INIBIOMA   20415
INSTITUTO DE INVESTIGACIONES EN BIODIVERSIDAD Y MEDIOAMBIENTE
Unidad Ejecutora - UE
congresos y reuniones científicas
Título:
General Tools to Model Movement in Highly Fragmented Patch Networks
Autor/es:
MORALES, JUAN M.
Lugar:
Seattle
Reunión:
Conferencia; International Statistical Ecology Conference; 2016
Resumen:
For many theoretical and applied questions it is important to understand how animals move in a patch network. However, this is no simple task because the probability that an individual moves from one particular patch to another one is affected in a non-trivial way by the characteristics and location of other patches in the network. Here we present some simple and flexible statistical movement models that take into account the spatial structure of a patch network. For this we follow Ovaskainen and Cornell (2003) derivation of the formulae for diffusion in highly fragmented landscapes but replace some of their analytical results with general "statistical" functions with parameters that need to be estimated from data, leaving diffusion as a special case. We start by considering a matrix H holding the probabilities of eventually going to patch j before dying or emigrating from the patch network given that the animal has just left patch i. The elements of this matrix are a function of the distance between patches dij but can also depend on other attributes such as patch size, quality, etc. These Hij values can be thought of as the probabilities that an individual starting at a distance dij from patch j will eventually reach it before dying considering that there are no other patches in the landscape. From matrix H we want to obtain the probabilities Pij of visiting next patch j given that the individual has just left patch i. If we assume that Pij depends only on the animal just leaving patch i but not on the full history of previous movements, we can write Hij as a combination of these Pij: H_ij=P_ij+∑_(k≠j)▒〖P_ik H_kj 〗. We then use a linear solver to obtain the Pijs. This is the main "trick" that we rely on for the purposes of estimating the effect of the spatial structure of the patch network. The Pij estimated in this way depend on movement properties and on the structure of the habitat network but are valid only for the case when animals forget where they were coming from once they arrive to a patch. However, after we have solved for Pij we can modify these probabilities using weights that depend on the visitation history.We present both a MCMC and ABC toolbox to estimate the relevant parameters for the case of data collected through tracking animals (MCMC) and mark-recapture protocols (ABC).