Conservation of Peary caribou based on a recalculation of the 1961 aerial survey

The estimate of 25 845 Peary caribou (Rangifer tarandus pearyi) on the Queen Elizabeth Islands (QEI) in the Canadian High Arctic in summer 1961 is the only nearly range-wide 'benchmark' for the past number of caribou. No variances or confidence intervals were calculated for this estimate and no estimates were calculated for Peary caribou on the three major islands of Ellesmere, Devon, and Axel Heiberg. We reexamined the 1961 raw data by grouping the QEI into five island-complexes ('eco-units') and calculating, for each unit, the estimated number of caribou and the standard error, and the 95% confidence interval of the estimate, using a 'bootstrap' technique with 100 000 replications. Our goal was to provide an ecological basis for evaluating subsequent changes in numbers rather than relying on single-island evaluations. Our bootstrap reanalysis produced an estimate of 28 288 ± 2205 SE with a 95% CI of 20 436—37 031 Peary caribou on the QEI in summer 1961. Substantial differences in density were apparent among the five eco-units, with about a 50-fold difference from 0.01 caribou • km 2 in the Eastern eco-unit to 0.5 caribou • km 2 in the Northwestern eco-unit. The 1961 findings, with our subsequent reexamination, are crucial to any evaluation of trends for the number of Peary caribou on the QEI and the relative importance of individual eco-units for these animals. These findings also allow a more accurate evaluation of the magnitude of the subsequent decline of Peary caribou on the QEI during the last four decades and may help predict future potential levels for caribou in each of the five eco-units.


Introduction
estimated there were 25 845 Peary caribou (Rangifer tarandus pearyi) on the Queen Elizabeth Islands (QEI) in summer 1961.This initial, range-wide, systematic aerial survey of Peary caribou has never been repeated.Tener found that the dis¬ tribution of caribou among the QEI was markedly uneven: based on island sizes he used, 94% (24 363) were on western QEI (WQEI), which makes up 24% of the total island landmass, and the remaining 6% (1482) were on eastern QEI (EQEI), with 76% of the landmass (Fig. 1).
Before 1961, the only information on distribution and numbers of Peary caribou across the QEI came from a few interested geologists in the mid-and late 1950s.After Tener's 1961 survey, only three compos-Rangifer, Special Issue No. 16,2005 ite multi-island caribou surveys were conducted in WQEI with adequate temporal separation (Miller et al., 1977a;Miller, 1987aMiller, , 1987bMiller, , 1988Miller, , 1989;;Gunn & Dragon, 2002).There have been no range-wide sys¬ tematic aerial surveys for Peary caribou on the EQEI or on any major portion of that region since 1961.Only parts of Ellesmere Island were subsequently aerially surveyed (Riewe, 1973;Case & Ellsworth, 1991;Gauthier, 1996;Manseau et al., 2004).
The 1961 survey of the three largest islands-Ellesmere, Devon and Axel Heiberg-was hindered by persistent bad weather in some sections (Tener, 1963).Survey coverage was low and fragmented.As a result, Tener (1963) did not quantitatively estimate the total number of caribou on those islands and his "intuitive guesses" for them were exceedingly small.The mean overall density of 0.003 caribou • km -2 was but a tiny fraction of what the usable range on those three islands could support (62% of the total usable range on the QEI occurs on these three islands).
Subsequent development of aerial survey tech¬ niques and statistical analyses allow us to reevaluate the 1961 data to determine standard errors and con¬ fidence intervals.We emphasize that the application of these statistical methods in no way betters Tener's original 1961 data set.The benefits of our approach are: 1) it refines Tener's (1963) estimate of abun¬ dance; 2) it permits more detailed comparisons with subsequent aerial surveys; and 3) it identifies five eco-units to serve as the best standard for evaluating future changes in Peary caribou numbers, distribu¬ tions, and movements.Our statistical analyses may also satisfy those individuals who place an unfounded degree of reliance on probability testing and statisti¬ cal significance instead of evaluating results in terms of biological significance (e.g., Yoccoz, 1991;Johnson, 1999;Steidl et al., 2000;Anderson et al., 2001).Our objective is to recalculate caribou numbers and densities using recently available accurate measures of island size, to place confidence intervals on those estimates using raw data in a bootstrap simulation, and to provide an ecological basis for the conservation of Peary caribou on the QEI.

