The Effects of Longlining on SeabirdPopulations: by-catch estimation, archival

tagging and population modeling

Dennis Heinemann

Neil Klaer

Tom Polacheck

Geoffrey Tuck

Tim Davis

Date 1 January 2002

FINAL REPORT AFMA 97/98

2

1 INTRODUCTION

1.1 Background

The incidental catch of seabirds in the Southern Ocean’s pelagic and demersal longliningfisheries has been identified as an issue of considerable importance to the sustainability ofseabird populations. Several albatross species, including the wandering albatross, Diomedeaexulans, have shown marked declines in abundance throughout their range (Weimerskirch andJouventin 1987, Croxall et al. 1990, Gales 1998). Wandering albatross and other species arefrequently taken as bycatch from longline fishing fleets, and mortality associated withlonglining has been implicated as the primary cause of the declines in abundance (Gales, 1993).With new and expanding fisheries form Patagonian toothfish, Dissostichus eleginoides, andincreasing pressure on southern tuna stocks, there is now a critical need to enhance ourunderstanding of the impact of seabird bycatch on bird populations through this internationallycollaborative project. This project builds upon the collaborative work begun under theAustralian Fisheries Management Authority (AFMA) contracts during 1995/1996, and1996/1997 and upon the international collaboration which developed between the CSIRO, theBritish Antarctic Survey (BAS) and the Centre National de la Recherche Scientifique (CentreD’etudes Biologiques de Chize, CEBC/CNRS). These latter two institutions are recognised asworld leaders in the study of albatross biology.

Since 1988, data on seabird bycatch have been collected by the Tasmanian Parks and WildlifeService and the AFMA observer program. These data, combined with fishery observer data andAFMA logbook data provide essential information on catch rates of seabirds. Estimates ofcatch rates have been determined in a collaborative project between CSIRO and the TasmanianParks and Wildlife Service (Klaer and Polacheck 1995) and have been presented to the first andsecond meetings of the CCSBT Ecologically Related Species Working Group (ERSWG; Klaer& Polacheck 1995, 1997a, 1997b). These catch and catch rate estimates for the AustralianFishing Zone had required updating on a yearly basis as new data have become available. Nowthat the Japanese are not fishing in the Australian Fishing Zone, updates are no longer needed.In the place of the AFMA observer program, a Pilot Observer Program to collect similar datafrom Australia’s domestic longline fisheries has been developed by CSIRO, AFMA and EA(Environment Australia). Catch rates currently provide the most feasible measure to analyse theprogress of mitigation and to base management objective upon. Long-term monitoring ofseabird catch and catch rates is of extreme importance in order to assess the impacts ofincidental takes and determine the effect of increased recent use of mitigation measures (e.g.night setting) and changes in gear on seabird catch rates. This was recognised at the secondCCSBT ERSWG meeting (Canberra 3-6 June, 1997).

Preliminary population modelling assessment results completed under previous AFMAcontracts indicate that seabird mortality associated with longlining can provide a reasonableexplanation for the historic decline in breeding pairs observed from the South Georgia andCrozet Island wandering albatross populations. The modelling framework produced can assessthe effects of the longline fishery on seabird populations, and this preliminary work was

3

presented to the CCSBT Ecologically Relation Species Working Group in Wellington, NewZealand in December 1995 (Tuck and Polacheck 1995). A necessary element in thedevelopment of the modelling framework was an understanding of the spatial and temporaldistribution of longline effort across the Southern Ocean. Effort employed by the varioussouthern hemisphere fisheries will have significant implications for those populations ofseabirds caught by tuna longline fleets. A paper describing trends in tuna longlining fisheries inthe Southern Ocean and implications for seabird by-catch was presented to the first meeting ofthe CCSBT-ERSWG (Wellington 1995), and an update was presented at the following meetingat Canberra in 1997 (Polacheck and Tuck 1997).

The model framework that has been developed considers a range of possible hypotheses for thespatial distribution of wandering albatrosses. As expected, preliminary results are sensitive toassumptions about the spatial distribution of the birds and so there is a need to reduce thisuncertainty and refine our understanding of wandering albatross foraging distributions. Anunderstanding of the spatial distribution is critical to the modelling process and will assist withdetermining the spatial and temporal interaction of the seabirds with fishery; this is vital forassessing impacts and designing mitigation measures. The at-sea distribution of wanderingalbatross remains largely unresolved, with limited information coming from banding, bycatchdata and satellite tracking. Short term information on the distribution of birds when breedingcan be provided by satellite tracking, however information of the distribution of non-breedingbirds is critical and archival tags currently provide the best potential technology to produce thisinformation.

Pilot work was initiated in 1995 jointly between CSIRO, BAS and CEBC/CNRS to evaluate thepotential for archival tags to provide information on the spatial distribution of seabirds.Analysis of the data from the tags deployed in 1996 confirmed that the archival tags have theability to give not only invaluable positional information, but also behavioural information(Tuck et al. 1998).

The abundance of several species of Southern Ocean seabirds has declined substantially overthe last three decades. The incidental catch of seabirds during oceanic longline operations hasbeen identified as a potential source of the mortality leading to the observed declines and assuch has been declared a key threatening process under Australia’s Endangered Species Act.With new and expanding fisheries for Patagonian toothfish and increasing and shifting fishingpressure on southern tuna stocks, monitoring and assessment research is now critical to minmiserisk to seabird populations. In order to evaluate the potential effect on seabird populations andmake informed management decisions to reduce incidental catches, the estimation of catch ratesand catches of seabirds in relation to season, area, fishing practices and mitigation measures iscritical. Analyses of the effects of mitigation measures on catch rates of seabirds and targetspecies is of extreme importance. A knowledge of the effect of night setting on catch rates ofSBT is vital if night setting is to be used effectively as a mitigation measure, as called for in theThreat Abatement Plan. Currently little is known about the effects of night setting on SBTcatch rates.

1.1 Objectives

1. Update estimates of the seabird catch and catch rates and their associated varianceswithin the Australian Fishing Zone by area and season.

4

2. Analyses of the effectiveness of various mitigation measures used to reduce seabirdbycatch. Explore the consequences of night setting on southern bluefin tuna catchrates, and evaluate the potential for wider adoption of night setting as a mitigationmeasure.

3. Continue work on the deployment and analysis of data from archival tags.

4. Update the wandering albatross population model in collaboration with BAS andCEBC/CNRS, and explore additional hypotheses regarding the impact of longliningon wandering albatross populations.

5. Extend the albatross model to include other species of Southern Ocean albatrosses.

2 RESULTS

2.1 Bycatch Rates

CSIRO Marine Research continued its efforts at providing annual updates to estimates ofbycatch rates of seabirds in the Japanese tuna fishery within the Australian Fishing Zone, upuntil the time they were excluded from the AFZ. Estimates for 1996/1997 were completed andpresented to second meeting of the CCSBT-ERSWG in 1997 (Klaer & Polacheck 1997). Thoseresults are reproduced in Appendix 1. Estimates for the 1997/1998 seasons were completed andpresented to the third meeting of the CCSBT-ERSWG, 9-12 June 1998, Tokyo, Japan (Klaer &Heinemann 1998). Those results are reproduced in Appendix 1.

In addition to these annual updates to Japanese bycatch data, other work on seabird bycatchestimation was completed. Also, presented to the 1998 CCSBT-ERSWG meeting in Toyko,were the results of observations carried out onboard two domestic vessels operating off the eastcoast of Australia (Young et al. 1998). Those results are reproduced in Appendix 2. In 1999CSIRO Marine Research completed a contract with AFMA and EA to develop a pilot observerseabird bycatch observer program for Australia’s domestic longline fisheries (Heinemann et al.1999). As part of that work and for the purposes of this contract, all of the available data onseabird bycatch in domestic longline fisheries was collated and analyzed. The relevant text andtables from Heinemann et al. 1999 were extracted and are presented in Appendix 3.

2.2 Mitigation & Night Setting

Results of analyses on the effects of environmental factors and mitigation measures on the by-catch rates in the Japanese fishery were presented to the third meeting of the CCSBT-ERSWG,9-12 June 1998, Tokyo, Japan and later published in the journal Emu (Klaer & Polacheck 1998)and is reproduced in Appendix 4. This paper and work by other researchers (Brothers et al.1999) have identified night setting as potentially the most effective mitigation measureavailable. However, resistance to the use of night setting has occurred in all tuna fisheries.Concerns range from the potential loss of catch, to inefficient and costly operating procedures,to crew safety issues. Data to assess the potential qualitative and quantitative effect of night

5

setting on the catches of tuna in the Japanese fishery were limited because little data met therequirements for inclusion in analyses. Therefore, plans were developed for the analysis ofadditional datasets, the use of hook monitors and other methods of determining the time atwhich tuna are caught, inferring catch patterns from information on the natural feeding patterns,and the design of a sampling experiment to collect new data on this specific question. Theseplans were presented to the third CCSBT-ERSWG meeting in Japan (Heinemann 1998), and arereproduced in Appendix 5. The plans were received favorably at the meeting. An ERSWGworking committee modified the plans and put them forward in a technical document withrecommendations to CCSBT and its Scientific Committee; the modified plan is reproduced inAppendix 6. Subsequently, some of these ideas were incorporated into research proposals toAFMA’s Tuna Ecologically Related Species Working Group (Appendix 7). One proposal wasfunded, but the project could not be undertaken because the Principal Investigator left CSIRO.

2.3 Archival Tags

Technical and logistic problems encountered during the period of this contract prevented thedeployment of additional archival tags on wandering albatrosses. The evaluation of positionalresolution of the existing archival tag data and analyses of the spatial distribution generatedfrom these data were not completed.

