Final Study Plan for Estimating the Size of the Pacific Walrus Population March 2006 Marine Mammals Management, U. S. Fish and Wildlife Service Alaska Science Center, U. S

Дата канвертавання24.04.2016
Памер158.48 Kb.
1   2

C. Russian approach: thermal imagery.

Thermal (infrared) imagery will be the primary data collection method, as for the U.S. portion of the survey. A systematic sample of strip transects will be surveyed with the thermal scanner in each survey block. The amount of heat, or thermal signature, of each detected group will be recorded.

C1. Thermal imagery equipment. The remote sensing system is a “Malachite” Airborne Multispectral Scanner (AMS). The system has a 1.5 milliradian instantaneous field of view (IFOV), and collects imagery across a sensor array 2048 pixels wide. The angle of view is 82 degrees. When flight altitude is 1500 m, the width of the thermal view band is 2400 m. The Malachite Scanner records a thermal infrared channel (f 8 – 13 µm) with12-bit radiometric resolution. Due to specific features of the thermal imagery system used by the Russian side (i.e., absence of absolute temperature referencing), an infrared radiometer will be installed. It will be used for continuous temperature calibration of the thermal scanner. Scanner data will be recorded on hard drives.
Thermal imagery will be collected from a turbine engine aircraft, the L-410. The aerial laboratory is a twin-engine aircraft with high wings and bubble windows that allow lateral and downward visibility on both sides.
C2. Details: thermal imagery. Thermal imagery data will be collected as the aircraft flies along transect lines within a survey block. Spatial resolution of the AMS system varies linearly with the altitude of the aircraft. Surveys will be conducted at 1,500 m (4,921 ft) Above Ground Level (AGL). The view band will be 2,400 m with spatial resolution of 2 m. A thermal signature will be recorded for each walrus group that is detected within the 2.4 km wide scanned strip.
C3. Calculation of thermal index. The thermal index will be calculated by the same method used by the U.S. side (see A3 above).
С4. Limitations and unknowns. To correct geometric distortions of thermal images, an additional digital 35-mm photo camera will be used, and data on spatial position of the aircraft will be recorded. Thermal images will then be adjusted by ratio-comparisons with photographs of the same areas.
D. Russian approach: aerial digital photography

Each thermally detected walrus group will be photographed with high resolution digital photography. Counts of aerially photographed walrus groups will be used to model the relationship between the amount of heat measured by the thermal scanner and the number of walruses in a group.

D1. Digital photography equipment. Two digital photo cameras will be mounted within the aircraft cabin. Both are Nikon D1X 6 megapixel format digital cameras that produce images with dimensions of 3008 х 1960 pixels. Digital photographs will be taken using 210 mm and 35 mm camera lenses from an altitude of about 1500 m AGL. The camera with the 35 mm lens will be mounted on a platform in a photo hatch in the floor of the aircraft. A Control PC (personal computer) will be connected to the local network in the immediate vicinity of the photo camera using a FIREWARE interface. The camera will be switched on at the command given by the leader of survey and will continuously take photos every 3-5 seconds. The second digital photo camera will be used within the guidance system, and will be used for detailed aerial photography of walrus groups. Aiming of the photo sight at walrus groups will be accomplished by using thermal images to locate walrus groups and indicate their current positions within the camera field of view. The operator will combine photo sight indicator with walrus heat spots on a thermal image and take a picture.

D2. Details: counting digital photographs Walruses on digital photographs will be counted using counting methods similar to those used by the U.S. side (see B3 above).
V. Group size assessment: correcting for walrus groups too small to be detected

Pilot studies in 2002 and 2003 suggested that the detection limit of the thermal scanner could be treated as fixed at a group size of 4 walruses, i.e., that group sizes of 3 or fewer walruses are not detectable. A subsequent pilot study in 2005 suggested that the detection limit is temperature dependant and groups of more than 4 animals may not be detectable at colder temperatures. Therefore, additional sampling and analytical procedures are being developed to allow estimation of the scanner detection limit as a function of temperature, and to estimate the proportion of the population that is hauled out but occurs in groups too small to be detected at the time of the survey.

This additional survey effort will involve dedicating the photography plane for some time each day to conduct low-level systematic line transect surveys of walruses hauled out on ice. All visually detected groups or single walruses will be photographed for later counting, to assess the distribution of group sizes. Data from areas that were both visually surveyed and thermally scanned will be compared to determine group size detection limits of the thermal scanner under varying temperatures, which will be measured from thermal images.
VI. Satellite radio-tagging: correcting for walruses in the water

The estimate of walrus numbers hauled out on pack ice derived from thermal scanning must be corrected for walruses that are in the water and unavailable for counting (sighting availability). As walruses spend the majority of their time in water (Jay et al. unpublished data), this correction factor is critical for an accurate estimate of population size.

