Ungulate

Recommended method(s): Total counts, Sample-based counts (see Section 4 for species-specific recommendations).

Determining absolute abundance requires estimation of both the bias and precision of the survey methods. Total counts and sample-based survey designs (e.g., stratified random block surveys) are the two basic methods for estimating absolute abundance. Total counts are often possible for species such as mountain goats, which occupy discrete mountain blocks. However, for other species such as moose and elk, which have a more contiguous distribution over the landscape, cost and logistic constraints favour using sample-based survey methods. These results are then extrapolated to the entire survey area to estimate absolute abundance with statistical confidence limits. Both methods require independent estimates of sightability, in order to correct survey bias for animals missed.

Surveys should be done when the target species is most restricted in distribution, or most visible. Generally, aerial surveys for ungulates must encompass areas of at least 200 km 2 to enclose discrete populations. In most cases ungulate survey areas will be much larger. When survey areas are less than 200 km 2, total counts should be used rather than sample counts. In open habitats, where sightability is high and search times can be reduced, areas up to 600 km 2 can be censused for reasonable cost using total count methods.

3.6.1 Total counts

Total counts are the simplest in principle. They are intended to enumerate all animals using 100% flight coverage of the study area. Alpine areas are usually small, and thus the technique is practical for surveying mountain sheep and goats, and sometimes caribou.

It is usually difficult or impossible to actually count every animal. Some animals are reclusive, and others are simply missed. Without some estimate of the numbers missed, or sightability of animals, the accuracy of total counts is always in question. Only in open habitats where sightability bias is low, will total counts approach an estimate of absolute abundance. Replicating surveys can help to determine how variable actual sightability may be.

Definition of survey sample units is recommended for total counts (TC). However, instead of randomly selecting sample units to survey, all the units are counted. Delimiting the sampling units provides clear definition of the areas to be surveyed. In-flight survey procedures for the sample units follow those outlined for sample-based counts.

3.6.2 Sample-based counts

Sample-based surveys are required wherever it is impractical to survey the entire area occupied by a population. In sampling surveys, a portion of the population is counted within defined sample units (e.g., quadrats or blocks). The results are then used to estimate animal abundance throughout the study area. Sightability can be increased by increasing search intensity, but for some low density populations or cryptic species living in dense cover (e.g., coast black-tailed deer), no aerial search methods are adequate.

In all survey methods where animals are counted in sample units, there is an associated edge effect. When an animal is near a sample unit boundary, a decision must be made to count it as "in" or "out". Keen surveyors may prefer to include those animals rather than ignore them, which biases the results upward. In stratified random block surveys, careful definition of sample unit boundaries using obvious features will reduce that problem. Where the problem persists, only 50% of such observations should be included as “in”.

If the animals being surveyed are expected to move appreciably, adjacent survey sub-units should be completed within the shortest possible time to minimize chances of double- or under-counting due to animals shifting between units. Choosing a larger sample unit size may also help to alleviate this problem.

Sample units

Sample units (SU’s) may include both blocks or quadrats. Blocks are irregularly shaped polygons (Gasaway et al. 1986:6-10), which are variable in size. Surveys should, however, attempt to keep block size as constant as possible, since large variations in block size will increase the variance of the survey estimate. They are particularly well suited to rough, mountainous terrain and clumped animal distributions, both conditions which apply well to ungulates in British Columbia. Quadrats are square (e.g., 5 x 5 km 2) or rectangular in shape (e.g., 3 x 8 km 2). The major disadvantage of quadrats are locating SU boundaries. However, with GPS and LORAN-C, this is becoming less of a problem.

Searching sample units is a dynamic process, and requires that the navigator define the flight pattern prior to initiating the search. The flight pattern must be appropriate to survey all terrain, and enable identifying, plotting, distinguishing, counting, and classifying each group of animals. Gasaway et al. (1986:29) provide a useful example of flight patterns in varied terrain. Large sample units may require subdivision into smaller units. This reduces the chance of double counting, since flight lines can be shortened, the time between passes is reduced, and animals located on adjacent lines can be more easily recalled and distinguished. That is especially important for deer and elk, and even sometimes moose, which may move long distances in a short time when disturbed. Goats and sheep may move quickly to different elevations.

