(from "The Report of the Scientific Committee", 1996 meeting)

The Group structured its discussions around the existing trial specifications as detailed in Appendix 3 of Rep. int. Whal. Commn 44: 120-44. It first reviewed the data available for 'conditioning' trials (i.e. choosing population parameters for the trials compatible with such data). These comprised sighting survey estimates of abundance; historical catches and their geographic distribution; conception data, isozyme, DNA and length frequency data; and CPUE information.

The greatest part of the Group's time was taken up by a detailed discussion of stock/sub-stock structure. Seventeen contributions dealing partly or wholly with these issues were presented. These addressed general questions about drawing inferences from genetic data, as well as presentations and/or analyses of genetic data, length and sex compositions of catches, whaling ground locations, and information on conception dates, morphology, parasites and pollutants. Both historical information, and that obtained from the whale research programme under special permit conducted in sub-area 9 (see Fig. 1 of Annex J) in 1994 and 1995 were considered.

The Group agreed that the available data and information were generally inconsistent with there being sub-stocks to the east of Japan characterised by different levels of latitudinal migration. Earlier, the presence of discrete whaling grounds had been argued to indicate site specific aggregations. However, SC/48/NP6 argued that this was rather the result of the discrete distribution of land stations, and the continuous distributions shown by sighting surveys do not support the existence of local aggregations. Instead, this coastal region seems to be a migratory corridor. Furthermore, the lack of mature females in sub-area 7 implies that a separate group of interbreeding animals is not present there. In contrast, much less information is available on the minke whales to the west of Japan, with consequential uncertainty about the linkage between those found to the east and the west of Korea.

In the light of these discussions, the original Implementation Simulation Trials for North Pacific minke whales were revised as reflected in Appendix 5 of Annex J. The most important changes are:

(a) the time step used in modelling whale migration has been shortened from two months to one, to better match model predictions and the further data now available for conditioning;

(b) sub-stock structure has been dropped;

(c) the catch mixing matrices (reflecting spatio-temporal changes in stock distribution) are now age and sex-structured.

The 144 trials agreed by the Group are listed in section G of Appendix 5 of Annex J.

The basis used to select the trials was by and large that used in the past, i.e. concentrating on cases with MSYR (mature) = 1% and factors likely to render the trials more difficult with regard to meeting risk-related criteria. The Group agreed that the relative plausibility of different trials needed to be taken into account in future discussions, but that this could not be done in the absence of consideration of the risk associated with the results for each.

Punt undertook recording of the trials, but was unable to complete this task before the end of the Scientific Committee meeting.

In noting the discussion of sub-stocks in Annex J, the Committee agreed that it was important to consider the general question of management units, and specifically questions associated with the conservation of the range of a species that may have a habitual component to its migration. This is discussed further under Item 20.2.

The Committee recommended that the Secretariat carry out computations of the trials listed in Appendix 5 of Annex J intersessionally. This Appendix includes a suggested order to conduct the large number of trials suggested. The reason for this is that results for the initial trials in this set may be sufficient to show that sensitivity to certain factors is minimal, and hence that there is no need to carry out some of the further trials suggested for that factor. The Committee appointed an intersessional Steering Group (comprising Allison, Butterworth, Hatanaka, Punt, Tanaka (E) and Taylor) to review results of these initial trials by correspondence, and to advise the Secretariat whether or not certain of the remaining trials need to be carried out in the light of these initial results. Since it was hot possible to complete the coding of the trials during the meeting, this Group was also authorised to effect minor amendments to Appendix 5 of Annex J if some specifications therein were shown in the coding process to be mutually inconsistent.

The Committee noted that the RMP specifications (Rep. int. Whal. Commn 44: 145-52) as presently drafted do not envisage temporal constraints within a season on a catch limit for a Small Area, although the specifications for the trials in Appendix 5 of Annex J (see section D thereof include such restrictions. The Committee agreed that should the results of the trials proposed indicate satisfactory performance, it would give attention to consideration of an appropriate amendment to the RMP specifications at a future meeting.

8.1.2 Future surveys
The Committee did not have sufficient time available to fully review the survey designs specified in SC/48/NP18 relative to the requirements specified under Item 7.1 and Annex K, particularly as the discussions under that item were not completed until the penultimate day. However, the Committee did provide general advice and comments about the design presented.