Walruses will be tagged with satellite radio transmitters to obtain continuous haul-out data from individual animals so that the proportion hauled out onto sea ice during a given time period can be estimated. The tags have conductivity sensors that record wet and dry intervals, corresponding to swimming/diving in seawater in contrast to hauling out (Figure 5). Knowing the proportion of walruses in the water and unavailable to the thermal scanner will allow correction of the population estimate derived from thermal scanning of walruses hauled out on ice. Satellite transmitters are necessary to obtain these data because walruses move about and are dispersed over vast areas of sea ice in the Bering Sea during the proposed survey period (spring). Successfully attached tags are designed to transmit for a minimum of 28 days after attachment.
A. Tagging equipment

Tagging will be done from two vessels, the 88 m Russian icebreaker Magadan (Figure 6) and the reinforced crabber operated by the Alaska Department of Fish and Game, the Stimson (Figure 7).

The Magadan can accommodate a scientific crew of up to 12 people. This vessel will board Russian scientists and 3 skiffs in Petropavlovsk-Kamchatski, Russia, and then sail to St. Paul, Alaska, where U.S. scientists will board. The Magadan will tag walruses in the U.S. waters of the St. Lawrence Polynya and in Anadyr Gulf, Russia.
The Stimson can accommodate a scientific crew of at least 8 people. This vessel will work out of Adak, and tag walruses from 2 skiffs in the southeastern Bering Sea, near Nunivak Island.
A shore-based twin-engine aircraft will conduct reconnaissance surveys for walruses in support of the ships’ tagging objectives. The aircraft crew will locate walruses for tagging and direct the ships to areas of unconsolidated ice to facilitate travel.
Tagging of walruses will take place from a total of 5 ship-based skiffs. Skiffs will be used to approach walruses within 15 m so that approximately 90 tags can be implanted with crossbows.
The tag is a puck-shaped transmitter that attaches to the walrus with a harpoon head mounted on a 6.0 cm × 0.6 cm stainless steel post (Figure 8). The harpoon head has broad flexible backward-projecting stainless steel fins and a cutting blade at the tip to facilitate entry through the skin and into the blubber layer. Each tag will be delivered with an arrow (2315 Lite Easton aluminum shaft filled with a solid fiberglass rod) that fits loosely into the back of the transmitter and is shot from a recurve crossbow (Excalibur, Exocet model, approximately 1.06 Joules work).
To infer whether the animal is in water or hauled out, the tag has a conductivity sensor that senses every 10 seconds whether the transmitter is submerged in seawater (Figure 5). At 30 minute intervals these sensor readings are summarized and recorded as one of two percentage classes (0-90% dry, and >90% dry). Each satellite transmission contains data of the most recent 240 30-minute intervals (5 days of data). Transmissions are suspended whenever the tag is submerged to conserve battery life. The tag currently has a battery capacity and transmission schedule that allows transmissions for at least 4 weeks. Planned modifications to the transmission schedule of the transmitters may extend longevity to ≥ 5 weeks.
B. Details: tagging

The satellite tags will be deployed on walruses from small skiffs. The skiffs will carry 3-person crews and approach walruses to within 15 m. The animal’s mid-dorsal line, slightly forward of the shoulders, will be targeted for tag attachment (Figure 8). Five skiffs will be used to deploy a total of about 90 tags in the three primary walrus breeding regions: the Southeastern Bering Sea, St. Lawrence Island Polynya, and Anadyr Gulf (Figure 9; Fay 1982, Mimrin et al. 1990). Tagging will take place in the Southeast Bering Sea and the St. Lawrence Island Polynya simultaneously from the two ships. These areas will then be aerially surveyed first, as conditions permit, while tagging takes place in Anadyr Gulf.