Regardless of whether blocks or quadrats are chosen as the SU, all SU’s should be pooled into strata of differing density, thereby assigning as much total variance as possible to differences among strata (Gasaway et al. 1986:7-19). In addition to increasing precision, stratification allows optimal allocation of sampling effort, thereby getting the most precision possible. Even a poor stratification will likely improve precision compared to unstratified random sampling.

In stratified random sampling, each sample unit is selected randomly without replacement from each strata using a computer or a table of random numbers. The standard procedure is to initially count at least 5 sample units from each stratum (e.g. 5 low, 5 medium, and 5 high sample units).

Gasaway et al. (1986) describe the sampling method for stratified random block (SRB) surveys in six basic steps, which should be applied to all SRB surveys in British Columbia.

1.Define the population of interest, the information required, and the survey area using reconnaissance surveys if needed.

2.Delineate all possible sample units on a topographic map or air photos of the survey area.

3.Divide the survey area into 3 to 5 strata, based upon the expected number of animals in each SU, determined from existing information or from a stratification survey.

4.Select a number of sample units randomly from each stratum, if the entire area cannot be surveyed with existing funds.

5.Fly surveys of selected sample units and employ standard methods to estimate the percentage of animals that were missed.

6.Calculate an estimate of population size and structure including confidence intervals around the estimates.

Following are some specific recommendations for in-flight block survey procedures (#5 above):

1.Animals are counted and classified using transects located within the sample unit. Transects are straight in flat terrain and follow contour lines in mountains. During the search, the observers are responsible for the field from the centre of the aircraft to half way to the next transect on their side of the aircraft. Animals spotted outside the transect width should be noted on the map and the pilot and navigator informed. The pilot and navigator should then decide based on the conditions and location of the animal(s) whether to proceed to count and classify the group or wait until flying the next transect.

2.Sample units in mountainous terrain should be surveyed from low to higher elevation, because animals move more slowly up-slope, and you will be less likely to miss them. The best survey pattern is usually along contours, starting from the lowest elevation and working systematically upward in jumps of 100 to 150 m, depending upon topography and cover. When tracks are sighted in fresh snow, circle the area to help locate the animals. Climbing slightly can assist in spotting animals in tall timber.

3.When a group of animals is seen, the pilot and navigator are directed to the area. If the group begins to move, make a mental note of where the animal(s) were first sighted. The group is then counted and classified as quickly as possible, using the helicopter to herd or hold the animals as necessary. Remember to scan for other groups as you approach. It is common to concentrate on one group and miss others. Following classification, you may need to return to the location where the animals were first spotted in order to collect the required auxiliary habitat-related information.

For standard stratified random block surveys, program MOOSEPOP can be used to calculate population estimates, confidence intervals, and estimates of numbers in each sex/age class. For stratified random quadrat surveys, a slight increase in precision may be obtained by using the statistical procedures outlined by Caughley (1977:28-30).

Optimal allocation of sampling effort

Optimal allocation is a means to use available funds most efficiently to attain the highest possible level of precision. Optimal allocation allocates sampling effort among strata on a daily basis during the survey, based on observed sample variance estimates of the strata (Gasaway et al. 1986). Five sample units in each stratum (e.g., high, medium and low) is generally considered the minimum sample size, prior to initiating optimal allocation. Further sampling effort is then applied to the strata based on the results of the allocation formula.

For large survey areas with many blocks, optimal allocation may limit your ability to preplan the order in which blocks will be surveyed. Adjacent units may have to be done several days apart. Ferry time will be substantially increased and refueling requirements cannot be as accurately preplanned. Nonetheless, optimal allocation will still provide better precision, than equal sampling of strata for a fixed cost.

The recommended procedure for calculating optimal allocation of sampling effort in SRB surveys is described by Gasaway et al. (1986:43-52), and should be consulted. These calculations can be performed with the program MOOSEPOP, which is available from Daniel J. Reed, Alaska Department of Fish and Game, 1300 College Road, Fairbanks, Alaska 99701. The program also performs optimal allocation of effort between standard and intensive searches (see Section 3.6.3.).

