A carregar...
Publicação
A Hidden Markov Model approach to analyze fish behavior
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.34 MB | Adobe PDF |
Orientador(es)
Resumo(s)
Hidden Markov models are a popular tool for analyzing animal movement data. In order to infer Wels catfish (Silurus glanis) behavior from observations spaced in 40-second time intervals including depth, water temperature and accelerometer coordinates, a 3-state hidden Markov model was implemented, with depth as a state-dependent variable following a zero-inflated gamma distribution. The three states presented clearly distinct mean values of depth (0.9, 2.3 and 5.7 meters for states 1, 2 and 3, respectively). The transition probability matrix based on mean covariable values was similar to the identity matrix, with transition probabilities within the same states very close to one and close to or equal to zero (in both transitions between states 1 and 3) in the other matrix entries, suggesting an extreme difference in activities in these two states. State 2, a candidate for the intermediate state connecting states 1 and 3, was the most common state for catfish, and stationary state probabilities showed that it was resistant to almost any change in covariable values, except for very warm water temperatures (a rare occurrence), where state 1 thrived. States 1 and 3, on the other hand, were sensitive to different levels of activity and periods of the day. State 1 lost relevance in high activity and during night (in which Wels catfish is known to be more active), while state 3 gained relevance in those territories. The model analysis led to the conclusion that state 1 was identified as the resting state for catfish, state 2 was identified as a low to moderate activity or swimming, and state 3 was considered a prey hunting or burst movement activity.
Descrição
Tese de mestrado, Bioestatística, 2025, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Hidden Markov models transition probability matrix hidden states Wels catfish