(from "Report of the Scientific Committee", the version distributed at 2002 meeting)

Comparing these estimates to other years in which parts of these areas were surveyed was difficult due to different longitudinal coverage. The Committee agrees that it is important to assess the longitudinal consistency of density estimates between surveys, and to gauge the scale at which different densities could expected to be found, i.e. would a high density area seen one year on a longitudinal scale of 30° (say) be expected to be detected in another year, on the same scale, or it is more appropriate to compare densities on a smaller or larger longitudinal scale?

It was noted that there appeared to be a difference in the estimated detection functions in IO mode between the Shonan Maru (SM1) and the Shonan Maru No. 2 (SM2). In principle, there should be little difference between these vessels since they have the same design and engine. It was not possible to determine the cause of this difference, nor indeed whether the difference should be of concern at all, given the fairly low sample sizes, different observers and different conditions encountered during the survey imparting a composite effect on the fitted detection function. Some members suggested it would be useful to look at the noise profiles of the two vessels from acoustic tests at different operational speeds to investigate whether the differences might be due to more vibration on the SM2 at typical survey speed (11.5 knots). This is of particular importance as it could provide evidence of responsive movement being the cause of the differences between vessels seen in the analyses.

10.2.2 Updated estimates by area
Using the estimates reported under 10.2.1, the third circumpolar series of surveys now encircles the Antarctic, although the surveys in Area V do not extend completely from the ice-edge northwards to 60°S. The percentage of ice-free area between 60°S and the ice edge covered by CPI, CPII and CPIII is 63.1%, 79.5% and 91.1% respectively. The updated estimate of abundance for CPIII is 312,000 (CV=0.086). Abundance estimates for comparable areas indicates the CPIII estimate is now 46% of that in CPII. This is statistically significant at the 5% level.

10.2.3 Inter-year comparisons and trend
10.2.3.1 METHODOLOGICAL ISSUES THAT MAY INFLUENCE ABUNDANCE ESTIMATES AND THEIR TRENDS
10.2.3.1.1 EFFECTIVE SEARCH HALF-WIDTH AND SCHOOL SIZE ESTIMATION
The effects of different pooling strategies on the estimation of abundance of Antarctic minke whales, using data from the CPII and CPIII were explored in several papers, and details can be found in Annex G. SC/54/IA13 found that whilst stratum had a significant effect on both perpendicular sighting distance and school size, there were no significant differences between the two primary survey vessels. SC/54/IA15 reported revised minke whale abundance estimates for the 1998/99 circumpolar cruise in Area IV, by adopting the pooling strategy suggested in SC/54/IA13. The revised estimate increased by 47% to 10,470 (CV=0.35). SC/54/IA5 also examined the effect on the abundance estimate for the 1998/99 survey of different pooling strategies. The changes in the abundance estimates from the estimates of a standard analysis, which was presented in SC/53/IA3, ranged from 0% to 23% in closing mode and from -3% to 18% in IO mode. Although SC/54/IA15 and SC/54/IA5 analysed the same data, the abundance estimates were substantially different because of the different ways in which mean school size was estimated. SC/54/IA32 presented Akaike Information Criterion (AIC) estimates for different pooling options for the IDCR/SOWER surveys. Using the authors・preferred north vs south strategy only increased the ratio of abundances from CPIII/CPII by a factor of 1.04 compared to the estimates using the standard method.

The purpose of estimating the Closing:IO mode density ratio (R) is to correct for bias caused by surveying in Closing mode. SC/54/IA26 investigated possible changes in R for minke whales on the IDCR/SOWER surveys with: (i) time; (ii) the inclusion of like-minke sightings; and (iii) whale density. It was shown that estimates of (average) R for minke whales (which ignore any whale density effect) were generally identical for CPII and CPIII, but lower (though not significantly so) for CPIII when like-minke sightings were also taken into account. For pragmatic reasons, the authors suggested that adopting different R estimates for specific time periods (e.g. CPII and CPIII) avoids the problem of continual updating of the historic abundance estimates as more data by which to calculate this ratio become available. The Committee concurs.

The above papers came to the conclusion that pooling by stratum fits the data better than pooling by vessel. Pooling by stratum may be preferred a priori on biological grounds, especially if the two survey vessels are operationally similar.

The Committee concluded that whilst different analysis options may produce different point estimates, it would be unlikely that these differences would generally be substantial, even if they happened to be so for a particular dataset. It endorses the view that decisions regarding general methodology should be based upon data from more than one survey, and pooling should not solely be based upon statistical criteria, but also on biological or environmental evidence, when available. Where appropriate, this strategy should be adopted even if it leads to higher variance estimates.

10.2.3.1.2 POTENTIAL COVARIATES
SC/54/IA17 investigated the effects of sighting conditions (school size, sighting cue, latitude and sea state) on Antarctic minke whale abundance estimation parameters (effective search half-width, sighting forward distance and mean school size). As school size decreased, the effective search half-width and the sighting forward distance decreased. Cues from schools of size 1 or 2 were usually 'body', which is generally considered more difficult to find in high sea states as compared to 'blow' cues. The proportion of schools of size 1 or 2 increased with decreasing latitude. Small school sizes and bad weather conditions prevailed in the northern part of the survey area. Because the survey area was extended northwards in the third circumpolar survey, the effects of small school size and bad weather conditions were substantial.

