Density estimate

Density estimates have two components, and both cause difficulties to biologists: the amount of area surveyed and the number of animals. As opposed to an index, the density estimate relies on a measure of area surveyed and not on effort. If the survey area is a true island, then the measurement is straightforward. If the survey area has an arbitrary boundary between "inside" and "outside," assumptions have to be made as to how the animals move with respect to the boundary. If the area surveyed is a mosaic of favorable and unfavorable habitat for a species, sampling protocol must take this into account. Placing survey lines to estimate the number of animals within the favorable area while using both favorable and unfavorable habitat to estimate the survey area will overestimate the number of animals. Unfavorable habitat may not contain a large number of animals during the survey, but may be used at other times of the year or when environmental conditions change.

It is difficult to estimate density from a line of traps, as it is difficult to estimate the distance that animals are drawn into the traps. Removal trapping suffers from this handicap, as removal of animals creates a vacuum that will eventually be filled by immigrating animals. Most studies that rely solely on trapping to estimate density construct trapping grids or webs to estimate the size of the survey area. Without prior knowledge of an animal's home range size and movements, it is difficult to accurately calculate the area sampled using most capture techniques.

The second component of the density estimate is the number of animals. To obtain a density estimate, a species usu-

ally needs to be observable or caught on a regular basis. It also helps if natural or added markings allow individuals to be identified. The two main techniques to estimate absolute density without changing the density are either mark/recapture or distance sampling. Mark/recapture uses an initial capture period that results in a known number of marked animals. A second capture period then looks at the ratio of marked to unmarked animals and estimates how many animals the population contains. Mark/recapture is based on the two assumptions that all animals can be captured and capture does not influence the probability of future captures and that animals can be marked and the marks will not fall off or influence the probability of the animal to be recaptured. There is also an assumption that no mortality, natality, or immigration occurs between capture sessions, but it is possible to relax this assumption and still estimate density.

For species that are readily captured, such as terrestrial rodents, it is a matter of trapping the animals in live-traps on a regular basis and estimating density. Commercial traps are available and techniques are well developed for small mammals. With all trapping, it is important to minimize trauma

A veterinarian examines a sedated chimpanzee (Pan troglodytes). (Photo by © Martin Harvey/Gallo Images/Corbis. Reproduced by permission.)
Scientists study rodents as surrogates for assessing vegetation condition in conservation areas. (Photo by Rudi van Aarde. Reproduced by permission.)

to the animal not just because it is humane but also so that the initial capture does not influence the probability of recapturing the animal. When the intent is to recapture the animal, some period of pre-baiting (with the trap baited but locked open) will increase the probability of capture. By providing adequate food and bedding and checking traps regularly, most small mammals can be trapped repeatedly and their density estimated. If the first priority is to ensure that all species have been detected, then using a variety of traps, both live and kill, would result in a better survey.

There is no terrestrial mammal that cannot be captured and released in a humane manner. However, for many animals such as large predators, the first trapping is traumatic enough to preclude the animal being recaptured by the same method. This does not necessarily preclude density estimates, as there is no assumption in the capture/recapture model that the capture method is the same for each period. The animals can be captured with snares, affixed with a mark, and then recaptured by cameras at bait stations. The first capture can be as a neonatal in the nest, and the recapture in a trap. For instance, an animal's DNA can be obtained in a blood sample attained at the initial capture in a snare, and the DNA later recaptured in a hair sample snagged on barbed wire strung around a bait station. If the animal has unique markings, as do many large cats, both the initial capture and subsequent recaptures can be with trip-cameras.

Removal sampling allows the observer to estimate a population's density post-hoc. The number of animals removed for each unit of effort, or a change in the ratio of two classes of animals (i.e., antlered and antlerless deer), is the basis for most surveys of harvested large mammals. For both calculations, the advantage is that the researcher is not necessarily the one doing the removal. Large areas and large populations can be estimated using volunteer hunters that agree to follow simple rules such as harvesting only antlered animals. The disadvantage is that there are limited species that attract a large number of volunteer hunters and the removal is limited to populations that are robust enough to sustain a harvest. It is possible to use the catch-per-unit-effort technique without harvesting animals, but the observer must be able to identify animals that have been previously sighted. There are several statistical programs that are commonly used to analyze this type of data and details of the programs and their assumptions are available at <http://www.cnr.colostate.edu/~gwhite/ software.html>.

For animals that are readily observable, or at least easier to observe than catch, strip transects or distance sampling techniques are used. In a strip sample, one counts all animals along a strip of known length and width. An assumption of this method is that all animals within the strip are counted. For most species, if the strip is wide enough to encounter a significant number of animals it is also wide enough to miss animals at the boundaries of the strip. There are ways to estimate the number of animals missed, such as double counting, but they do not solve the problem. Aerial surveys often involve strip samples where all animals within a strip along a side of the plane are counted by observers. These surveys are diurnal and usually occur in grassland or open habitats. Thermal imaging allows the heat signature of animals to be used to count animals at night. Use of thermal imaging for

Natalie Fobes
Scientists put a radio collar on a caribou (Rangifer tarandus). (Photo by © Natalie Fobes/Corbis. Reproduced by permission.)

density estimates is problematic because abiotic objects can radiate heat and be misidentified as animals. Also, canopy closure in forest settings is often dense enough to block the sensor's view of animals.

Distance sampling takes into account the assumption that animals can be missed along the transect, and the farther the animal is from the transect the more likely it is to be missed. If the researchers know how far they surveyed, the number of animals observed, and their perpendicular distance to the transect line, they can estimate the area that they surveyed and the number of animals that they did not see within that area. The number of animals seen plus the number of animals missed per unit area is a density estimate. The assumptions are that all animals along the line are observed; animals are randomly distributed relative to the line; and no animals are counted twice along the same line. This technique was originally developed to survey marine mammals, but has quickly been adapted for many large terrestrial mammals. There are many nuances to distance sampling, and it is difficult to cover them all in this space, but additional information is available <http://www.ruwpa.st-and.ac.uk/distance>.

0 0

Responses

  • Agamennone
    How to calculate species density estimates?
    4 years ago

Post a comment