2.4 Update Wandering Albatross Model

The modelling framework for the analysis of the population dynamics of wandering albatrossesand their interactions with longline fisheries, first described in Tuck and Polacheck 1995, hasundergone continued development. That development has included significant improvements todata inputs, model structure and the fitting algorithm used in the model. The latest version ofthe model and the modelling results for wandering albatross populations on Bird Island and theCrozet Islands have been submitted for publication (Tuck et al. In Press). That manuscript isreproduced in Appendix 8.

2.5 Extend Albatross Model

Work started on the application of the albatross model to other species. This process involvedobtaining data for those species, and modifying the model to handle differences in thedemographic system. Specifically, the model was being modified to handle annual breeders, i.e.species that breed every year, and to address more complicated spatio-temporal albatross-distribution hypotheses. This work was not completed.

On Bird Island, sufficient data are available to model the population dynamics of the annuallybreeding black-browed (Diomedea melanophrys) and biennially breeding grey-headed albatross(D. chrysostoma), both species that have shown population declines possibly related to longlinefishing mortality (Croxall et al. 1998, Croxall pers. com.). On the Crozet Islands, sufficientdata appear to be available for 3 species: biennially breeding sooty (Phoebetria fusca),biennially breeding light-mantled sooty (P. palpebrata ) and annually breeding black-browedalbatross (Weimerskirch and Jouventin 1998, Weimerskirch pers. comm.). All of these

6

populations have also shown population declines in recent decades, and longline fishing isbelieved to have been an important factor.

ACKNOWLEDGEMENTS

Portions of the work reported here build upon previous research sponsored by AFMA and thework on the design of an observer program (Appendix 5) was done as part of a contract toEnvironment Australia. This Appendix has been included so that the current report provides acomprehensive set of documents on seabird research. We also wish to acknowledge and thankall of the observers who have contributed data that have been utilized in this report.

REFERENCES

Brothers, N., R. Gales, and T. Reid. 1999. The influence of environmental variables andmitigation measures on seabird catch rates in the Japanese tuna longline fishery within theAustralian fishing Zone, 1991-1995. Biological Conservation 88: 85-101.

Croxall, J.P., P. Rothery, S.P.C. Pickering, and P.A. Prince. 1990. Reproductive performance,recruitment and survival of wandering albatross Diomedea exulans at Bird Island, SouthGeorgia. Journal of Animal Ecology 59: 775-796.

Croxall, J.P., P.A. Prince, P. Rothery and A.G. Wood. 1998. Population changes in albatrossesat South Georgia. In: Albatross Biology and Conservation. G. Robertson and R. Gales(Editors). Surrey Beatty & Sons, Chipping Norton, Australia. Pp. 69-83.

Gales, R. 1993. Co-operative mechanisms for the conservation of albatross. Australian NatureConservation Agency. Tasmanian Government Printer, Hobart, Australia, 132 pp.

Gales, R. 1998. Albatross populations: status and threats. In G. Robertson and R. Gales (eds),Albatross Biology and Conservation. Surrey Beatty and Sons, Chipping Norton, Australia.Pp 20-45.

Klaer, N. and D. Heinemann. 1998. Japanese longline seabird by-catch in the AustralianFishing Zone, April 1997 – March 1998. Third meeting of the Council for the conservationof southern bluefin tuna ecologically related species working group; 9-12 June 1998, Tokyo,Japan. CCSBT-ERS/9806/24, 9 pp.

Klaer, N. and T. Polacheck. 1995. Japanese longline seabird by-catch in the Australian FishingZone April 1991 – March 1994: catch and catch rates by area and season and an evaluationof the effectiveness of mitigation measures. CSIRO Division of Fisheries Report.

Klaer, N. and T. Polacheck. 1997a. Japanese longline seabird bycatch in the Australian FishingZone April 1995 – March 1997. CCSBT-ERS 1997 Working Group Document.

Klaer, N. and T. Polacheck. 1997b. By-catch of albatrosses and other seabirds by Japaneselongline fishing vessels in the Australian Fishing Zone from April 1992 to March 1995.Emu 97: 150-167.

7

Klaer, N. and T. Polacheck. 1998. The influence of environmental factors and mitigationmeasures on by-catch rates of seabirds by Japanese longline fishing vessels in the Australianregion. Emu 98: 305-316.

Tuck, G. and T. Polacheck. 1997. Trends in tuna longline fisheries in the southern oceans andimplications for seabird bycatch: 1997 update. Second meeting of the Council for theConservation of Southern Bluefin Tuna Ecologically Related Species Working Group; 3-6June 1997, Canberra, Australia. CCSBT-ERS/9706/35, 9 pp.

Tuck, G., T. Polacheck, J. Croxall, H. Weimerskirch, P. Prince and S. Wotherspoon. 1998. Thepotential of archival tags to provide long-term movement and behaviour data for seabirds:first results from wandering albatross (Diomedea exulans) of South Georgia and the CrozetIslands. CSIRO draft document.

Weimerskirch, H. and P. Jouventin. 1987. Population dynamics of the wandering albatross,Diomedea exulans, of the Crozet Islands: causes and consequences of the population decline.Oikos 49: 315-322.

Weimerskirch, H. and P. Jouventin. 1998. Changes in population sizes and demographicparameters of six albatross species breeding on the French sub-Antarctic islands. In:Albatross Biology and Conservation. G. Robertson and R. Gales (Editors). Surrey Beatty &Sons, Chipping Norton, Australia. Pp. 84-91.

8

APPENDIX 1

Japanese longline seabird by-catch in the Australian FishingZone, April 1997 – March 1998

Neil KlaerDennis Heinemann

15 May 1998

Document prepared for the third meeting of theCCSBT Ecologically Related Species Working Group,

9-12 June 1998, Tokyo, Japan

9

Introduction

This working paper presents tables of estimates of seabird by-catch by Japaneselongline vessels operating in the Australian Fishing Zone from April 1997 to March1998. The tables supply updated information to that given in Klaer and Polacheck1995, 1996 and 1997. The methods used to derive the estimates are the same as thoseused and documented in Klaer and Polacheck (1995) to derive estimates from April1991 to March 1994. Results are presented here for comparable area and time strata.No analysis of catch and catch rate by species is presented, as information on thespecies identification of seabirds retained by observers was not available to us.

These results underestimate the total number of seabirds actually killed because afraction of birds captured are lost before the line is hauled, and therefore are notobserved. Captured seabirds can be eaten by sharks or other fish before the longline ishauled and some are torn from the line during hauling (Brothers 1991). Some birds areknown to escape after being hooked (Weimerskirch & Jouventin 1987; Huin & Croxallin press) and may subsequently die. In addition, crew on some vessels dispose ofcaptured seabirds during hauling by quickly cutting them away, and some of these maybe missed by the observers. Brothers (1991) reports that the estimated number of birdsobserved to have been captured during a set was 27% more than the numbersubsequently observed to be hauled on board (this includes birds hooked but whichescape before being hauled underwater). As such, the estimates of incidental catchbased on the observed numbers recovered during hauling operations, as in this paper,may substantially underestimate actual seabird kills.

Data Processing and Analyses

Observer data was filtered to remove records that appeared to be recorded incorrectly,and those that could not be matched with records on the Japanese AFZ longline logbookdatabase. In 1997, 17% of the sets or 9% of the birds were removed, which was notunusual when compared to the previous two years (Table 1).

Table 1 . Observer data filtering for by-catch and by-catch rate analyses. The numbers of records rejectedby category are given to indicate the importance of different problems, but are not additive, and, thereforecannot be summed to give the total number of invalid records.

Sets Birds

1995 1996 1997 Total 1995 1996 1997 Total

Total records 338 414 358 1110 70 122 34 226

Reason for rejection

Bad position or outside AFZ 5 45 44 94 0 6 3 9

mis-match with logbook 6 39 28 73 0 4 0 4

Other1 24 27 41 92 2 4 0 6

Number valid 304 342 300 945 68 112 31 245

Percent valid 90 81 83 85 97 89 91 92

1e.g. observed greater than total number of hooks, insufficient data to calculate variances

10

Changes in Data Reporting

Monitoring of the Japanese fleet within the AFZ moved in 1996 from the collection ofpaper logbook records and reporting of fishing effort via radio to real-time reporting ofcatch and effort via satellite (known as the Vessel Monitoring System or VMS). In thepast, there was a considerable delay in the receipt of vessel logbooks by the AustralianGovernment as these were often sent when the vessel returned to Japan. Delays weremore than a year in some cases. Due to this delay, daily radio reports of numbers ofhooks set and noon position of Japanese vessels were often used for the calculation oftotal effort recently expended by Japanese vessels in the AFZ by area and time. Therehave always been problems in completely matching observer records with vessellogbook or radio report data for individual vessel sets. A match is required to allow thedata to be used in catch calculations such as those presented in this paper. In particular,matching observer records with radio report data was problematic as radio reports givethe vessel noon position and hooks set since the previous radio report. Observers recorddetails on start and end times and position for setting and hauling. As Japanese vesselsnormally carry out one fishing operation per day, to match observer and radio reportdata it was normally assumed that the radio report refers to the haul completed in theprevious 24 hours as recorded by the observer. This was not always the case, and somemis-matching was expected. The introduction of VMS has continued this problem, asVMS reports are made in the same way as the old radio reports. A noon position isreported, as well as catch and effort details for the previous 24 hours. Effort is beingmade to change this standard to reporting a time and position for some part of the lastfishing operation (start of set, end of haul, etc.), which would allow simple cross-reference of observer records. Such a change would greatly increase the utility of thedata collected via VMS.