Animals and study area
The Canadian High Arctic Islands form the northern apex of the North American continent and collec¬ tively comprise a landmass of 419 061 km 2 (Fig. 1).The QEI are those islands entirely north of ca.74°N latitude.Although there are 2126 islands in total, only 34 of them each exceed 136 km 2 in size and are considered 'large' islands.The 22 islands surveyed in 1961 differ in size from 412 km 2 to 196 236 km 2 and total 410 765 km 2 or 98% of the entire QEI.The largest is Ellesmere Island, the tenth largest island in the world.The 12 large islands not surveyed in 1961 had a collective landmass of 5975 km 2 .
1961 Aerial survey Macpherson's (1961) summary of wildlife observa¬ tions by geologists in the mid-and late 1950s was available as an internal report in 1960 and provided some factual information about the numbers and distributions of Peary caribou or other animals on the QEI.However, Tener (1963: 8) believed that those surveys were "not detailed enough in their animal sightings to be of real value in selecting the method of survey."Discussions with P. Larkin (University of British Columbia) led Tener to choose system¬ atic sampling as the best design for obtaining both distributions and numbers.Although systematic surveying on the Canadian Arctic Archipelago has become a method used by most biologists for aerial surveys of caribou and muskoxen (Ovibos moschatus), Tener (1963) did not apply any statistical analysis to his 1961 survey results because he and P. Larkin concluded that a useful statistical approach was not available at the time.
The survey transect flights in 1961 employed two survey crews, each operating with a Piper Super Cub aircraft.The planes were equipped with large, low-pressure tires ('tundra tires') to allow them to take-off and land on unprepared ground.When possible, the crews flew the surveys at about 152 m (500 ft) above ground level and 137 km • h -1 (85 miles • h -1 ) air speed.The single observer in each plane sat directly behind the pilot and limited his observations to the right side of the aircraft.Transect width was about 402 m (0.25 mi).On each island, the first transect was selected at random and drawn on a survey map; additional transects were then drawn parallel to it at fixed intervals.
Aerial coverage varied by island and was small, averaging 4% (see Tener, 1963: Table 1).The inter¬ val between transects was determined by the chosen sampling intensity.Tener (1963) based his survey intensity on the belief that using a large number of small samples was better than using only a few large ones and that a large number of systematically spaced transects probably ensured adequate representation of range types and the geographical distribution of the animals within them.
Observers used previously prepared forms to record animals within the survey strip as being 'on-transect' and those seen beyond as being 'off-transect.' Tener (1963: 9) believed that the selected transect width and flight altitude facilitated the sighting of all ani¬ mals on each transect and accurate identification of them by sex and age classes.The exception to this was on parts of Devon Island where, to take advan¬ tage of the good weather while the Piper Super Cubs were not ready for service, he used a deHavilland Beaver aircraft.They surveyed two transects each about 805 m (0.5 mi) wide by flying the Beaver air¬ craft at about 305 m (1000 ft) above ground level.
During the 70-d period between 10 June and 18 August 1961, prevailing weather at the time permitted survey flights on only 34 (49%) days.The Piper Super Cubs flew about 500 h and the Beaver flew about 100 h, but only about 205 h were on survey transects.

