Bayesian zero-inflated models for stream fish distribution and abundance in counts and biomass: spatially correlated data under a distance based on the channelling of water flow and its confluences

• Grant number: 15/03739-2
• Support Opportunities: Scholarships abroad - Research
• Start date: August 01, 2016
• End date: January 31, 2017
• Field of knowledge: Biological Sciences - Ecology - Applied Ecology
• Principal Investigator: Júlio César Pereira
• Grantee: Júlio César Pereira
• Host Investigator: Martin Spencer Ridout
• Host Institution: Centro de Ciências e Tecnologias para a Sustentabilidade (CCTS). Universidade Federal de São Carlos (UFSCAR). Sorocaba , SP, Brazil
• Institution abroad: University of Kent, England

Abstract

The quantitative modelling of abundance and fish species distribution in streams can be difficult due to some particularities in the data, such as many zeros, which are caused by catch abscence, and correlation between data collected from different points. In this way, the project target is to propose Bayesian models which fit well and provide precise predictions of abundance and species distribution. Also, this model will take into account the zero-inflation and spatial correlation determined by a distance depending on the channelling of water flow and its confluences, because the Euclidean distance may not be suitable in these cases. Considering the catch in number of fish, to achieve the target a random componnent will be added to the zero-inflated Poisson and to the zero-inflated negative binomial models to accommodate the spatial correlation which depends on the non-Euclidean distance. Considering the catch in biomass, the correlation will be added to the compound Poisson model, which belongs to the Tweedie distribution family. Simulations will be performed to assess the fitting and predictive capacity of the proposed models and to compare them to simpler versions of themselves. In order to illustrate the proposed methodology an application of the models to a real data set will be presented considering data sampled by experimental fishing in streams of the Sorocaba and Paranapanema river basins and in streams of Corumbataí river basin, in the São Paulo state. (AU)