Plenary Address - 1:15-1:55 PM

Modeling Spatially Referenced Capture-Recapture and Related Data

J. Andrew Royle

Many avian monitoring efforts generate spatially referenced data relevant to avian abundance, survival and related quantities. Data collection from such efforts is commonly based on networks of constant effort mist net stations or various counting protocols such as removal, multiple observer or distance sampling. A common objective is to relate demographic quantities (abundance, survival) to landscape or environmental covariates.

The data resulting from such studies can be generally characterized as a collection of multinomial random variables with spatially referenced multinomial index (N, or ``local abundance'') or parameters which determine the multinomial cell probabilities (e.g., survival rate). An important statistical consideration in many of these studies is that only a small number of individuals are captured and resighted at each spatial sampling location, resulting in sparse data that renders conventional modeling approaches ineffective. A common solution is to pool data from a number of samples (e.g., within geographic strata), resulting in a loss of spatial information and a general inability to make inferences about variation at the resolution of sample locations.

In this paper, I present a general strategy for modeling spatially referenced multinomial data that preserves information available in the spatial attribution of the data and that facilitates modeling of ``site level'' structure in sparse capture-recapture data. In the context of estimating and modeling abundance effects, the proposal is to retain site-specific abundance parameters as random effects in the model, where spatial structure is incorporated into the abundance prior distribution. I will consider several applications that focus on modeling spatial structure in abundance.



Individual Papers

1:55-2:20 PM

Population dynamics of the White Stork in the Netherlands: assessing life-history and behavioural traits using data collected at large spatial scales - Blandine Doligez, David L. Thomson & Arie van Noordwijk

Following the decline of its populations all over Europe after 1945, the White Stork Ciconia ciconia has been the object of several successful reintroduction programs. As a consequence of the development of these programs, White Stork populations have been monitored over large spatial scales. Despite these intense efforts, very few reliable estimates of life-history traits of the White Stork are however currently available, in particular very little is known about their variation with age. Such general knowledge however constitutes a prerequisite for investigating the consequences of conservation measures in terms of population biology. Ultimately, this knowledge is required to assess population dynamics expected under different scenarios, allowing adequate further conservation measures to be taken.

In the Netherlands, the reintroduction of the White Stork consisted of a captive breeding program coupled with intensive supplementary feeding of captive and free-flying birds throughout the year. Identification of breeding adults and ringing of young in the growing population has been performed over the last 20 years throughout the country by many volunteers. Using this large-scale and long-term ringing and resighting data set, we constructed capture-recapture models to investigate the variation of life-history traits underlying the dynamics of this population, and the consequences of the reintroduction program on these traits.

In a first step, we fully describe the effects of age, time, cohort and trap-dependence on White stork survival and resighting rates, thus providing precise estimates of these traits. A gradual increase in both survival and resighting rates with age is described for the first time in this long-lived species. Interestingly, survival rate decreases with time, paralleling the increase in the proportion of free-flying individuals in the population. These results also allow us to account for time and age effects in investigating the consequences of supplementary feeding on life-history traits, while keeping capture-recapture models tractable in these further analyses.

The exact causes of the White Stork population decline were not clear, but mortality on migration is thought to have increased (exacerbated by the proliferation of high-voltage power lines across Europe). Adverse weather on the wintering grounds could also have depressed survival. Supplementary feeding may then have ameliorated survival either by giving the birds a better body condition prior to migration, or by encouraging them to stay behind, surviving the European winter on artificial food supplements. In a second step, we thus used detailed data on the location of breeding attempts with respect to the food sources over the country (food was provided at the breeding stations) to assess the consequences of supplementary feeding on survival, resighting rate and migratory behaviour. More specifically, we investigate whether survival rate to the next year, resighting rate and migrating probability depend on food availability during the breeding season. Food availability was considered to decrease with increasing distance from the nest to the nearest breeding station. Further, we analyse potential differences in survival rate according to individual migratory status (i.e. migratory vs. resident individuals). This status is defined using resighting data collected all over Europe and Africa. Preliminary analyses suggest that survival does not differ according to food availability, but the probability of migrating decreases with increasing food availability.

Overall, our results emphasize the importance of the large scale population monitoring in obtaining life-history trait estimates that will prove useful for managers to make efficient decisions for future strategies of conservation for the White stork in the Netherlands. Building an integrated demographic population model with the precise estimates obtained via our analyses, based on data collected at large spatial scales, will indeed eventually allow reliable predictions on the long-term population dynamics to be made under different conservation scenarios.