It was suggested that a 'synthetic sightability' variable that is some, perhaps non-linear, combination of other variables might be a better covariate than any single covariate. It was agreed that the IDCR/SOWER data should be used to determine which covariate or group of covariates provides the best estimate of optimal sighting conditions and should be included in any method to estimate abundance. The Committee recommends that a synthetic sightability variable be investigated.

SC/54/IA16 explored the effect of different factors on the probability of a duplicate sighting of several species, and investigated any change in these probabilities between CPII and CPIII. The effect of school size on the probability of duplicate sightings was statistically significant for minke, humpback and southern bottlenose whales. For minke whales, the probability of a duplicate sighting for schools of size 2 was 1.56 times larger than for solitary animals (95% CI=1.21-1.78), and for schools of 3 or more animals, it was 1.84 times larger (95% CI=1.68-1.93). This provides strong evidence that g(0) depends on school size. For minke whales, the duplicate sighting probability was about 20% smaller in CPIII than CPII, with this effect being more appreciable for the northern strata. This may be because solitary whales constitute a greater proportion of the schools in the northern regions, and survey effort was extended northwards in CPIII. A similar reduction was found for humpback whales; this suggests the possibility that humpback and minke whales have similar habitat preferences. In addition, it was shown that as visibility improves, the probability of a duplicate sighting tends to increase significantly. There is also a similar decreasing non-significant trend with sea state and a weaker trend with sightability, for which variation may not change in a systematic manner. In addition, when inexperienced observers (those with experience of less than 5 surveys) were observing from the IO platform, the duplicate sighting probability was 20% less than when observers with at least 5 years experience were there. This effect was not statistically significant, but sample sizes were quite small.

Whilst recognising that differential observer experience is a plausible explanation for changes in the duplicate sighting probabilities, the Committee noted that the effect is difficult to tease out from the data. The Committee concluded that the hypothesised observer experience effect contributes less to differences in g(0) than does school size, but should be considered further.

The Committee cautioned that prior to the completion of CPIII, it is difficult to interpret some of these results because they may be strongly influenced by a particular survey area. For example, Area V East, which has not yet been surveyed, contributes a large proportion of the sightings.

The Committee agrees that these papers provide evidence supporting the assertion that g(0) could be less than 1, at least in some of the circumstances investigated, and covariates that should be considered in future analyses include sighting cue, observer experience, survey mode, sightability (synthetic or that recorded in the field), Beaufort sea state, school size, distance from ice and stratum (North or South).

10.2.3.1.3 METHODS TO INCORPORATE COVARIATES AND ESTIMATE G(0)
SC/54/IA29 explored the use of the covariate estimation framework available in the program Distance 4 (Thomas et al., 2001) for improving detection function estimation in the computation of minke whale abundance estimates for the IDCR/SOWER surveys. Estimation from pooled data including environmental and other covariates was attempted for the CPII. Annex G, Appendix 4 examined data from CPII, and showed that including covariates when estimating the search half-width improved the fit in all cases except when weather was included. In all cases, the AIC value was lower for the hazard-rate model than for the half-normal. A correlation matrix was obtained between all the covariates; although most factors were significant, correlations > 0.2 were only obtained for sightability with weather and wind speed, and confirmed sightings with closing mode.

SC/54/IA23 investigated generalisations of size bias regression methods for estimating school size from line transect data.

The methods were illustrated using data from the 1999/2000 and 2000/01 circumpolar surveys. There appeared to be little gained from applying GLMs (generalised linear models) or GAMs (generalised additive models) in the perpendicular distance based model. However, the inclusion of a regional effect in the detection function did provide improvement over the perpendicular distance only approach, yielding increased precision, and possibly more reliable estimates. Including unconfirmed sightings was not successful and possible reasons for this were suggested.

The Committee agrees that this method is a promising way to extend the standard method that could produce less biased abundance estimates and could be relatively easily incorporated into DESS. However, it requires evaluation with respect to its robustness and accuracy.

The 'Big Beautiful Model' estimates abundance by accounting for variations in g(0) caused by size bias, changes in school size distribution and changes in sighting conditions (SC/54/IA21). The method provides estimates of g(0) independent of double-platform data and so avoids some of the complexities and sensitivities associated with double platform estimation. The Committee welcomed this new approach and recommends its further development (see Item 21).

Spatial modelling potentially provides a better insight into true uncertainty than stratified estimation, since the latter tends to yield fluctuating estimates of variance. However, in practice, developing well-behaved spatial estimators can be difficult. Five linked issues requiring further work were raised:

(1) estimating variance (dealing with clustered sightings, and incorporating uncertainty from earlier steps in the analysis, e.g. in effective search half-width);

(2) choosing the scale of smoothing, when sightings are clustered on small spatial scales;

(3) allowing for spatial variation in the scale of smoothing, perhaps linked to localised environmental factors;

(4) changing the form of the underlying smoother, e.g. to accommodate non-smooth changes in density;

(5) restricting the fitted density surfaces to realistic values in regions of rectangular strata that are far from the zigzag track.

