Estimating the harvest of Pacific walrus,
Odobenus rosmarus divergens, in Alaska
Joel L. Garlich-Miller
Douglas M. Burn
Marine Mammals Management
U.S. Fish and Wildlife Service
1011 East Tudor Road, Anchorage, Alaska 99503
E-mail address (for J. L. Garlich-Miller): Joel_GarlichMiller@mail.fws.gov
Manuscript accepted 13 J anuary 1999
Fish. Bull. 97(4):1043–1046 (1999).
F or thousands of years, walrus
hunting has been an important
component of the economy and cul-
ture of Native communities along
the Bering and Chukchi Sea coasts
(Ray, 1975). Today, the Pacific wal-
rus (Odobenus rosmarus divergens)
remains a valuable resource to
coastal natives in Alaska (United
States) and Chukotka (Russia) as
a source of food and raw materials
for traditional equipment and
Accurate information regarding
the number of animals removed
annually from the population is
fundamental for the conservation
and management of any species. As
the agency responsible for manag-
ing Pacific walrus in U.S. waters,
the U.S. Fish and Wildlife Service
(FWS) gathers data on the size and
composition of the subsistence wal-
rus harvest in Alaska. The FWS
presently administers two separate
harvest monitoring programs: the
Walrus Harvest Monitoring Project
(WHMP) and the Marking Tagging
and Reporting Program (MTRP).
The WHMP is an observer pro-
gram carried out at select walrus
hunting villages in Alaska. E ach
spring, as the pack ice recedes
northward, hunters from coastal
communities in the Bering Strait
region have access to herds of wal-
rus as they migrate to their sum-
mer range. Historical harvest infor-
mation indicates that approxi-
mately 80% of the annual reported
walrus harvest in Alaska occurs in
this region (F ay and Bowlby1).
WHMP monitors stationed at the
primary walrus hunting villages in
the Bering Strait region (Gambell,
Savoonga, L ittle Diomede and
Wales; Fig. 1) collect information on
the size and composition of the wal-
rus harvest. Harvest monitors meet
boats as they return from walrus
hunting trips in order to collect bio-
logical samples and harvest infor-
mation at the boat landing site
(Garlich-Miller2). The goal of the
WHMP is to identify and record the
gender and age class of every wal-
rus retrieved by hunters from these
villages during the monitoring pe-
riod. Although there is no way of
evaluating the degree to which this
goal is achieved, WHMP monitors
meet most of the returning boats,
and the number of retrieved ani-
mals not recorded during the har-
vest monitoring period is believed
to be small (Dickerson3).
The MTRP is a Federally man-
dated year-round, statewide pro-
gram (Fig. 1). The marking and tag-
ging rule requires that all hunters
certify (tag) walrus ivory (tusks)
and report all walruses that are
taken. The objectives of the MTRP
are to collect harvest information
and to certify specified marine
mammal parts to help control ille-
gal harvests and trade. Hunters are
required to bring walrus tusks to a
MTRP tagger within 30 days of the
kill. The tagger attaches individu-
ally numbered wire tags to the
tusks and records the numbers on
a tagging certificate. MTRP tags
are not attached to calf walruses (or
other walruses that may be miss-
ing tusks); however, hunters are
required to report all animals
taken. The age class, gender, kill
date, and kill location of each wal-
rus are recorded on the certificate
(Stephensen et al.4).
These two programs indepen-
dently provide information on the
size and composition of the harvest.
Except in the case of the village of
Wales, WHMP and MTRP staff are
different people. E ach of the two
monitoring programs has its
strengths and weaknesses. The
WHMP benefits from the presence
of on-site staff to collect accurate
biological information from every
walrus retrieved in a community
during the monitoring period. Un-
fortunately, the monitoring period
is seasonal (restricted to the spring
hunt) and operates only in four
coastal villages. The MTRP is a
statewide, year-round program;
however hunter compliance with
the MTRP rule is variable and ani-
mals lacking tusks (e.g. calves,
yearlings, and animals with broken
tusks) often go unreported (Burn,
1Fay, F. H., and C. E . Bowlby.
harvest of Pacific walrus, 1931–1989.
U.S. Fish and Wildlife Service, Marine
Mammals Management, Anchorage,
AK . Technical report MMM 94-2,43 p.
2Garlich-Miller, J .1997.
reproductive status of Pacific walrus har-
vested in the Bering Strait region, 1994–
1996.U.S. Fish and Wildlife Service, Ma-
rine Mammals Management, Anchorage,
AK.Technical Report MMM 97-1,25 p.
3Dickerson, L .1989.
Wildlife Service, Marine Mammals Man-
agement, Anchorage, AK .
4Stephensen, W. D., D. Cramer, and D.
Burn.1994.Review of the Marine
Mammal Marking Tagging and Reporting
Wildlife Service, Marine Mammals Man-
agement, Anchorage, AK .
port MMM 94-1, 49 p.