Reexamination of 1961 data
Our approach We relied on the observations plotted and tran¬ sect lines drawn on the original 14 field survey maps, as the original data forms no longer exist (J. S. Tener, pers. comm., 2001).During his initial analysis, Tener plotted his observa¬ tions on the maps from the data forms, but after he completed the survey.We compared our raw data tabu¬ lations from those maps to the text, tables and figures in Tener (1963).To reconcile minor discrepancies between data on the maps and Tener (1963), Frank L. Miller and Samuel J. Barry each inde¬ pendently tabulated observa¬ tions of Peary caribou from each data map.They repeated this process four more times over the course of several months, leaving sufficient time between repetitions to decrease 'remembering' their previous interpretation.They then did one joint count to confirm that their reasoning on each observation was the same.Finally, they compared the totals for each island and for each smaller survey land unit (stratum) used by Tener (1963) on those islands.
Ca lc ulation of estim ated nu mbers a nd con fidence intervals We now know that many, perhaps most, caribou on the QEI make seasonal inter-island migrations (e.g., Miller et al., 1977aMiller et al., , 1977b;;Miller, 1990aMiller, , 1995aMiller, , 2002)).The objective of this evaluation of the 1961 aerial survey findings (Tener, 1963) was to recal¬ culate the estimates by grouping the 22 individual islands in a more ecologically sound manner into five island-complexes and to attach a measure of vari¬ ability and confidence to the estimates.At the level of our five eco-units, we used a 'bootstrap' procedure (Krebs, 1999) to estimate the number of caribou, standard errors, and confidence intervals.We then summed the eco-unit statistics to obtain values for WQEI, EQEI, and the entire QEI.
Our bootstrap procedure is as follows.Using only those observations of caribou seen on-transect, we simulated a 'resurvey' of each island 100 000 times.On each resurvey we selected the same number of transects that Tener had flown in 1961 randomly and with replacement from the set of transects for that island.For each of these 100 000 resurveys we summed the number of caribou 'observed' and the area 'surveyed' from the randomly selected transects.For each of these resurveys, we calculated the number of caribou within each eco-unit by first pooling these island results (caribou observed and area surveyed) among the islands in the eco-unit then multiplying the resulting caribou density (total pooled caribou observed divided by total pooled area surveyed) by the area of usable range within the eco-unit.Our bootstrap estimate of caribou for each eco-unit is the mean of 100 000 of those estimates.We measured bootstrap variability for each eco-unit as the variance of the bootstrap estimates (reported as standard error) and our bootstrap 95% confidence interval as the 2.5th and 97.5th percentiles of the distribution of all bootstrap estimates of caribou numbers.Hereafter, 'caribou' is synonymous with 'estimated caribou' in all cases.
Although Tener (1963) used the best information available at the time for the sizes of the 22 islands he surveyed, the areas of those 22 islands have been recalculated recently (http://atlas.gc.ca/english/facts/ islands.html)and the reported sizes for 21 of them have changed.Only Little Cornwallis Island, the smallest island surveyed in 1961, remained at its 1961 measured size of 412 km 2 .Therefore, we used sizes for 13 of 21 islands that were smaller than the values used by Tener (1963) by 1% to 44% and for the eight islands that were larger by 1% to 22%.Thus, our estimates of numbers of caribou reflect these changes in island sizes, which cause some devi¬ ation from Tener's (1963) original 1961 estimates.
Ellesmere, Devon, and Axel Heiberg islands have large glaciated areas and high elevations that are not suitable caribou range.Additionally, Melville Island has a small west-central permanent ice field.Tener (1963) excluded these areas from his calcula¬ tion of caribou numbers (but his calculations were based on different island sizes).We recalculated the size of permanently glaciated areas using GIS-based 1:250 000 topographic maps with 150-m elevation contours and the recently revised island sizes (Table 3).As a result, we excluded all land > 750 m above mean sea level (amsl) and the glaciated areas < 750 m amsl from the land area we used for computing estimates for the number of caribou present.This removed all large areas of unsuitable range including all permanent snow and ice fields.We also excluded areas on Axel Heiberg and Ellesmere islands that received no aerial coverage.

Results
We tallied 393 observations of 1583 caribou on-transect on the QEI from Tener's 1961 field maps.Cari-bou in groups of two or more individuals constituted 78% (306) of the observations and the remaining 87 (22%) were individual caribou.By our calculations, Tener flew 27 872 km of transect lines and surveyed 11 213 km 2 (ca.4%) of the usable range within the 1961 survey area.The 100 000 estimates of caribou numbers formed by randomly resampling transect counts within each eco-unit yielded 28 288 ± 2205 SE and a 95% CI of 20 436-37 031 Peary caribou on the QEI in summer 1961.Our estimate for the QEI is nearly 9% greater than Tener's (1963) original esti¬ mate (Table 2).Calves made up 20% of all caribou observed throughout the QEI.