3.6.3 Estimating numbers missed

Whenever a wildlife population is surveyed, some animals are inevitably missed (referred to as sightability or visibility bias). That results in an underestimate of the population size. Vegetation cover is probably the single, most important factor affecting the sightability of animals in British Columbia. This is because much of the ungulate winter range in the lower two-thirds of the province consists of closed or semi-closed canopy coniferous forests.

There are a number of ways of improving sightability when conducting an aerial survey:

1.Use slower flying aircraft with good visibility.

2.Use as many good observers as the aircraft and conditions will allow.

3.Use higher search intensity (min/km 2) in undulating terrain and dense forest cover.

4.Circle observed animals to detect animals nearby.

5.Use overlapping transects to ensure adequate coverage in dense forests.

6.Use smaller sample units which require less overall search time, to reduce observer fatigue and thus reduce sightability bias.

7.Choose a time period in the annual cycle during which the highest proportion of the population is in open or relatively open habitats.

8.Choose the time of day when animals are most active and/or least likely to be using thick cover.

9.Choose weather/light conditions that maximize animal activity and visibility.

All those methods help to improve sightability, but some animals will still be missed. While increasing sample size can improve the precision of a population estimate, improving accuracy requires an estimate of the number of animals missed.

If sightability is not accounted for, the population estimate will be biased. That bias usually means that the reported confidence intervals will not include the true population density the specified percentage of the time (Gasaway and Dubois 1987). Thus, missing 20% of moose in the surveyed sample units could result in a confidence interval being closer to 50% rather than the stated 90%. Many past surveys in British Columbia have not provided an estimate of sightability. Rather, in most cases, sightability has either been ignored or an informed guess, based on sightability estimates determined from other studies, has been used to adjust the population estimate. Guesses should be discouraged, as they do not provide objective measures that can be used consistently over time.

There are three methods recommended for estimating sightability: two stage sampling, mark /resight of animals, and sightability models. The latter has only recently been developed for use in Idaho (Unsworth et al. 1994), and still requires more testing for elk and moose under conditions in British Columbia. This method is included, however, as it will likely become the desired estimation method following additional testing.

Two stage sampling

Two stage sampling uses a repeated survey of animals in a given sample unit. The "two samples" can then be used to estimate a ratio of how many animals are being missed in the survey. There are two basic approaches used. One resurveys a sample unit or portion thereof, immediately after the initial survey (Gasaway et al. 1986). The other uses two observers simultaneously, and compares the counts to estimate sightability (Magnusson et al. 1978; Cook and Jacobsen 1979; Caughley and Grice 1982).

The Gasaway method (Gasaway et al. 1986:30-36) is the recommended two stage sampling method. It uses an intensive re-survey of a portion of a sample unit in a stratified random block count. It is flown in medium and high density blocks using a 2 to 3-fold increase in search intensity (min/km 2). The number of blocks to be resurveyed intensively is calculated by the optimal allocation method (Gasaway et al. 1986:48-49). The sightability correction factor is calculated by dividing the number of animals seen during the intensive search, by the numbers seen in the standard search. Minimizing time between the standard and intensive search is essential to reduce the chances of animals moving in or out of the block.

The two stage sampling method is most useful in relatively open habitat for large species, such as moose in the northern 1/3 of the province. In much of the rest of British Columbia and for smaller, more cryptic ungulates, the "intensive" survey will still miss a substantial number of animals. Under these conditions, mark/resight methods are preferred. Program MOOSEPOP should be used to perform the calculation for two stage sampling.

Mark and resight

Although relatively costly, and at times limited in accuracy, mark and resight surveys often provide the best method for adjusting survey results for sightability bias. Mark-resight is a modification of mark-recapture methods. Many texts on animal sampling, such as Caughley (1977) and Krebs (1989), provide excellent reviews and discussion of mark-recapture techniques, and at least one of these texts should be consulted prior to initiating a mark-resight survey.