The amount of time spent hauled out on ice can vary among walruses due to factors related to foraging effort, thermoregulation, and animal movements. These factors may be influenced by regional differences in prey abundance, weather, ice dynamics, and walrus age, gender, and reproductive status (e.g., pregnant, lactating, non-pregnant, breeding). It is assumed (see Section IX, Statistical Assumptions) that tagged walruses are a random sample of walruses in the population, or at least, that the haul-out behavior of the tagged sample is representative of the population.
To meet this assumption, tags should be distributed throughout the range of the population and in proportion to the size of age and sex classes in the population. This assumption cannot be fully met because walruses tend to have a clumped distribution, animals of all age, sex, and reproductive classes are not equally accessible, and only animals ≥ 6 years old have blubber and skin that is thick enough for tagging.
In an effort to obtain a sample of haul out behaviors that is representative of the population, three spatial scales will be considered in selecting animals for tagging within a region: 1) selection among several aggregations of walruses that can be separated by ≥1 day transit by ship, 2) selection among groups of walruses on individual ice floes within an aggregation, and 3) selection of an individual walrus among members of a group.
If possible, larger aggregations of walruses will be allocated a greater proportion of tags; however, time constraints may preclude deployments in some aggregations. Aerial reconnaissance will be used to determine the approximate sizes and locations of aggregations in a given region (e.g., Figure 10). This information will be communicated to the ship and used to determine roughly the number of tags to deploy in each aggregation within the constraints of available ship time. Time constraints will probably limit tagging to only a few aggregations within each region.
Tags will be deployed among different groups of walruses as they are encountered throughout each day, with an effort to spread the tags across an aggregation; however, some groups will not be accessible from a skiff. Within a group, there are usually only 1-2 animals that are properly positioned for tagging because the skiff must approach from downwind and the animal’s dorsal surface must be exposed and perpendicular to the shooter. Tagging one animal usually results in the entire group of animals fleeing into the water.
Daily tag allocations may be modified slightly to target animals in under-represented sex, age, and reproductive classes as determined from cumulative tallies of animals from opportunistic observations from the skiffs each day. Only walruses at close range will be classified, and only into very broad age categories. Age is estimated primarily from the ratio of the length of the tusk to the width of the muzzle, and sex is determined by tusk and animal morphology (Fay and Kelly 1989). Some classification of local walruses may be possible from the ship, but shipboard classification often competes with tagging efforts for ship time and access to walruses.
The ship may support other projects on the same research cruise if these additional research efforts are not in conflict with the primary tagging mission. However, success of the walrus tagging crew will have precedence over other data collection efforts.
C. Aircraft/ship coordination

Reconnaissance will not be required every day, but on an as-needed basis depending on the needs of the ship to locate new walrus groups. Aerial reconnaissance for tagging in U.S. waters will be based out of Bethel and Nome during the two tagging cruises. In Russian waters, the Russian aerial survey team will provide one or two days of reconnaissance, based out of Anadyr.

Communication protocols between the reconnaissance and tagging crews, including satellite phone, marine radio, e-mail, and data logging, are detailed in cruise plans for work accomplished in the spring of 2004 and 2005 (Jay and Fischbach 2004, 2005). Each day, the aerial reconnaissance team will collect walrus and ice information over the prescribed study area. In the evening, the data will be transmitted to the ship by e-mail in text format. On the ship, the data will be imported into a GIS database. Each morning, the reconnaissance information from the previous evening and current weather and ice conditions will be assessed to plan the day’s tagging mission.
VII. Timing of survey efforts

The timing of the survey will be critical for success. There are four main issues constraining the timing:

1) The survey must take place during a period of clear skies (no or minimal cloud cover) when air temperatures are above -12º C (10º F).
2) A sample of walruses throughout the study area must be tagged with satellite radio transmitters immediately before the aerial portion of the survey.
3) The survey must take place while walruses are associated with the pack ice, before they begin their migrations toward summering areas or before they begin using terrestrial haulouts. Migration and movement toward land have historically begun in April (Fay 1982).
4) Both the U.S. and Russia portions of the survey must take place as synchronously as possible to minimize walrus movements between regions during the survey.
Assessment of Bering Sea air temperatures in March and April from 1980 to 2005 indicates that temperatures continue to rise through March and April (Figure 12). Temperatures from approximately 01 to 21 April are warm enough, on average, to ensure a high likelihood of successfully completing the survey within a three-week period. Analyses of cloud cover and wind chill factor are ongoing, and the final determination of survey timing is not yet complete.
The chronological order of the survey components is fixed. Walruses must be tagged before the aerial survey takes place, so that tags are recording ice haul-out behavior during the aerial survey. The tags have a life expectancy of approximately 4 to 5 weeks. For example, in order for the tags to be functioning during a U.S. aerial survey from 01 to 21 April, both ships must leave their respective ports on about 22 March (Figure 13). Walruses will be tagged in U.S. waters from approximately 23 March to 28 March. Once walruses in U.S. waters are tagged, the U.S. aerial survey will begin as the Magadan heads to Russia for further tagging in the Gulf of Anadyr (Figure 13). Under this scenario, aerial surveys in Russia will begin in early April after completion of tagging.
One aircraft will assist with searching for walruses and assessing ice conditions for the ship-based tagging crew while it is in U.S. waters (approximately 22 to 28 March). Once tagging operations in U.S. waters are complete, the same aircraft will then begin duty as the thermal scanning plane. The second aircraft will arrive in Nome or Bethel to begin the aerial photography at the same time. By beginning aerial surveys as soon as possible after walruses are tagged, the number of walruses whose tags will still be working will be maximized, since some tags fail earlier than others.

