Author: Darryl Mackenzie, Proteus Research & Consulting Ltd., P.O. Box 5193, Dunedin, New Zealand, Darryl@proteus.co.nz.

In some circumstances, rather than monitoring the number of individuals of a target species (abundance) in an area, it may be advantageous to monitor what fraction of the area is occupied by the species, i.e., the distribution of the species across a landscape. For example, what fraction of wetlands is inhabited by a certain frog species, or what proportion of a forest is used as nesting sites by a particular bird? Sometimes this is referred in as the frequency of occurrence, although here the phrase “proportion of area occupied” (PAO) is used. The logic being that estimating abundance can be expensive and difficult in some situations (individuals need to be identifiable and counted accurately), but that changes in abundance will likely correspond in a change in the PAO by the species (which may be more easily measured with presence/absence surveys), hence it may be suitable as a coarse surrogate.

PAO may be most appropriate in mid-level monitoring programs where a lower resolution measure could provide adequate information about changes in species distributions without consuming resources that might be better used in more intensive monitoring (i.e., abundance estimation) of more “valuable” species. Such a philosophy is employed by the USGS Amphibian Research and Monitoring Initiative.

The term “area” refers to the region or collection of units that are of primary interest to the monitoring program. This could be a refuge, national park, forest, or a collection of discrete habitat units such as ponds, vegetation patches, or potential nesting sites. A number of monitoring sites are selected from within the area of interest (using a rigorous study design), and the presence or absence of the species at the sites determined through any appropriate field techniques.

The problem is, however, that few species are likely to so conspicuous that they will always be detected when present, hence an “absence” under these circumstances may be due to the species genuinely being absent, or the species not being detected due to random chance: a false absence. The consequences of imperfect detectability is that a naïve estimate of PAO (based upon a simple count of the number sites where the species was detected) will likely underestimate the real number of sites occupied by the species due to false absences. Further, comparing two or more naïve estimates collected at different times or regions will only be valid if the ability to detect the species is identical among all those sites and throughout the time period when the surveys were conducted, or if the amount of bias is low compared to the standard errors of the estimates. These arguments are similar to those used by various authors when discussing the pitfalls of using indices (simple counts) as a measure of absolute and relative abundance (Skalski and Robson 1992, Skalski et al. 1983, Yoccuz et al. 2001, Pollock et al. 2002, MacKenzie and Kendall 2002).

This issue with imperfect detectability has been long recognized by field biologists who have used repeated surveys of sites within a relatively short time period to maximise their probability of detecting the species if it really resides in the location. By taking repeated samples at least some of the sites it is possible to estimate both PAO and detectability (Giessler and Fuller 1987, Azuma et al. 1990, MacKenzie et al. 2002, Tyre et al. in press, Royle and Nichols in press), which provides a much more robust basis for making management decisions rather than simply relying on a simple count or index. Recently, new statistical models have also been developed by MacKenzie et al. (in press) that allow multiple year datasets to be modeled, enabling unbiased estimation of the rate of change in PAO across a landscape.

As always, the level of effort required for an efficient study design will vary with each application, and expert advice should be sought early in the design phase. However based upon the results of simulation studies by MacKenzie et al. (2002) and MacKenzie et al. (in press), there are some general rules of thumb that outline the circumstances for when the PAO estimators work reasonably well.

The number of repeat surveys should be sufficient so that if the species is present at a site, there is at least a 70% chance of detecting it at least once. For example, even if the species is moderately detectable (say it is detected 5 times out 10 surveys) then 2 or 3 repeated surveys of a site may be acceptable, with the number of required surveys increasing as detectability decreases.

The number of sites to monitor depends upon what the true PAO is likely to be and the desired level of precision in the estimator. If PAO is low, then more sites will be required than if it is high, and if greater precision is required, obviously, more sites should be sampled. Generally, at least 20 sites will probably be required, but depending upon the above factors, and also the objectives of the monitoring program, many more maybe required.

There is also a great deal of flexibility in how the repeated surveys might be conducted, for instance, they could be once a week, on consecutive days or multiple surveys might be conducted during a single visit to a site, either by the same or different observers. See MacKenzie et al. (2002) and MacKenzie et al. (in press) for some suggestions. The overriding consideration about how to conduct the multiple surveys is to ensure that the biology of the species, the sampling frame and the assumptions of the statistical model are all in agreement.

As always, it is recommended that a proposed study design should be given to a friendly statistician or biometrician for comment before embarking on any field work. This may save considerable resources in the long run caused by inefficient designs, or designs that do not collect the appropriate data to address the monitoring objectives.

Advantages

• Relatively low-cost as individual animals do not need to be identified (either by capture or other means).
• Recent methods of analysis use a statistically robust modeling framework (MacKenzie et al. 2002, MacKenzie et al. in press, Royle and Nichols in press, Tyre et al. in press).
• Explicitly accounts for detectability issues unlike indices and simple counts.

Disadvantages

• May be difficult to identify changes in the overall abundance of a species in the area of interest if there’s no corresponding change in distribution, i.e., the species continues to occupy the same fraction of the area, but at a different density.
• Cost. Repeat visits are required and many sites may be required to achieve moderate precision in the estimates.

Relevant web sites

• USGS Amphibian Research and Monitoring Initiative
• Proteus Research & Consulting Ltd.

Literature Cited

Azuma, D. L., J. A. Baldwin, and B. R. Noon. 1990. Estimating the occupancy of spotted owl habitat areas by sampling and adjusting bias. USDA Forest Service General technical Report PSW-124.

Geissler, P. H., and M. R. Fuller. 1987. Estimation of the proportion of area occupied by an animal species. Proceedings of the Section on Survey Research Methods of the American Statistical Association 1986:533-538.

MacKenzie, D.I., and W.L. Kendall. 2002. How should detection probability be incorporated into estimates of relative abundance. Ecology 83: 2387-2393.

MacKenzie, D.I., J.D. Nichols, J.E. Hines, M.G. Knutson and A.D. Franklin. Estimating site occupancy, colonization and local extinction probabilities when a species is not detected with certainty. Ecology (in press).

MacKenzie, D.I., J.D. Nichols, G.B. Lachman, S. Droege, J.A. Royle and C.A. Langtimm. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83: 2248-2255.

Pollock, K.H., J.D. Nichols, T.R. Simons, G.L. Farnsworth, L.L. Bailey and J.R. Sauer. 2002. Large scale wildlife monitoring studies: statistical methods for design and analysis. Environmetrics 13: 105-119.

Royle, J.A., and J.D. Nichols. Estimating abundance from repeated presence-absence data or point counts. Ecology (in press).

Skalski, J. R., and D. S. Robson. 1992. Techniques for wildlife investigations. Academic Press, San Diego, California, USA.

Skalski, J. R., D. S. Robson, and M. A. Simmons. 1983. Comparative census procedures using single mark–recapture methods. Ecology 64:752–760.

Tyre, A. J., B Tenhumberg, S.A. Field, D. Niejalke, K. Parris and H. P. Possingham. Improving precision and reducing bias in biological surveys by estimating false negative error rates in presence-absence data. Ecological Applications (in press).

Yoccuz, N. G., J. D. Nichols, and T. Boulinier. 2001. Monitoring of biological diversity in space and time. Trends in Ecology and Evolution 16:446–453.