Two major zones
Our recalculation suggests that 90% of Peary cari¬ bou on the QEI occurred on the WQEI and only 10% on the EQEI in summer 1961 (Table 2).The number of caribou on WQEI is only 4% greater than the original number estimated there by Tener (Table 2).Tener (1963) did not calculate an estimate for the islands surveyed in the Eastern eco-unit but used only "intuitive guesses"-collectively 650 caribou.When that 650 is combined with the 832 caribou estimated by Tener (1963) on the four islands in the North-central eco-unit there are 1482 caribou on the EQEI in 1961.Therefore, our estimate is 95% greater than the original 1961 collective island estimate plus guesses, although our 95% CI includes Tener's 1961 combined estimate for the EQEI (Table 2).The WQEI and the EQEI represent 24% and 76% of the collective landmass of the 22 islands surveyed in 1961 and 32% and 68% of the potential range for caribou, but the WQEI held nine times as many caribou as the EQEI.Caribou exhibited a strong pref¬ erence for WQEI with a density 18-fold greater than that on the EQEI in summer 1961.

Five eco-units Southwestern
There are five islands from 549 to 42 149 km 2 in size that make up this eco-unit (Table 1; Fig. 1).However, Byam Martin Island (1150 km 2 ) was not surveyed in 1961.Therefore, the eco-unit estimate is based on the other four islands minus 159 km 2 for ice fields on western Melville Island (Table 1).There was a strong preference for Peary caribou in this eco-unit, most (58%) of the caribou on the QEI and 64% of those on WQEI occurred there (Table 2).

Northwestern
There are three islands from 764 to 5048 km 2 in size that make up this eco-unit (Table 1; Fig. 1).Thus, the eco-unit estimate is based on the collective landmass of Mackenzie King, Borden, and Brock islands, excluding 325 km 2 from Brock Island due to fog (Table 1).Caribou within this eco-unit made the third largest contribution to the estimate on the QEI (Table 2: 15%).Although the number of caribou contributed from this eco-unit to the QEI estimate was only moderate, the density was exceptionally high (Table 2).

Eastern
There are 10 islands from 137 to 196 236 km 2 in size that make up this eco-unit (Table 1; Fig. 1): Ellesmere (196 236 km 2 ), Devon (55 247 km 2 ), Axel Heiberg (43 178 km 2 ), Graham (1378 km 2 ), Meighen (955 km 2 ), North Kent (590 km 2 ), Coburg (344 km 2 ), Stor (313 km 2 ), Hoved (158 km 2 ), and Buckingham (137 km 2 ).Only the three largest islands of Ellesmere, Devon, and Axel Heiberg were surveyed in 1961.Aerial coverage was low and fragmented due to poor weather.The potential range for caribou on the three islands is 63% of their collective landmass because 108 303 km 2 are covered by glaciers (Table 3).Only 27% of the land area on the three islands is <300 m amsl; in contrast, about 90% of the entire WQEI lies below the 300-m amsl level (Table 3; Miller et al., 1977a).More importantly, about 60% of the land¬ mass within the Eastern eco-unit is at higher eleva¬ tions than those found anywhere else on the QEI.Caribou within this huge eco-unit composed only 7% of the total in the QEI, but 74% of the caribou in the EQEI (Table 2).The density of caribou within 70 this eco-unit was exceptionally low, based on usable range for caribou (Table 2).Calves were appreciably lower in this eco-unit than in the other four eco-units (Table 1: % 2 = 21.55, 4 df; P < 0.005).

