Browsing by Subject "Bayesian"

Integrated population models (IPMs) are a promising approach to assess and manage wildlife populations in dynamic and uncertain conditions. By combining multiple data sources into a single, unified model, they enable the parametrization of versatile, mechanistic population models that can predict population dynamics in novel circumstances. This is in contrast to traditional approaches where independent empirical estimates for demographic parameters are typically incorporated into a population projection matrix such as a Leslie matrix. A major limitation of conventional methods is their inability to fully utilize all available information, as the synergies between different data sources are not exploited. The Baltic ringed seal (Pusa hispida botnica) presents an example illustrating the limitation of conventional monitoring approaches. Despite the availability of long-term monitoring data, population assessment is hindered by dynamic environmental conditions, varying reproductive rates, and the recently re-introduced hunting, thus limiting the quality of information available to managers regarding, for example, hunting quotas. In particular, population counts of ringed seals from aerial surveys have exhibited unexpected trends and large fluctuations during the last decade, making it impossible to obtain reliable estimates of population growth from survey data alone. This thesis presents a Bayesian IPM for the ringed seal population inhabiting the Bothnian Bay in the Baltic Sea. The central aim of this work is to outline an approach that can overcome some of the challenges that have crippled Baltic ringed seal monitoring efforts during the last decade, and support science-based management decisions. The thesis broadly consists of three parts. First, a state-space model is presented for the Bothnian Bay ringed seal population. Demographic processes are described through a stochastic age and sex structured population model that includes both hunting mortality and the hypothesized effects of environmental variables such as pollution and sea ice cover on demographic parameters and seal behaviour. Next, the model is fit to census and various demographic and reproductive data, as well as hunting statistics, from 1988 to 2023 under a Bayesian framework where posterior samples of model parameters are obtained using Markov Chain Monte Carlo methods. Finally, posterior estimates of model parameters are used to construct a Leslie matrix, and model behavior is analyzed using methods developed for matrix projection models. Future population dynamics are also simulated under alternative management scenarios to inform ringed seal management decisions. In general, this thesis demonstrates the value of mechanistic IPMs for monitoring and managing natural populations under changing environments, and supporting science-based management decisions.

Unsupervised learning techniques can detect clinically relevant structure in population cohort data of human gut microbiota. While the gut microbiota composition is influenced by individual factors such as diet, medication, and development of the immune system during early childhood, it is proposed that individuals maintain a relatively stable microbiota ecosystem throughout adulthood. This stability allows to distinguish individuals into subgroups based on their gut microbiota characteristics, which define the key features of microbiota community types within the population. For this, I compared three probabilistic unsupervised learning techniques, optimization-based Non-negative Matrix Factorization, and Bayesian modelling techniques, Dirichlet Multinomial Mixtures and Latent Dirichlet Allocation, with a naive benchmark clustering based on dominant taxa. I used an all-cause mortality association strength as a quantitative metrics to distinguish biologically relevant structure in a large Finnish population cohort with almost 18 years follow-up. The techniques defined microbiota assemblages as either discrete enterotypes, which assigned each sample to a single community type, or continuous enterosignatures, which identified patterns of co-occurrence of microbiota community types within each sample. I found five rather robust community types, characterized by Bacteroides, Alistipes, Agathobacter, Escherichia, and Prevotella bacterial genera. Latent Dirichlet Allocation detected the strongest early mortality signal using Cox regression, outperforming all other techniques. The replicability of Latent Dirichlet Allocation was assessed using cross validation. The predicted community types uncovered similar ecological landscape on the data with the community types obtained using the entire data, confirming the clinical relevance, robustness, and scalability of the technique.