The Committee encourages further intersessional work on these issues.

SC/54/IA10 introduced a relatively simple hazard probability model using perpendicular and forward distance data for double-platform line transect surveys in which the independent observers have the same visual searching area. This model can provide an abundance estimate of diving animals without the assumption of g(0) = 1, and takes account of unmodelled heterogeneity derived from the whales・ surfacing behaviour. The proposed model may be viewed as an integration of the Skaug and Schweder (1999) and Cooke (2001) approaches, but is more general and tractable. It may be easily extended to a model with only perpendicular distance data or to include data from 'incompletely independent' observers.

A simulation study and an application to actual IDCR/SOWER sightings data suggested that the proposed method performed well and could be useful for the future analyses. In particular, simulation tests indicated that using other available data, such as incompletely independent observer sightings data and mean surfacing rate estimated from external data, can improve the performance of the model. The ability to utilise the former is important because such data can easily and economically be collected. The provisional results from applying the model to actual sightings data when g(0) is assumed to be 1 and when it is not, yielded considerably different abundance and trend estimates.

In discussion of SC/54/IA10, it was noted that the results presented were considered preliminary by the authors, the main aim of the paper being to introduce the methods. Key points raised were: (i) the improvement in precision obtained when the IO Platform data were added to the model was likely to be an artefact of increasing sample size (since a common detection function was assumed across all platforms); (ii) school size was not included in the current simulation (although this is planned ・with simultaneous surfacing of all animals in a school); it is however presently incorporated as a covariate in the model; (iii) the reversal of the trend in abundance estimates seen when comparing those from the proposed methods and those from Branch and Butterworth (2001b) was not fully understood ・in particular, neither the difference in the methods used to estimate school size, nor g(0) estimation were thought to be responsible ・it seemed most likely that the reversal was a result of different stratification options used in the analyses; and (iv) the apparent gain in precision using the proposed methods compared to the standard methods was due to assuming independence between sightings, which is invalid in the presence of clustering.

SC/54/IA30 examined the discrepancy in whale density between the true value and an estimate assuming that g(0)=1, using a simple mathematical equation. Whale density was defined as the product of the school density estimate and mean school size. Estimates of g(0) from the 1989/90 and 1993/94 IDCR/SOWER surveys (SC/54/IA10), stratified by school size class (1, 2, 3+), were assumed to apply for all years in the 2^nd and 3^rd circumpolar surveys respectively. This approach was taken for illustrative purposes only; clearly it is preferable to estimate g(0) separately for each survey. The discrepancy between density estimates which assumed that g(0)=1, and those which used g(0) stratified by school size class, ranged from -18% to -39%.

SC/54/IA1 reported progress on (i) estimating strip width without the assumption g(0)=1; and (ii) estimation of additional variance. Estimation of strip width was based on the method described and tested in Cooke (2001), which the Committee had recommended last year (IWC, 2002k, p.199). The two tasks were addressed together using the integrated modelling framework outlined in Cooke and Leaper, 1998. The method for treating the three platforms on IDCR/SOWER cruises was similar to that used in SC/54/IA10. The data collected in IO mode are necessary for the estimation of g(0), but school sizes estimated in this mode appear to be biased low, hence a method was developed for allowing for school size bias in the estimation.

The Committee noted that this method was promising because it would probably be able to be used to extrapolate to unsurveyed areas, simultaneously accounting for all the potential covariates that have been discussed. Technical questions included: how was the changing ice-edge in the non-symmetrical Antarctic study area incorporated into the Fourier series; was there a conflict between estimating the overdispersion term and automatically selecting the terms for the smoothing function; what should the most appropriate scale for smoothing be; and should that scale vary depending on the amount of variability in a region? After discussion, it was concluded that further work should resolve these technical issues.

The Committee welcomes all the new methods, encourages further development and simulation testing of these (and other) methods and looks forward to seeing the results from applying them to the IDCR/SOWER data.

The Committee agrees that g(0) differs by school size, but the difficulty remains as to how to address this in analyses. It is complicated by the fact that in IO mode (which has traditionally been the mode used to estimate g(0)), there is negative bias in the size of recorded schools - a large proportion of which are unconfirmed. It was noted that data from Closing mode, for which the recorded school sizes are thought to be unbiased, could be used to calibrate the recorded IO mode school sizes.

The Committee agrees that the performance and robustness of any new method should be evaluated. Some new simulation datasets will have to be created for this purpose, since the original datasets were primarily designed for testing North Atlantic minke whale analysis methods. Some of these changes will be minor: a detection function to simulate upper or front bridge observers' sightings, for example. Other changes will require more thought e.g. determining an appropriate way to generate Antarctic minke whale school size distributions and surfacing patterns (as well as detectability based on school sizes and other covariates). To further complicate matters, the fact that recorded school sizes in IO mode tend to be negatively biased estimates of true school size was raised, and its effects should be considered and if possible, accounted for. A more complete description of factors to be included in the simulated datasets was developed, and a timetable was established to ensure results from the simulations could be presented next year.