Discussion
Tener was greatly limited in any interpretation or evaluation of his results, with almost no on-site background information to draw from.In 1961, it appeared that there was a relatively large number of Peary caribou available on the QEI to support the establishment of Inuit settlements there.Unfortu¬ nately, that condition has not persisted and Peary caribou on the QEI are now 'Endangered' (Gunn et al., 1981;Miller, 1990b;Miller & Gunn, 2003a, b).Peary caribou populations on the WQEI now occur as mere remnants of their 1961 sizes (Tener, 1963;Miller et al., 1977a;Miller, 1995bMiller, , 1998;;Gunn et al., 2000;Gunn & Dragon, 2002).With the benefit of 4 decades of fragmentary information, we can extend our assessment of the 1961 findings.
In summer 1961 the relative mean densities and distribution of Peary caribou populations on the QEI indicated strong preferences for the Northwestern, Southwestern and South-central eco-units, low occur¬ rence in the North-central eco-unit, and extremely low occurrence in the Eastern eco-unit.Since then only the Southwestern and South-central eco-units have served as 'heartlands' for Peary caribou.

Two major zones
In summer 1961 there were 18-fold more Peary caribou per unit area of usable range on the WQEI than on the EQEI (Fig. 1).Subsequent aerial sur¬ veys suggested that the highest estimates of Peary caribou were still in the Southwestern and the Southcentral eco-units (e.g., Miller et al., 1977a;Miller, 1990b;Gunn & Dragon, 2002).The high numbers of caribou estimated in summer 1961 within the Northwestern eco-unit and on Lougheed Island in the South-central eco-unit have not been observed again (Miller et al., 1977a;Miller, 1990b;Gunn & Dragon, 2002).
We believe the observed distribution of Peary caribou on the WQEI in 1961 reflected a summer¬ time redistribution of the relatively high density of caribou in the Southwestern eco-unit beyond the eco-unit boundaries to 'out islands' in the North¬ western eco-unit.Thus, the relatively high density of caribou in the Northwestern eco-unit only reflects temporary ingress from elsewhere, not a persistent year-round state.It is also most likely that summer range extended in 1961 to Lougheed Island, with some spillover to the North-central eco-unit islands.
Whether this summertime redistribution involved caribou from both the Southwestern and Southcentral eco-units is not clear, but Lougheed Island appears to be more closely linked over time with changes in caribou numbers and distributions within the South-central eco-unit than with changes in the Southwestern eco-unit.Lougheed Island could not support caribou at a density of 1.1 • km -2 year-round for more than a few years without overutilization of the range.The same applies to Borden Island at 0.6 • km -2 and Mackenzie King Island at 0.5 • km -2 , where much of the range is on the poorly vegetated Beau¬ fort Formation (Tener, 1963;Edlund & Alt, 1989).However, surveys in winter have shown that sparsely vegetated but wind-blown range on the Beaufort formation can be important to the survival of Peary caribou (Miller et al., 1977a).
Five eco-units Southwestern After Tener (1963) completed his analyses, we learned that many, if not most, caribou within this eco-unit make seasonal inter-island migrations, moving from summer ranges on eastern Melville and Byam Mar¬ tin islands to winter ranges on Prince Patrick and Eglinton islands (Miller et al., 1977a(Miller et al., , 1977b)).West¬ ern Melville Island appears to be important to Peary caribou only during times of high numbers, such as in summer 1961.The small size of Emerald Ile appar¬ ently relegates it to transitional range for some few caribou, mainly during spring and autumn.
Inter-island movements within this eco-unit must have persisted over time, as Parry (1821) noted Peary caribou moving off eastern Melville Island in autumn.Therefore, multi-island surveys of at least Melville and Prince Patrick islands are necessary to estimate the number of caribou and any changes in the number of caribou there over time.Ideally, Byam Martin, Eglinton, and Emerald islands should be included in those surveys.

