Recent Applications of Multi-State Models to Capture-Recapture and Occupancy Studies

Symposium
ROOM: Room 120 – Dona Ana
SESSION NUMBER: 21
 
Multi-state models allow demographic parameters to vary over time and space with probabilistic transitions between states defined by geography, life history, disease status, habitat, etc. Estimating transition probabilities among states (e.g., breeding), accounting for unobservable states, and improving survival estimates are some of the realized benefits from using these models. Additionally, one can explore species interactions, habitat and species dynamics, and disease and host-pathogen relationships. Applications of these models span a broad range of vertebrate species and this session will demonstrate this breadth. All talks will demonstrate how this class of models is used in real-world systems for improving estimates of demographic parameters useful to researchers and managers.

1:10PM The Multistate Model as an Omnibus Framework for Robust Inference from Mark-Recapture Studies
  William Kendall
Wildlife population studies can involve multiple complexities, resulting in transitions of individuals among various states. States can arise from different sources of interest, including multiple life stages (e.g., pre-breeder, breeder, skipped breeder), choices of locations during migration (breeding, stopover, wintering), disease states (e.g., susceptible, infected, recovered), affinity to an area (transient, resident), dispersal (e.g., natal, breeding), size classes, etc. Multiple states can also be imposed by the practicalities of study design. When the home range of an individual is not fully covered by sampling effort, it can be available or unavailable to capture effort. Tag loss is a nuisance process that can also create multiple states (e.g., two tags, one tag, zero tags). Multistate models provide a framework for evaluating hypotheses related to state-specific survival, or to transition probabilities among states. In addition to imperfect detection, other complicating factors can influence multistate studies. One or more states could be unobservable (e.g., non-breeders do not attend a breeding colony). For a given detected individual, its state membership could be uncertain (e.g., an infected individual might show no clinical sign). I will give an overview of multistate models, including some challenges and partial solutions, using select examples. Fortunately user-friendly software is available for analyzing such data (e.g., MARK, E-SURGE, R-package MARKED).
1:30PM Estimating Nest Density and Productivity of Grassland Birds with Multi-State Capture-Recapture Models
  Adrian P. Monroe; Richard B. Chandler; Loren W. Burger; James A. Martin
Habitat quality for grassland birds is often assessed through nest searching and monitoring to obtain metrics of individual fitness (e.g., nest survival), and these can be complemented with inferences at the population level by measuring density. However, to date most studies have used relative or apparent indices of nest density that may be confounded by detectability that varies with nest stage and treatment. In a study evaluating dickcissel (Spiza americana) use of exotic forages and native warm-season grass (NWSG) pastures in Mississippi (2011-2012), we applied a multi-state capture-recapture model in a Bayesian framework to estimate nest density while accounting for nest stage-specific variation in survival and detection. A novel feature of our model is the use of an additional state for nests that fledged ≥1 nestlings, allowing us to estimate dickcissel productivity among our study pastures. Nest density and productivity were consistently greater in non-grazed NWSG than grazed exotic pastures for this tall structure specialist, whereas productivity in grazed NWSG treatments was intermediate and declined between years. These results suggest dickcissels respond positively to NWSG in this system, but livestock grazing may reduce nest site abundance, and consequently productivity. Additionally, variation in survival and density was not always consistent with productivity estimates, and therefore relying only on either parameter may be misleading when evaluating habitat quality. Using multi-state capture-recapture models offers the ability to jointly use data on density and demographics to estimate grassland bird productivity, and thus evaluate habitat quality while addressing detection bias.
1:50PM The Use of Multi-State Models to Explore Relationships between Changes in Body Condition, Habitat, and Survival of Grizzly Bears
  John Boulanger; Gord Stenhouse; Marc Cattet
One of the principal goals of wildlife research and management is to understand and predict relationships between habitat quality, the health of individuals, and their ability to survive. Infrequent sampling, non-random loss of individuals due to mortality, and variation in capture susceptibility create potential biases with conventional analysis methods. To account for such sampling biases, we used a multi-state analytical approach to assess relationships between habitat, health, and survival of grizzly bears over a 10-year period along the east slopes of the Canadian Rockies in Alberta, Canada. Bear health states were defined by body condition estimated from the relationship between weight and body length. We used a sequential model building process to first account for potential sampling biases, and then explored changes in body condition relative to habitat use and survival. Bears that used regenerating forest habitats (mostly due to forest harvesting) containing a diversity of age classes were more likely to see gains in body condition, while bears that used older forests were more likely to see reductions in body condition. Survival rate was reduced most by road densities which in turn were positively correlated with regenerating forest habitat. Human activities that promote young regenerating forests, such as forest harvesting, therefore promotes improved health (increased body condition) in bears, but are offset by reductions in survival rates. Multi-state analyses represents a robust analytical tool when dealing with complex relationships and sampling biases that arise from dynamic environments