The Committee recommends that simulated datasets be used to evaluate any method that might be used to analyse the IDCR/SOWER data, including the standard method. In addition, the Committee recommends that further development of the standard method is, at this time, lower priority than evaluating the performance of the new methods. The Committee also agrees that when all the circumpolar data are available, these data are analysed by the standard method, and by any new methods found to be robust to the heterogeneities introduced into the simulated data. The Committee also recommends that results of methods when applied to both simulated and actual IDCR/SOWER data be presented at next year's meeting.

10.2.3.2 TIMING OF THE SURVEYS
The JSV, JARPA and IDCR/SOWER sightings data in Areas III, IV and V were used to examine the migration pattern of Antarctic minke whales (SC/54/IA12). The analysis included JSV data from 1971/72 to 2000/01 and JARPA and IDCR/SOWER data from 1987/88 to 2000/01. Five factors were considered as covariates in a GLM: latitude, longitude, time, season and Beaufort wind force. AIC was used to select the final model. It was concluded that in the 1970s, the peak migration was in late January; in the 1980s, it was bimodal, in mid- January and mid-February; and in the 1990s, it was also bimodal, in late December and mid-February. CPUE data from the 1970s showed a consistent peak of migration in January (Shimadzu, 1980), which is somewhat different than the results of this analysis.

A potential problem with the analysis is the differing nature of the three data sources. It was suggested that including a covariate for JSV vs. IDCR vs. JARPA might make the analysis more robust. Nevertheless, because of these data source problems, any results would necessarily be rough.

A GLM analysis of daily minke whale density estimates from IDCR/SOWER surveys from 1978/79 to 1997/98 was used to determine migration patterns. The effects of survey mode, latitude, year and Management Area were included. A 20% drop in abundance was seen only after mid-February, although a larger real drop may be masked by a possibly confounding year effect. Inclusion of a year-date interaction term might help with this.

SC/54/IA7 noted that this shift in timing of migration was in the December to January period when the annual decrease in sea ice extent is most rapid and divergent. The author indicated that inter-annual climatological oscillations that affect variability in ice coverage may have indirectly affected the timing of the minke whale migration through changes due to, for example, prey availability or changes in the relative amounts of open water, ice edge, and pack ice habitats. A more complete discussion of climatological indices and their relationship to pack ice and minke whale abundance and distribution is given under Item 10.2.3.4 and in SC/54/IA7.

The reason for the investigation into the timing of the minke whale migration was to determine if the recent change in timing of the IDCR/SOWER surveys affects the estimated abundance (IWC, 2002k, p.199). Given the new results presented, the Committee agrees that changes in survey timing probably had only a small impact on abundance estimates.

10.2.3.3 USING JSV DATA TO EXTRAPOLATE TO UNSURVEYED REGIONS
SC/54/IA12 extrapolated densities from the IDCR/SOWER (south of 60°S) area to northern areas covered by the JSV cruises (30°S-60°S). The IDCR/SOWER data were from CPII, 1985/86 to 1990/91; the JSV data were collected from 1971/72 to 2001/02. Assuming no difference in conditions, the extrapolation rate was calculated as the ratio of the index abundance between the northern area and the area south of 60°S. This ranged from 5% to 18%. Using estimates from CPII, the population in the northern area is estimated to be about 320,000 in Areas III, IV and V. This may be slightly positively biased, because the sightings in the northern area may include some dwarf minke whales.

10.2.3.4 COMPARING JARPA DATA TO IDCR/SOWER DATA
SC/54/IA14 compared trends in Antarctic minke whale abundance between JARPA and IDCR/SOWER surveys in Areas IV and V. It reported that the trends in abundance estimates from the JARPA surveys (Hakamada et al., 2001) in both Area IV (6 surveys examined for the period 1989/90 to 1999/2000) and Area V (6 surveys examined for the period 1990/91 to 2000/01) were not significantly different from zero. In contrast, abundance estimates from IDCR/SOWER surveys showed a negative trend in these Areas. The IDCR/SOWER surveys started several years earlier (1978/79 for Area IV; 1980/81 for Area V) but were conducted less frequently. It was suggested that JARPA surveys may better reflect the trend in abundance for minke whales for the following reasons: (i) the estimates of abundance from the IDCR/SOWER surveys in CPIII are negatively biased estimates of true minke whale abundance and thus the downward trend is exaggerated; (ii) there have been more JARPA surveys in these Areas than IDCR/SOWER surveys; and (iii) there is more consistency among JARPA surveys than among IDCR/SOWER surveys in terms of the survey method, and geographical and temporal coverage of the surveys.

Childerhouse commented that the first five JARPA surveys in Area IV indicated a downward trend similar to the IDCR/SOWER surveys, and that the nonsignificant result quoted above was a reflection of a particularly high value for the sixth survey within the series. In response, it was pointed out that the low results for the fifth survey (1997/98) were particularly influential in suggesting an initial negative trend. There were reasons related to different ice conditions and a smaller proportion of mature females in the survey region in 1997/98 which strongly suggested that the 1997/98 survey had covered a smaller than usual proportion of the overall population.

The Committee considered that the temporal incomparability and survey design differences between the IDCR/SOWER and JARPA surveys rendered direct comparisons between their trend estimates inappropriate. However, the greater consistency of the JARPA surveys suggests that they may provide information on additional variance.