Age, sex, and
U.S. F ish and
U.S. F ish and
1044Fishery Bulletin 97(4), 1999
Location of Alaska villages where walrus harvest information is collected through the Walrus Harvest
Monitor Project (✪) and the Marking Tagging and Reporting Program (G). Note that tusks may be
tagged in selected locations outside the walrus’ range.
Owing to funding and logistical constraints that
prevent the year-round, statewide implementation
of the WHMP and imperfect hunter compliance with
the MTRP, not all harvested walruses are recorded.
The FWS must therefore rely on analytical methods
to estimate the size of the total annual harvest based
on the best available information. Here we describe
a new method for estimating the size of the annual
walrus harvest in Alaska using a correction factor
for noncompliance with the MTRP rule.
Materials and methods
Prior to the initiation of the MTRP, the statewide
walrus harvest was approximated by using a predic-
tion equation applied to WHMP data. The predic-
tion equation was based upon the historic relation-
ship of the size of the spring harvest at WHMP-moni-
tored villages in relation to the remaining hunting
villages in Alaska. This information was obtained
from the Alaska Department of Fish and Game,
which administered a statewide harvest monitoring
program between the years of 1960 and 1978. To es-
timate the statewide harvest, the FWS used a least-
squares regression to describe the relation between
the Gambell, Savoonga, and Diomede harvests to the
statewide totals. The resulting prediction equation
(Equation 1, r2=0.239) was subsequently used to es-
timate the total statewide harvest on the basis of
data collected annually after 1978 through the
WHMP at Gambell, Savoonga, and Diomede.
N = 459.7 + 1.26 (WHMPTotal),(2)
WHMP Total= total number of retrieved walruses re-
corded during monitored
spring hunts at select villages
(Diomede, Gambell, and Savo-
onga) for a given year.
N = total statewide harvest estimate; and
The evolution of the MTRP as a year-round, state-
wide monitoring program provided the opportunity
to improve the reliability of harvest estimates by
using current data. Because there is spatial and tem-
poral overlap between the two programs, the WHMP
can serve as a baseline for evaluating compliance
with the MTRP program. A correction factor for non-
compliance can be estimated as a proportion by us-
ing the following equation:
1045 NOTEGarlich-Miller and Burn: Estimating the harvest of Odobenus rosmarus divergens in Alaska
whereR = correction factor for noncompliance;
MTRPWHMP= a subset of the number of walruses
recorded through the MTRP with re-
ported kill dates and locations that
coincide with WHMP operations (pre-
sumed to have been recorded through
both the MTRP and the WHMP).
The standard error of this ratio estimate is calcu-
lated according to Snedecor and Cochran (1967, p.
537) by treating each of the three villages as inde-
pendent estimates of overall compliance. Data from
the village of Wales were not used in the analysis
because of small harvests and a lack of independence
between the WHMP and MTRP programs in that
Assuming the estimated compliance with the
MTRP is uniform throughout the full range of wal-
rus hunting villages, the total MTRP harvest can be
adjusted as follows:
N = MTRPTotal × R, (3)
where MTRPTotal= the total number of walrus re-
corded through the MTRP.
Data from the MTRP and WHMP are currently avail-
able for the years 1992–97 (Table 1). Estimated com-
pliance with the MTRP (the reciprocal of R) ranged
from 63% to 99%. Harvest estimates based on the
prediction equation and the proportional equation
differed depending on the proportion of animals re-
corded through each of the two monitoring programs.
The difference between the two methods was not sta-
tistically significant for any year. Comparisons of the
1992 MTRP and WHMP harvest data suggests that
the WHMP monitors were unable to account for all
the retrieved walruses landed during the monitoring
period. In this instance, the MTRPTotal was considered
the minimum estimate of the statewide harvest.
Prior to 1997, the FWS used WHMP data collected
annually at Gambell, Savoonga, and Diomede as an
index for estimating the total statewide harvest of wal-
rus. One drawback of this method has been that the
equation assumes that the relation between the hunt-
ing success of WHMP and non-WHMP villages is con-
stant over time (Fay et al., 1997). In fact, recent har-
vest data have shown that the relative hunting suc-
cess of each village is highly variable. The annual har-
vest at each village is subject to large interannual varia-
tion, presumably as a result of weather and ice pat-
terns affecting the availability of walruses to hunters
at a given geographical location (Garlich-Miller2).
One advantage of the proportional equation (Eq. 3)
over the prediction equation (Eq. 1) is that it can
account for variability in the relative success of hunt-
ing villages. For example, on the basis of comparable
numbers of walrus recorded through the WHMP in
Gambell, Savoonga, and Diomede in 1995 and 1996,
the prediction equation produced remarkably simi-
lar harvest estimates for the two years (1974 and
1994 walrus, respectively). Although MTRP compli-
ance rates were also comparable for these years, the
proportional equation, in which statewide MTRP data
were used, produced harvest estimates that were mark-
edly different (1681 and 2481 walrus, respectively). We
Walrus harvest data from the Walrus Harvest Monitor Project (WHMP) and Marking, Tagging, and Reporting Program (MTRP),
E quation 1
E quation 3
95% CI of
E quation 3
1The calculated estimate was less than the value of MTRPTotal. The MTRPTotal was therefore used as a minimum estimate of the statewide walrus
1046Fishery Bulletin 97(4), 1999 Download full-text
believe that this difference, which reflects inter-
annual variability in the hunting success of non-
WHMP villages, more accurately reflects statewide
harvest levels for these two years. For the reasons
stated above, the FWS has adopted the use of the
proportional equation to estimate the size of the
In 1992, the assumption that all retrieved walrus
were recorded during WHMP operations was not met.
