1998) including stochasticity in the model led to slightly higher estimates of risk than those made by Martien et al. (1999). We also made more specific predictions about risk to individual populations. The North Island west coast population was recently listed as Critically Endangered by the IUCN (Hilton-Taylor

Burkhart & Slooten–Population viability of Hector’s dolphin 2000) and is clearly at serious risk of extinction. Population unit 4 was identified as having the greatest risk of extinction, and two of the other North Island populations also appear to be declining sharply. Other Hector’s dolphin populations at a relatively high risk of population decline are those on the west coast of the South Island (units 7 and 8), Marlborough and North Canterbury (units 10 and 11), Otago (unit 13), and one of the Southland populations (15). In a few cases (units 1 and 15) very low population size meant that even relatively low levels of fishing effort caused a high risk of extinction. However, in most cases the risk of population decline was high in areas with high fishing effort. Martien et al. (1999) suggested that conservation measures were urgently needed for the highly threatened North Island population. Our model finetunes this recommendation and identifies the populations at the northern and southern end of this range (units 1, 3 and 4) as at highest risk. Some of the management options for this population, recently under discussion, were focused on protecting only population unit 2. This could have led to a reduction in the geographic range of the North Island Hector’s dolphin and to a further reduction in population size. Such very small populations have a high extinction risk simply due to stochastic factors (Shaffer 1981; Gilpin & Soulé 1986; Belovsky et al. 1994). A protected area for North Island Hector’s dolphin created by the Minister of Fisheries in 2003 protects all of populations 1 and 2 and most of population 3.

ACKNOWLEDGMENTS We are very grateful to the following people for advice, information, discussions, and feedback on this work: Tracy Blair, Corey Bradshaw, Steve Dawson, Dagmar Fertl, David Fletcher, Tony Laidler, Karen Martien, Ralph Riley, Shaun Stephenson, Maria Struzak, Colin Sutton, Barbara Taylor, and Ross Thompson. Stephanie Burkhart was supported by a Fulbright Fellowship.

REFERENCES Baird, S. J.; Bradford, E. 2000: Estimation of Hector’s dolphin bycatch from inshore fisheries, 1997–98 fishing year. Published Client Report on Contract 3024, Funded by Conservation Services Levy. Wellington, Department of Conservation. 20 p. (www.csl.doc.govt.nz/CSL3024.pdf)

565

Belovsky, G. E.; Bissonette, J. A.; Dueser, R. D.; Edwards, T. C.; Luecke, C. M.; Ritchie, M. E.; Slade, J. B.; Wagner, F. H. 1994: Management of small populations: concepts affecting the recovery of endangered species. Wildlife Society Bulletin 22: 307–316. Beudels, R. C.; Durant, S. M.; Harwood, J. 1992: Assessing the risks of extinction for local populations of roan antelope Hippotragus equinus. Biological Conservation 61: 107–116. Boyce, M. S. 1992: Population viability analysis. Annual Review of Ecology and Systematics 23: 481–506. Bräger, S.; Dawson, S. M.; Slooten, E.; Smith, S.; Stone, G. S; Yoshinaga, A. 2002: Site fidelity and alongshore range in Hector’s dolphin, an endangered marine dolphin from New Zealand. Biological Conservation 108: 281–287. Burgman, M. A.; Ferson, S.; Akcakaya, H. R. 1993: Risk assessment in conservation biology. London, Chapman and Hall. 314 p. Clark, I. N.; Major, P. J.; Mollett, N. 1988: Development and implementations of New Zealand’s ITQ management system. Marine Resource Economics 5: 325–349. Data Description Incorporated 1992: Data desk handbook, Version 4.0. Volume 2: the new power of statistical vision. New York, Ithaca. Dawson, S. M.; Slooten, E. 1988: Hector’s dolphin Cephalorhynchus hectori: distribution and abundance. Reports of the International Whaling Commission, Special Issue 9: 315–324. Dawson, S. M. 1991: Incidental catch of Hector’s dolphin in inshore gill nets. Marine Mammal Science 7(3): 283–295. DOC & MFish 1994: Review of the Banks Peninsula Marine Mammal Sanctuary: a paper for public comment. Canterbury Conservancy Miscellaneous Report Series No. 3. Christchurch, Department of Conservation. Frontline Systems Incorporated 1994: Solver. In: User’s guide: Microsoft Excel Version 5.0. United States, Microsoft Corporation. Pp. 562–563. Gilpin, M. E.; Soulé, M. E. 1986: Minimum viable populations: the processes of species extinctions. In: Soulé, M. E. ed. Conservation biology: the science of scarcity and diversity. Sunderland, Massachusetts, Sinauer Associates. Pp. 19–34. Ginzburg, L. R.; Slobodkin, L. B.; Johnson, K.; Bindman, A. G. 1982: Quasiextinction risk probabilities as a measure of impact on population growth. Risk Analysis 21: 171–181. Hilton-Taylor, C. (comp.) 2000: 2000 IUCN red list of threatened species. Gland, Switzerland and Cambridge, United Kingdom. IUCN, xviii + 61p.