10.2.3.5 ANIMALS WITHIN THE PACK ICE
The survey ships used to collect IDCR/SOWER data cannot survey within the pack ice at normal survey speed (11.5 knots), so the ice-edge defines the southern border of the survey area. Although it is known that minke whales are within the pack ice during the surveys, it is not known what the order of magnitude nor inter-annual variability of the abundance in the pack ice is. Attempts were made to quantify these unknowns as detailed in Annex G.

One attempt to directly quantify the density of minke whales in the pack ice involved the analysis of data collected on the annual Southern Ocean Cetacean Ecosystem Program (SOCEP) and Australian APIS cetacean survey. This shipboard survey was conducted from 3-28 December 1999 within the pack ice starting at (64°23S 150°50E) and travelling westwards through the pack ice to Davis Base (68°34S 77°50E). The target species were seals, not cetaceans. In addition to the usual line transect data, they collected an extensive description of patterns of ice distribution. The potential to make inferences about densities in the pack ice is limited due to timing of the surveys and potential violations of line transect assumptions. The Committee recommends that methods to obtain unbiased, precise abundance estimates using data collected in the pack ice be developed. Progress on this matter was being developed in the Standing Working Group on Environmental Concerns (Annex J, Appendix 3).

In the absence of Antarctic minke whale abundance estimates within the pack ice for the same times and Areas as the IDCR/SOWER or JARPA surveys, other more indirect ways were explored (see Annex G). Two sources of data were used: JARPA data and previously published estimates of animals in the pack ice.

10.2.3.5.1 JARPA DATA
Based on the observation that mature female minke whales tend to be found further south, near the ice-edge, the sexual maturity rates and ice coverage from years with low abundance estimates in open water were compared to those from years with high open water abundance estimates. A coarse estimate of the number of whales in the pack ice was calculated by multiplying the observed proportions of mature females from the JARPA samples by the abundance in open water as estimated from the JARPA data.

On the 2001/02 JARPA cruise, a substantial number of minke whales were sighted in the SE stratum and in Prydz Bay, and the sexual maturity rate of sampled females was 74%. In contrast, in the SE stratum in the 1997/98 JARPA survey there were few minke whale sightings, and the sexual maturity rate of sampled females was only 5%. According to the NIC satellite information, there were many ice-free areas south of the ice-edge in 1997/98, where research vessels could not enter. In fact, in Area IV in 1997/98, the ice coverage was higher than in other years in the JARPA series (Annex G, Appendix 6). Using these data, it was estimated that the number of mature females south of Area IV in the pack ice in 1997/98 was 2,876 (Annex G, Appendix 7). It was noted that this figure was probably an underestimate of the total number of minke whales in the pack ice, as the estimate does not include males or immature females.

The Committee spent some time discussing the possible whereabouts of minke whales during years when the open water abundance estimates were low and the various hypotheses included: into the pack ice; north of 60°S; and east or west.

SC/54/IA18 presented the same pattern between ice coverage and abundance from the 1988/89 (CPII) and 1998/99 (CPIII) IDCR/SOWER surveys in Area IV. The 1998/99 season was colder and the estimated abundance of minke whales was very low; it is not clear where they were. On either side of the study area, the Ross Sea and Prydz Bay were both closed in January 1998/99, although they are well known to be high-density areas of Antarctic minke whales. It is possible that the whales scattered widely in the northern stratum, although the abundance in the northern stratum was lower than the southern stratum (Burt and Stahl, 2001). Therefore, the authors concluded that many minke whales could have gone into the pack ice region in 1998/99.

It was further noted that the proportion of mature males in Area IV (greater than ca 0.6) always exceeds that observed in Area V (ca 0.5) (Annex G, Appendix 8). This is consistent with the differences in ice conditions in Area IV and V. There was also substantial variability in the proportion of mature males by year (i.e. a difference of about 0.25 between the maximum and minimum proportions). In Area IV this appears to be inversely proportional to the JARPA abundance estimates, while in Area V there is little or no relationship. The latter, despite similar levels of variation in the sex ratios in Area IV and V, suggests that either the change in ice cover is not the sole explanation for the observed data in Area IV or that different factors are the source of the variation in Area V. The estimates of the number of mature and immature animals in Areas IV and V, by sex and year, lead to the conclusion that the pattern of abundance estimates cannot be explained simply by increasing proportions of females within the ice, and that if increase in ice is the explanation for the declines, then both males and females must be entering into the ice zone.

During further discussion, it was noted that the analyses in Annex G, Appendix 8 estimated abundance by using Area-specific proportions of mature females. It was agreed that using estimates of abundance by stratum within an Area would provide further information for the Committee to consider, should such data be available.