South-central
Caribou within this eco-unit also make seasonal inter-island migrations (Miller, 1995a(Miller, , 1998(Miller, , 2002)); however, some remain year-round on Bathurst Island (Miller & Barry, 2003) and on other smaller islands (Miller, 1997(Miller, , 1998)).The primary linkage appears to be among Bathurst, Alexander, Marc, Massey, Vanier, and Cameron islands.Ile Marc was not surveyed in 1961: it is small (56 km 2 ) but important caribou summer range, lying between Alexander and Massey islands.There is some evidence of movement between Bathurst and the other lesser satellite islands as well as Cornwallis and Lougheed islands (Miller, 1997(Miller, , 1998)).
Rangifer, Special Issue No. 16,2005 This eco-unit is of particular importance because it is readily accessible by hunters.The Inuit settlement of Resolute Bay on Cornwallis Island was established in the mid-1950s.Bathurst Island became a primary hunting area and remained so until the major die-off of caribou there in winter and spring 1973-74 (Parker et al., 1975;Miller et al., 1977a).Caribou abundance had apparently been adequate to satisfy annual har¬ vest needs from at least the time of settlement until 1973, when they were already decreasing in number.Based on the past, this eco-unit could once again support a harvestable number of caribou, but prob¬ ably only if the hunters from Resolute Bay give them time to recover.The three most recent annual die-offs between 1994 and 1997 saw the number of caribou reduced from about 3000 to < 100 (Miller, 1998;Gunn & Dragon, 2002;Miller & Gunn, 2003a).

Northwestern
This eco-unit is particularly remote from human settlement and the prevailing weather does not favor systematic aerial surveys.Mackenzie King, Borden, and Brock islands all lie beyond the northern limit of prostrate shrubs, and sedges and vascular plants are sparse, completely herbaceous, and least diverse of the QEI (Edlund & Alt, 1989;Edlund, 1990).This eco-unit has served on occasion as a 'spillover area' for caribou from the Southwestern eco-unit and possibly the South-central eco-unit when the caribou on one or both eco-units uncommonly reached a relatively high density.However, it is the smallest of the five eco-units and there is no reason to believe that it could sustain anything more than a relatively small low density 'reservoir area' for Peary caribou.
North-central This eco-unit lies mainly in the environmentally stressful northern zone of plant growth (Edlund & Alt, 1989;Edlund, 1990).The ability of this range to support a large number of caribou year-round is highly questionable given its moderate size and mea¬ ger flora.In all likelihood, this eco-unit serves mainly as a rarely used 'spillover area' for caribou from the Southwestern eco-unit and possibly the South-central eco-unit when caribou on one or both eco-units infre¬ quently reach relatively high densities.This eco-unit should be viewed as a low density reservoir that can help foster the long-term persistence of Peary caribou on the QEI.The islands are not readily accessible to Inuit hunters and could never provide any appreciable level of sustained annual harvest.

Eastern
It is a mystery why there are not many more Peary caribou in this eco-unit.Its vastness makes Tener's (1963) "intuitive guess" of only 650 Peary caribou there in 1961 appear ultra-conservative.This is especially true as there are at least 186 000 km 2 of potential range within this eco-unit.Our estimate of 2127 ± 571 SE requires only 0.01 caribou • km -2 and our estimate minus 1 SE (1556) would require, on average, only 0.008 caribou • km -2 of usable range.
Based on the 1961 results and subsequent lim¬ ited information, we view the Eastern eco-unit as a huge, low-density reservoir for Peary caribou.In the foreseeable future, caribou there can make only a relatively small contribution to annual harvests of Peary caribou on the QEI.However, because of the Eastern eco-unit's large and rugged land base, it has great benefit to the persistence of Peary caribou on the QEI.Also, given the paucity of data for the three major islands in this eco-unit and their large size, there is a chance that this area may have more caribou than anyone is aware of.
Ironically, when the scenarios of 'global warming' are considered, we could speculate that the Eastern eco-unit may become the most important region for Peary caribou on the entire QEI.Because of the ruggedness of its terrain and its elevation relief (Table 3), many micro niches will be available that could foster Peary caribou survival in that region compared to survival in the relatively low-lying WQEI and the North-central eco-unit of the EQEI.If global warm¬ ing advances as predicted by some, this change could occur within the next century.Even more important¬ ly, it is possible that virtually only the Peary caribou on the Eastern eco-unit may survive the transitional phase of greater snowfall and more frequent icing events during global warming.The probability of this may be tied to the rapidity of warming processes which may cause micro niches to become critically important to the caribou's long-term survival on the QEI.