VIII. Survey design and statistical analysis

A. Survey transect design

Transect midlines will be spaced 12 km apart, to give a strip width that matches the area covered by the thermal scanner, and will be oriented in a north-south direction so they are predominantly perpendicular to the ice edge (see Figure 14 for example). Transects will be grouped into survey blocks of the approximate area that can be surveyed in one day of good survey conditions at the specified intensity, when flying time to and from the survey block is considered. A systematic sample of transects will be surveyed within each block. The final configuration of transects and survey blocks will be determined shortly before the survey takes place, when the extent of the pack ice is known. Surveys will be done only in areas where the sea floor depth is ≤200 m.

B. Statistical methods

Statistical analysis of the data collected in the U.S. and Russia will be done cooperatively, so that a single total population estimate is calculated.

Estimating the size of the Pacific walrus population from the survey data will involve the following sequence of steps: (1) estimating the relation between thermal signature intensities and numbers of walruses in thermally detected groups, (2) estimating the proportion of the population hauled out on the ice during the survey, (3) estimating the total number of walruses on surveyed transects, (4) estimating the total number of walruses in each survey block and summing to estimate the total population size, (5) adjusting this estimate to account for walruses hauled-out on the ice but in groups too small to be detected by the scanner, and (6) estimating the variance and confidence intervals for the population estimate. Each of these steps is described in more detail below.
B1. Relating thermal signature intensity to group size. Data from photographed groups will be used to develop a regression model relating thermal signature intensity to group size (Figure 4). This model will be used to estimate the number of walruses in each thermally detected group that was not photographed. Pilot studies conducted near St. Lawrence Island in 2002 and 2003 indicated that variances of the photographic counts were proportional to the mean, or the square of the mean, counts. Therefore, we will use a generalized linear model (McCullagh and Nelder 1989) with an identity link and a Poisson or gamma distribution to estimate the relation between numbers of individuals and thermal signatures in the photographed groups. Notation is summarized in Appendix I. The form of this model is
with the Poisson distribution, or
with the gamma distribution. Here, hgtb is the thermal index for group g on transect t of block b, α is the minimum size group that can be detected by the scanner, and β and φ are parameters that are estimated by maximum likelihood. The estimators for these parameters and their variances do not have explicit forms, so they have to be determined numerically. Estimation will be accomplished with SAS PROC GENMOD software (Statistical Analysis Systems, Cary, North Carolina, USA).
B2. Estimating proportion hauled out. The proportion of the population that was hauled out on the ice and therefore potentially available to be detected during the thermal scanner survey will be estimated with the satellite telemetry data from the tagged walruses as
where pi is the proportion of the relevant time period that walrus i was available to be detected. In practice, we expect to estimate separate availability proportions for different sets of blocks using only data from restricted areas or time periods. Here we consider only a single overall estimate that will be used for all blocks. Modifications required for using separate estimates are straightforward.

B3. Estimating totals for surveyed transects. The number of walruses on a surveyed transect will be estimated by summing the counts of individuals in all the photographed groups and the estimated counts in all the detected groups that were not photographed and then dividing this total by the estimated proportion of the population that was available to be detected. For transect t in block b, we have:
where ygtb is the number of walruses in group g on transect t of block b, photographed groups are indexed 1, ... , ctb, and groups that were not photographed are indexed ctb+1, ..., Gtb. If there are no photographed groups on a transect, then ctb = 0.

B4. Estimating totals for blocks and the population. The total population size is estimated as a sum of separate ratio estimators (Cochran 1977) of the totals for each block:

Atb is the area of transect t in block b, Tb is the number of transects in block b, tb is the number of surveyed transects in block b, and B is the number of blocks.
We anticipate that estimates for Russian survey blocks will be obtained using similar methods, modified as necessary to accommodate any differences in scanner function. Estimated totals for Russian survey blocks will be included in sum (6) to obtain the total population estimate.
B5. Accounting for walruses on ice in groups too small to be detected by the scanner. The population estimator (6) assumes that the number of walruses hauled-out on the ice but in groups too small to be detected by the infrared scanner was negligible. We will use data from a low level line-transect survey conducted with the photography plane on a subsample of the transects covered in the scanner survey to assess this assumption and to adjust the population estimate as necessary. The adjusted population size will be estimated as

where is the estimate of the number of hauled-out walruses in groups of size < sb obtained with standard line-transect methods (Buckland et al. 1993), and sb is the detection limit for the scanner (i.e., the smallest group size that can be detected in the scanner survey) for block b.

We note that to adjust the total population estimate, it is not necessary to estimate the numbers undetectable hauled-out walruses from the low-level survey in each block. Rather, it would be sufficient to estimate just the ratios ub/db, where db is the number of detectable hauled-out walruses in block b. If enough groups were observed, these ratios could estimated directly from the observed groups in narrow strips where it could be assumed that all groups were detected in the low-level survey. The adjusted population size could then be estimated as
Methods for estimating this ratio directly while accounting for the likely dependency of detection probability on group size at greater distances would have to be developed.