SC/48/NP18 presented the plans for sighting surveys in the North Pacific for the summer and winter seasons of 1996/1997. Eight surveys with six vessels will be carried out between 15 May 1996 and 28 March 1997 in four different areas. All surveys will be conducted with experienced observers using the IDCR closing mode approach. For each survey a primary species or group of species has been identified but sightings of all species observed will be recorded. Distance and angle data will be collected with reticule binoculars and angle boards, respectively. Training in distance and angle measurement will be conducted prior to the trip and testing will be conducted during the trip. All except one vessel (Survey 8) have a top barrel for sightings. For the survey of Baird's beaked whale (survey 4) and the winter small cetaceans survey (survey 9), only 1 person will observe from the barrel but in the remaining surveys there will be 2 observers in the barrel. For survey 8 in winter, sightings will be made from the upper bridge. Only results from surveys 2, 3, 4 and 5 will be considered with respect to the RMP implementation.

Discussion was limited by time and the information available. It was noted that abundance estimates will be derived from the survey data assuming that g(0)=1. However, it was encouraged that attempts be made to collect independent observer data and to assess the feasibility of applying methods of estimating g(0) based on such data. The use of a single observer in a survey because of limited vessel capacity was discussed. It was recognised that g(0) will most likely be less than unity and that estimates from these surveys would most likely underestimate abundance.

The Committee noted the potential importance of the sightings surveys outlined in SC/48/NP18 for its work, and notes the importance to their success of the granting of research or entry permits into the EEZs of relevant countries.

8.2 Southern Hemisphere minke whales
The Committee noted the discussion within the Commission last year regarding the implementation of the RMP (Rep. int. Whal. Commn 46:25). A number of countries believed that adoption of the Sanctuary precluded the adoption of the RMP in the area, at least for the time being. They considered that the Committee should not consider Southern Hemisphere minke whales in this context unless advised to do so by the Commission.

Japan had opposed this majority view because of its objection to the Sanctuary and the advanced state of the Committee's work on minke whales.

Item 9.2 considers other aspects of Southern Hemisphere minke whales.

8.3 North Atlantic minke whales
8.3.1 Abundance estimates from work of AEWG
CHAIRMAN'S SUMMARY OF SC/48/REP1
After the failure of the Scientific Committee to agree on an acceptable estimate of abundance for the Northeast Atlantic minke whales last year (RIWC 46:61-2), an intersessional abundance estimation working group (AEWG) was established with a developed workplan (RIWC 46:216-17)

'to 'maximise' the probability that the Scientific Committee would be able to agree on an estimate at its next annual meeting'.

The multi-national AEWG was comprised of biologists and statisticians with extensive experience in the conduct and analysis of line transect surveys: Polacheck (Chair); Butterworth; Cooke; Hammond; Laake; 懃en; Palka; Schweder and Smith. The work plan included two inter-sessional meetings and an extensive set of identified tasks and working papers which needed to be completed. The identified work included analyses and verification of:

(1) the survey and experimental data that had been collected in 1988-1990;

(2) the survey conducted during the summer of 1995; and

(3) the specifics of the hazard probability method as it had been developed for the analyses of these data.

As agreed in the Workplan, the AEWG had extensive intersessional communication using electronic mail for the automatic distribution of correspondence to all members (e.g. over 250 messages were exchanged); completed to their satisfaction all specific task identified in its terms of reference; produced forty one working papers; met twice; and produced an agreed report as a record of its work over the year (SC/48/Rep1). This report documents the AEWG's consensus on acceptable estimates of abundance of northeast Atlantic minke whales.

As recommended by the Committee, the AEWG agreed to use the hazard probability approach as the basic analytical method for obtaining abundance estimates from the survey data. The hazard probability approach estimates the density of whales by integrating estimates of the surfacing rate with estimates of the conditional probability of detecting a whale that surfaces at a particular position relative to the survey platform, given that the whale has not previously been observed. The hazard probability is modelled and estimated as a function of sighting angle, radial distance and covariates which can include environmental, observer, platform and whale heterogeneity effects. The basic data available for estimating the parameters of the hazard probability model are the initial sighting positions of whales and the success/failure of Bernoulli trials data collected from independent observer teams tracking surfacings of whales. The estimation procedure is based on a maximum likelihood approach.

The AEWG reviewed the sighting survey plans of the Institute of Marine Research (Bergen) and provided advice on details of the data collection procedure prior to the 1995 survey. Four members of AEWG participated in the 1995 Norwegian Independent Line Transect Survey (NILS95) and summary reports from seven survey participants describing their experience were presented to AEWG. These reports dealt with potential problems in the data collection and implementation of the survey protocol that could have had implications for the analyses. The AEWG reviewed these reports and the data collected, including a large number and variety of summary statistics. Based on this review, the problems identified were either (1) corrected in the data prior to the finalisation of the analyses; (2) taken account directly in the analyses; or (3) judged to be of minimal consequence in the estimation. The AEWG agreed that the data were acceptable for use in estimating minke whale abundance.