Management and conservation implications
Future aerial surveys of Peary caribou on the QEI should be at the eco-unit level to maximize both the accuracy of the estimated number of caribou and the probability of detecting real changes in caribou num¬ bers and distribution over time.Surveys should be flown in July, at three-year or shorter intervals as long as the number of caribou remains small and especial¬ ly if they are being harvested annually.Regardless of when the last survey was carried out, a July survey should be flown in the summer immediately after a winter or spring with exceptionally severe weather to determine the impact, if any, on the overwinter survival of caribou and the production and early sur¬ vival of calves.Once adequate aerial surveys indicate that the number of caribou has increased to a size where the annual level of harvest would not exceed 5% (cf.Kelsall, 1968), based on the mean estimate minus 1 SE, minus the percentage of calves present, the interval between surveys could be extended to 5 years.In all cases, however, an aerial survey still should be carried out following any winter or spring with exceptionally severe weather to prevent a major decline in number going undetected.
The past and current status of Peary caribou in the EQEI remains debatable.No current information exists for caribou in the North-central eco-unit.In the Eastern eco-unit, recent information from north¬ ern Ellesmere Island suggests a stagnant condition with a minimum of only 45 Peary caribou estimated there between 1988-2002(Manseau et al., 2004)).However, recent information from Inuit hunters from Grise Fiord on southern Ellesmere Island suggests that caribou are increasing there, based on sightings of a low number of caribou in places where they have been absent for several decades.Both M. Raillard, Canadian Wildlife Service, and J. England, Univer¬ sity of Alberta, believe that while conducting other field activities they have seen numbers of caribou on parts of Ellesmere and Axel Heiberg islands in summers 1999-2001 that suggest increases are probably occurring on some relatively well-vegetated sections.The accuracy of island-wide extrapolations of caribou numbers from such casual observations cannot be measured with confidence.Thus, such extrapolations form a poor basis for making decisions regarding rates of annual harvest, particularly for an 'Endangered Species.'Only prescriptions for annual harvest that are based on actual counts of caribou that yield a total minimum number adequate to sus¬ tain desired harvests should be employed.Even then, following exceptionally severe weather years, harvest of Peary caribou may have to be temporarily halted by emergency action through cooperation with Inuit hunters.
The vastness and concomitant prohibitive costs of aerial surveys on the Eastern eco-unit require a dif¬ ferent approach.We recommend the establishment of at least five permanent survey plots, 100 km by 100 km on areas of known importance to Peary caribou and where annual harvests are most likely to occur.These plots should be flown in early July when weather is most stable, and should be flown each time as close as possible to the same days of this month to record the percentage of calves, and ide¬ ally yearlings, present and the phenological state of plants for comparisons among years.This procedure would provide a relative estimate of the number of caribou, calving success, yearling recruitment, and annual variation in new plant growth within those 'food patches' that would allow the setting of flex-ible maximum annual harvest levels.Accurate calf and yearling counts should be made, as a measure of performance to eventually provide a measure of their trend over time to better prescribe the maximal allowable level of annual caribou harvest, especially when annual harvest appears high.After the first year of plot surveys, both the previous number of animals counted and the size of the desired future annual harvest would dictate the needed frequency of those aerial searches.In the absence of a better mea¬ sure, a maximum of a 5% annual harvest should be adhered to, especially as long as Peary caribou remain an 'Endangered Species.' It is most unlikely that the concepts of 'carrying capacity' and 'sustained yield' have any real applica¬ tion to Peary caribou on the QEI.To date, everything suggests that the caribou on the QEI function in a 'nonequilibrium grazing system' where abiotic fac¬ tors, mainly snow and ice, control their fate through infrequently occurring, sporadic and unpredict¬ able 'exceptionally hard weather years' (Caughley & Gunn, 1993;Behnke, 2000;Miller & Gunn, 2003a, b).At such times, extremely unfavorable snow and ice conditions prevent the animals from getting to food or cause caribou to use more energy accessing it than they recover in forage intake.In years when such snow and ice conditions are prolonged and wide¬ spread, a large number of caribou (and muskoxen) will die from starvation.This happens because the carrying capacity of those caribou ranges will change, both quickly and unpredictably, with the prevailing severity of the snow and ice cover in each year.The more severe the snow and ice conditions and the longer they persist, the greater the loss of animals regardless of their densities or nutritional state at the beginning of the die-off period.Once the snow and ice melts, the carrying capacity of the range instantly experiences a many-fold increase-the absolute for¬ age supply is still there and is once again readily available to the animals.Therefore, caribou range in early winter that could support several thousand animals under favorable snow and ice conditions can be rapidly reduced to one that will support only several hundred or fewer ani¬ mals.Then, following snowmelt, it will return imme¬ diately to a relatively favorable state.Such capricious range conditions will not allow a predictable fixed sustainable level of annual harvest of any appreciable size for more than a short series of years, if that.The lack of adequate periodic monitoring programs and the uncertainty of 'climate change' leave the fate of Peary caribou on the QEI unpredictable.Intervals of > 5 years between estimates of caribou on the QEI could run the risk of merely documenting serious declines after the fact, especially if continual annual Rangifer, Special Issue No. 16, 2005 harvests had been at relatively high levels.Although there is no evidence that harvesting has caused Peary caribou declines on the QEI, it could have accelerated and deepened them in some instances.

