Analysing Spatial Structures

Abstract

In many cases, multivariate data are collected for entities that are geographically located (i.e., georeferenced). This chapter describes several techniques to incorporate the spatial information in multivariate methods using packages sp, spdep and adespatial. 12.1 Introduction Spatial data are commonly used in Ecology due to the development of technologies for their gathering (e.g., global positioning system, satellite imagery) and management (e.g., geographic information system). Hence, sampled entities (e.g., sites) are described by the measurements of environmental variables and/or species abundances as well as geographical attributes. Since the early work of Goodall (1954), a major concern of Ecology is the identification and explanation of the spatial patterns of ecological structures. Answering these questions leads to the notion of spatial autocorrelation and requires multivariate methods that consider explicitly the spatial information. Whereas traditional approaches used polynomial of geographical coordinates (trend-surface analysis) or distances (Mantel-based approaches), this chapter focuses on recent methods that introduce space using a Spatial Weighting Matrix (SWM). 12.2 Managing Spatial Data The sp package provides classes and methods to manage spatial data in R (Bivand et al. 2013; Pebesma and Bivand 2005). It allows to deal with raster (grid of cells) and vector (lines, points or polygons) data with or without attributes. This chapter focuses on vector data stored using the SpatialPoints and SpatialPolygons classes. Usually, spatial data are managed in Geographic Information System (GIS) and the maptools package (Bivand and Lewin-Koh 2017) contains functions to import these data directly in R. For instance, the © Springer Science+Business Media, LLC, part of Springer Nature 2018 J. Thioulouse et al., Multivariate Analysis of Ecological Data with ade4, https://doi.