B6. Estimating variance. The variance and confidence intervals for the population estimate will be obtained with a bootstrap procedure based on the general approach of Booth et al. (1994) for finite populations. The procedure involves generating a series of simulated populations, estimating statistics of interest by resampling from each simulated population, and then averaging these statistics over the simulated populations.
We generate simulated populations of transects (with associated walrus observations and low level line-transect survey segments) for each block by first replicating the complete set of surveyed transects in the block as many times as possible without exceeding the total number of potential transects in the block. We then add a random sample without replacement from the surveyed transects to complete the population of potential transects. Bootstrap survey samples are obtained by drawing random samples without replacement from the simulated populations to give the same number of transects as in the original survey. For each bootstrap survey sample, we also obtain a bootstrap sample of photographic counts for fitting the regression model by taking a random sample with replacement from the photographed groups to give the same sample size as for the original survey. We also obtain a bootstrap sample of transmitter data to go with each bootstrap survey sample by taking a random sample with replacement from the tagged walrus data to give the same number of tagged walruses as in the original sample. If there is any stratification used for estimating region or time-specific regression parameters or haul-out proportions in the original survey, then bootstrap resamples are obtained with the same stratification.
Estimation for each bootstrap sample follows the same procedure as for the original sample. We obtain K1 bootstrap samples and the associated estimates of population size for each simulated population and then calculate the standard error and 2.5 and 97.5 percentiles of these K1 estimates. We repeat this process for K2 simulated populations, taking the averages of the standard errors and 2.5 and 97.5 percentiles as our estimates of standard errors and 95% confidence limits for the estimates from the original survey. K1 and K2 need to be large enough for the estimates to converge. Pilot study results suggest that K1 = 100 and K2 = 500 should be sufficient, but this should be checked and the number of simulated populations and bootstrap samples per population increased as necessary to obtain convergence in any particular application.
We may also consider a second type of approach for estimating variance based on modeling the spatial structure of the data (Cressie 1991). Properly accounting for this spatial structure may ultimately give more precise estimates of the population size.
IX. Statistical assumptions

The following is a partial list of assumptions required for unbiased estimation of population size and its variance for this study design. Also included is some discussion of survey design features related to these assumptions.

1. The data collected cover the entire range of the population at the time of the survey. A preliminary delineation of this area will be based on historical walrus distributions and bathymetry, but a final delineation will be made immediately before the survey based on the current ice distribution.
2. Survey blocks form a complete partition of the study area and strip transects form a complete partition of each block. This means that blocks completely cover the study area with no overlap and that strip transects completely cover the area of each block with no overlap. Blocks will be the smallest areas beyond the surveyed transects for which we will estimate population size. Each block will be sized so it can be surveyed in a single day at the specified survey intensity. Transect widths are determined by the width of the scanner coverage, so block boundaries will have to be adjusted to the nearest whole transect. One other criterion to be considered in designing blocks is that variance of the population estimate will be minimized by making transects within a block as homogenous as possible with respect to walrus density.
We use the terms “transect” and “strip transect” interchangeably. The survey aircraft will fly along the midlines of the selected strip transects.
3. There is no net movement of walruses among transects or blocks while they are being surveyed. Block boundaries will be located in areas expected to have relatively low walrus density and the survey will be completed as rapidly as possible to minimize potential for net movement.
4. Surveyed transects are a random sample of transects in each block. This assumption is required to minimize bias of the variance estimators. We will actually use a systematic sample of transects because that will likely provide a more accurate estimate of population size, but this may result in the variance estimates being somewhat conservative (Wolter 1984). An approximate (100/k)% systematic sample of transects in a block is obtained by randomly selecting one out of the first k transects and then every k’th transect beyond that. Bias due to systematic sampling of transects would not be a problem for spatial estimates of variance.
5. Satellite radio-tagged walruses are a random sample of walruses in the population or at least, the haul-out behavior of the tagged sample is representative of the population. To meet this assumption, transmitters will need to be distributed throughout the range of the population and in proportion to age and sex class proportions in the population. We expect that, at minimum, this will require separate transmitter deployment efforts in three regions, the Southeast Bering Sea, the St. Lawrence Island Polynya, and Anadyr Gulf (Figure 9).
6. Photographed groups are randomly distributed throughout the survey area and time period or at least, are obtained under conditions representative of those during the survey.
7. All groups detected by the scanner contain only walruses. Other ice seals are not likely to form groups large enough to be detected by the scanner.

Literature Cited

Booth, J.G., Butler, R.W., and Hall, P. 1994. Bootstrap methods for finite populations. Journal of the American Statistical Association 89(428):1282-1289.