In the past, our estimates of seabird by-catch and by-catch rate have been stratified byseason, area, and gear type (Klaer and Polacheck 1995, 1996 and 1997). To makestratified estimates, it must be possible to stratify both the observed and unobserveddata according to the factor of interest. Gear type, actually mainline material(traditional kuralon or lighter monofilament) has been determined to be important forthe calculation of seabird by-catch rates. The paper logbook system contained a geardetail sheet in which was recorded a description of the gear used by a Japanese vesselwhile fishing in the AFZ. Unfortunately, the VMS system that replaced the logbooksystem does not have a facility for the reporting of gear used by all vessels. Our onlyrecord of gear used now comes from those vessels that were observed. In the past, thenumber of vessels using monofilament has been small, and they were specificallytargeted for observer coverage. Up to 1996, we are confident that most or all of thevessels that used monofilament were identified. For 1997, however, this was notpossible. We know that some vessels that used monofilament in the AFZ in 1997 werenot observed, and that no data was collected on the gear used on non-observed vessels.This means that we now do not have the data required to make estimates of seabird by-catch and by-catch rates in the same manner as for previous years, ie. Using the geartype stratification. Accordingly, calculations are provided in this paper without use ofthe gear type stratum for 1997. For comparative purposes, we present data for theprevious years stratified in the same way, ie. By season and area only.

In 1997, a number of vessels fished in the zone surrounding Norfolk Island, and anobserver was placed in that area. Historically there has not been a large amount of

11

Japanese longline effort expended in Australian remote territories. Therefore, tomaintain comparability with results produced previously, the fishing effort, and thesmall number (3) of birds caught in that zone are not included in these analyses.

Results

The distribution of Japanese longline sets within the AFZ during winter and summer1997 are shown in Figure 1. Fishing effort was much lower in summer than winter, andthe distributions differed, with virtually no fishing occurring on the east coast north ofTasmania and the effort on the west coast shifting north. These seasonal changes werevery similar to those in previous years (see figures in Klaer and Polacheck 1995, 1997).

Compared to 1996, the total effort in 1997 that could be used for estimating seabird by-catch and by-catch rates was considerably higher in winter (103 versus 78 cruises, and2238 versus 1884 sets), but the summer effort declined significantly (1 versus 8 cruises,and 5 versus 174 sets).

The distributions of fishing effort with observers in 1997 (Figure 2) showed a verysimilar pattern to the total distribution of effort (Figure 1). The proportion of cruisesand sets in 1997 that were included in the analysis with observers was similar to that in1996; 36% versus 43% and 13% versus 16%, respectively.

All seabird by-catch and by-catch rates presented in previous reports have beencalculated using a stratification procedure which has included area, season and geartype. Stratification by gear type for the 1997 data was not possible for the reasonsdescribed above. Therefore, estimates for 1997 are stratified by area and season only(Table 2); comparable estimates for 1995 and 1996 are shown in Appendix 1.1.

12

Figure 1. Japanese longline fishing effort in the AFZ April 1997 to March 1998 for winter (left panel)and summer (right panel). Each point represents a single set. Regions used for stratification (North WestAustralia, North East Australia, South East Australia, Eastern Tasmania, Western Tasmania, SouthernAustralia and South East Indian Ocean) are overlayed on the maps.

Figure 2. Observed Japanese longline fishing effort in the AFZ April 1997 to March 1998 for winter (leftpanel) and summer (right panel). Each point represents a single set. Regions used for stratification(North West Australia, North East Australia, South East Australia, Eastern Tasmania, Western Tasmania,

Southern Australia and South East Indian Ocean) are overlayed on the maps.

Of the 31 seabirds observed killed in 1997, 29 were caught in the Eastern Tasmaniaarea during the winter season. The winter by-catch rate in that area (0.15 seabirds per

NW Aus NE Aus

SE Aus

SE Ind

S Aus W Tas E Tas

120 E

120 E

140 E

140 E

160 E

160 E

40 S 40 S

20 S 20 S

JOP 97 WIN

NW Aus NE Aus

SE Aus

SE Ind

S Aus W Tas E Tas

120 E

120 E

140 E

140 E

160 E

160 E

40 S 40 S

20 S 20 S

JOP 97 SUM

NW Aus NE Aus

SE Aus

SE Ind

S Aus W Tas E Tas

120 E

120 E

140 E

140 E

160 E

160 E

40 S 40 S

20 S 20 S

OBS 97 WIN

NW Aus NE Aus

SE Aus

SE Ind

S Aus W Tas E Tas

120 E

120 E

140 E

140 E

160 E

160 E

40 S 40 S

20 S 20 S

OBS 97 SUM

13

1,000 hooks ) was, with the exception of 1994 (see below), higher than in previousyears (0.09 in 1996, 0.10 in 1995, 0.04 in 1993, 0.05 in 1992, and 0.04 in 1991).Because all the cruises had observers on board, this by-catch rate scaled-up to anestimated 84 seabirds killed (not all sets were observed). The remaining two birds werecaught the South East Indian Ocean area in the winter. Because there were 23 sets withobservers and 177 without, the relatively low by-catch rate (0.04 seabirds per 1,000hooks) scaled up to an estimated 28 seabirds killed. In the past, when observers havebeen deployed in this area, by-catch rates were considerably higher (0.59 in 1996 and0.75 in 1993). The lack of seabird by-catch in the North East Australia area is the sameas what happened in 1994 – 1996, although high by-catch rates occurred there in 1991-1993. This pattern was similar as what occurred in the South East Australia area – noby-catch in 1997, low rates for 1994 – 1996 (0.06 – 0.09), and high rates for 1991 to1993 (0.48 – 0.67).

Table 2 . Estimated by-catch and by-catch rates of seabirds in the AFZ for 1997. The shaded values hadto be estimated indirectly because of insufficient sampling effort.

Area Season Cruises Sets ‘000 Hooks Birds Variance S.D. C.V. Catch

Tot Obs Tot Obs Tot Obs Obs Tot Set Cruise Total Rate

NE Aus Summer 1 0 5 0 15 0 0 0.25 0 0 0.28 0.02

E Tas Winter 8 8 171 76 565 194 29 84.42 603 0 603 25 0.29 0.15

NE Aus Winter 37 12 924 110 2977 254 0 0.00 0 0 0 0 0.00 0.00

SE Aus Winter 49 15 943 90 2957 218 0 0.00 0 0 0 0 0.00 0.00

SE Ind Winter 9 2 200 23 616 47 2 26.24 180 182 362 19 0.72 0.04

Observed strata 103 37 2238 299 7114 712 31 110.66 782 182 964 31 0.28 0.02

Southern strata 66 25 1314 189 4138 458 31 110.66 964 31 0.28 0.03

Total strata 104 37 2243 299 7130 712 31 110.91 964 31 0.28 0.02

These comparisons have been made without stratifying by gear type, because of thedata limitations in 1997. The effect of stratifying by gear type can be examined in asimple manner by looking at the estimates of overall seabird by-catch rate. Thesouthern areas were selected to make this comparison because that is where most by-catch has occurred. As expected, differences were found in overall by-catch rateestimates depending on whether stratification by mainline material was included in theanalysis (Figure 3).Figure 3. Overall by-catch rate of seabirds in southern areas and 95% confidence intervals calculatedwith (+ mono) and without stratification (- mono) by gear type (monofilament and traditional mainlinematerial); the stratification could not be performed in 1997 for the reasons described earlier.

14

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

1992 1993 1994 1995 1996 1997

Catchrate(per'000hooks)

mean + mono

95% ci + mono

mean - mono

95% ci - mono

During 1992, monofilament was not used, and so the estimates are identical. In thesummer of 1993, a small number of observed vessels that used monofilament caught alarge number of seabirds in comparison with other observed vessels using traditionalgear in the same areas. In response to this situation, and in consideration of thepossibility that monofilament gear increased the time that baits were accessible toseabirds, gear (mainline) type was included subsequently as a stratum in the estimationof by-catch and by-catch rates (Klaer and Polacheck 1995). This results in a muchlower estimate of by-catch rate variance, compared to that when the data were notstratified by gear type (Figure 3), and reduces the uncertainty in the overall estimate ofby-catch. Stratifying reduced the overall by-catch rate estimate, in this case, because ofthe particular distribution of observer effort and by-catch.

In 1994, observed vessels using monofilament were still few in number, but, unlike1993, did not catch comparatively larger numbers of seabirds, so the two estimatesproduced by the different stratifications and their variances are nearly the same. At thetime, this change was interpreted as probably due to the scattered and clumped natureof seabird catches. The same situation continued for estimates made for 1995 (Table 2;Klaer and Polacheck 1997). In 1996, in the South East Indian Ocean area duringwinter, observers were placed on two vessel using monofilament and one usingtraditional gear (Klaer and Polacheck 1997). This observation effort was unbalanced,as the number of vessels actually using traditional gear was much greater than thoseusing monofilament in that area/time stratum (13 traditional versus 2 monofilament).Contrary to expectations, the vessel using traditional gear caught a large number ofseabirds consistently over a number of separate sets. This meant that the high by-catchrate for the single vessel using traditional gear was scaled to a large proportion of theeffort in the stratum, leading to the calculation of a high catch rate overall. It was notedat the time that the overall AFZ by-catch rate was heavily influenced by theobservations made from this single cruise. Calculations made by pooling the data forthat area/time stratum leads to a lower overall AFZ 1996 by-catch rate estimate, butmuch larger variance (as shown in Figure 3).

15

Table 2. Comparison of observed by-catch rates of seabirds by gear type pooled annually for all southernstrata from 1992 to 1997.

