Statistical Design and Analysis

General Design Considerations

We believe that the design of an effective monitoring program for amphibians requires substantial thought about program objectives (why?) and implementation (how?) relative to those objectives.  Objectives can usually be classified as either scientific or management, and clear specification of objectives facilitates development of an efficient program design.

 Our monitoring program designs must incorporate two major sources of variation and uncertainty in animal counts: (1) spatial variation and (2) detectability. Spatial variation arises in the typical situation where the investigator(s) cannot apply survey or monitoring techniques over the entire area of interest. In such situations, sample units must be selected from the entire area of interest and sampled in such a manner that permits inference about the entire area. Detectability refers to the near-universal situation in animal population monitoring in which survey methods do not detect all animals present in the sampled area.  Even on the areas that are selected as part of the spatial sample, we cannot count all animals present. Our monitoring program must thus incorporate methods for estimating or removing effects of detectability, so that estimated changes in animal abundance, or a related quantity, reflect true changes and not differences in detectability.

The above considerations are extremely general. However, we fully recognize that the development of a good monitoring program involves many detailed considerations that go far beyond these general suggestions. These detailed considerations will depend on the specifics of the objectives, the areas to be sampled, and the taxa of interest. Amphibians of the southeastern United States clearly will provide a major challenge and it will take joint cooperation and understanding between field biologists and biometricians to design and implement a realistic and robust monitoring program.

Inventory strategy

A robust initial inventory of species at both our index sites and subsequent extensive sites is key to establishing a monitoring program to assess status and trends of amphibians in the southeast. Our inventory objectives are to determine: 1) what species occur at the site, 2) what we may be missing as we sample because of variation in detectability of species, 3) where species occur over the landscape and in what habitats and 4) how presence/absence varies seasonally.  Multiple sampling techniques will be employed to maximize detection of species. Research is being conducted to determine the best field techniques. The new statistical approach of estimating species richness with associated estimates of probability of detection, is being evaluated as a tool to analyze data and make comparisons among plots and sites.

Monitoring strategy

Long-term monitoring will be implemented after the inventory. Clear objectives will be defined at each site and a monitoring program will be designed to meet those objectives. The major statistical approaches will focus on 1) monitoring changes in community composition with the estimation of species richness and 2) monitoring changes in the abundance of selected species with techniques designed to include some measure of detectability and spatial variation.

Species Richness

The number and diversity of amphibians in the southeast makes monitoring all species difficult, if not impossible.  Nonetheless, amphibian diversity is a hallmark of ecosystems in the southeastern U.S. and changes in ecosystems through disturbance, human development, environmental contaminants, or other factors could negatively impact the composition and richness of amphibian communities. Estimating variation in species richness through time and among different locations is one means of tracking the status of amphibians as a group, and may be more effective than focusing on abundance measures of individual species, which have been shown in most studies to lack statistical power.

In the past the main hindrance to making valid inferences about variation in species richness has been the inability to count all species present in an area during a survey. Weather conditions, the behavior of different species, cryptic coloration, and observer skill are just some factors affecting detection. Invariably some species will be missed, thus biasing the estimates (Boulinier et al. 1998). However, methods are now available which account for variation in detection probabilities and which estimate species richness, standard error, and 95% Confidence Intervals (Nichols and Conroy 1996).  These methods have been extended to estimate several important vital rates in animal communities, which would be useful to assessing status, e.g., rates of local species extinction, turnover, and colonization (Nichols et al. 1998a). And they have been used to test hypotheses concerning factors affecting temporal (Boulinier et al. 1998) and spatial variation (Nichols et al. 1998b.) in species richness as well.

The application of these estimation methods to amphibian survey data is promising, not only because they can address important questions, but they may easily be applied to inventory surveys, intensive monitoring at index sites, and extensive surveys initiated by partners at other sites. Furthermore, detection of a change in species richness can alert biologists and managers to potential problems that may require more focused study.  SE ARMI is evaluating these estimation methods for use with amphibians, particularly with regard to proposed field protocols and issues of spatial scale.

Boulinier, T., J. D. Nichols, J. E. Hines, J. R. Sauer, C. H. Flather, and K. H. Pollock.  1998. Higher temporal variability of forest breeding bird communities in fragmented landscapes. Proceedings of the National Academy of Sciences  95:7497-7501. (Abstract)

Boulinier, T., J. D. Nichols, J. R. Sauer, J. E. Hines, and K. H. Pollock. 1998.  Estimating species richness: the importance of heterogeneity in species detectability. Ecology 79:1018-1028. (Abstract)

Nichols, J. D., and M. J. Conroy.  1996. Estimation of species richness.  Pp. 226-234 in Measuring and Monitoring Biodiversity.  Standard Methods for Mammals. Wilson, D.E., F. R. Cole, J. D. Nichols, R. Rudran, and M. S. Foster (eds). Smithsonian Institution Press, Washington, D.C. 

Nichols, J. D., T. Boulinier, J. E. Hines, K. H. Pollock, and J. R. Sauer. 1998a. Estimating rates of local species extinction, colonization, and turnover in animal communities. Ecological Applications 8:1213-1225. (Abstract)

Nichols, J. D., T. Boulinier, J. E. Hines, K. H. Pollock, and J. R. Sauer. 1998b. Inference methods for spatial variation in species richness and community composition when not all species are detected. Conservation Biology12:1390-1398.