10.2.3.5.1 PREVIOUSLY PUBLISHED DATA
SC/54/IA19 presented estimates of the abundance of minke whales in unsurveyed regions within the pack ice for three scenarios of the proportion of whales in the pack ice. The closing mode density of Antarctic minke whales in open water was obtained from the southern stratum of each sector in the IDCR/SOWER data from 1978/79 to 1997/98 (Branch and Butterworth, 2001b). This density was applied to areas of open waters inside the pack ice using data from Kasamatsu et al. (2000). Scenario 1 was based on 1979 data from Naito (1982); scenario 2 combined the 1979 and 1981 data in Naito (1982); and scenario 3 used data from Ainley (1985). The estimated abundance (for each scenario) within the pack ice in CPII was 18%, 102% and 51% of that in the ice-free area; in CPIII the percentages were 28%, 159% and 72% respectively. The area within the pack ice region in CPII was 123% of that in CPIII. Estimated total abundances in CPII were 100%, 99% and 110% of that in CPIII. This suggests that there could be a substantial difference between CPII and CPIII.

The Committee noted that there are very few surveys on densities of minke whales within the pack ice, and that this paper was useful in presenting calculations of the possible proportion of minke whales in the pack ice. However, certain incompatibilities (the density estimates in the pack ice were not from the same Areas as the open water density estimates; the timing of surveys within the pack ice did not overlap the entire time period of the IDCR/SOWER open water surveys) made interpretation of the results difficult.

SC/54/IA7 noted that the distribution of minke whales may be affected by interannual climatological variability that affects the extent and coverage of the sea ice. To try to examine the mechanism behind the inverse correlation between open water abundances and ice coverage, the Committee attempted to link the patterns of proportions of mature females with patterns of pack ice. A list of environmental covariates that might be useful in quantitatively investigating this correlation include: (a) Southern Oscillation Index (Kwok and Comiso, 2002); (b) SAO (Reuter, 1936; Van Loon, 1967); (c) Antarctic Oscillation Index (AOI) (Jong and Wang, 1999); (d) El Niño-Southern Oscillation (ENSO); (e) sea ice-edge e.g. monthly mean sea ice-edge, equator-most position of the 30% isopleth of ice concentration in each degree of longitude (Yuan and Martinson, 2000); (f) sea ice edge anomaly - monthly, removes seasonal cycle, contains interannual and longer-term variability as well as linear trends (Yuan and Martinson, 2000); (g) ice concentration ・25 km x 25 km grids; (h) ice area; (i) sea ice motion; and (j) ice thickness.

This information introduces some hypotheses for changes in open water abundance of minke whales; however, more research and data from the pack ice is needed to test the hypotheses. In the meantime, the Committee agrees that correlation analyses between open water abundance estimates and environmental factors may help to develop functional/biological connections between climate, ice, productivity (e.g. chlorophyll) and Antarctic minke whale abundance.

One attempt to develop such a functional link between the patterns of proportions of mature females with patterns of ice conditions is given below. In Area IV, the pattern of ice melt is characterised by the southward recession of the ice edge and (based on satellite predicted estimates) the development of ice-free areas inside the pack ice, near the Antarctic coast. In a 'normal' year, the main ice edge recedes southward, and especially at longitudes corresponding to the ice-free areas inside the pack ice. During the survey period, some, but not all, of the ice-free areas inside the pack ice become contiguous with the open sea to the north, and characteristically they form bays in the main ice edge. The pattern of ice melt was considered 'normal' in the following seasons: 1989/90, 1991/92, 1993/94 and 1999/2000. The proportions of mature females in the catch data from the southern stratum for these years were 31.1%, 39.1%, 27.6% and 42.8%, respectively. The pattern of ice melt was 'abnormal' in 1997/98 when the ice edge was farther north than usual. Based on satellite predicted estimates extensive ice-free areas were present inside the pack ice this year. The percentage of mature females in the catch was substantially lower (7.1%).

These analyses and observations support the hypothesis that the 'missing' mature females are possibly distributed in the pack ice, but do not preclude other hypotheses, such as longitudinal movement out of the Areas. More work is needed to fully explore these ideas. Despite the difficulties in interpreting the variety of information, the Committee agrees that there could be large numbers of minke whales within the pack ice, quite possible some tens of percent of the open water IDCR/SOWER abundance estimates, at least for certain areas. However, potential biases and paucity of surveys in the pack ice make it difficult to be more definitive. The Committee recommends the ice edge information in the analyses of SC/53/IA15 be used to give a simple annual index of the week or month of greatest ice cover and the percent ice coverage for the times and areas of the IDCR/SOWER and JARPA surveys. In addition, it recommends that the IDCR/SOWER and JARPA data (which covered a longer time series) be used to investigate if there is a correlation between open water abundance and ice coverage.

10.2.3.6 WORK PLAN
In the light of current uncertainty about the density and distribution of minke whales in the pack ice, and to what extent this might affect interpretations about trend from the IDCR/SOWER and JARPA surveys, the Committee agrees that little further progress can be made on this issue without new data and analyses. It therefore sought to identify potential existing sources of data, and to make recommendations on what data would be useful to collect from vessels that operate in the pack ice.