Year Traditional Gear Monofilament Gear Ratio

Sets Birds Rate Sets Birds Rate

1992 561 229 0.41 − − − −

1993 799 270 0.34 35 148 4.23 12.51

1994 430 185 0.43 34 1 0.03 0.07

1995 220 60 0.27 71 8 0.11 0.41

1996 273 107 0.39 27 1 0.04 0.09

These differences across years in the relative by-catch rates of vessels using the twogear types can be shown to be statistically, significantly different, at least in some years(see Klaer and Polacheck 1995). However, we also recognise that seabird distributionsare clumped and that seabird captures are relatively rare events (Klaer and Polacheck1998), so the significance of these differences is over-estimated if random distributionis assumed and sample sizes are not very large. This aside, the data suggests a changein the effect of monofilament gear from being associated with large by-catch in 1993, tosmaller by-catch from 1994 to 1996. Anecdotal reports from AFZ observers and ourcalculations suggested that monofilament did increase chances of catching seabirdswhen it was first introduced. We cannot give a reason for this change, except to notethat Japanese operators may have modified their fishing methods in a way that offset theeffect (eg. Better construction of tori poles, thawing of baits, use of bait throwers, moresetting at night, line weighting, etc.), or the hypothesised link between the use ofmonofilament line and higher seabird by-catch rates may have been incomplete. Athorough scrutiny of the written observer reports for monofilament vessels during 1994to 1996 with regard to these issues, or directed trials may provide some information onthis question, but neither has yet been undertaken. One new adjustment reported byobservers was the practice of attaching heavy (2kg) weights to the buoy lines. Suchpractices are not included in the standard set of details collected by observers for entryinto the observer database. A modification of the type of data collected routinely byobservers on the gear used on longline vessels deserves consideration – particularly inrelation to how weighting is carried out, or generally on how fast the line sinks.

Given that there is evidence to support the case that monofilament and traditional geartype catch rates are different (greater or less depending on the year), estimates based onthe full stratification by area, season and gear type will have more validity. Therefore,by-catch rates calculated using the full stratification and previously presented for theperiod 1992 to 1996 are the most precise. Unless information becomes available toallow us to continue to stratify the data by gear type, our best estimates from 1997onwards will, of necessity, be less precise. Nonetheless, the overall by-catch rate in wasso low that stratification by gear type, in addition to area and season, would not make alarge difference.

16

References

Brothers, N. 1991. Albatross mortality and associated bait loss in the Japanese longlinefishery in the Southern Ocean. Biological Conservation 55, 255-268.

Huin, N. & Croxall, J.P. In press. Fishing gear, oil and marine debris associated withseabirds at Bird Island, South Georgia, 1993/94. Marine Ornithology.

Klaer, N. and Polacheck, T. 1995. Japanese longline seabird bycatch in the AustralianFishing Zone April 1991 – March 1994: catch and catch rates by area and seasonand an evaluation of the effectiveness of mitigation measures. CSIRO Division ofFisheries Report.

Klaer, N. and Polacheck, T. 1996. Japanese longline seabird bycatch in the AustralianFishing Zone April 1994 – March 1995. CCSBT ERS 1996 Working GroupDocument.

Klaer, N. and Polacheck, T. 1997. Japanese longline seabird bycatch in the AustralianFishing Zone April 1995 – March 1997. CCSBT ERS 1997 Working GroupDocument.

Klaer, N. and Polacheck, T. 1998. The influence of environmental factors andmitigation measures on by-catch rates of seabirds by Japanese longline vessels inthe Australian region. Paper modified for publication from the ERS 1997 WorkingGroup Document by the same title.

Weimerskirch, H. & Jouventin, P. 1987. Population dynamics of the WanderingAlbatross, Diomedea exulans, of the Crozet Islands: causes and consequences ofthe population decline. Oikos 49, 315-322.

17

Appendix 1.1

Estimated by-catch and by-catch rates of seabirds in the AFZ, stratified by area and season, for 1995.The shaded values had to be estimated indirectly because of insufficient sampling effort.

Area Season Cruises Sets ‘000 Hooks Birds Variance S.D. C.V. Catch

Tot Obs Tot Obs Tot Obs Obs Tot Set Cruise Total Rate

E Tas Summer 6 0 139 0 452 0 0 39.92 71 8 0.21 0.09

E Tas Winter 26 17 1152 169 3662 368 38 378.07 2403 3863 6265 79 0.21 0.10

NE Aus Summer 1 0 25 0 74 0 0 7.18 2 2 0.21 0.10

NE Aus Winter 31 1 301 13 947 31 0 0.00 0 0 0 0 0.00 0.00

NW Aus Summer 1 0 16 0 50 0 0 4.59 1 1 0.21 0.09

S Aus Summer 8 0 113 0 332 0 0 32.45 47 7 0.21 0.10

S Aus Winter 6 0 39 0 115 0 0 11.20 6 2 0.21 0.10

SE Aus Summer 2 0 14 0 42 0 0 4.02 1 1 0.21 0.10

SE Aus Winter 46 13 1563 116 4981 262 23 437.80 1679 26698 28377 168 0.38 0.09

SE Ind Summer 10 1 158 15 475 40 8 95.66 407 1828 2236 47 0.49 0.20

SE Ind Winter 4 0 12 0 37 0 0 3.45 1 1 0.21 0.09

W Tas Summer 3 0 20 0 59 0 0 5.74 1 1 0.21 0.10

W Tas Winter 14 0 47 0 148 0 0 13.50 8 3 0.21 0.09

Observed strata 113 32 3174 313 10065 700 69 911.53 4489 32389 36878 192 0.21 0.09

Southern strata 125 31 3257 300 10302 669 69 1021.81 37012 192 0.19 0.10

Total strata 158 32 3599 313 11373 700 69 1033.58 37015 218 0.21 0.09

Estimated by-catch and by-catch rates of seabirds in the AFZ, stratified by area and season, for 1996.

Area Season Cruises Sets ‘000 Hooks Birds Variance S.D. C.V. Catch

Tot Obs Tot Obs Tot Obs Obs Tot Set Cruise Total Rate

E Tas Summer 2 1 45 3 131 7 4 79.79 35 74 109 10 0.13 0.61

E Tas Winter 11 9 392 93 1216 239 22 112.08 362 381 743 27 0.24 0.09

NE Aus Winter 16 5 405 34 1307 75 0 0.00 0 0 0 0 0.00 0.00

NW Aus Summer 2 0 78 0 227 0 0 33.96 180 13 0.40 0.15

NW Aus Winter 1 0 1 0 3 0 0 0.44 0 0 0.40 0.16

S Aus Winter 1 0 1 0 3 0 0 0.44 0 0 0.40 0.15

SE Aus Summer 1 0 5 0 17 0 0 2.18 1 1 0.40 0.13

SE Aus Winter 27 15 778 156 2475 344 27 194.00 2625 5482 8107 90 0.46 0.08

SE Ind Winter 15 3 260 26 690 51 30 405.76 452 99764 100216 317 0.78 0.59

W Tas Summer 3 0 46 0 134 0 0 20.02 63 8 0.40 0.15

W Tas Winter 7 4 47 22 145 77 25 47.24 169 738 907 30 0.64 0.33

Observed strata 78 37 1927 334 5963 793 108 838.87 3642 106439 110082 332 0.40 0.14

Southern strata 67 32 1574 300 4810 718 108 861.51 110145 332 0.39 0.18

Total strata 86 37 2058 334 6348 793 108 895.90 125557 354 0.40 0.14

18

APPENDIX 2

The Ocean Colour Project:fishing effort, and shark and seabirdby-catch on three domestic tuna longlinersin eastern Australian waters

Jock W. YoungDennis HeinemannRosemary BaileyDon C. McKenzieJessica Farley

2 June 1998

Document prepared for the third meeting of theCCSBT Ecologically Related Species Working Group,

9-12 June 1998, Tokyo, Japan

19

Introduction

In 1993 a field study supporting ocean colour satellite imagery in Australian waters wasfunded jointly by Australia’s Fisheries Research and Development Corporation andEastern Tuna Management Advisory Committee (Parslow and Lyne 1994). The aims ofthe project were to provide ground-truth data to calibrate the satellite imagery and toexamine the relationship, if any, between tuna and billfish and ocean colour, a measureof ocean productivity, off eastern Australia. Delays in the launch of the ocean coloursatellite meant that the start of fieldwork was postponed until early 1997.

As a part of the project, CSIRO observers spent time aboard domestic longline fishingvessels operating on the east coast of Australia. They collected catch data and watersamples for analysis. A summary of the fishing practices and catch data, including by-catch, are included in this report.

Methodology

The field program centred around three domestic longlining vessels operating in twovery different oceanographic environments and fisheries on the east coast of Australia,one working out of Mooloolaba, a coastal port 90km north of Brisbane, and the othertwo operating out of the port of Bermagui on the south coast of New South Wales. Allboats are ~20m in length and fish for southern bluefin tuna, yellowfin tuna, bigeye tunaand broadbill swordfish depending on the season and area. Each boat is capable ofsetting over 1000 hooks on 60km of monofilament mainline. Trip duration for these andother boats in the area is for 2 to 3 days, and trips are generally restricted to waterswithin 50 - 150nm of the coast (on the continental shelf).

Each vessel was equipped with an underway sampling device developed by CSIROMarine Research to collect measurements of salinity, fluorescence (a measure ofphytoplankton concentration) and temperature at the sea surface. Each month, a CSIROobserver spent up to 4 days on each boat. Their duties included collecting samples ofseawater for determination of salinity and chlorophyll, recording the fishing practices ofthe vessels and recording details of all fish taken during each set.

Results

Nine observer trips, 4 out of Bermagui and 5 out of Mooloolaba, have been completedbetween May 1997 and February 1998 (Table 1), and a further 6 are planned. A total of5,700 hooks were observed in Bermagui and 16,100 hooks in Mooloolaba. The numberand relative proportions of the major fish species caught are shown in Table 2. In total,four seabirds, none of which were albatrosses, were caught during the observer trips. Aswell, 46 sharks were caught.