The basic mark and resight procedure is to randomly mark a portion of the animals and then resurvey the area. The proportion of the marked animals missed during standard surveys can be used to estimate the proportion of the population missed in each survey unit. A number of methods can be used to mark animals including ear tags, collars, and paint ball marking. Applying visual marks such as ear tags and collars requires capture and handling. Since the cost of handling large mammals is high, it is best to install radiocollars on captured animals, as radios help to ensure that the assumptions of mark-resight sampling are being met. Paint marking using an air-powered paint gun allows animals to be marked without handling them, and many animals can be marked in a short time (e.g., 20 per hour for goats in the Babine Mountains, Cichowski et al. 1991). Currently, the paint marking technique is only recommended for mountain goats, although research on other ungulates is encouraged. Where paint marking is used, animals should be paint marked on both sides or on the back to make later identification of marked animals more certain.

The Petersen mark-recapture method is the simplest and probably the most commonly used method in aerial survey work. It involves marking animals in one session and surveying the area again in another session.

The Peterson method relies on several statistical assumptions:

1.all animals have equal probability of being marked;

2.animals are not leaving (emigration/mortality) or entering (immigration/natality) the survey population between marking and recapture;

3.marking animals will not affect their future catchability (i.e., probability of detection of a marked animal is the same as an unmarked animal);

4.no marks are lost between the marking and recapture (resight) periods;

5.all marked animals counted during the survey are identified as marked; and

6.animals are not counted more than once during a sample.

Because different age/sex classes often occupy different habitats, the assumption of equal marking probability is unlikely to be met under most conditions. Also, the assumption of equal probability of sighting marked and unmarked animals may be violated, if marked animals have a greater tendency to detection from aircraft.

Both visual marks and radio-collars can be used, however the latter are recommended because they enable validation of four of the assumptions stated above:

1.radio-collared animals can be monitored after the survey to determine whether marked animals left or entered different survey areas;

2.the movements and behaviour of collared animals can be monitored to assess their detectability relative to unmarked animals;

3.loss of marked animals can be detected; and

4.by relocating marked animals immediately after a survey area is completed, it is often possible to determine the reasons that animals were missed, including failure to spot marks.

Program NOREMARK (White 1996) should be used to compute estimates of population size for an ungulate population with a known number of marked animals and one or more resighting occasions. This program allows for resightings or recaptures of marked animals, and the number of unmarked animals observed on one or more replicate surveys are used to compute population estimates. Four different estimators are provided in the program, each with different assumptions about individual heterogeneity and immigration and emigration to and from the study area. The program is available through the internet at the following WWW (World Wide Web) site: http://www.cnr.colostate.edu/~gwhite/software.html.

Sightability models

Sightability models were initially developed for elk in Idaho (Samuel 1984), but more recently have been applied to mule deer, moose, mountain sheep, and caribou (Unsworth et al. 1994). In open and semi-closed habitats, they offer a suitable alternative to double-stage sampling or mark-resight for estimating visibility bias. While this technique holds considerable promise for ungulate inventory in British Columbia, it still requires more testing before adopting as a standardized methodology, and thus is not recommended at this time. We do, however, recommend and encourage further testing and refinement of this methodology within the province. We also encourage collection of % vegetative cover, % snow cover and activity as User Statistics for all absolute abundance surveys. Once sightability models have been developed and verified, standardized sightability corrections incur little additional cost to the survey, and greatly improve the value of census information. We expect that following the required research, sightability models will become the survey method of choice for correcting ungulate visibility bias from aerial surveys in British Columbia.

Software needed to estimate population size and composition for several ungulates in Idaho has already been developed. The software (program AERIAL SURVEY, version 4.0) is available from: Oz Garton and Fred Leban, Department of Fish and Wildlife, University of Idaho, Moscow, Idaho 83844-1136. The manual accompanying the software provides detailed instructions for data recording, data entry and running the program.

It should be noted that the “Idaho” sightability models do not consider survey unit size in developing its population estimate. The “Gasaway” SRB method calculates a density per unit area and applies the density to estimate numbers per unit area in unsurveyed sub-units. Larger sub-units will always have higher predicted numbers than smaller sub-units in the same stratum. The Idaho models assume that sub-units are defined to enclose similar numbers of animals, irregardless of sub-unit size. This method may be more appropriate for elk, sheep, goats and bison which are more highly clumped and predictable in distribution.

[Back to TOC] [Previous] [Next]

Published by the Resources Inventory Committee