2:10PM Multi-State Models Demonstrate Hoary Marmots Are Not Biennial Breeders
  Vijay Patil
Biennial breeding is a rare life-history trait observed in animal species living in harsh, unproductive environments. This reproductive pattern is thought to occur in 10 of 14 species in the genus Marmota, making marmots useful model organisms for studying its ecological and evolutionary implications. Biennial breeding in marmots has been described as an obligate pattern which evolved as a mechanism to mitigate the energetic costs of reproduction (Evolved Constraint hypothesis). However, recent anecdotal evidence suggests that it is a facultative pattern controlled by annual variation in climate and food availability (Environmental Constraint hypothesis). Finally, in social animals like marmots, biennial breeding could result from reproductive competition between females within social groups (Social Constraint hypothesis). We evaluated these three hypotheses using mark-recapture data from an 8-year study of hoary marmot (Marmota caligata) population dynamics in the Yukon. Annual variation in breeding probability was modeled using multi-state mark recapture models, while other reproductive life-history traits were modeled with generalized linear mixed models. Hoary marmots were neither obligate nor facultative biennial breeders, and breeding probability was insensitive to evolved, environmental, or social factors. However, newly mature females were significantly less likely to breed than older individuals. Annual breeding did not result in increased mortality. Female survival and, to a lesser extent, average fecundity were correlated with winter climate, as indexed by the Pacific Decadal Oscillation. Hoary marmots are less conservative breeders than previously believed, and the evidence for biennial breeding throughout Marmota, and in other arctic/alpine/Antarctic animals, should be re-examined. Prediction of future population dynamics requires an accurate understanding of life history strategies, and of how life history traits allow animals to cope with changes in weather and other demographic influences.
2:30PM The Use of Multi-State Models with State Uncertainty to Estimate Survival and Movement Probabilities of Florida Manatees
  Catherine A. Langtimm; William L. Kendall
After 44 years on the U.S Endangered Species list, the Florida manatee (Trichechus manatus latirostris) was downlisted this year to threatened by the U. S. Fish and Wildlife Service. Multi-state capture-recapture demography models applied to photo-identification data provided information to manage and evaluate the subspecies. Now with significant gains in population abundance and changing environments and threats, management objectives and questions are being revised. Answers require information on an increasing number of state-specific population parameters, including where there is uncertainty as to state identity of individuals. We demonstrate modeling efforts to assess movement where states are geographic, defined by wintering area. Some wintering areas are not covered by sighting effort, but these individuals could be detected at other times of year. Wintering area assignment is certain in some cases, but not in others. We use a multistate Barker robust design framework, which allows for uncertainty about location assignment and models survival specific to location and state transition among monitoring sites. Additional movement parameters of interest in the model include temporary emigration and permanent dispersal and site fidelity. Exposure to threats or management actions varies along coastlines and responses may not be evident in region-wide analyses. We can now examine smaller scale, local events such as hurricane strikes, red-tide algal blooms, and cold air outbreaks with location-specific analyses.
2:50PM Refreshment Break
3:20PM Overview of Multi-State Models in Occupancy Estimation
  Larissa L. Bailey
Investigating spatial and temporal processes is fundamental to understanding many of the drivers of species dynamics. This is especially true for ‘meta-populations’, as the interconnectedness of local populations through colonization and local extinction processes determine species distributions over time. Often in these systems there is a desire to include more than one occupied state. For example, there are many biological systems where it is advantageous to classify different subcategories of species occurrence (e.g. breeding/non-breeding or relative abundance classes) or simultaneously model the joint dynamics of habitat and species occurrence or species interactions. Here, I give a brief overview of static and dynamic multi-state occupancy models using a diversity of existing applications. I emphasize the flexibility of these models, setting the stage for the four subsequent talks in this section of the symposium. Finally, I will mention model assumptions and the importance of defining key terms (e.g. site, survey, season) with respect to the biological question(s) of interest.
3:40PM Identifying Drivers of Local Extinction in an Amphibian-Pathogen System Using Multi-Species Occupancy Models
  Brittany Mosher; Larissa L. Bailey; Erin Muths; Kathryn P. Huyvaert
Emerging infectious diseases are an increasingly common threat to wildlife, and emergence is often driven by changes in host susceptibility, pathogen infectivity, or the environment. Chytridiomycosis, caused by the fungal pathogen Batrachochytrium dendrobatidis (Bd), is an emerging infectious disease that has been linked to amphibian declines around the world. Few studies exist that explore amphibian-Bd dynamics at the landscape scale, limiting our ability to identify factors influencing variation in population susceptibility and to develop effective in situ disease management. Declines of boreal toads (Anaxyrus boreas boreas) in the southern Rocky Mountains are largely attributed to chytridiomycosis but variation exists in local extinction of boreal toads across this metapopulation. Using a large-scale historic dataset, we explored several potential factors influencing disease dynamics in the boreal toad-Bd system: geographic isolation of populations, amphibian community diversity, climate differences, and habitat permanence. We found evidence that boreal toad extinction risk was highest at low elevations where temperatures may be optimal for Bd growth and where large boreal toad populations facilitate density-dependent pathogen transmission. We illustrate a framework that will be useful to natural resource managers striving to make decisions in amphibian-Bd systems, and provide evidence that the physiological tolerances of Bd may interact with habitat features to shape amphibian declines.