Bowden, D.C., White, G.C., Franklin, A.B., and Ganey, J.L. 2003. Estimating population size with correlated sampling unit estimates. Journal of Wildlife Management 67(1):1-10.
Burn, D.M., Webber, M.A., and Udevitz, M.S. In press. Application of airborne thermal imagery to surveys of Pacific walrus. Wildlife Society Bulletin.
Cochran, W.G. 1977. Sampling techniques, third edition. John Wiley & Sons, New York.
Cressie, N. 1991. Statistics for spatial data. John Wiley & Sons, New York.
Efron, B. 1982. The jackknife, the bootstrap and other resampling plans. Society for Industrial and Applied Mathematics, Philadelphia, Pennsylvania.
Fay, F.H. 1982. Ecology and biology of the Pacific walrus, Odobenus rosmarus divergens Illiger. North American Fauna 74:1-279.
Fay, F.H. and Burns, J.J. 1988. Maximal Feeding Depth of Walruses. Arctic 41(3):239-240.
Fay, F.H., Eberhardt, L.L., Kelly, B.P., Burns, J.J., and Quakenbush, L.T. 1997. Status of the Pacific walrus Population, 1950-1989. Marine Mammal Science 13(4):537-565.
Fay, F.H., and Kelly, B.P. 1989. Development of a method for monitoring the productivity, survivorship, and recruitment of the Pacific walrus population. Final Report, OCSEAP Study MMS 89-0012. Minerals Management Service, Anchorage, Alaska. 51 pp.
Fay, F.H., Kelly, B.P., Gehnrich, J.L., Sease, J.L., and Hoover, A.A. 1984. Modern populations, migrations, demography, trophics, and historical status of the Pacific walrus. Final Report R.U. #611. NOAA Outer Continental Shelf Environmental Assessment Program, Anchorage, Alaska. 142 pp.
Fay, F.H., Kelly, B.P., and Sease, J.L. 1989. Managing the exploitation of Pacific walruses: a tragedy of delayed response and poor communication. Marine Mammal Science 5:1-16.
Garlich-Miller, J. and Jay, C.V. 2000. Proceedings of a workshop concerning walrus survey methods. USFWS R7/MMM Technical Report 00-2. 92 pp.
Gilbert, J.R., Fedoseev, G.A., Seagars, D., Razlivalov, E., and LaChugin, A. 1992. Aerial census of Pacific walrus, 1990. USFWS R7/MMM Technical Report 92-1. 33 pp.
Gilbert, J.R. 1999. Review of previous Pacific walrus surveys to develop improved survey designs. In Garner, G.W., Amstrup, S.C., Laake, J.L., Manly, B.F.J., McDonald, L.L., and Robertson, D.G. (eds), Marine Mammal Survey and Assessment Methods. A. A. Balkema, Rotterdam. 287 pp.
Hills, S. and Gilbert, J.R. 1994. Detecting Pacific walrus population trends with aerial survey - a review. Transactions of the North American Wildlife and Natural Resource Conference.
Hunt, G.L., Jr. and Stabeno, P.J. 2000. Ecosystem responses to weather patterns in the Southeastern Bering Sea. Arctic Research of the United States 14:17-24.
Jay, C.V. and Fischbach, A.S. 2004. Cruise Plan for Walrus Tagging Aboard the P/V Stimson Bering Sea, April 2004. U.S. Geological Survey, Alaska Science Center, Anchorage, Alaska. Unpublished internal document. 22 pp. Available at
Jay, C.V. and Fischbach, A.S. 2005. Cruise Plan for Walrus Tagging Aboard the P/V Stimson Bering Sea, March 2005. USGS, Alaska Science Center, Anchorage, Alaska. Unpublished internal document. 19 pp. Available at
Jay, C.V., Heide-Jørgensen, M.P., Fischbach, A.S., Jensen, M.V., Tessler, D.F., and Jensen, A.V. In press. Comparison of remotely deployed satellite radio transmitters on walruses. Marine Mammal Science 22(1):000-000.
McCullagh, P. and Nelder, J.A. 1989. Generalized linear models, second edition. Chapman and Hall, Boca Raton, Florida.
Mimrin, N.I., Smirnov, G.P., Gaevsky, A.S., and Kovalenko, V.E. 1990. The distribution and abundance of walruses in the Anadyr Gulf of the Bering Sea. Zoologichesky Zhurnal 69:105-113. (Translated from Russian).
Scribner, K.T., Hills, S., Fain, S.R., and Cronin, M.A. 1997. Population genetics studies of the walrus (Odobenus rosmarus): a summary and interpretation of results and research needs. In Dizon, A.E., Chivers, S.J., and Perrin, W.F. (eds). Molecular Genetics of Marine Mammals. Marine Mammal Science 3:173-184.
Seber, G.A.F. 1982. The estimation of animal abundance, second edition. Oxford University Press, New York, New York.
Udevitz, M.S. 2005. Statistical procedures for estimating Pacific walrus population size. Technical report, Alaska Science Center, U.S. Geological Survey, Anchorage, Alaska. 9 pp.
U.S. Fish and Wildlife Service. 2001. Stock assessment of the Pacific walrus (Odobenus rosmarus divergens).
Wolter, K.M. 1984. An investigation of some estimators of variance for systematic sampling. Journal of the American Statistical Association 79(388):781-790.

