Ecological Research 1
Presenter: Adam Butler
When: Monday, July 11, 2016 Time: 11:00 AM - 12:30 PM
Room: Salon C Carson Hall (Level 2)
Session Synopsis:
SIMULATION-BASED APPROACHES FOR DESIGNING STUDIES OF WILDLIFE POPULATIONS USING ELECTRONIC TAGS
Electronic tags, such as accelerometers and GPS tags, are increasingly widely used within ecology. The use of electronic tags is allowing large quantities of data to be collected automatically, at relatively low cost, and is providing important new insights into key aspects of animal behaviour and movement. There are interesting statistical questions relating to the design of studies that use electronic tags - such as how many tags should be used and how often should the tags collect records? These questions are of practical, as well as methodological, interest, because constraints on data collection mean that trade-offs have to be made. For example, limited battery life often means that there is a choice between collecting high resolution data over a relatively short period of time or low resolution data over a much longer period. What is the optimal sampling interval to use in such situations? We use the example of a standard autoregressive time series model to demonstrate that the answers to such questions are not always obvious, and can depend upon the properties of the underlying process in subtle ways. In realistically complex applications analytic approaches to the identification of optimal sampling designs are not feasible. We therefore outline a simulation-based approach and illustrate this approach by application to two real ecological data sets. We conclude by highlighting connections between this approach and the simulation-based approaches that are often used to perform power calculations or (in the context of estimation rather than design) as a basis for likelihood-free inference.
Ecological Research 1
Presenter: Lisa-Marie Harrison
When: Monday, July 11, 2016 Time: 11:00 AM - 12:30 PM
Room: Salon C Carson Hall (Level 2)
Session Synopsis:
A hurdle mixed model for linking Antarctic Krill presence/absence and density to phytoplankton and environmental factors
Antarctic Krill (Euphausia superba) are a vital part of the Southern Ocean food web and ecosystem processes. Despite this, much remains unknown about the species, especially in relation to drivers of distribution. We present a hurdle model for predicting krill presence/absence and density in relation to environmental and biological variables. The data used are from the 2006 Eastern-Antarctic survey, Baseline Research on Oceanography, Krill and the Environment. Krill were sampled using active acoustics and a Fluorometer/Conductivity Temperature Depth (CTD) was used to collect environmental information and phytoplankton fluorescence. The hurdle model has two components: 1) a binomial generalized linear mixed model to model krill presence/absence in relation to environmental variables; 2) A mixed model to estimate krill density, given presence, from phytoplankton density and dissolved oxygen levels. A random effect for CTD station was included in both stages of the model due to sampling at multiple stations. Depth, temperature, salinity and light levels were predictors for krill presence/absence and a linear relationship was found between log krill, phytoplankton density and dissolved oxygen. By using a hurdle model, we were able to model krill density with high accuracy (presence/absence cross-validation: sensitivity = 0.69, specificity = 0.65; density model: conditional R2 = 0.69). This method allows us to quantify the environmental and biological patterns behind krill presence and may be a useful tool for understanding the future distribution of krill in a rapidly changing environment.
Ecological Research 1
Presenter: Julio Pereira
When: Monday, July 11, 2016 Time: 11:00 AM - 12:30 PM
Room: Salon C Carson Hall (Level 2)
Session Synopsis:
Predicting counts of fish based on zero inflated data in a stream network
When modelling counts of a species of fish sampled along streams, there are usually a large number of zeros counts that arise either because of species happens by chance not be present at time of sampling, or because the environment at the location is not suitable for the species. In addition, counts in a stream network may be correlated. However the dependence may be better explained by hydrological rather than Euclidean distance. Also the topological configuration of the network can affect how the counts are correlated. In this work we use hydrological distance and number of confluences between sampled locations in order to model the spatial dependence of the counts. Hence a random effect was included into the linear predictor of a zero-inflated Poisson regression model to accommodate the spatial correlation. A Bayesian approach was used for the inference procedure. The proposed method is illustrated using a data set composed of counts of fish and environmental covariates that were collected in streams of the Sorocaba and Paranapanema river basins in the state of Sao Paulo, Brazil.
