Search USGS⌕ Search

Geology topics

William A. Link

Publications and source records attributed to William A. Link.

54 records · Page 3Linked to original sources
USGS2005

Using the North American Breeding Bird Survey as a tool for conservation: A critique of Bart et al. (2004)

Bart et al. (2004) develop methods for predicting needed samples for estimation of long-term trends from Count survey data, and they apply these methods to the North American Breeding Bird Survey (BBS). They recommend adding approximately 40% more survey routes ill the BBS to allow for estimation of long-term (i.e., 20 year) trends for a collection of species. We critique several aspects of their analysis and suggest that their focus on long-term trends and expansion of the present survey design will provide limited benefits for conservation because it fails to either enhance the credibility of the survey or better tie the survey to regional management activities. A primary innovation claimed by Bart et al. (2004) is the incorporation of bias in estimation of study planning. We question the value of this approach, as it requires reliable estimates of range of future bias. We show that estimates of bias used by Bart et al. (2004) are speculative. Failure to obtain better estimates of this bias is likely to compromise the credibility of future analyses of the survey. We also note that the generic analysis of population trends that they provide is of questionable validity and is unlikely to be relevant for regions and species of management concern.

Journal of Wildlife Management↗
USGS2004

Estimating population trends with a linear model: Technical comments

Controversy has sometimes arisen over whether there is a need to accommodate the limitations of survey design in estimating population change from the count data collected in bird surveys. Analyses of surveys such as the North American Breeding Bird Survey (BBS) can be quite complex; it is natural to ask if the complexity is necessary, or whether the statisticians have run amok. Bart et al. (2003) propose a very simple analysis involving nothing more complicated than simple linear regression, and contrast their approach with model-based procedures. We review the assumptions implicit to their proposed method, and document that these assumptions are unlikely to be valid for surveys such as the BBS. One fundamental limitation of a purely design-based approach is the absence of controls for factors that influence detection of birds at survey sites. We show that failure to model observer effects in survey data leads to substantial bias in estimation of population trends from BBS data for the 20 species that Bart et al. (2003) used as the basis of their simulations. Finally, we note that the simulations presented in Bart et al. (2003) do not provide a useful evaluation of their proposed method, nor do they provide a valid comparison to the estimating- equations alternative they consider.

Condor↗
USGS2003

Demographic analysis from summaries of an age-structured population

Demographic analyses of age-structured populations typically rely on life history data for individuals, or when individual animals are not identified, on information about the numbers of individuals in each age class through time. While it is usually difficult to determine the age class of a randomly encountered individual, it is often the case that the individual can be readily and reliably assigned to one of a set of age classes. For example, it is often possible to distinguish first-year from older birds. In such cases, the population age structure can be regarded as a latent variable governed by a process prior, and the data as summaries of this latent structure. In this article, we consider the problem of uncovering the latent structure and estimating process parameters from summaries of age class information. We present a demographic analysis for the critically endangered migratory population of whooping cranes (Grus americana), based only on counts of first-year birds and of older birds. We estimate age and year-specific survival rates. We address the controversial issue of whether management action on the breeding grounds has influenced recruitment, relating recruitment rates to the number of seventh-year and older birds, and examining the pattern of variation through time in this rate.

Biometrics↗
USGS2002

Of bugs and birds: Markov Chain Monte Carlo for hierarchical modeling in wildlife research

Markov chain Monte Carlo (MCMC) is a statistical innovation that allows researchers to fit far more complex models to data than is feasible using conventional methods. Despite its widespread use in a variety of scientific fields, MCMC appears to be underutilized in wildlife applications. This may be due to a misconception that MCMC requires the adoption of a subjective Bayesian analysis, or perhaps simply to its lack of familiarity among wildlife researchers. We introduce the basic ideas of MCMC and software BUGS (Bayesian inference using Gibbs sampling), stressing that a simple and satisfactory intuition for MCMC does not require extraordinary mathematical sophistication. We illustrate the use of MCMC with an analysis of the association between latent factors governing individual heterogeneity in breeding and survival rates of kittiwakes ( Rissa tridactyla ). We conclude with a discussion of the importance of individual heterogeneity for understanding population dynamics and designing management plans.