Figure 1. Approximate ice coverage in the Bering Sea in late spring. Shown are the approximate coverages for a heavy ice year (1997, top, 835,000 km2) and a light ice year (2002, bottom, 521,000 km2). The blue line indicates the U.S.-Russian border.

Figure 2. Example of an aerial digital photograph of walrus groups hauled out on ice and the matching thermal image of the same walruses. Data were collected in April 2004.

Figure 3. Example of frequency histogram of AMS thermal imagery showing temperature threshold between the background environment and walruses that occurs at

-2.81° C. Pixels to the right of the threshold value have some portion of their area covered by walruses.

Figure 4. Example gamma regression of walrus group size as a function of thermal signature for AMS thermal imagery at 4 m spatial resolution (Burn et al. in press). Dashed lines represent 95% confidence intervals. Data were collected in April 2002.

Figure 5. Walrus haul-out chronology data from 2004, collected from satellite tags which are of similar design to those that will be used in the 2006 walrus population survey (* = anomalously long dry period at the end of the haul out chronology, which suggests the tag may have malfunctioned or became detached from the animal on land or ice).



Nationality (Flag state):



Russian Federation

Owner Address:

103685,6, Nikolskiy pereulok, Russia


Far Eastern Shipping Company

Operator Address:

15, Aleutskaya Street, Vladivostok, 690990, Russia

Overall length (meters):


Maximum draught (meters):


Displacement (metric tons):


Main Engines:

WARTSILA SULZER Type, 8R 32 4 x 2685 kW

Cruising & Maximum speed (knots):

14, 17.5

Call sign:


Method and capability of communication:


Name of master:

Roman Nikolayevich Mogilyan

Number of crew:


Number of scientists on board:


Figure 6. Specifications of the icebreaker Magadan.


PV Stimson

Nationality (Flag state):



State of Alaska


Dept of Public Safety, Fish and Wildlife Protection

Overall length (feet):


Maximum draught (meters):


Displacement/Gross tonnage:

443 tons


Diesel Electric

Cruising & Maximum speed:

8, 11

Call sign:

PV Stimson

Method and capability of communication (including emergency frequencies):

AV Radio: channel 129

Marine VHF: channel 16

Sat phone: 011- 881- 631- 459 - 136

Name of master:

Troy Magnuson

Number of crew:

1 state trooper, 4 crew

Number of scientists on board:


Onboard helicopter support:


Figure 7. Specifications of the Patrol vessel Stimson.


Telonics Inc., Mesa, Arizona, U.S.A.



Transmitter dimensions

(excluding antenna)

5.2 × 2.8 cm (puck-shaped disc)

Approximate transmitter weight in air

91 g

Transmission duty cycle (daily)

18 hrs on, 6 hrs off

Transmission output power

0.25 W

Transmission repetition interval

60 sec

Argos message length

31 bytes

Argos transmission frequency

401.618 to 401.680 MHz

Modulation (BPSK)

-1.1 to +1.1 ±0.1 Radians

Quiescent current (typical)

3 μA

Spurious emissions

-45 dB

Transmitter delivery system


Figure 8. Specifications of the walrus satellite tag that will collect haul-out behavior data for estimating sighting availability during the walrus aerial survey.

Figure 9. The three study areas (Anadyr Gulf, St. Lawrence Island Polynya, and Southeastern Bering Sea) where tagging of walruses will take place.

Figure 10. Examples of aerial reconnaissance track lines, plotted during walrus tagging in April 2004.

Figure 11. Example of ice conditions during walrus tagging from skiffs in spring 2004 and 2005, Southeastern Bering Sea.

Figure 12. Percentage (y axis) of temperatures for April and May (x axis) from 1980-2005 that falls within each of 5 temperature categories: green, >30º F; blue, 20-30º F; yellow, 10-20º F; orange 0-10º F; red, <0º F.

Figure 13. Chronological order of the tagging and aerial survey missions, and estimated life expectancy of the satellite tags.

Figure 14. Example of transect and survey block layout used for pilot surveys in 2005. Transects were spaced 12 km apart, and numbered blocks defined areas that could be surveyed within one day. Surveys took place only over areas where the depth of the sea floor was ≤200 m.