Ecological Research 1
Presenter: Lasantha Premarathna
When: Monday, July 11, 2016 Time: 11:00 AM - 12:30 PM
Room: Salon C Carson Hall (Level 2)
Session Synopsis:
MLE and Bayesian Approach to Partial Stratification in Two-Sample Capture-Recapture Experiments of a Closed Population with Heterogeneity
Capture heterogeneity is known to cause bias in estimates of abundance in capture-recapture experiments. Often this heterogeneity is related to observable fixed characteristics of the animals such a sex. If this information can be observed for each handled animal at both sample times, then it is straightforward to stratify (e.g. by sex) and obtain stratum-specific estimates. In many fishery experiments, it is difficult, for example to sex all captured fish because morphological differences are slight or because of logistic constraints. In these cases, a sub-sample of the captured fish at each sampling occasion is selected and further, more costly, measurements are made. Our data now consists of three types of marked-animals; animals whose value of the stratification variable is unknown, and sub-samples at each occasion where the value of the stratification variables are determined. In this talk we develop and apply new methods for these types of experiments. Furthermore, given the relative costs of sampling for a simple capture and for processing the sub-sample, optimal allocation of effort for a given cost can be determined. We also develop methods using Bayesian implementation to account for additional information (e.g. prior information about the sex ratio) and for supplemental continuous covariates such as length. These methods are applied to a problem of estimating the size of the walleye population in MilleLacs, MN. Keywords: Capture heterogeneity, abundance, survey-design and analysis, MLE, Bayesian analysis, MCMC, Metropolis-Hasting.
Ecological Research 1
Presenter: Hideyasu Shimadzu
When: Monday, July 11, 2016 Time: 11:00 AM - 12:30 PM
Room: Salon C Carson Hall (Level 2)
Session Synopsis:
Rarefaction techniques and a bias in observed richness
The number of species, often called species richness, is a key aspect quantifying biodiversity in ecological communities, but also one of difficult indicators to deal with, in spite of our familiarity. Difficulties stem from the fact that the number of species we acquire, observed richness, is non-linearly related to other variables such as those, the number of individuals, the size of study area and the completeness of survey. In other words, observed richness increases or decreases in magnitude non-linearly with those variables. When we compare observed richness of different sites, it becomes a challenging task, since the extent to which the survey actually covers in terms of space and species may vary over different study sites. A standardisation method is therefore needed for a fare comparison of richness, and a technique called rarefaction has widely been utilised in quantitative ecology; it compares observed richness conditioning on an equalised number of individuals. This talk addresses an issue that the rarefaction technique induces, in spite of its popularity, a bias in observed richness and likely leads us to an unfair comparison of richness, although its conditional feature can eliminate sampling effects in a certain situation. Particular emphasis is given to highlighting the mechanism, how the extra bias is actually induced to observed richness by the rarefaction technique, with example data sets from biodiversity research.
Ecological Research 1
Presenter: Paul van Dam-Bates
When: Monday, July 11, 2016 Time: 11:00 AM - 12:30 PM
Room: Salon C Carson Hall (Level 2)
Session Synopsis:
New Zealand master sample using Balanced Acceptance Sampling
The Department of Conservation (DOC) has been progressively implementing the Biodiversity Monitoring and Reporting System (BMRS) to ensure DOC has effective monitoring in place to report on the status of New Zealands natural heritage and the impact of management. The sampling design to report on management success needs to be adaptable to changing management regions, a diverse number of ecosystems, and needs to be spatially scalable to inform at the local and national level. For this purpose DOC has created a master sample of New Zealand; a census of points across the country. These points have a visitation order such that any set of points that are included in order are spatially balanced. This allows DOC to sample at the national level choosing the first n points that appear in managed areas. It also allows for an increase in sampling effort (points) in a specific management unit to report at the local level. Balanced Acceptance Sampling (BAS), which utilizes a quasi-random number sequence called a Halton-Sequence, was used to create the sample. We then compared this with Generalized Random Tessellation Stratified sampling (GRTS) which is frequently used in this context.