Fishing Practice and by-catch

Bermagui

The longline vessels observed in this area generally worked within 100nm of the coastaround 36°S, but have worked as far north as 28°S. In previous seasons they haveworked as far south as SE Tasmania and up to Mooloolaba. They usually set one

20

longline per day. Generally lines are set during daylight hours, usually from first thingin the morning until ~10am, and "soak" for the remainder of the day. The vessel returnsto the start of the set to begin retrieving after dark. However, setting times can varydepending on weather conditions. Between 550 and 1000 hooks are set on 60km ofmainline during each shot. Bait consists of live bait (mackerel and yellow tail), freshbait (rat tails) or frozen bait (mackerel, pilchards or squid plus glow sticks) dependingon the latitude and the season. The vessels targeted yellowfin, bigeye and southernbluefin tuna depending on the season.

There appears to be a potential for seabird interaction in the Bermagui area becauselines are set sometimes during daylight and there can be large numbers of seabirdspresent (mostly shearwaters). Shearwaters, albatrosses and other seabirds have beenobserved following the vessels whilst shooting away. Shearwaters are the main speciesfor potential interaction as they can, at certain times of the year, aggregate around thestern of the longliners during setting. Seabirds were observed flying and/or feedingduring the setting of all lines observed, and varied between 100 birds. Fourshearwaters (muttonbirds) were caught during the trips observed in this area, all duringone haul in November. Given the small amount of fishing effort observed (only 2vessels, 4 trips and less than 6,000 hooks), no useful inference can be made aboutseabird by-catch rate.

The setting of lines during daylight appears to be in response to the presence of sharks,which can be a significant problem at times. Hooks are soaked during the day andwinched up after dark to try to avoid the loss of hooked fish to sharks and minimisedamage to fishing gear. The total observed catch of sharks was 13 in the Bermagui area.

Mooloolaba

The vessel working out of this port usually works a wider area than the main domesticfishery off Mooloolaba, generally in the area of the Morton Seamount. The vessel isequipped with a fuel bladder to enable it to work up to 150nm offshore. It aims to setone longline per day, and, in contrast to boats working in mores southern waters, sets itsline immediately after dark. The longline “soaks” overnight, during which time thevessel returns to the start of the set to begin retrieving, usually starting at 7am. Between500 and 1150 hooks are set on each shot. Depending on the month, three maincommercial species are caught out of Mooloolaba – yellowfin and bigeye tuna andbroadbill swordfish. Bait consists of squid (Loligo spp.) aided by glow sticks.

There appears to be less potential for seabird interaction in the Mooloolaba area becauselines are usually set during the night (or very early morning), and there are generallyfewer birds present in the area. Birds were recorded as flying and/or feeding during only6 of the 17 shots observed. Shearwaters, albatross (very rare) and other seabirds havebeen observed following the vessel but in smaller numbers compared to the Bermaguiarea to the south. Again, shearwaters are the main species for potential interaction. Noseabirds of any species were caught during the trips in this area.

The total observed catch of sharks was 36 in the Mooloolaba area.

21

Discussion

The small number of vessels in this study strongly constrains the conclusions that canbe drawn about interactions between the domestic longline fleet and ecologically relatedspecies. Periodically, muttonbirds and other seabirds crowd the stern of boats workingout of Bermagui, however, four muttonbirds were the only mortalities observed, andthey were likely to have been short-tailed shearwaters, a species not listed as vulnerableor endangered. Anecdotal observations suggest that the presence of seabirds decreasesto the north of Bermagui. Bird sightings were low off southern Queensland incomparison with waters further south. These results are consistent with previousobservations from Australian domestic longline vessels, in which seabird by-catch ratesin Queensland have been zero, and some seabirds have been caught off New SouthWales (Brothers 1995; Polacheck 1995; Whitelaw 1995, 1997).

Our observations show that ~6% of the longline catch is composed of sharks, and, assuch, they suggest that the by-catch of sharks may not be insignificant. However,outside of a stock assessment context, it is not possible to evaluate the magnitude of anypotential impacts.

References

Brothers, N. and Foster, A. (1995) Seabird catch rates: an assessment of causes andsolutions in Australia’s tuna longline fishery. CCSBT-ERS/95/19.

Parslow, J. and Lyne, V. (1994) Development, application and evaluation of the use ofremotely sensed data by Australian fisheries (FRDC project number 94/045)

Polacheck, T. (1995) Summary tables of available information on seabird by-catch byAustralian tuna longline vessels. CCSBT-ERS/95/40.

Whitelaw, W. (1995) Some observations on seabird by-catch by Australian tunalongline fishing vessels. CCSBT-ERS/95/46.

Whitelaw, W. (1997) Some observations on seabird by-catch from Australian longlinefishing vessels 1994-1996. CCSBT-ERS/97/.

22

Table 1. Summary of the number of cruises and shots observed in the Ocean Colourstudy between May 1997 and February 1998.

Area Month No. of sets Total no. hooks

Mooloolaba July 1997 3 2600

Aug 1997 3 2600

Oct ‘1997 4 3900

Nov 1997 3 3000

Feb 1998 4 4000

Bermagui May 1997 1 700

Sept 1997 2 1800

Nov 1997 2 1300

Feb 1998 2 1900

Total 9 cruises 24 21800

23

Table 2. Summary of species caught by longliners involved in the Ocean Colour studybetween May 1997 and February 1998.

Species Count %

Albacore tuna 86 10.80

Bigeye tuna 101 12.69

Yellowfin tuna 253 31.78

Southern bluefin tuna 1 0.13

Black marlin 3 0.38

Blue marlin 5 0.63

Striped marlin 6 0.75

Broadbill swordfish 162 20.35

Lancet fish 24 3.02

Manta ray 3 0.38

Rudder fish 19 2.39

Blue whaler shark 11 1.38

Mako shark 13 1.63

Ocean white tip shark 8 1.01

Tiger shark 3 0.38

Bronze whaler shark 2 0.25

Unidentified shark species 9 1.13

Others 87 10.93

Total 795 100.00

APPENDIX 3

Pilot Seabird Bycatch Observer Programfor Australian Domestic Longline Fisheries

Design

Guidelines

Options and

Recommendations

Dennis Heinemann

Jessica Farley

Naomi Clear

John Gunn

Neil Klaer

Wade Whitelaw

29 January 1999

25

2. FISHERIES DESCRIPTIONS

2.1 Tuna Fisheries

2.1.5 Potential for seabird interactions

The domestic and Japanese longline fisheries within the AFZ generally target the same species,fish the same areas and use the same fishing approaches. However, some features of thedomestic fisheries suggest that they may have a lower seabird bycatch rate than the Japanesefishery. Those features generally include:

1) shorter mainlines and fewer hooks per set,

2) smaller vessels, which are lower to the water,

3) slower setting speeds,

4) the use of live bait in some cases, and

5) lack of offal discharge.

0

20

40

60

80

100

120

140

160

Fre

qu

en

cy

0 500 1000 1500 2000 2500 35003000

Number of hooks per shot

a

0

500

1000

1500

2000

2500

Number of hooks per shot

Fre

qu

en

cy

0 500 1000 1500 2000 2500 35003000

b

Figure 1. Number of hooks per set (shot) in the domestic (a)SBT (longline sector), and (b) non-SBT tuna fisheries, betweenJanuary 1997 and June 1998.

26

However, other factors could produce higher bycatch rates, such as:

1) the use of monofilament mainlines (which sink more slowly than Kuralonmainlines),

2) weighting lines less, and

3) the lower prevalence and inconsistent use of mitigation measures such as nightsetting, bird lines, bait casters and bait thawing.

Very few data are available from the domestic tuna fisheries to compare to the large body ofbycatch statistics from the Japanese SBT fishery (Klaer & Polacheck 1995, 1997, Klaer &Heinemann 1998, Brothers et al. 1998a, b). What is available for domestic fisheries, has beencollected since 1994 by CMS and Tasmania Parks and Wildlife Service observers on a largelyad hoc basis on a small number of longline fishing vessels. The results from most of thesecruises, which are provided in Whitelaw (1997), Young et al. (1998), Brothers et al. (1998a, b),Brothers & Foster (in press), are summarised in Table 4.

For comparative purposes the range of bycatch rates for the Japanese fishery are provided inTable 5. No birds were caught on any of the 28 trips sampled in Queensland over 4 years.Within the Japanese fishery only two birds have been observed caught in that same area andperiod.

Table 4. Seabird bycatch rates in the domestic tuna fisheries on the eastcoast during summer and winter, combining data for 1994-1996.

Area Season Trips Sets Hooks(x103)

Bycatch

(birds)

Bycatch Rate(birds per 103

hooks)

Summer 12 77 28.6 0 0.0NorthernQueensland1 Winter 10 48 26.1 0 0.0

Summer 3 11 10.9 0 0.0SouthernQueensland2 Winter 3 9 7.8 0 0.0

Summer 7 11 7.4 8 1.1New SouthWales Winter 4 6 4.3 0 0.0

Summer 11 30 34.4 8 0.2Tasmania

Winter 4 12 16.0 1 0.06

1 Cairns area, 2 Mooloolaba area

27

Table 5. Observed seabird bycatch rates (birds per 1,000 hooks) and observed fishingeffort (in parentheses; 1,000’s of hooks) for the Japanese Fishery off the east coast, inwinter and summer, for the years with the most reliable data, 1992 – 1997.