566

New Zealand Journal of Marine and Freshwater Research, 2003, Vol. 37

McCarthy, M. A.; Burgman, M. A.; Ferson, S. 1996: Logistic sensitivity and bounds for extinction risks. Ecological Modelling 86: 297–303. Manly, B. F. J. 1998: Review of the impact of incidental catch on Hooker’s sea lions Phocartos hookeri. In: Hilborn, R.; Pascual, M.; Gerber, L.; Gulland, F. 1997. Report commissioned by the New Zealand Squid Company Limited, Wellington, New Zealand. Marmontel, M.; Humphrey, S. R; O’Shea, T. J. 1997: Population viability analysis of the Florida manatee (Trichechus manatus latirostris), 1976–1991. Conservation Biology 11(2): 467–481. Martien, K. K.; Taylor, B. L.; Slooten, E; Dawson, S. 1999: A sensitivity analysis to guide research and management for Hector’s dolphin. Biological Conservation 90: 183–191. Milner-Gulland, E. J. 1994: A population model for the management of the saiga antelope. Journal of Applied Ecology 31: 25–39. Pichler, F.; Baker, C. S.; Dawson, S. M.; Slooten, E. 1998: Geographic isolation of Hector’s dolphin populations described by mitochondrial DNA sequences. Conservation Biology 12: 676–682. Possingham, H. P.; Lindenmayer, D. B.; Norton, T. W. 1993: A framework for the improved management of threatened species based on population viability analysis (PVA). Pacific Conservation Biology 1: 39–45.

Schaefer, M. B. 1954: Some aspects of the dynamics of populations important to the management of commercial fisheries. Bulletin of the Inter-American Tropical Tuna Commission 1: 27–56. Shaffer, M. L. 1981: Minimum population sizes for species conservation. BioScience 31: 131–134. Shaffer, M. L. 1990: Population viability analysis. Conservation Biology 4: 39–40. Slooten, E.; Lad, F. 1991: Population biology and conservation of Hector’s dolphin. Canadian Journal of Zoology 69: 1701–1707. Slooten, E.; Fletcher, D.; Taylor, B. L. 2000: Accounting for uncertainty in risk assessment: case study of Hector’s dolphin mortality due to gillnet entanglement. Conservation Biology 14: 1264–1270. Smith, T. D.; Polachek, T. 1979: Analysis of a simple model for estimating historical population sizes. Fishery Bulletin 76: 771–779. Soulé, M. 1987: Introduction. In: Soulé, M. ed. Viable populations for conservation. Cambridge, Cambridge University Press. Pp. 1–10. Tamaya Technics Incorporated 1996: PLANIX 7 Digital planimeter instruction manual. Tokyo, Japan. 36 p. Wade, P. R. 1998: Calculating thresholds to the humancaused mortality of cetaceans and pinnipeds. Marine Mammal Science 14(1): 1–37.