Spatio-Temporal model structures with shared components for semi-continuous species distribution modelling
David Conesa (University of Valencia)
Friday 3rd November, 2017 15:00-16:00 Seminar room 311B
Understanding the spatio-temporal dynamism and environmental relationships of species is essential for the conservation of natural resources. Many spatio-temporally sampled processes result in continuous positive [0, ∞) abundance datasets that have many zero values observed in areas that lie outside their optimum niche. In such cases the most common option is to use two-part or hurdle models, which fit independent models and consequently independent environmental effects to occurrence and conditional-to-presence abundance. This may be correct in some cases, but not as much in others where the detection probability is related to the abundance. The aim of this work is to infer the spatio-temporal dynamism of ecological processes and to fit more robust environmental effects in two-part models. On the one hand we propose different spatio-temporal to infer the fundamental spatio-temporal behaviour of the process under study. On the other hand, we propose the use of shared component modelling (SCM) to estimate more robust model effects in related semi-continuous datasets by combining information from occurrence and conditional-to-presence abundance. We use a simulation study to test the application of shared components in two different types of semi-continuous datasets. Lastly, we implement all the proposed model structures in a case study on the recruitment of hake in the western Mediterranean.