Area Season Fishery 1992 1993 1994 1995 1996 1997

Domestic 0.0(5.6)

0.0(23)

0.0(12)

Summer

Japanese 0.0(18)

0.0(76)

0.0(27)

Domestic 0.0(20)

0.0(13.3)

Queensland

Winter

Japanese 0.01(368)

0.01(151)

0.01(182)

0.0(31)

0.0(75)

0.0(254)

Domestic 2.41(0.8)

0.61(6.6)

Summer

Japanese 0.0(31)

0.0(6)

Domestic 0.0(1.0)

0.0(0.8)

0.0(2.5)

New SouthWales

Winter

Japanese 0.05(210)

0.05(289)

0.06(124)

0.09(262)

0.08(344)

0.0(218)

Domestic 0.36(6.4)

0.29(17)

0.0(11)

Summer

Japanese 0.22(23)

0.0(19)

0.53(107)

Domestic 0.0(2.5)

0.07(14)

Tasmania

Winter

Japanese 0.05(551)

0.04(854)

0.06(644)

0.12(368)

0.1(239)

0.15(194)

28

Both fisheries have recorded seabird bycatch in New South Wales. A few birds have beencaught in the Japanese fishery in all but one winter. None have been recorded in the domesticfishery in winter, but little fishing effort has been observed during winter in New South Wales(6 sets over 4 trips). In the two years that data have been collected during summer in theJapanese fishery, no bycatch was recorded. By contrast, in the summer of 1994 a substantialbycatch rate was observed in the domestic fishery (2.41 birds per 1,000 hooks). Again,however, this was based on very little observer effort (3 sets totaling only 830 hooks over 2trips). In the summer of 1997, a moderate bycatch rate of 0.61 birds per 1,000 hooks wasrecorded, this time from a somewhat larger sample (approximately 7,000 hooks observed).

The observed bycatch rate for Tasmania in summer looks very much the same in the twofisheries. Both have recorded one year with no birds caught, and in the other two years withdata, bycatch rates were in the order of one bird caught per 2,000 to 5,000 hooks. In winter theJapanese fishery recorded bycatch rates of 0.04-0.06 in the early years, and 0.10-0.15 in the lastthree years. The only winter data from the domestic fishery in Tasmania recorded one birdcaught in 12 sets over 4 trips.

For the most part, there are few obvious differences between the bycatch rates recorded in thedomestic and Japanese fisheries. The largest differences, which occurred in New South Walesduring summer, may have been artifacts of very low sampling levels, and, perhaps, because theobservations came from different areas off New South Wales. Therefore, in light of thesimilarities between the fisheries in the other areas and the sample size limitations, it would beinadvisable to ascribe too much weight to these differences. We conclude from an examinationof the little domestic data available, that there is little evidence to suggest that incidental seabirdmortality in the domestic tuna fisheries off eastern Australia is lower overall than in theJapanese SBT fishery in the AFZ. It is important to recognize that because the domestic data donot provide a representative sample, the bycatch rate estimates presented here cannot becombined reliably with effort data from the fishery to give unbiased estimates of the totalnumber of seabirds killed in the domestic fishery.

In our interviews with operators of domestic tuna longliners, they stressed that the bycatch ofseabirds, especially albatrosses, is low. However, they have reported that shearwaters can be aproblem at certain times of the year. Many stated that the Australian vessels are significantlydifferent to the Japanese vessels, and consequently catch fewer seabirds. They suggest that theuse of live bait may also reduce the bycatch of seabirds, as properly hooked live bait will swimrapidly down, thus reducing the time the bait is available to seabirds.

2.2 Southern Shark (demersal longline sector)

…

2.2.5 Potential for seabird interactions

We were unable to find any reports of seabird interactions or bycatch on shark longline vessels.N. Brothers, Tasmania Parks and Wildlife Service (pers. comm.) spent roughly 30 days on ashark demersal longliner off the west coast of Tasmania in Jan-Feb in the late 1970’s. Duringthat period he reported that they made one 1400-hook, daytime shot each day, and that he didnot observe any seabirds caught. Other demersal longline fisheries are known cause significant

29

levels of incidental seabird mortality. Seabird bycatch rates in demersal longline fisherieselsewhere in the world range from approximately 0.1 to 0.7 birds per 1000 hooks (Ashford et al.1998, Bergin 1997, Moreno et al. 1996, Neves and Olmos 1998, Stagi et al. 1998, Williams &Capdeville 1996, CCAMLR 1997) Similarities between these fisheries and the Australiandemersal longline fishery in the gear used, suggests the potential for a significant seabirdbycatch in the domestic fishery.

Operators of shark longline vessels stated that very few seabirds are caught while fishing. Theyattribute this to the fast sinking rate of the mainline during setting, which reduces the probabilityof a seabird being hooked. The line sinks quickly because it heavily weighted at both ends andat intervals along the line. Some fishers do not discharge offal and set during the night whichmay also reduce the bycatch of seabirds.

2.3 South East Non-Trawl Fishery

…

2.3.5 Potential for seabird interactions

Observations of seabird interactions have not been made on vessels using droplines or trotlines.Given the size of the weights and vertical deployment of droplines, it is unlikely that they wouldhave a significant bycatch of seabirds.

Demersal longlines and trotlines both have a greater potential for seabird bycatch because of theway they are deployed, and what is know from other fisheries that use similar gear. The onlydata from this fishery come from two demersal automatic longliner cruises observed by aseabird biologist from the Tasmania Parks and Wildlife Service (Brothers et al. 1998a, b).Observations were made of 12 sets (57,000 hooks) over 5 days in 1996, and of 10 sets (60,500hooks) in 1997, also over 5 days. No seabirds were observed to have been caught or killed oneither cruise. The crew and master on the 1996 cruise stated they had caught only “a few birdsduring the setting of three million hooks” (Brothers et al. 1998a).

Longline operators in the SENTF who we interviewed indicated that they very rarely, or never,catch seabirds. They suggested this is due to the vertical (or near vertical) setting of the gear,which is especially true in the case of droplines, which descend vertically very fast duringsetting. The most extreme example reported to us by any fisher was a bycatch rate of 3 seabirdsin 25 years of fishing. The practice of setting at night and the lack of offal discharge by manyvessels were other reasons given by skippers for their low seabird bycatch.

30

Paper prepared for submission to Emu, June 1998 v 2.00

APPENDIX 4

The influence of environmental factors and mitigationmeasures on by-catch rates of seabirds by Japaneselongline fishing vessels in the Australian region.

Neil Klaer and Tom Polacheck

CSIRO Division of Fisheries, GPO Box 1538, Hobart, Tasmania 7001

Summary

Most seabirds caught and killed by longline fishing are captured during line setting. Datacollected by Australian observers on Japanese longline vessels from April 1992 to March1995 were used to determine the influence of various environmental factors andmitigation measures on seabird catch rates. Generalised linear models were used todetermine the significance of the effect of each factor on the seabird catch rate. Resultsshow that the environmental factor which most influences the seabird catch rate iswhether line setting was carried out at night or during the day. For the data examined, thenight catch rate was 9% of the day catch rate. Seabird catches at night are significantlygreater when the moon is on the full half-phase. The catch rate at night during the newmoon was 2% of the day catch rate. Other environmental factors with significant effectswere the area and season fished. Wind, cloud and sea conditions were not found to havea significant influence. Considerable variation in the seabird by-catch rate among vesselswas found, which is probably due to differences in their implementation of mitigationmeasures, as well as the clumped distribution of seabirds by area and time. Inter-annualvariation in the by-catch rate was found to be statistically significant but the differencesamong years was small in comparison to other factors.

An examination of the influence of mitigation measures for sets made during the day insummer in the Tasmanian area showed that the level of bait thawing and unidentifiedfactors related to individual vessels were most significant in determining the seabird by-catch rate, followed by the use of a bait throwing device. It was not possible to examinethe influence of the use to bird scaring tori poles and lines as these were used during allsets examined in detail. For this data set, the amount of cloud cover had an influence,while moon phase, sea conditions and wind strength did not.

Introduction

Measured recent declines in the population size of a number of albatross species (e.g.Tomkins 1985; Weimerskirch and Jouventin 1987; Croxall et al. 1990) and estimates ofseabird by-catch from commercial fishing operations (e.g. Brothers 1991; Murray et al.1993; CCAMLR 1994; Klaer & Polacheck 1995) have led to the recognition that longlinefishing is a threat to the long-term viability of some seabird populations, notablyalbatrosses. A number of international treaties, conventions and agreements have

31

acknowledged this threat and recommended action to reduce albatross by-catch bylongline fisheries (e.g. Bonn Convention 1995, 1997; International Union forConservation of Nature and Natural Resources 1996). In 1995 the AustralianGovernment listed pelagic tuna longlining as a key threatening process for seabirds(particularly albatrosses) under the Australian Government Endangered SpeciesProtection Act.

Analysis of Australian Fishing Zone (AFZ) observer records for the years 1992 to 1995(the same time period as this study) indicated that at least 2,800 to 3,600 seabirds werecaught annually by Japanese longline vessels in the southern Australian region, of which78% were albatrosses (Klaer & Polacheck 1995, 1997). Black-browed Albatross Diomedeamelanophrys and Shy Albatross D. cauta were caught in the greatest numbers, and catchesof Yellow-nosed Albatross D. chlororhynchos, Wandering Albatross D. exulans and Grey-headed Albatross D. chrysostoma were also significant (Klaer & Polacheck 1995, 1997).

Mitigation measures designed to reduce the threat of longline fishing to albatrosses havealso been developed and increased employment recommended by various groups (e.g.Brothers 1991; CCAMLR 1994; Alexander et al. 1997). Mitigation measures normallyconsidered or recommended include setting lines at night, bird scaring lines andstreamers trailed behind fishing vessels during line setting (known as tori poles and lines),machines to cast baits clear of the vessel wash during line setting, weighting lines moreheavily so that they sink more quickly, thawing bait and using bait which more readilysinks, fishing area or season closures and avoidance of dumping of offal near the fishinglines during setting and hauling.