Journal of Wildlife Management↗
USGS2001

Suggestions for presenting the results of data analyses

We give suggestions for the presentation of research results from frequentist, information-theoretic, and Bayesian analysis paradigms, followed by several general suggestions. The information-theoretic and Bayesian methods offer alternative approaches to data analysis and inference compared to traditionally used methods. Guidance is lacking on the presentation of results under these alternative procedures and on nontesting aspects of classical frequentists methods of statistical analysis. Null hypothesis testing has come under intense criticism. We recommend less reporting of the results of statistical tests of null hypotheses in cases where the null is surely false anyway, or where the null hypothesis is of little interest to science or management.

Journal of Wildlife Management↗
USGS1998

Estimating relative abundance from count data

Much of the available information on large-scale patterns of animal abundance is based on count surveys. The data provided by such surveys are often influenced by nuisance factors affecting the numbers of animals counted, but unrelated to population size. Temporal and spatial patterns in nuisance factors may exist, causing simple summaries of counts to give a misleading view of patterns in population size. We develop models for count data that allows the incorporation of such factors, and describe methods for estimating spatial patterns of relative abundance from counts. We carry out spatial analyses of North American Breeding Bird Survey data, in which observer ability is a nuisance parameter nested within sites. In light of evidence that new observers tend to count more birds than the observers they replace, we model observer ability as a random effect with mean depending on observer initiation year.

Austrian Journal of Statistics↗
USGS1998

Estimating population change from count data: Application to the North American Breeding Bird Survey

For birds and many other animal taxa, surveys that collect count data form a primary source of information on population change. Because counts are only indices to population size, care must be taken in using them in analyses of population change. Temporal or geographic differences in the proportion of animals counted can be misinterpreted as differences in population size. Therefore, temporally or geographically varying factors that influence the proportion of animals counted must be incorporated as covariables in the analysis of population parameters from count data. We describe the North American Breeding Bird Survey (BBS) for illustration. The BBS is a major, landscape-level survey of birds in North America; it is typical of many count surveys, in that the same sample units (survey routes) are sampled each year, and change is modeled on these routes over time. We identify covariables related to observer ability, the omission of which can bias estimation of population change from BBS data. Controlling for observer effects or other potential sources of confounding requires the specification of models relating counts to population size. We begin with a partial model specification relating expected counts to population sizes; we describe estimators currently in use in relation to this partial specification. Additional assumptions lead to a class of over-dispersed multinomial models, for which we describe estimators of population change and procedures for parsimonious model selection. We illustrate the use of over-dispersed multinomial models by an application to data for Carolina Wren (Thryothorus ludovicianus).

Ecological Applications↗
USGS1997

New approaches to the analysis of population trends in land birds: Comment

James et al. (1996, Ecology 77:13-27) used data from the North American Breeding Bird Survey (BBS) to examine geographic variability in patterns of population change for 26 species of wood warblers. They emphasized the importance of evaluating nonlinear patterns of change in bird populations, proposed LOESS-based non-parametric and semi-parametric analyses of BBS data, and contrasted their results with other analyses, including those of Robbins et al. (1989, Proceedings of the National Academy of Sciences 86: 7658-7662) and Peterjohn et al. (1995, Pages 3-39 in T. E. Martin and D. M. Finch, eds. Ecology and management of Neotropical migratory birds: a synthesis and review of critical issues. Oxford University Press, New York.). In this note, we briefly comment on some of the issues that arose from their analysis of BBS data, suggest a few aspects of the survey that should inspire caution in analysts, and review the differences between the LOESS-based procedures and other procedures (e.g., Link and Sauer 1994). We strongly discourage the use of James et al.'s completely non-parametric procedure, which fails to account for observer effects. Our comparisons of estimators adds to the evidence already present in the literature of the bias associated with omitting observer information in analyses of BBS data. Bias resulting from change in observer abilities should be a consideration in any analysis of BBS data.

Ecology↗
USGS1996

Extremes in ecology: Avoiding the misleading effects of sampling variation in summary analyses

Surveys such as the North American Breeding Bird Survey (BBS) produce large collections of parameter estimates. One's natural inclination when confronted with lists of parameter estimates is to look for the extreme values: in the BBS, these correspond to the species that appear to have the greatest changes in population size through time. Unfortunately, extreme estimates are liable to correspond to the most poorly estimated parameters. Consequently, the most extreme parameters may not match up with the most extreme parameter estimates. The ranking of parameter values on the basis of their estimates is a difficult statistical problem. We use data on 401 species from the BBS and simulations to illustrate the potential misleading effects of sampling variation in rankings of parameters. We describe empirical Bayes and constrained empirical Bayes procedures that provide partial solutions to the problem of ranking in the presence of sampling variation.