It was noted that the most useful data were those from dedicated cetacean observer platforms, particularly those platforms on which the observations were made both in open water and in the pack ice, such as the 1995/96 BROKE survey (Nicol, 2000) and SO-GLOBEC cruises with cetacean observers onboard (SC/54/E12). Other potential data sources include: APIS data; data collected during transits to and from Antarctic bases; surveys in Area III conducted by Australian SOCEP; and data collected during ice navigation modes during the IDCR/SOWER and JARPA surveys. Concerns were expressed regarding the APIS aerial survey data, given that these surveys were primarily aimed at surveying seals (hauled out on to the ice) and hence would not have consistent search effort for whales. Some of these aerial surveys also had dedicated cetacean observers onboard so cross-validation with corresponding dedicated shipboard cetacean observations data would be useful.

Different views were expressed regarding the applicability of conventional line transect estimation methods to data from within the pack ice. However, it was agreed that sighting distance and angle data may have a role to play in any future analyses, and that these data should be collected where possible. The Committee also recommends that consistent and regular descriptions of the ice characteristics (see Annex J, Appendix 6) and other factors affecting the detectability of minke whales be collected. Furthermore, the Committee recommends that data corresponding to where the IDCR/SOWER 訴ce edge・would have been located should be recorded where possible. It was also suggested that satellite telemetry should be used to provide information on the movement of minke whales between pack ice and open water (the IDCR/SOWER survey area).

The Committee also recommends the following items to facilitate progress on identifying available data sources and to encourage the collection of new data within the pack-ice region:

(1) the Secretariat should make an official request to the APIS coordinators inquiring about availability and access to any cetacean data that their member countries may have collected within and outside the pack ice;

(2) relevant members of the Scientific Committee should approach individuals whom they know to have been involved with the APIS surveys with the same enquiry as in (1) above (Thiele and Gales will coordinate this effort);

(3) requests should be made to countries that use ice breakers in the Antarctic to conduct dedicated cetacean observations from their vessels (e.g. Australia, France, Germany, Japan, South Africa, the UK and the USA). Thiele will coordinate this effort, and ensure that data are collected in a standardised format.

10.2.3.7 TRENDS IN ABUNDANCE USING POPULATION DYNAMIC MODELS
In SC/54/IA25, the ADAPT VPA methodology of Butterworth et al. (1999) was refined and/or extended with catch-at-age and abundance information (from both the IDCR/SOWER and JARPA programmes) in Areas IV and V. The results suggest statistically significant increases in recruitment until a peak in the late 1960s, followed by a drop and then stabilisation for more recent years. Total (1+) population trends over the past two decades show slight non-significant declines of about 1% per annum. The Area IV population is, however, estimated to have declined over the 1970s. Estimates of M for the two Areas range from 0.046-0.070 per year, and are statistically compatible. These results are consistent with supercompensation (with the population having expanded beyond its current carrying capacity) and subsequent reductions in that carrying capacity (the population first having rapidly increased towards an increased carrying capacity earlier in the last century).

The authors noted that the suggestion of only a slight decline in abundance (about 1% per year, or 20% over the timeframe of the IDCR/SOWER surveys) initially appears to contradict the results from the IDCR/SOWER surveys that show an appreciable decrease in abundance estimates between CPII and CPIII. Further examination of the abundance estimates showed that estimates from Areas IV and V in CPIII was 58% of the estimates from CPII. This supports the results from this model. However, the estimates from the remaining Areas I, II, III and VI in CPIII was 36% of that in CPII, a much greater reduction. The difference in circumpolar estimates between CPII and CPIII is therefore not completely addressed by the ADAPT VPA model. However, the IDCR/SOWER survey estimates have wide confidence limits, and are not statistically incompatible with the VPA results.

In discussion, several members argued that other plausible scenarios existed to explain these results. In particular, the analyses presented implicitly require some subjective decisions to be made, fixing some parameters to enable estimation of others. For example, some concerns were expressed about the plausibility of the extent of the increase in population until the 1970s; such an increase would not follow if the natural mortality rate (M) was as high as 0.1. Opinions differed on how realistic a rate of 0.1 for M was, but Butterworth pointed out that it exceeded the confidence limits of the estimates of M provided by the VPA.

Concerns were raised about the selectivity constraints in the VPA model and lack of fit to the age distribution in the 'plus' group. Although some concerns about the selectivity constraints had been addressed in Butterworth et al. (1999), further consideration of fit to the age data for older animals was needed in Area IV. Another concern raised was the need to consider the implication of uncertainty about stock structure on the VPA results.

Possible reasons for the estimated decrease in abundance in the late 1960s were discussed. It was noted that the large harvests in Area IV probably contributed. However, it is unlikely that these trends could be explained by supercompensation alone; the postulated decrease in carrying capacity was also essential to explain the observed patterns (unless the current estimates of abundance from the IDCR/SOWER surveys were very negatively biased). Other suggested explanations for the observed changes in population size include an inertial dynamics model, and changes in carrying capacity that were climatologically driven.

It was pointed out that the stock recruitment relationships estimated by the VPA for minke whales in Areas IV and V (particularly the recent large drops in per capita recruitment) do not appear to be supported by some of the biological evidence available from commercial catch and JARPA data. If recruitment estimates from the VPA are truly representative, and pregnancy rate and age at first parturition are without trend, then some other mechanism(s) must be postulated to account for a variable stock/recruitment relation. One possible explanation is a decrease in juvenile survival rate (lactation having higher demands on females than pregnancy). It was noted that an estimate of body condition, perhaps derived from data on whale length, weight and blubber thickness, as well as apparent pregnancy rate and age at first ovulation, might be a useful diagnostic for future results from VPA or other integrated models. The historical catch records also include data (e.g. oil yield, blubber thickness) that might be useful in determining an appropriate condition factor.