There has been minimal analysis of the effect of different mitigation measures andenvironmental factors on the catch rates of seabirds by Japanese longline vessels in theAustralian region. It has been determined that setting longlines at night rather thanduring the day is an effective method which may be used to reduce incidental catches ofseabirds (Brothers 1991; Murray et al. 1993; Klaer & Polacheck 1995). Cross-tabulationsof seabird by-catch rates by environmental conditions and the use of seabird by-catchmitigation measures were presented in Klaer & Polacheck (1995). However, no analysishas been attempted to determine the relative importance of the various factors ininfluencing seabird by-catch rates. This paper presents an integrated analysis of theavailable Australian observer data from April 1992 to March 1995.

Methods

Allocation of time of capture

Almost all seabirds retrieved dead on hauling of a longline are hooked during line settingwhile the bait is close to the water surface (Brothers 1991). Baits normally sink out ofreach of seabirds within a short time, so it follows that the time of capture for mostseabirds is close to the time that the hook was set. To examine the effect that the time ofday has on seabird catches and catch rates, it is necessary to assign a time of capture foreach bird observed attached to a hook as the line is hauled. The actual capture is notusually observed and recorded as the line is set, so this time must be estimated.

32

The method used to estimate time of capture (t) was the same as that applied by Murrayet al. (1993). Using the time of an event during the haul tev (in this case the observedlanding of a seabird) in relation to the time of haul start (ths) and end (the), the time of theevent during the set is estimated as a proportion of the time between sets start (tss) andset end (tse) as follows:

−−

−−+=)()(

)(hshe

hsevsssess tt

ttxtttt

The value for x was set to 1 if the line was hauled from the end of the line last set, and 0if from the start of the set. Most hauls (93% of all observed hauls) were made from theend of the set.

The main assumption in using this method is that the line is both set and hauled at areasonably constant rate. A number of events during hauling may lead to delays,including line breaks, line tangles, and retrieval of large tuna. When a line break isencountered, the vessel may recommence hauling on a section of line that is out ofsequence with the line set (C. Ramirez, pers. comm.). As such, this assumption is notalways met, which means that some fishing effort and incidental kills may be mis-classified as to whether they took place during night or day. The consequence of anysuch mis-classifications will be to reduce the probability of detecting a significantdifference in catch rates between day and night and to under-estimate the magnitude ofany effect if detected. In order to minimise the amount of mis-classification due to gearproblems, sets in which it took more than 10 hours to set the longline or more than 15hours to retrieve it were excluded from the analyses.

Time of day calculations

From first light to sunrise and sunset to last light are periods of twilight. Preliminaryexamination of seabird catch rates have shown that catch rates during the twilight periodare usually more similar to catch rates during the day than at night (Klaer & Polacheck1995). For this study, times of capture have been defined only in terms of night or day,delimited by the time of evening and morning nautical twilight. The times of nauticaltwilight were calculated using algorithms given in Doggett et al. (1990). Given a position(latitude and longitude) and a date, these algorithms produce the times of astronomicalevents in Greenwich Mean Time (GMT).

For each observed set, the start and end time for each observation period during the haulwas recorded. Often there was more than one observation period for a single haul. Usingthe method described previously, start and end times for observation periods during thehaul were translated to corresponding times during the set. Using start and end times forobservation periods during the set, and the calculated times for nautical twilight, it waspossible to estimate for any observed hook, whether the hook was set during the night orday.

The number of observed hooks set within a time period was estimated using theproportion of total observation time that occurred during the period multiplied by totalobserved hooks.

33

Position of all observed sets during 1992 to 1994 are shown in Figure 1. Onlyobservations from –southern regions (south-east Indian Ocean, southern Australia,western Tasmania, eastern Tasmania and south-east Australia) are included in theanalyses in this paper. This is because previous work has shown that seabird catch ratesare very much lower in the Northern regions - mainly due to the absence of albatrossesin these Northern waters (Klaer and Polacheck, 1995, 1997). As such, the sets fromthese Northern waters can provide little information on factors effecting seabird by-catchrates in areas where they do occur.

In addition to excluding observations from these more Northern regions, sets were alsoexcluded for sets in which critical information was either not recorded or where therewas an unresolvable inconsistency. All together, about 30% of the observed sets wererejected (Table 1).

Table 1. Observer data filtering for time of capture analysesReason for rejection1 Sets Birds

1992 1993 1994 Total 1992 1993 1994 Total

Total records 634 884 532 2050 252 429 196 877

Time zone not recorded 223 61 20 304 60 36 0 96

Incomplete observation period durations2 8 24 39 71 0 0 2 2

Monofilament gear 0 32 32 64 0 140 1 141

Haul time > 15 hours 19 22 22 63 20 4 5 29

Mismatch of calculated observed hooks3 11 27 8 46 3 17 5 25

Haul start or end time not recorded 17 7 16 40 1 0 0 1

Observation period overlap4 9 6 5 20 0 1 1 2

Observed hooks = 0 9 7 2 18 0 0 0 0

Set time > 10 hours 1 3 0 4 3 1 0 4

Valid 337 695 388 1420 165 230 182 577

Percent valid 53.15 78.62 72.93 69.27 65.48 53.61 92.86 65.79

Notes: 1 Only the first reason for the rejection of a record is included in this summary, so the table does not show the total number of records failing each criterion2 After adding times for each observation period, total time 1.44 Times of start and end of observation periods overlap

34

Figure 1. Position of observed Japanese longline sets from 1 April 1992 to 31 March 1995

Most records were rejected due to lack of information of the time zone used forrecording. Such information is critical for estimating the time that hooks were set, as theAustralian Fishing Zone spans 4 hours GMT in longitude (GMT +7.00 hours to GMT+10.00 hours). On Japanese longline vessels the observers usually record set timesaccording to Japan time (GMT +9.00 hours), but sometimes may use local or other timezones. It was only during the 1992 fishing year that observers began recording the timezone used on the vessel. As noted above, the criteria that total set time must be less than10 hours and total haul time must be less than 15 hours were also introduced in an effortto exclude records where large delays may have occurred during the set or haul.

Since mis-recording of information for a set is likely to be independent of factorsaffecting seabird catch rates, such filtering should have produced an unbiased randomsample of the available records. A valid time of capture was subsequently obtained for allbirds recorded captured for valid sets, avoiding the introduction of a bias throughrejection of records based on this criterion.

Generalised linear models (GLMs)

NW Aus NE Aus

SE Aus

SE Indian Ocean

S Aus W Tas E Tas

100 E

100 E

120 E

120 E

140 E

140 E

160 E

160 E

40 S 40 S

20 S 20 S

35

The response variable under consideration in the statistical analyses presented here iswhether a seabird is caught on a hook or not. The response is therefore binary. As such,a logistic regression procedure (see McCullagh & Nelder 1983) was chosen to model theprocess of capturing seabirds. In this case, an event was defined as the setting of alongline and a trial was the setting of a single hook. The response variable to beexamined was the number of seabirds caught per hook in a set. The distribution of theresponse variable was assumed to be binomial, and the logit link function (McCullagh &Nelder 1983) was used. For the purposes of the GLM, the 1,420 valid sets detailed inTable 1 were further divided into portions of the set which were made during the nightand day. This division produced 2,291 valid events for the analysis. In this paper, theGLM carried out using all available data is referred to as the full GLM.

To simplify the analysis and to allow presentation of cross-tabulations, all factors ofinterest were reduced to a small number of discrete classes as shown in Table 2.

The linear form of the model used was:

logπ

πβ

1 1−

==

∑ jj

p

where π = probability of catching a seabird on a hook;

p = total number of factors;

β = the value of a factor;

j = factor index.

36

Table 2. Observed factors included in the analysis and discrete classes defined for each.

Factor Classes Explanation

Vessel Various unique identifiers for each fishing vessel

Year 1992 1 April 1992 to 31 March 1993

1993 1 April 1993 to 31 March 1994

1994 1 April 1994 to 31 March 1995

Time of capture night nautical dusk to nautical dawn

day nautical dawn to nautical dusk

Moon phase full from mid-phase to mid-phase through full

new from mid-phase to mid-phase through new

Area SE Australia

Tasmania

Southern Australia

SE Indian Ocean

Season winter April to September

summer October to March

Wind low Beaufort scale 0 to 3

medium 4 to 6

high over 6

Cloud low 0 to 35% coverage

medium 36 to 65%

high 66 to 100%

Sea low 0 to 2.5 m swell

medium 2.6 to 4.5 m

high over 4.5 m

Tori pole* yes tori pole in use during line setting

no no tori pole

Bait thawing* poor bait not thawed

fair bait partly thawed

good bait well thawed

Bait thrower* yes bait thrower in use during line setting

no no bait thrower

Note: * information for these factors was not recorded in a largefraction of the data set (see Appendix), so these factors were notexamined using the full GLM. Bait thawing and bait throwerfactors were examined using the sub-set GLM.

37

A fishing year was defined to commence in April and finish in March the following year,with the winter season generally occurring from April to September and the summerseason from October to March. This timing of seasons delimits the seasonal movementsof the Japanese fishing fleet (Sainsbury et al. 1994; Klaer & Polacheck 1995). Theseasonal distribution of seabirds differs by species, but many have distinct summer andwinter distributions (see Marchant & Higgins 1990).

Results

Generalised linear models

A cross-tabulation of the sample sizes (in sets) used for the GLM analyses is given in theAppendix. Sample numbers are highest in the Tasmanian region generally. As observereffort is deployed to obtain approximately similar coverage rates by area and time strata,this is mainly due to higher total fishing effort in this region by Japanese vessels duringthe years examined.