Ecology↗
USGS1996

Empirical Bayes estimation of proportions with application to cowbird parasitism rates

Bayesian models provide a structure for studying collections of parameters such are considered in the investigation of communities, ecosystems, and landscapes. This structure allows for improved estimation of individual parameters by considering them in the context of a group of related parameters. Individual estimates are differntially adjusted toward in overall mean, with the magnitude of their adjustment based on their precision. Consequently, Bayesian estimation allows for a more reliable ranking of parameters and, in particular, a more credible identification of extreme values from a collection of estimates. In Bayesian models, individual parameters are regarded as values sampled from a specified probability distribution, called a prior. The requirements that the prior be known is often regarded as an unattractive feature of Bayesian analysis and may be the reason Bayesian analyses are not frequently applied in ecological studies. Empirical Bayes methods provide an alternative approach that incorporates the structural advantages of Bayesian models while requirng a less stringent specification of prior knowledge. Empirical Bayes methods require only that the prior be in a certain family of distributions, indexed by hyperparameters that can be estimated from the available data. This structur is of interest per se, in addition to its value in allowing for improved estimation of individual parameters; for example, hypothese regarding the existence of distinct subgroups in a collection of paramet ers can be considered under the empirical Bayes framework by allowing the hyperparameters to vary among subgroups. We describe the empirical Bayes approach in application to estimation of proportions, using data obtained in a community—wide study Brown—headed Cowbird paratism rates for illustration. Empirical Bayes estimates identify those species for which there is the greatest evidence of extreme parasitism rates. Subgroup analysis of our data on cowbird parasitism rates indicates that parasitisms rates for neotropical migrants as a group are no greater than those of resident/short—distance migrant in this forest community. Our data and analyses demonstrate that the parasitism rates for certain neotropical migrant species (Wood Thrush and Rose—breasted Grosbeak) are remarkably low while those for others (Ovenbird and Red—eyed Vireo) are remarkably high.

Ecology↗
USGS1996

Coexistence and community structure of tropical trees in a Hawaiian montane rain forest

We measured the diameter at breast height of all trees and shrubs > 5 meters in height, including standing dead trees, on 68 0.04-hectare study plots in a montane, subtropical rain forest on Mauna Loa, Hawai`i. The canopy species consisted of 88 percent Metrosideros polymorpha (ohia) and 12 percent Acacia koa (koa). Negative associations were found between the densities of koa and ohia, the density of koa and the total basal area of ohia, and the total basal areas of koa and ohia. The two-species lottery competition model, a stochastic model in which the coexistence of two species in a space-limited community results from temporal variation in recruitment and death rates, predicts a quadratic-beta distribution for the proportion of space occupied by each species. A discrete version of the quadratic-beta distribution, the quadratic-beta binomial distribution, was fit to the live koa and ohia densities and assessed with goodness-of-fit tests. Likelihood ratio tests provided evidence that the mean adult death rates of the two species were equal but that the relative competitive abilities of the two species favored ohia. These tests were corroborated by a contingency table analysis of death rates based on standing dead trees and growth rate studies which report that koa grows much faster than ohia. The lottery model predicts a positive covariance between death rates and ohia recruitment when mean death rates are equal and koa has a higher growth rate than ohia. We argue that the competitive advantage of ohia is due to its superior dispersal ability into large gaps, which would yield the positive covariance described above, and it is this positive covariance term that skews the occupation of space in favor of ohia.