The hazard rate estimation procedure as applied to the northeastern Atlantic minke whale survey data is complex and comprises a large number of components. Responses to concerns raised at last year's Committee meeting resulted in substantial new developments in methodology. This resulted in an increased workload for both the AEWG and the NCC (Norwegian Computing Centre). Nevertheless, the AEWG was able to complete extensive examination of the various components (including new developments) and agreed on an acceptable approach for each of these relative to the estimation of abundance from the available data. At its January meeting, the AEWG concluded that there was nothing in principle to prevent the adoption of abundance estimates based on the agreed procedures, given that the result yielded satisfactory performance as demonstrated by suitable diagnostics. The second meeting of the AEWG was delayed by six weeks until the end of April in order to allow the completion of the implementation of the new developments and the finalisation of the analyses of the data.

The specific methodology and analytic issues for which the AEWG developed agreed approaches with respect to the analysis of the 1988-90 and 1995 data included: (1) the specific definition of the components of the likelihood function; (2) the specification of the inverse logit model as the functional form of the hazard probability model; (3) the use of only linear terms in the link function in the bias correction procedure; (4) the estimator for the parameters for the Neyman-Scott process model used to model the spatial distribution of whales; (5) the need for an interactive procedure for the estimation of the effective strip half width and the parameters of the Neyman-Scott process used in the bias correction procedure; (6) the decision on whether a single iteration of the Neyman-Scott parameters was sufficient for estimating abundances; (7) the set of covariates and levels of stratification to be used in the estimation of the hazard probability function; (8) the decision on whether the use of the estimated 'whale' presentation angle or 'fisk' categories at the time of initial sighting should or should not be included at this time as a covariate; (9) the definition of the truncation region; (10) the procedure to use for combining the data from the independent observer teams; (11) the procedures for accounting for measurement errors; (12) the specific functional form and parameter values used for modelling measurement errors; (13) rules for duplicate identification; (14) the surfacing rate model; and (15) the method for estimating the variances of abundance estimates.

The extensive documentation of the data, the analytical methods and the software implementation were reviewed by the AEWG and the AEWG concluded that they were sufficient. With respect to verification of the NCC software implementation used to calculate the abundance estimates, the AEWG recognised that several approaches had been undertaken. These included: (1) an extensive and independent review of software by computer scientists from the University of Aarhus, Denmark and (2) use of simulated data to compare estimates with data where the correct answer is known. The AEWG agreed that the software had been sufficiently validated for use in estimating minke whale abundance.

The AEWG undertook extensive review, testing and comparison of the automated procedures that had been developed for the classification of surfacings from independent observer teams in terms of whether they were duplicates. Analyses were also developed for estimating the probability of missing a surfacing when tracking a whale and the probability of losing a track altogether. Comparison of the automated procedures when applied to the 1995 survey data and to simulated data indicated that the different rules are in general agreement with each other and yield reasonable agreement to the true proportions in the simulated data. The AEWG noted that the form and application of the duplicate rule is an area where it is difficult to examine sufficient diagnostics and the sensitivity of the estimation procedure to the specific rule used. Because of this, it agreed that it would be necessary to examine the effect of this uncertainty by using at least two different duplicate rules to evaluate the robustness of the bias correction procedure. Such a comparison was provided for the final estimates considered by the AEWG at its April meeting and the resulting estimates were found to be in close agreement.

An important component of the identified intersessional work was the testing of the procedure with simulated data. The AEWG agreed on a small set of simulation tests to be conducted for verification given limitations imposed by the computing time requirements of the software developed by the NCC. In addition, more extensive simulation testing of the general performance of the methods was conducted with the less computationally intensive implementation developed by Cooke. Based on the results of these simulation tests, the AEWG agreed that they constituted confirmation of satisfactory estimation performance in the context of the analyses of the 1988/89 and 1995 survey data.

The implementation of the abundance estimation procedure assumed no whale movement. The AEWG agreed that, although accounting for whale movement would be desirable, insufficient information was available at this time to allow specification of other than arbitrary models. It agreed that there was no evidence for attractive movement towards the vessel under survey conditions and that there was some suggestion of avoidance behaviour for whales with radial sighting distances less than 1,000m based on the estimated swim directions observed at the time of initial sighting. However, because of large estimation errors in angular swimming direction measurements and because the ability of observers to classify swim direction may be a function of sighting angles, the results need to be interpreted with caution. The AEWG agreed that while additional research in this area was needed, an estimate that did not account for whale movement would be acceptable.

