A shiny R app for spatial analysis of species distribution models

• Published in: Ecological informatics Vol. 80; p. 102542
• Main Authors: Figueira, Mario, Conesa, David, López-Quílez, Antonio
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01-05-2024
• Subjects: Geostatistics; INLA; Preferential models; Shiny; Species distribution models; Species distribution models; INLA; Preferential models; Shiny; Geostatistics
• ISSN: 1574-9541
• DOI: 10.1016/j.ecoinf.2024.102542

In ecology, Species Distribution Models (SDMs) are a statistical tool whose use has expanded considerably over the last two decades. As their use has grown, so has the complexity of the data analysed and the structures of the models used.
This has led to the development of various tools to facilitate the incorporation and use of these new data and statistical methodologies, mostly embodied in new R packages and Shiny applications that allow different types of SDMs to be solved. However, the Integrated Nested Laplace Approximation (INLA) approach, which has become increasingly popular in the field of ecological sciences, has not yet been integrated into an application that can synthesise the complexity of its code into a user-friendly interface for continuous spatial modelling.
To overcome this shortcoming, we present in this paper a novel application, called BAYSPINS (BAYesian SPatial INla for SDMs), which allows the use of INLA for those who are not very experienced, or for those who are experienced and prefer a tool that allows them to carry out an initial analysis quickly, avoiding the process of writing code.
BAYSPINS allows both geostatistical and preferential modelling, as well as a mixture of the two. It integrates the complex and hard coded SPDE-FEM (Stochastic Partial Differential Equation, along with the Finite Elements Method) approach to perform continuous spatial analysis with a visual interface. It also allows the use of default settings that automate the process or the customisation of a large number of elements that drive the modelling process. In this way, quick initial evaluations or more rigorous studies of the data provided by the user can be carried out, depending on the user's skill and understanding of the fundamentals underpinning the application.