2:20-2:45 PM

Estimating correlates of survival rates from nationally coordinated ringing data on owls in Finland - Pertti Saurola & Charles Francis

Since 1974, bird ringers in Finland have been encouraged to ring both nestlings and adults of many species of birds of prey, especially owls. This coordinated effort, involving several hundred ringers, now results in more than 30,000 potential nest sites for owls being checked annually, and has led to over 200,000 owls being ringed in Finland up until 2002. Many of these owls are subsequently recaptured as breeding adults, by the same or different ringers, while others are recovered dead by the general public. All of the ringing and encounter data, as well as many biometric data are centrally computerized, allowing for large-scale analyses of geographic and temporal variation in such parameters as survival and dispersal rates. These can then be related to external covariates, such as winter snow depth and prey abundance, as well as individual covariates such as biometrics. In this presentation, we illustrate both the potential and the challenges of working with these types of data, using data from 9000 recaptures and recoveries from over 30,000 Tawny Owls ringed as nestlings or adults. Average survival rates of this species increased from 30% over the first year post-fledging, to 60% in the second year, and 74% in subsequent years, but varied among years in response to winter severity and vole abundance. Survival rates were lowest after winters with deep snow accumulations, and in years when vole populations crashed. Similar variation occurred in dispersal, with greater dispersal distances in years when voles crashed just after the breeding season. Combining data from many different contributors to the national scheme, in addition to greatly increasing the sample size, allows one to address questions that could not be addressed using data from a single study area. For example, it is possible to examine the impact of dispersal on survival rate estimates, and the relative influence of survival and dispersal on population dynamics of this species. It is also possible to test for geographic variation, such as north-south clines in demographic parameters. However, combining data from many sources also presents some challenges, such as dealing with geographic variation in the intensity of coverage.

2:45-3:10 PM

Population dynamic and temporal variation in recruitment and survival of 14 common passerines - Romain Julliard

At large spatial scale, variation of breeding abundance of a given species results from variation of survival of established adults and variation of recruitment of new individuals. Which of these two parameters is the most variable and the best predictor of variation of abundance is a central issue of population dynamic (Saether et al. Science 2002). Yet the problem looks simple, an important pitfalls is the difficulty to obtain independent estimates of the different parameters.

In many countries, large scale breeding bird monitoring based on counts are established to monitor population abundance at the scale of the country. They are often coupled with standardized mist-netting scheme set up for the specific purpose of monitoring demographic parameters for the same populations. From these capture-recapture data, variation of survival and recruitment (using Pradel's approach) may be estimated, independently of population size variation estimated from count survey. In addition, captures of young individuals are used to estimate variation of reproductive success (using the young adult ratio as an index). Because survival and recruitment are not independently estimated, their temporal variation may not be compared directly. Rather, the amount of temporal variation as well as the synchronization of temporal variation across sites, and how temporal variation were correlated with changes in abundance and productivity index were compared between survival and recruitment for a given species. This was done for 14 species with high enough recapture rate to estimate demographic parameters. The data comes from the French common bird monitoring scheme settled in 1989, and included capture-recapture data from 20 to 40 sites per year for 13 years (about 35,000 captures).

3:10-3:35 PM

Population dynamics of small mammals at three spatial scales: a 5-years study of 122 trapping grids - Nigel Yoccoz & Rolf Ims

Small mammals are known to exhibit a large variety of dynamics, both in time (multi-annual cycles vs. seasonal variation only) and space (regional synchrony, travelling waves). Small mammals have therefore been the focus of a large number of studies, that used mostly trapping indices. These studies do not therefore take into account differences in trappability that may indeed be confounded with the phenomenon we want to explain (for example, that trappability is lower in the increase than in the decrease phase of the cycle). In this paper, we use our own study investigating population dynamics and demography of small mammals at three spatial scales (0.1, 10 and 100 kms) to address some methodological and practical issues. The study is based on an unusually large sampling effort (122 grids trapped twice a year in 5 years, in spring and fall, each trapping session being 4 days long), with more than 10,000 individuals captured of three different species. The main methodological issue we address is: what is the consequence of using a too simple/complex model for capture probabilities on the estimates of spatial patterns in population densities, growth rates and demography, and how should we model these probabilities in practice. We explore in particular the use of Bayesian models that include hierarchical random effects.