The AEWG reviewed estimates of g(0) for minke whales from other surveys in other areas with respect to their relevance to the estimation of abundance from the 1988/90 and 1995 data. The AEWG concluded that current quantitative comparisons from other surveys did not provide a basis for an estimate or for setting an upper bound for g(O) for the northeastern Atlantic minke whale surveys.

At the AEWG's April meeting, working papers were presented and reviewed which detailed the application of the hazard probability method to the two sets of survey data for the Northeastern Atlantic minke whales based on the methods and procedures agreed on by the AEWG. These documents presented abundance estimates, results of intermediate calculations and extensive sets of diagnostic and descriptive statistics. The diagnostic and descriptive statistics presented were based on the set developed by the AEWG and which it had agreed were sufficient for evaluating the adequacy of the fit. The AEWG concluded that the diagnostics demonstrated an adequate fit to the data. However, it did note some lack of fit of the estimated hazard probability model for the Bernoulli success data with respect to the marginal distribution of radial sighting distances. In particular, there was an underestimation of the proportion of observed successful trials within 400m of the vessel. The AEWG recognised a number of factors that could be contributing to this lack of fit and concluded that the overall effect of this lack of fit on the estimates of abundance in terms of possible bias were (i) of unknown direction and (ii) of second order. Immediately following the April meeting, further diagnostic statistics were discussed via e-mail that confirmed this conclusion.

The AEWG also agreed on the specific and implementation details of the method for calculating the variance for these abundance estimates. Time did not permit finalisation of these calculations during the April meeting. Preliminary calculations indicated CVs on the order of 20% for the 1988-90 data and 11% for the 1995 data.

Based on its careful examination of the hazard probability method and approach, the AEWG agreed that the application of the procedure as described in its report and the supporting working papers submitted to it is acceptable for estimating the abundance of northeastern Atlantic minke whales from the 1988-90 and 1995 data. The application of the method resulted in estimates of 67,531 and 118,299 whales for 1989 and 1995 respectively. Specific estimates by survey strata are given in SC/48/Rep1 (table 5). The AEWG recommended the acceptability of the estimates given in Table 2 (from SC/48/NA1).

The AEWG agreed that the data collection and analysis methods on which these estimates are based are a substantial advance on previously used methods for North Atlantic minke whales. Nonetheless, as is always the case, the method could be further improved for use in the northeastern Atlantic and elsewhere. During the course of its work and discussions, the following areas were identified as warranting further investigation:

(1) improved methods of obtaining representative dive time data;

(2) methods of measuring whale movement and a possible responsive behaviour, and of incorporating such information in the estimates;

(3) methods of estimating presentation angle more accurately, and of incorporating this and other heterogeneity;

(4) methods of estimating the bias and variability in positional estimates that are taken during the survey, with special focus on how uncertainty in these estimates would affect the abundance estimate uncertainty;

(5) methods for incorporating encounter rate information directly into the likelihood for estimating the parameters of the hazard probability function;

(6) models of and methods for estimating spatial distribution patterns, especially the suitability of the Neyman-Scott process at different spatial scales.

Noting that the methods developed in SC/48/NA1 for North Atlantic minke whales are a substantial advance on methods based on perpendicular distances, the AEWG recommended that a programme of experimental comparison of methods of analysis of double-team sighting survey data be conducted to compare the performance of this approach with that of perpendicular distance based estimation methods.

At the end of his presentation of SC/48/Rep1, Polacheck thanked all the members of the AEWG for their long hours of hard work and effort. From his perspective, he noted that the report of the AEWG represented a truly collaborative and co-operative effort. All members of the AEWG were active participants throughout the process, making substantial contributions to the actual analyses and contributing to at least one of the 41 working papers presented to the meeting. These contributions were combined and synthesised into the final report and estimates of abundance. Finally he expressed his gratitude and thanks on behalf of the AEWG and the Scientific Committee to the staff of the NCC for the logistical support, co-operation and good humour it had displayed throughout the process.

DISCUSSION
The Chairman of the AEWG noted that the calculation of the variance estimates for the abundance estimates had been completed prior to the Committee meeting and the results were, presented in SC/48/NA1. The resulting CVs were 19% and 10% for the 1984 and 1995 estimates, respectively.

The discussion of this item in the Committee was dominated by consideration of a working paper submitted by Cooke, a member of the AEWG. That working paper is reproduced as Annex N1. Its contents can be separated into two general concerns: (1) the way in which the AEWG had conducted its business; and (2) scientific questions about the conclusion the AEWG had reached. This report of the Committee's deliberations below follows that structure. More general issues of how the Committee can improve its working practices are considered under Item 20.2.