Conclusions
The Peary caribou on QEI is a unique part of Canada's High Arctic natural heritage and is recog¬ nized as an 'Endangered' form of wildlife in Canada.The number of caribou now on the QEI is a vestige of the number estimated there in 1961; yet, Peary caribou continue to be considered a usable renewable resource!Our recalculation of the 1961 aerial survey is the best benchmark for subsequent evaluations of changes in numbers, distributions, and movements of Peary caribou on the QEI.The division of the entire QEI into five eco-units provides a standard for evalu¬ ating the above changes over time on an ecologically sound basis for prescribing and assessing conserva¬ tion measures for Peary caribou on the QEI.
Continued annual harvesting of Peary caribou-an 'Endangered Species'-requires that special safe¬ guards and an adequate monitoring program be established.Cooperation among wildlife agencies and Inuit hunters is essential, because precautionary measures will require the participation of Inuit users in setting the restrictions and complying with them.No more than a few Peary caribou should be taken from any island within any eco-unit without first obtaining actual counts to determine that enough caribou exist there to sustain the desired level of annual harvest.Otherwise, we run the risk of being found negligent in our stewardship responsibility to this unique resource.

Table 1 .
(Tener, 1963)mpiled from 14 observation maps for the 1961 aerial survey of Peary caribou(Tener, 1963)and the most recent values for the 22 islands in the 1961 aerial survey, Queen Elizabeth Islands, Canadian Arctic Archipelago.Includes only potentially usable range for Peary caribou.b Island sizes available from http://atlas.gc.ca/english/facts/islands.html.c Excludes 11 376 km 2 of usable range that did not receive aerial coverage. a

Table 2 .
Tener (1963)imates of Peary caribou fromTener (1963)compared to the new estimates, associated standard errors, and confidence intervals calculated from the distribution of 100 000 bootstrap replications for each of five eco-units, Queen Elizabeth Islands, Canadian Arctic Archipelago.95%CIbased on the 2.5 th and 97.5 th percentiles of 100 000 bootstrap estimates of caribou numbers.cEstimatesfor Ellesmere, Devon, and Axel Heiberg islands were not calculated in 1961 but based only on "intuitive guesses" byTener (1963).
a N equals number of aerial survey transects flown.b

Table 3 .
Tener's (1963)f divisions of the three mountainous islands of Ellesmere, Devon, and Axel Heiberg, within the Eastern eco-unit, eastern Queen Elizabeth Islands, Canadian Arctic Archipelago.)werenotsurveyed in 1961.Therefore, the eco-unit estimate is based on the collective landmass of Bathurst, Cornwallis, Lougheed, Vanier, Cameron, Alexander, Massey, and Little Cornwallis islands (Table1:Tener's (1963)"Bathurst Island" included Bathurst and the islands of Vanier, Cameron, Alexan¬ der, and Massey).Caribou within this eco-unit made the second largest contribution to the estimate for the QEI (Table2: 17%).