Hawaii↗
USGS1994

Estimating breeding proportions and testing hypotheses about costs of reproduction with capture-recapture data

The proportion of animals in a population that breeds is an important determinant of population growth rate. Usual estimates of this quantity from field sampling data assume that the probability of appearing in the capture or count statistic is the same for animals that do and do not breed. A similar assumption is required by most existing methods used to test ecologically interesting hypotheses about reproductive costs using field sampling data. However, in many field sampling situations breeding and nonbreeding animals are likely to exhibit different probabilities of being seen or caught. In this paper, we propose the use of multistate capture-recapture models for these estimation and testing problems. This methodology permits a formal test of the hypothesis of equal capture/sighting probabilities for breeding and nonbreeding individuals. Two estimators of breeding proportion (and associated standard errors) are presented, one for the case of equal capture probabilities and one for the case of unequal capture probabilities. The multistate modeling framework also yields formal tests of hypotheses about reproductive costs to future reproduction or survival or both fitness components. The general methodology is illustrated using capture-recapture data on female meadow voles, Microtus pennsylvanicus. Resulting estimates of the proportion of reproductively active females showed strong seasonal variation, as expected, with low breeding proportions in midwinter. We found no evidence of reproductive costs extracted in subsequent survival or reproduction. We believe that this methodological framework has wide application to problems in animal ecology concerning breeding proportions and phenotypic reproductive costs.

Ecology↗
USGS1994

Within-site variability in surveys of wildlife populations

Most large-scale surveys of animal populations are based on counts of individuals observed during a sampling period, which are used as indexes to the population. The variability in these indexes not only reflects variability in population sizes among sites but also variability due to the inexactness of the counts. Repeated counts at survey sites can be used to document this additional source of variability and, in some applications, to mitigate its effects. We present models for evaluating the proportion of total variability in counts that is attributable to this within-site variability and apply them in the analysis of data from repeated counts on routes from the North American Breeding Bird Survey. We analyzed data on 98 species, obtaining estimates of these percentages, which ranged from 3.5 to 100% with a mean of 36.25%. For at least 14 of the species, more than half of the variation in counts was attributable to within-site sources. Counts for species with lower average counts had a higher percentage of within-site variability. We discuss the relative cost efficiency of replicating sites or initiating new sites for several objectives, concluding that it is frequently better to initiate new sites than to attempt to replicate existing sites.

Ecology↗
USGS1993

Simulation testing of unbiasedness of variance estimators

In this article I address the evaluation of estimators of variance for parameter estimates. Given an unbiased estimator X of a parameter θ, and an estimator V of the variance of X , how does one test (via simulation) whether V is an unbiased estimator of the variance of X? The derivation of the test statistic illustrates the need for care in substituting consistent estimators for unknown parameters.

American Statistician↗
USGS1993

Optimal allocation of point-count sampling effort

Both unlimited and fixed-radius point counts only provide indices to population size. Because longer count durations lead to counting a higher proportion of individuals at the point, proper design of these surveys must incorporate both count duration and sampling characteristics of population size. Using information about the relationship between proportion of individuals detected at a point and count duration, we present a method of optimizing a point-count survey given a fixed total time for surveying and travelling between count points. The optimization can be based on several quantities that measure precision, accuracy, or power of tests based on counts, including (1) mean-square error of estimated population change; (2) mean-square error of average count; (3) maximum expected total count; or (4) power of a test for differences in average counts. Optimal solutions depend on a function that relates count duration at a point to the proportion of animals detected. We model this function using exponential and Weibull distributions, and use numerical techniques to conduct the optimization. We provide an example of the procedure in which the function is estimated from data of cumulative number of individual birds seen for different count durations for three species of Hawaiian forest birds. In the example, optimal count duration at a point can differ greatly depending on the quantities that are optimized. Optimization of the mean-square error or of tests based on average counts generally requires longer count durations than does estimation of population change. A clear formulation of the goals of the study is a critical step in the optimization process.

The Auk↗
USGS1991

Social foraging and feeding environment of the black-crowned night heron in an industrialized estuary

A nesting colony of more than 300 pairs of Black-crowned Night-Herons in the Patapsco River estuary, Baltimore Harber, Maryland was monitored from May to July 1988. The departure direction and departure time of each Heron was recorded during mornjng and evening observations at the colony. In addition, individual Herons were followed to landing sites in the estuary with small boats and aircraft. Herons appear to be quite social in foraging and may concentrate their feeding in some of the most industrialized parts of the estuary. Night-Herons may prosper in urban environments because some of their prey are attracted to intense shoreline illumination at night. New quantitative methods were developed to cope with the lack of statistical independence when dealing with social species.

Maryland↗