WORKING OF THE GROUP
In response to a question from the chair of the Committee as to why he had changed his opinion since report had been finalised, Cooke noted that although the AEWG had been adopted by consensus, he stated that he had expressed reservations on many issues at the AEWG's meeting, some of which he thought were mentioned in the report at the time. He had not, however, dissented from the agreed report. He noted the psychological fact that if one was in a minority of one, one tended to go along with one's colleagues. He noted that with the benefit of hindsight he should not have agreed to all of the decisions made at and between the AEWG meetings. He took full responsibility for this, stating that his concern was that the Committee was once again being pressured into premature acceptance of an estimate of abundance for northeast Atlantic minke whales.

The other members of the AEWG expressed complete surprise at this response and to Annex N1 in general. They noted that the AEWG had had extensive intersessional communication including the distribution of electronic mail to all members of the AEWG in addition to the two meetings that had been held. The AEWG Chairman had been extremely careful to structure the work to ensure, in accordance with its remit, 'that all decisions are based on consensus'. All members of the AEWG (except Cooke) and in particular its Rapporteur, noted that they had not detected any lack of consensus around any of the many decisions taken by the AEWG. They noted that all members of the AEWG had entered into serious discussion throughout the process that involved addressing reservations from many members about technical aspects of the estimation procedure. All of these reservations were resolved in coming to consensus of each of the many decisions taken. In the end no concerns were expressed to the Rapporteur or to other members of the AEWG that they held reservations which were not expressed in the report.

Despite the impression created by Annex N1, the other members of the AEWG noted that Cooke had participated fully in the work of the AEWG, entering into all discussions and debates, participating in the NILS-95 cruise and contributing many important ideas that had been incorporated into the estimation method adopted. They noted that there had been ample opportunity for members of the AEWG to indicate disagreement throughout the year long process. Indeed one member of the AEWG, relatively new to the Committee, had expressed some concern about one aspect of the analysis after the final report had been agreed at the April meeting. All members of the AEWG, including Cooke, had communicated during this period and additional calculations had been completed and agreement reached.

After this agreement the Rapporteur had e-mailed the AEWG with the hope that they 'would continue to communicate freely prior to the meeting so that we may all have time to consider any new issues that may arise in the meantime'.

Members of the AEWG believed that the impression Annex N1 created about the way the AEWG worked was in complete contrast to the spirit in which the AEWG had been established and had carried out its work. Cooke's concerns were not indicated to any other member of the AEWG at any time during the meetings either before or after the finalisation of the report. They were particularly disappointed and surprised that Annex N1, which criticised some consensus agreements reached in January, was not circulated to the Committee until a few hours before discussion of the report, nine days after the start of the Committee meeting and one month after it had been agreed by consensus. In addition, Annex N1 does severe injustice to the other members of the AEWG and is misleading to readers who did not participate in the process.

Whilst strongly supporting the right of scientists to express and alter their views, the AEWG believed that the manner in which it was done in this case to be unacceptable and an impediment to the work of the Scientific Committee.

This has serious implications for the future work of the Committee, which must be carried out in an atmosphere of trust and legitimate scientific debate, if it is to meet its responsibilities.

Some members of the AEWG found Cooke's explanation unconvincing given his previous contributions to the work of the Committee. Indeed in Annex N1 he refers to his own previous minority statement on the subject of northeastern Atlantic minke whales. They believed it was particularly unfortunate that this should have involved the IUCN representative, an organisation that has close links and special status within the Scientific Committee.

Cooke responded that, while he understood what he perceived were negative reactions of some members towards those who try to check or reproduce accepted estimates, he considered such checking to be part of the normal process of peer review which is within the Committee's remit. He regretted that not more members of the Committee were willing to take on the thankless task of cross-checking abundance estimates. De la Mare commented that he believed that the timetable established for the Working Group had been too severe.

During review of the meeting report, the other members of the AEWG expressed their appreciation for Cooke's work in helping to validate abundance estimates and affirmed that they also considered such checking to be part of the peer review process. They regretted he feels that such negative reactions exist towards his contribution, but re-assured him that they did not hold such feelings.

SCIENTIFIC ASPECTS
In discussion of SC/48/Rep1, a number of issues were raised in a scientific context. The most important can be summarised as follows:

(1) the adequacy of the simulation testing element in (a) verifying the software and (b) determining whether the hazard probability method is acceptable for estimating abundance and its variance;

(2) the reasons for differences in estimates from the 1989 and 1995 surveys;

(3) the differences between the NCC and Cooke estimates for 1989;

(4) an apparent inconsistency between estimates for part of the North Sea for three different surveys and years.

The report is structured to present a brief summary of the issues raised and the responses given.