In this instance we considered the uncorrected MTRP
data as a minimum harvest estimate for that year.
The WHMP was re-initiated in 1992 after a two-year
hiatus; the lack of experienced personnel likely con-
tributed to the poor monitoring results that season.
Since that time, program managers have attempted
to improve the program by hiring experienced moni-
tors and additional village assistants to meet all re-
turning boats and by emphasizing the importance of
recording all retrieved animals (Dickerson3).
Another assumption of the proportional equation
is that compliance in the villages of Gambell,
Savoonga, and Diomede is representative of non-
WHMP villages. Compliance was variable between
villages and years, suggesting that there was little
correlation between the two programs. Since conduct-
ing analyses of MTRP compliance, information and
education efforts in these villages appear to have
been effective at increasing compliance. The accu-
racy of this method could be improved by increasing
the number of villages in the WHMP.
In both the prediction and proportional equations,
estimates are based on the number of walrus that
are retrieved by hunters. Fay et al. (1994) estimated
that 42% of walrus struck by hunters are not re-
trieved and subsequently die at sea. In order to esti-
mate total human-caused removals from the Pacific
walrus population, harvest estimates are adjusted
to account for animals struck and lost.
Accurate harvest data are vital to the management
of the Pacific walrus population. Harvest data are
incorporated into stock assessment reports that chart
the status and trend of the population. The stock
assessment process compares estimates of human-
caused mortality with a calculated potential biologi-
cal removal (PBR) level to determine the status of a
stock. One of the reasons a stock may be designated
as “strategic” depends upon whether or not its level
of human-caused mortality exceeds the calculated
PBR level (Wade, 1998). Since 1992, most human
caused mortality affecting the Pacific walrus popu-
lation has been associated with walrus hunting ac-
tivities in Alaska and Chukotka. Between 1992 and
1996 the combined annual take of walrus in the U.S.
and Russia averaged 4869 walrus per year (Gorbics
et al.5). Russian harvest data are currently unavail-
5Gorbics, C. S., J . L. Garlich-Miller, and S. L. Schliebe.
Draft Alaska marine mammal stock assessments 1997: sea otter,
polar bear and walrus. U.S. Fish and Wildlife Service, Marine
Mammals Management, Anchorage, AK.
Admin. Report, 129 p.
able for 1997. Because annual estimates of human
caused mortality have been lower than the calculated
PBR of 7533 the population has been classified as non-
strategic (Gorbics et al.5). It is essential that harvest
monitoring in both nations be maintained in order to
accurately assess the impact of the harvest to this stock.
In summary, this new method of harvest estima-
tion uses data from both harvest monitoring pro-
grams to account for interannual and intervillage
variability in hunting success and applies a correc-
tion factor to adjust total harvest estimate to account
for noncompliance with the MTRP. The accuracy of
harvest estimates is therefore dependent upon the
degree of hunter compliance with the MTRP rule. It
is hoped that through ongoing information and edu-
cation efforts that explain the importance of accu-
rate harvest data, the understanding of and compli-
ance with monitoring programs will improve.
We would like to acknowledge the hard work and
dedication of the MTRP taggers and WHMP harvest
monitors. Harvest monitoring activities were coor-
dinated by Larry Dickerson, Polly Hessing, Dana
Seagars, and Wells Stephensen. This manuscript was
improved by constructive comments offered by the fol-
lowing reviewers: Susan Lapkass, Larry Dickerson,
Susan Hills, and Mark Udevitz.
Burn, D. M.
1998.E stimation of hunter compliance with the Marine
Mammal Marking, Tagging, and Reporting Program for
walrus. Wild. Soc. Bull. 26(1):68–75.
F ay, F. H., J . J . Burns, S. W. Stoker, and J . S. Grundy.
1994. T he struck-and-lost factor in Alaskan walrus
harvests. Arctic 47(4):368–373.
F ay, F. H., L . L . E berhardt, B. P. K elly, J . J . Burns, and
L . T. Quakenbush.
1997.Status of the Pacific walrus population, 1950–1989.
Mar. Mamm. Sci. 13:537–565.
R ay, D. J .
1975. The eskimos of the Bering Strait, 1650–1898.
Washington Press, Seattle, WA, 305 p.
Snedecor, G. W., and W. G. Cochran.
1967.Statistical methods, sixth ed.
Press, Ames, IA, 593 p.
Wade, P. R .
1998.Calculating limits to the allowable human-caused
mortality of cetaceans and pinnipeds.
The Iowa State Univ.
Mar. Mamm. Sci.