Results for full GLM are given in Table 3. The maximal model included vessel, time ofcapture, moon phase, time of capture/moon phase interactions, area, season,area/season interactions, wind, cloud, sea condition and year as factors. After exclusionof non-significant factors, the final model included vessel, time of capture/moon phaseinteractions, time of capture, area, season and year.

In assessing the goodness of fit of the model, the log-likelihood ratio statistic or deviancewas scaled to approximate a chi-squared distribution by dividing by a scaling value(McCullagh & Nelder 1983). The scaling value was calculated by dividing the Pearsonchi-square value for the maximal model by the degrees of freedom for the maximalmodel (2809/2187 = 1.284). A calculated scale value greater than 1.0 indicates either apoorly fitting model, or over-dispersion in the data. A scale value close to 1.0 indicatesthat the data are consistent with the underlying model and error structure. Over-dispersion arises frequently in GLM applications (McCullagh & Nelder 1983) andindicates a clumping of events. The scale value obtained here of 1.284 suggests thatsome degree of over dispersion exists in these data. Essentially, over-dispersion in ourcase indicates that for some reason, if a seabird is caught on a hook during a set, then thecapture of further birds is more likely than a random distribution of events wouldindicate. Over-dispersion is not surprising as the spatial/temporal distribution of seabirdsis highly clustered within the spatial time strata used for the GLM. Murray et al. (1993)examined small-scale differences in seabird capture rates by area and concluded thatcatch rates differed among areas that were closely spaced (within 100 nautical miles).

38

Table 3. Measures of the effect of vessel and environmental factor inclusion on modeldeviance, and the significance of exclusion of each factor in alternative models for thefull GLM.

Model Nominal Change Chi sq

df deviance df deviance scaled deviance

Null model 2290 2316.21

Maximal model 2187 1082.47

- Vessel 2272 1610.00 85 527.53 410.72 < 0.001

- Time*Moon 2188 1096.17 1 13.70 10.67 < 0.01

- Time 2189 1239.07 2 142.90 111.26 < 0.001

- Moon 2189 1096.19 2 0.02 0.02 ns

- Area*Season 2190 1084.32 3 1.85 1.44 ns

- Area 2193 1121.63 6 37.31 29.05 < 0.001

- Season 2191 1128.33 4 44.01 34.26 < 0.001

- Wind 2189 1082.90 2 0.43 0.34 ns

- Cloud 2189 1084.01 2 1.54 1.20ns

- Sea 2189 1083.16 2 0.69 0.54 ns

- Year 2189 1090.90 2 8.43 6.56 < 0.05

Final model 2196 1086.87 9 4.40 3.43 ns

Note: df = degrees of freedom.

Results given in Table 3 indicate that the environmental factors which most affectseabird catch rates are time of day (day/night sets) (p

39

the available data per longline set includes information on all mitigation measures used(see Appendix). Anecdotal accounts from fisheries observers also report that the qualityof implementation of various mitigation measures such as bird scaring poles and linesvaries considerably among vessels (Klaer & Polacheck 1995).

About 53% of the null model deviance is explained by the full model. This indicates thatover half of the variation in seabird by-catch can be explained by the six significantfactors included in the model and suggests that these factors are important variables thatneed to be considered in assessing by-catch rates. White substantial variation remainsunexplained by the model, this is not surprising given the high degree of spatial andtemporal clustering of seabirds.

Transformed values shown in Table 4 are presented as odds ratios, with each valueexpressed as a probability in the range 0.0-1.0. For each factor, a base probability of 0.50was assigned arbitrarily to one factor class, and values for all other classes are expressedin relation to the base class. The interpretation for time of day is that with lines set duringthe day as the base (p=0.5), night setting produces lower catch rates of seabirds (p=0.17).There is little influence of the moon on day catch rates by moon phase (p=0.5 full,p=0.51 new). For sets at night, with the full moon as the base, night sets on the newmoon have lower catch rates (p=0.15). For areas, with Tasmania as the base, SouthernAustralia produces a much higher catch rate (p=0.90) and SE Australia (p=0.53) and SEIndian Ocean (p=0.51) produce similar catch rates to the Tasmanian area. For season,with winter as the base, summer produces higher catch rate (p=0.81). For year, with1994 as the base, 1992 produced a slightly lower catch rate (p=0.42), while 1993 wasslightly higher (p=0.54).

40

Table 4. Parameter estimates produced by the final model of the full GLM, standarderrors of the estimates, transformations of the estimates to odds ratios using an inverselogit function and 95% confidence intervals for the odds ratios using +/- 1.96 times thestandard error.

Factor Class Estimate Est s.e. TransformedEstimate

min 95% max95%

Time day 0.00 0.50

night -1.58 0.29 0.17 0.10 0.27

Time * Moon day - full 0.00 0.50

day - new 0.05 0.11 0.51 0.46 0.57

night - full 0.00 0.50

night -new

-1.70 0.62 0.15 0.05 0.39

Area S Aus 2.20 0.47 0.90 0.78 0.96

SE Aus 0.11 0.38 0.53 0.34 0.70

SE Ind 0.04 0.89 0.51 0.15 0.86

Tas 0.00 0.50

Season summer 1.13 0.20 0.76 0.67 0.82

winter 0.00 0.50

Year 1992 -0.33 0.21 0.42 0.32 0.52

1993 0.18 0.22 0.54 0.43 0.65

1994 0.00 0.50

Parameter estimates for each vessel are not shown in Table 4 as there were 86 vessels inthe analysis, and estimates for each are not important except for recognition that there ismuch variation among vessels in their ability to catch seabirds. This partially relates todifferences among vessels in mitigation strategies that may have been employed.

As there are strong area, season and time of day effects, in order to exclude suchinfluences and examine the effects of mitigation measures implemented by individualfishing vessels in detail, a sub-set GLM was conducted using a sub-set of the full datafrom a stratum where all of these factors were constant. The stratum chosen was theTasmanian region during the day in summer, as this stratum had the largest number of

41

observations suitable for analysis in combination with a relatively high catch rate ofseabirds (see Appendix). This combination improves the ability of the GLM to detectsignificant differences in mitigation effects. Results from the GLM using this sub-set ofthe data are given in Table 5. The sample size is small for this analysis (141 observedsets), as a requirement was that all classes for each factor were recorded - i.e. no factorwas recorded as unknown. To keep the sample size as large as possible, the year effectfound to be just significant in the full GLM was ignored, allowing examination of datafrom this stratum for all years.

Mitigation measures able to be examined did not include tori poles as all sets during theTasmanian summer used them. The maximal model did not include a vessel effect in thiscase, as it was expected that differences among vessels would be captured by observeddifferences in mitigation measures employed. As the type and quality of mitigationmeasures used normally remain fairly constant during the observation for each vessel,there would be confounding due to the high correlation between mitigation measuresemployed and the vessel. There are insufficient available data to examine interactionsbetween vessel and mitigation effects.

Table 5. Sub-set GLM (Tasmanian region day sets during summer) - contribution ofvarious environmental and mitigation factors to the model fit.

Model Nominal Change Chi sq

df deviance df deviance scaleddeviance

Null model 140 316.48

Maximal model 130 262.55

- Moon 131 264.30 1 1.75 1.45 ns

- Bait thawing 132 276.47 2 13.92 11.56 < 0.001

- Bait thrower 131 273.23 1 10.67 8.87 < 0.01

- Wind 132 265.53 2 2.98 2.48 ns

- Cloud 132 271.07 2 8.52 7.08 < 0.05

- Sea 132 264.03 2 1.47 1.23 ns

+ Vessel 114 147.06 16 115.49 95.95 < 0.001

Bait thawing is the most significant measured factor (p

42

catch rates of individual vessels that have not been measured and included in the model.This may, for example, be due to differences among vessels in the quality of mitigationmeasures employed and the fine-scale spatial clustering of seabirds within this region. Itis of interest to note that the scaling value for the sub-set GLM was 269/130 = 2.07,which is more over-dispersed than the full GLM. This suggests that spatial and temporalvariability in the distribution of seabirds may be higher in this stratum than others.

Table 6. Parameter estimates produced by the final model of the sub-set GLM, standarderrors of the estimates, transformations of the estimates to odds ratios using an inverselogit function and 95% confidence intervals for the odds ratios using +/- 1.96 times thestandard error.

Factor Class Estimate Est s.e. TransformedEstimate

min 95% max95%

Baitthawing

fair -0.68 0.24 0.34 0.24 0.45

good -0.68 0.36 0.34 0.20 0.51

poor 0.00 0.50

Baitthrower

no 0.70 0.32 0.67 0.52 0.79

yes 0.00 0.50

Cloud high -0.44 0.22 0.39 0.29 0.50

med 0.00 0.50

low -0.35 0.28 0.41 0.29 0.55

Odds ratios given in Table 6 show that in comparison with poorly thawed bait as thebase (p=0.5), fair and good thawing decrease the seabird by-catch rate (p=0.34). With theuse of a bait thrower as the base (p=0.5), not using a bait thrower increases the seabirdby-catch rate (p=0.67). For cloud cover, in comparison with medium cover (p=0.5), bothhigher (p=0.39) and lower (p=0.41) levels of cover decrease the seabird by-catch rate. Ofthe significant effects, cloud cover was the least significant, and was not found to besignificant at all in the full GLM. This combined with the fact that there was not aconsistent trend in the effect of cloud cover, suggests that the result for cloud coverrequires further investigation and should be treated with caution.

Examination of the Tasmanian winter stratum in a similar manner was unsuccessful dueto even greater dispersion shown by the results and smaller number of birds captured.

43

Cross-tabulations of seabird by-catch by significant environmental factors

Tables 7 to 10 present cross-tabulations of seabird by-catch information in relation to theenvironmental factors found to be significant using the full GLM.

Table 7. Total numbers of observed sets carried out under various environmentalconditions for the full GLM data set.

Area Season D