(1) SIMULATION TESTING
With respect to simulation testing in Annex N1, Cooke had commented that his estimation procedure had been run on the full suite of simulation trials developed in SC/A96/AE18 and AE20 whereas the NCC method had only been run on 4 simulated data sets. A number of other members of the Committee requested clarification of this issue.

In response it was noted that simulation testing is only one tool, albeit important, in addressing two quite different questions (1(a) and (b) above). With respect to the verification of the software, the process followed by NCC and reviewed by the AEWG was to develop and check the software during all stages of its development and after its completion and included:

(1) detailed documentation and specification of the statistical, computing and mathematical procedures;

(2) the internal procedures used within the NCC for writing and checking the codes;

(3) the numerical internal testing conducted by the NCC for each of the components or modules in the estimation procedure;

(4) the use, where possible, of pre-existing and well established sub-routines and procedures for performing specific computation tasks;

(5) the development within the NCC of an alternative set of software for producing approximate numerical solutions that yielded similar numerical results;

(6) the independent review and evaluation of the software performed by the University of Aarhus (Denmark);

(7) the agreement in the results from the simulation tests for the NCC and Cooke implementation;

(8) the extensive comparative work that previous working groups of the Committee on abundance estimation had done between the NCC and Cooke implementations (e.g. Annex I - IWC, 1996). This had resulted in an agreement in the numerical results to within two decimals when the methodological differences in the implementation were removed.

More specifically, the four simulations conducted were carefully selected to cover the range of important components of the estimator. It was recognised that 100% certainty in the verification process is not possible. While additional simulations would have strengthened the process of verification in conjunction with the results of the other extensive checks, the level of scrutiny and testing of the software far exceeded that performed on most software used by the Committee. The AEWG had agreed that the software had been adequately verified for use in estimating minke whale abundance.

(1B) ACCEPTABILITY OF THE HAZARD PROBABILITY METHOD
With respect to the question of whether the hazard probability method is acceptable, the AEWG conducted extensive simulation testing, the results of which are presented in Polacheck and Cooke (1996). These tests were carried out using Cooke's software which was agreed by the AEWG to be a sufficiently similar mathematical estimator to the NCC software for the purpose of addressing this question. These results confirmed the previous conclusion of the Committee (Rep. int. Whal. Commn 46:61) that the hazard probability method is an acceptable approach to estimating abundance. Specifically, the simulation testing showed that the hazard probability method has, at most, small statistical bias and is robust to a range of factors. The simulation testing was more limited with respect to investigating the variance of the point estimate but did indicate that it was low, as expected for sample sizes typical of those actually achieved for the NILS-95 surveys. More importantly, independent empirical bootstrap methods confirmed this.

Despite the explanation given above, De la Mare, Lankester and Slooten did not consider that four single simulation trials provided an adequate basis for validating the statistical properties of the NCC estimation procedure.

(2) THE DIFFERENCE BETWEEN THE 1989 AND 1995 ESTIMATES
Annex N1 had commented that the increase in the point estimates from 67,531 in 1989 to 118,299 in 1995 will be 'puzzling to outside observers'. There was considerable discussion over this in the Committee.

In response a number of plausible reasons for this were put forward that are not mutually exclusive:

(1) the extrapolation from the 1990 experiment to the 1989 survey data may introduce a negative bias;

(2) the 1989 estimate may have been negatively biased because the covariates it was agreed should be included in the 1995 estimate could not be incorporated in 1989 to the same extent;

(3) the number of minke whales in the survey area may have increased in 1995 from immigration;

(4) a natural rate of increase in the population during the 6 year period - for the lower 95% confidence interval for the 1995 abundance estimate this represents an annual increase of around 2%;

(5) the 1989 and 1995 estimates may not be different statistically if the CV is under-estimated from unaccounted sources such as potential greater variability in dive sequences than is exhibited in the data from the eight minke whales used in the analysis.

(6) the 1995 estimate of minke whale abundance is a more reliable estimate because it was derived from a designed survey with independent teams of observers and did not depend on extrapolation from independent observer data collected in a different year as was the case for the 1989 estimate. The Committee last year (Rep. int. Whal. Commn 46) had agreed that analysis of the 1989 data was problematic.

(3) THE DIFFERENCE BETWEEN THE 1989 COOKE AND NCC ESTIMATES
Annex N1 had noted that the unstratified abundance estimate for the 1988-90 data given in SC/48/NA14Rev of 42,500 was different to the unstratified abundance estimate presented in SC/48/NA1 (56,700). SC/48/Rep1 had identified a number of differences in the implementations (p.29) and the AEWG bad concluded that the estimates were not comparable. However, some members of the Committee believed that the source of the differences should have been investigated more fully and that they cast doubt on the validity of the 1995 estimate. Cooke stated in Annex N1 that it was 'not worth going to the trouble of calculating an estimate for 1995' but in discussion noted that part of this decision was that he did not have the resources available to him to produce the software for calculating the estimates as quickly as the NCC.

