Spatio-temporal modelling in ecology using INLA and inlabru

Jafet Osuna & Janine Illian (University of Glasgow)

Friday 8th March 13:00-14:00 Maths 311B

Abstract

Topic 1: Complex spatio-temporal modelling in practice - working and communicating with users

Ecological spatio-temporal modelling serves as a crucial tool for understanding the dynamic interactions within ecosystems. However, the inner complexity of ecological systems along with complex observational processes, make the analysis of ecological data challenging. Moreover, the rise of new technologies and emerging data submission platforms over recent years, has had provided access to large volumes of ecological data. As the volume of ecological data continues to grow, so does the need for computationally efficient methods for fitting complex spatio-temporal ecological models. In this sense, the Integrated Laplace approximation (INLA) has gained popularity as a Bayesian inferential framework for many ecological models, as it allows an accurate and computationally efficient estimation of the flexible class of Latent Gaussian Models (LGMs). INLA’s increasing popularity is also owed to its ongoing development and derivative projects that enhance and expand upon INLA's capabilities/ An example of such is inlabru, a wrapper around R-INLA which facilitates model construction and predictions and adds extra functionality to the already powerful machinery that is available in the R-INLA software.

This talk will give an overview of recent development in ecological space-time models and discusses some of the challenges of effectively communicating such developments to a larger class of audiences. The aim is to make these advanced modelling techniques more accessible and beneficial for a wider range of users.

 

Topic 2: Spatial occupancy models using Integrated Nested Laplace Approximation

Modern methods for quantifying, predicting, and mapping species distribution have played a crucial part in biodiversity conservation and management. Occupancy models, in particular, have gained popularity for analyzing species occurrence data due to their ability to distinguish between observation errors caused by imperfect species detection and biases affecting the occupancy process. However, in many applications, the spatial and temporal variation in occupancy that is not accounted for by environmental covariates is often ignored or modelled through simple spatial structures because of the computational costs of fitting explicit spatio-temporal models. In this work, we demonstrate how INLA can be an efficient Bayesian inferential approach to fit such complex models. We show how the R-INLA package can provide an interface that allows for different structures such as spatio-temporal random effects and smooth terms to be included.

Add to your calendar

Download event information as iCalendar file (only this event)