Ohsumi noted that the 1997 JARPA Review meeting had confirmed that blubber thickness had decreased since the JARPA programme began in 1987 (IWC, 1998b). He also expressed the view that blue whales and humpback whales were increasing in the Antarctic, and that consequent competition effects might be restricting the carrying capacity of minke whales.

The Committee noted that whilst the combined three-year-three-age catch-at-age data from these Areas is published, the sensitivity and robustness of the VPA model could only be independently investigated if the corresponding data on a one-year-one-age basis were also available. Kato reminded the Committee that the Institute of Cetacean Research, which owns the JARPA data, established data policies during the JARPA Review Meeting that are applicable in this case. Furthermore, it was noted that details on how the catch-at-age matrices are constructed were also necessary, of which some were provided in Butterworth et al. (1999). To address some of the concerns about this VPA analysis, Polacheck agreed to convene an intersessional e-mail group to: (1) request the required summary data following the established data policies; (2) make a list of the concepts that need to be addressed in further analyses (i.e. alter the model structure, incorporate other biological data and stock structure alternatives, and investigate model robustness); and (3) coordinate individuals to pursue analyses to address these concerns.

The Committee recommends that the power of different approaches to detecting a trend be investigated, including a simple regression on abundance estimates, the integrated approach suggested in SC/54/IA1 and the VPA approach of Butterworth et al. (1999).

10.2.4 Other
10.2.4.1 UPDATE ON MTDNA ANALYSES
An update of the mitochondrial DNA (mtDNA) RFLP analysis in Antarctic minke whales from Areas V and VIW was presented (SC/54/IA9). The analysis used all the available samples from these Areas (1988/89-2000/01 JARPA surveys). Samples were divided arbitrarily as follows: Area V Western (130°-165°E); Area V Eastern (165°E-170°W); and Area VI Western (145°W-170°W), with two temporal periods (Early and Late). A total of 2,228 samples was examined in six longitudinal/temporal groups. Following an examination of yearly variation, a hierarchical analysis by AMOVA was conducted for the total samples. Overall, no significant mtDNA heterogeneity was found in Areas V and VIW. Each of the longitudinal/temporal groups in Areas V and VIW differed significantly from an out-group sample from Area IVWE in both Fst and PHIst statistics.

It was suggested that information on any geographical and temporal differences in stock structure in the Antarctic Areas could be incorporated into integrated models such as the VPA. Pastene informed the Committee that a new study covering Areas III-VI is underway using nuclear DNA in addition to mtDNA. The Committee recommends that attempts be made to collect samples from lower latitudes, recognising that the exact locations of the putative breeding grounds are at present unknown. In this regard, it recommends that the use of satellite telemetry to track whales between the Antarctic and lower latitudes should also be investigated.

10.2.4.2 RESPONSE TO RESOLUTION 2001-7
Resolution 2001-7 'requests the Scientific Committee to provide to the Commission at IWC 54 (IWC, 2002c):

(i) a list of plausible hypotheses that may explain this apparent population decline;

(ii) the possible implications that such a decline in abundance may have for the management of minke whales in the Southern Hemisphere, and for ecologically-related species, in particular other cetaceans, and the state of the Antarctic marine ecosystem.

The Committee noted the estimates of abundance using the 'standard methods' of Branch and Butterworth (2001b) for the third circumpolar set of surveys are appreciably lower than estimates for the second circumpolar set of surveys. Last year, after coarsely quantifying many of the factors affecting abundance estimates, there remained evidence of a decline in abundance estimates from CPII to CPIII, although it was not clear how this reflected any actual changes in minke abundance. This year, many contributions were submitted suggesting refinements to these coarse quantifications (Annex G, Appendix 10), but it is premature to attempt an update of last year's computations, at least until the results of further work scheduled for the 2003 meeting could be reviewed.

Given the wide range of plausible hypotheses identified above, the Committee respectfully informs the Commission that it believes it is premature to comment on the equally wide range of potential management implications. The Committee agrees that the most appropriate time to fully address this issue will be after completing its work on reviewing the IDCR/SOWER abundance estimates.

10.2.5 Plans for completion of the Antarctic minke whale review
Annex G, Appendix 9 details the tasks identified by the Committee to further the review of Antarctic minke whale abundance estimates, together with an indication of priorities for the next year. Noting the need to explain why the estimates of abundance using the standard methods for CPIII are appreciably lower than for CPII, the Committee strongly recommends that substantial progress be made on all high priority tasks by next year's Scientific Committee meeting.

To successfully complete its review of the IDCR/SOWER abundance estimates and trends, and to address Resolution 2001-7, resources are required to complete the last two years of the IDCR/SOWER survey and to develop and test new analytical methods that result in less biased abundance estimates and trends. Financial details of the IDCR/SOWER cruises are discussed in Annex G, Appendix 2; and financial details of the method development and testing are discussed under Item 21.

_