Butterworth (Annex N2) examined one of the several factors identified in SC/48/Rep1 (different measurement error models) and noted that this difference alone would increase the estimate from SC/47/NA14Rev by some 12%. Because of substantial differences such as these, the AEWG had noted that the two estimates are not directly comparable and as such do not provide any information with respect to the verification process. The AEWG recalled that the main focus of its work was to validate methods for analysing data from 1995, where double-team data were collected as part of the survey (Rep. int. Whal. Commn 46: 216-17). The AEWG reaffirmed its view that because of differences in methodology, apparent differences between the NCC and Cooke estimates for 1989 cannot be used as a justification for questioning the validity of the 1995 estimate.

The AEWG had hoped and expected to see alternative estimates of abundance presented based on two conceptually comparable, but technically different implementations of the hazard probability approach as part of its verification process. However, it was not within its resources to produce alternative estimates and it was dependent upon the efforts of its individual members. The AEWG appreciated the efforts of Cooke in this regard, and it was unfortunate that no such alternative calculations were able to be produced. The AEWG had agreed that 'given the additional verification that had been accomplished that the results presented in AE17 [SC/48/NA1] had been verified to its satisfaction'.

(4) APPARENT INCONSISTENCIES IN ESTIMATES FOR THE NORTH SEA
Annex N1 referred to differences in point estimates of abundance for some parts of the North Sea between surveys carried out in 1989 (5,580 - SC/48/NA1) in 1994 (3,000-4,000 - SC/48/NA17) and 1995 (20,294 SC/48/NA1), noting that such differences might be implausible to lay people and that the high 1995 estimate from this area would have been substantially reduced if a vessel specific estimate from the distance experiments had been applied. This was discussed at some length by the Committee.

In discussion a number of points were made indicating that such comparisons were inappropriate.

(1) The North Sea area was only one sub-area in the large area covered by the NILS-95 survey. It is not surprising that the distribution of whales within a large survey area might change seasonally, with resultant changes in abundance in sub-areas. This is well known from other surveys (e.g. Palka, 1996) and is related to environmental factors such as food availability. This is supported by the fact that changes in estimated abundance between 1989 and 1995 are apparent in other survey blocks in the opposite direction. For example the abundance estimate for the Kola region was more than 10 times higher in 1989 than in 1995 (SC/48/NA1).

(2) Important differences in the objectives, methodology and survey blocks are particularly relevant to the comparison between the SCANS 94 and NILS'95 surveys (SC/48/NA17), rendering interpretation of differences difficult, if not impossible, even if Item 1 above is not considered.

(3) An average instead of a vessel specific bias correction factor for radial sighting distance was used in the abundance estimate because analyses of the distance experiments indicate that, when the variance between days for individual observers was taken into account, there was no significant added variance due to observed, vessels or platform (SC/A96/AE34).

(4) If abundance in the North Sea was overestimated in 1995 as a result of the use of an average distance bias correction factor, underestimation of abundance in other survey blocks would be expected and the net effect will be approximately unbiased.

After addressing these factors the Committee considered how to interpret the abundance estimate endorsed by the AEWG and presented in SC/48/NA1. It was noted that the objective of the work in recent years had been to obtain an estimate of abundance adequate for use in the CLA of the RMP, not to obtain a 'perfect' estimate. The latter is not an achievable scientific goal. Indeed, as noted above, SC/48/Rep1 had identified a number of areas for future work.

Some members believed that there were some parallels with the 1992 situation when the Committee had accepted an abundance estimate for this area without a full set of simulation tests and without independent recalculation of the results. They noted that the fact that the estimate subsequently turned out to be erroneous had led the Committee to decide that it was determined to avoid its previous mistake of prematurely accepting an abundance estimate (RIWC 46:62). They further noted that despite the earlier discussion (issue 3 above), they remained concerned about the considerable differences between the estimate of 42,500 in SC/47/NA14 rev for 1988-90 for the case of no covariates, and the estimate for this case of 56,721 in SC/48/NA1. Although they could accept the conclusion that the estimates would be adequate for implementation of the RMP, they did not consider the estimates scientifically validated until the tasks identified below had been satisfactorily completed.

Most members disagreed that the current situation paralleled that in 1992, noting that the process undertaken by the AEWG had substantially advanced the Committee's work since 1992, thanks in no small degree to contributions by Cooke. This work had greatly reduced the possibility of the Committee accepting an erroneous estimate and had ensured that the estimates had been adequately validated taking into account the procedures identified in the Report of the Verification Working Group (RIWC 47: Annex O)

The Committee agreed that the estimates of abundance SC/48/NA1 presented in are adequate for use in the RMP.