Appendix I. List of statistical notation used in this report.
N = walrus population size
α = smallest group of walruses detectable with the scanner (i.e., detection limit)
β = slope parameter for the generalized linear model relating thermal signature intensity to walrus group size
φ = dispersion parameter for the generalized linear model relating thermal signature intensity to walrus group size
w = number of transmitters deployed (i.e., number of tagged walruses)
pi = proportion of time during the survey period that tagged walrus i was hauled out on the ice and therefore available to be detected
p = proportion of the walrus population available to be detected during the survey
B = number of blocks
Ab = area of block b
Nb = number of walruses in block b
ub = number of walruses hauled-out on the ice but in groups to small to be detected by the scanner in block b
db = number of walruses hauled-out on the in groups large enough to be detected by the scanner in block b
sb = smallest group size that can be detected by the scanner
Tb = total number of transects in block b
tb = number of surveyed transects in block b
= average area per transect for surveyed transects in block b
Rb = ratio of total number of walruses on surveyed transects to total area of surveyed transects in block b
Atb = area of transect t in block b
Ntb = number of walruses on transect t of block b
Gtb = number of walrus groups detected by the scanner on transect t of block b
ctb = number of walrus groups photographed on transect t of block b
ygtb = number of walruses in group g on transect t of block b
hgtb = thermal signature intensity for walrus group g on transect t of block b

Appendix II. List of cooperators by agency.
Marine Mammals Management, U.S. Fish and Wildlife Service, 1011 E. Tudor Road, Anchorage, AK 99503, U.S.A. Phone: (907) 786-3800
Rosa Meehan, Ph.D. Chief.

Management of marine mammals, Alaska Native relations, human impacts on marine systems, coordination with Russia

Brad Benter, B.S. Biological Science Technician.

Aerial digital photography, subsistence harvest assessment

Douglas Burn, M.S. Supervisory Wildlife Biologist.

Thermal imagery systems design and development, survey design, sea otters, marine mammal conservation

Charles S. Hamilton, M.S. Wildlife Biologist.

Management of marine mammals, regulatory development, international relations and conservation

Joel Garlich-Miller, M.S. Wildlife Biologist.

Pacific walrus life history, ecology, conservation, and management

Suzann G. Speckman, Ph.D. Wildlife Biologist.

Development of Study Plan, coordination among U.S. and Russian specialists, marine ecology, oceanography, fisheries acoustics

John Trent, Ph.D. Supervisory Wildlife Biologist.

Pacific walrus management, Alaska Native relations, subsistence harvest assessment

Alaska Science Center, U.S. Geological Survey, 4230 University Drive, Suite 201, Anchorage, AK 99508, U.S.A. Phone: (907) 786-3512
Chadwick V. Jay, Ph.D. Research Ecologist.

Walrus natural history and ecology, radio telemetry, benthic ecology

Anthony Saul Fischbach, M.S. Wildlife Biologist.

Arctic marine mammals, GIS, radio telemetry, aerial surveying, animal behavior

Mark S. Udevitz, Ph.D. Research Statistician, Biology.

Methods for sampling, estimating parameters, modeling dynamics of wildlife populations

GiproRybFlot, Research and Engineering Institute for the Development and Operation of Fisheries, 18-20 Malaya Morskaya str., St. Petersburg 190000, Russia

Chernook Vladimir, Ph.D. in Geography, Head of Department.,

Development of Study Plan, coordination among Russian and U.S. specialists, supervision of Russian aerial surveys, development of walrus aerial survey method, participation in air surveys and data analysis

Melentyev Vladimir, Ph.D. in Physics and Mathematics, Leading Researcher.

Ice parameters of the Bering Sea studies, walrus-ice correlation

Vasiliev Alexander, Chief Specialist,

Technical and software support of aerial surveys, participation in aerial surveys, software development for walrus survey data processing

Shubina Marina, M.S., researcher.

Aerial survey data processing

Chelintsev Nikita, Ph.D. in Biology, Leading Researcher.

Mathematic support of aerial surveys, mathematic processing and analysis of aerial survey data.

Krukova Natalia, Engineer.

Processing of images of walrus aggregations

ChukotTINRO, Pacific Research Institute of Fisheries and Oceanography, Laboratory of Marine Mammals Study, P.O. Box 29, Anadyr, Chukotka 689000, Russia
Myasnikov Vladimir, M.S. in Biology, Director.

Organization and financial support of aerial surveys in Chukotka

Kochnev Anatoliy, Head of Laboratory.

Organization of walrus tagging from the Russian side, walrus tagging onboard of icebreaker, processing and analysis of walrus aerial survey data

Litovka Denis, Researcher.

Participation in aerial surveys, preliminary aerial survey data processing

Kudryavtsev Alexander, Junior Researcher.

Participation in aerial surveys, preliminary aerial survey data processing

1   2

База данных защищена авторским правом © 2016
звярнуцца да адміністрацыі

    Галоўная старонка