4:00PM Joint Estimation of Habitat Dynamics and Species Interactions
  Courtney Davis
Disentangling the role that multiple interacting factors have on species responses to shifting climate poses a significant challenge. Our ability to do so is of utmost importance to predict the effects of climate change on species distributions. We examined how three species of wetland-breeding amphibians, which varied in life history requirements, responded to a six-year period of variable precipitation. This interval was punctuated by extensive drought and heavy precipitation and flooding, providing a natural experiment to measure community responses to environmental perturbations. We estimated occurrence dynamics using a discrete hidden Markov modeling approach that incorporated information regarding habitat state and predator-prey interactions. This approach allowed us to measure how metapopulation dynamics of each species was affected by interactions among weather, wetland hydroperiod, and co-occurrence with fish predators. The pig frog proved most resistant to perturbations, with colonization and persistence being unaffected by variation in precipitation or co-occurrence with fishes. The ornate chorus frog responded positively to drought owing to increased persistence and colonization during periods of low-rainfall. Low occurrence probabilities of the ornate chorus frog in long-duration wetlands were driven by interactions with predators due to low colonization when fishes were present. The mole salamander was most sensitive to shifts in water availability. In our study area, this species never occurred in short-duration wetlands and persistence probabilities decreased during periods of drought. At the same time, negative effects occurred with extreme precipitation because flooding facilitated fish colonization to isolated wetlands and mole salamanders did not colonize wetlands once fishes were present. We demonstrate that the effects of changes in water availability depend on interactions with predators and wetland type and are influenced by the life history of each species. Our dynamic modeling approach offers promise when the goal is to disentangle complex interactions that determine species responses to environmental variability.
4:20PM Multi-State Models Applied to Species Interactions to Inform Endangered Species Management
  Charles Yackulic
Estimating the impact of interspecific interactions relative to other stressors in the population dynamics of endangered species can play a vital role in informing management decisions. Here I discuss applications of multistate occupancy and multistate mark-recapture models to understand the joint population dynamics of barred owls and threatened Northern spotted owls in the Pacific Northwest and rainbow trout and endangered humpback chub in the Grand Canyon. I will begin by briefly discussing the definition of states in a two-species dynamic occupancy model and will then show how they have been used extensively to understand the impacts of the barred owl invasion on the territorial dynamics of Northern spotted owls throughout the Pacific Northwest, to estimate how barred owl removal has affected Northern spotted owl territorial occupancy, and to predict how competition dynamics may change under habitat management. Next, I will discuss joint modelling of mark-recapture data from two fish species. In our modelling, we use states defined in terms of location and fish size to model the humpback chub vital rates in terms of the abundance of rainbow trout, temperature and turbidity while accounting for uncertainty in rainbow trout abundances. Analyses using multistate models have brought some resolution to long-standing debate over the impacts of rainbow trout and environmental conditions on humpback chub dynamics, but have also raised important new questions.
4:40PM Detailed Dynamics: Using Multi-State Responses for Greater Specificity on Species Interaction Dynamics
  Jay E. Jones; Andrew J. Kroll
Biotic interactions such as competition, facilitation, and mutualism are important determinants of species turnover and fundamental in shaping community structure. In addition, quantifying spatial and temporal variation in species interactions has meaningful conservation and management applications, including control of invasive organisms, reduction of hyper-abundant populations, and conducting species reintroduction efforts. Although species co-occurrence models have long been available, recent methodological advances support extensions to incorporate temporal population dynamics with variation in the observation process. Furthermore, contemporary dynamic co-occurrence models often focus on simple presence/absence despite evidence that the influence of interactions may extend beyond species presence to behaviors influencing fitness such as mating success or recruitment. Here, we apply a recently proposed model for multistate co-occurrence dynamics to quantify state-specific interactions between the northern spotted owl (Strix occidentalis caurina) and barred owl (Strix varia). Given on-going programs to survey and monitor northern spotted owl populations, the efficient allocation of sampling effort to reduce uncertainty in state status is a critical applied outcome. We compare our results with those from contemporary spotted owl studies, and highlight consistency of findings while emphasizing the greater specificity multistate models provide for species interaction dynamics.

 
Organizers: William Gould, New Mexico State University, Las Cruces, NM
 
Supported by: TWS Biometrics Working Group

Symposium
Location: Albuquerque Convention Center Date: September 24, 2017 Time: 1:10 pm - 5:00 pm