The Committee also agreed that some analyses should be carried out during the coming year and completed by next year's meeting. These are:

(i) additional simulation tests to more fully explore the statistical properties of the NCC estimator and for the purpose of further confirmation that the software was adequately verified (Wall indicated that this work would be undertaken in Norway during the coming year with respect to the method described in SC/48/NA1, based on advice from the Steering Group, see below);

(ii) to reconcile any differences between comparable estimates obtained from the NCC (as described in SC/48/NA1) and Cooke (SC/47/NA14Rev) implementations by identifying the main causes of any such differences (Cooke undertook to complete a comparable analyses of the NASS89 and NILS '95 data for his implementation);

(iii) to further assess the implications, in terms of possible bias in the NCC estimates, of the lack of model fit to the Bernoulli data, with respect to the marginal distribution of radial distances (Wall indicated that this work will be undertaken in Norway).

The Committee agreed that specification of the details of these analyses should be determined by a Steering Group to be appointed by the Chairman. It was agreed that the intention of these analyses is not to preclude further development of the methods, but such work falls outside the terms of reference of the Group.

At the close of the discussion the Committee expressed its thanks and admiration to Polacheck, the AEWG Chairman. Under his wise guidance the AEWG had undertaken a considerable volume of work on the Committee's behalf.

8.3.2 Future surveys
The Committee did not have sufficient time available to fully review the designs specified in SC/48/NA4 relative to the requirements specified under Item 7.1 and Annex K, particularly as the discussions under Item 7.1 were not completed until the penultimate day. However, the Committee did provide general advice and comments about the design presented.

SC/48/NA4 presented an overview of a proposed six year sighting survey programme for northeastern Atlantic minke whales. The following issues relative to survey design and conduct were discussed in the context of previous northeast Atlantic minke whale surveys: (1) data quality variation between the 11 vessels during the 1995 large scale survey; (2) the fact that a series of smaller scale surveys would enable better control of training and selection of qualified personnel and would spread survey costs over years; (3) the importance of observer training and testing; (4) expansion of the survey region to include the entire CM block; (5) international participation; (6) annual reports to and reviews by the Committee on the series of surveys; and (7) further collection of dive data in 1996. Two different designs were considered in SC/48/NA4 using a series of small scale surveys across 6 years which would in total provide slightly more sampling effort than the 1995 survey: (1) to cover the entire survey region with 2 vessel months each year; and (2) to cover a different sub-area more intensely with 2 vessel months each year for 5 years and an additional year reserved to cover the area that had had the poorest coverage due to unfavourable sighting conditions. SC/48/NA4 had recommended the latter.

During discussions, the cost and problems of managing a large scale survey were recognised as very practical constraints. However, concerns were raised about problems that might be encountered interpreting survey results that were obtained from different areas in different years in light of the distributional shifts observed between 1989 and 1995. The Committee recommends further consideration be given to the survey design in light of the following ideas: (1) use a similar design as in 1995 with a higher priority and greater resources spent on training and co-ordination, if adequate funding was made available; (2) find a suitable compromise between the suggested 2 designs such as doubling the size of the areas with less effort per year but sample for 2 years in some rotation; (3) evaluate designs by examining observed spatial variability in the survey data and catch statistics; (4) evaluate whether the current block stratification is appropriate; and (5) use a different randomly chosen set (instead of a fixed set) of tracklines for each survey of the same area.

In addition, the following suggestions (in no particular order) were made relative to the survey protocol and data collection: (1) to the extent feasible, use the same vessels and observers each year; (2) consider rotation schemes for observers which maintain team integrity (i.e. pairs of observers remain constant); (3) avoid vessels with masts in front of the observation platforms; (4) develop standard methods for collection of whale angle and methods for measuring its bias and variance; (5) collection of dive data should be widened relative to geographic areas and time of day; (6) plans for collection of dive data should be considered as an integral part of the overall survey plan and the relative resources and effort expended between surveys and collection of dive data should be considered; (7) annual changes in ice extent in the northern blocks should be considered relative to the survey design and area calculations; (8) double platform effort should be used exclusively and all observers should track a whale after initial sighting as specified in the protocol; and (9) conduct of distance experiments should be re-evaluated on the basis of the 1995 experiment.

The Committee encourages individual members to submit comments and suggestions relative to the design and conduct of this particular survey to the Institute of Marine Research in Bergen, and in particular scientists who participated in the 1995 survey and in the review of the analysis.

_