The need to understand the dynamics of species distributions in space and time is essential in ecology, and in particular in biological invasions where the spatial pattern of invasion is specific to a time, space and spatial scale (Theoharides and Dukes,
2007). To understand plant species patterns (e.g. composition and richness) in relation to the spatial distribution of environmental factors, two phases are required: data gathering and analysis. In this chapter, I focus on the data gathering that takes
the form of a floristic survey that defines the area of investigation and the potential environmental factors that may characterize this area.
The Banks Peninsula comprehensive floristic survey was designed to find out as much as possible about the vegetation cover and species distribution patterns on the Peninsula (Wilson, 1992). But, if we are to fully understand species patterns within this area we need also to collect data on factors that might affect species distributions. The larger and more diverse the study area is the larger and the more variable is the dataset necessary for an adequate representation of its vegetation (so-called ecological representativeness;sensu Stohlgren et al., 2003b;Rew et al.,
2006;Role˘ceket al.,2007). Because factors underpinning plant species distributions patterns operate at different spatial and temporal scales (Collinghamet al.,2000), it is essential to identify their relative importance and emphasize a congruent spatial scale between the resolution of species and environmental factors. Bearing in mind this, there is often issues of discrepancy between map or layer resolution and spatial scales at which the ecological studies is conducted. In fact, environmental factors are usually available at a relatively coarse scale, but other factors (e.g. geological and lithological maps, soil moisture or soil nutrient content) may vary substantially on a much finer grain or be unknown (Role˘ceket al.,2007). To deal with this, we need to emphasize a congruent spatial scale between the resolution of species and environmental factors. Spatial scale is in fact important to both species distribution and related environmental data (Elith and Leathwick,2009). When studying their relationship, it is important that the grain (i.e. grid cell or polygon size) of the explanatory variables is consistent with the species data layers (Elith and Leathwick,
2009). However, in most cases around the world, as in Banks Peninsula, this kind of consistency is difficult to find.
For this, Geographic Information Systems (GIS), Remote Sensing (RS) and spatial analysis techniques are useful tools in landscape ecology and biogeography, as well as in invasion ecology. GIS and RS have emerged as distinct spatial data handling technologies with their own methods of data representation and analysis (Goodchild,1994). These technologies have attracted considerable interest in the field of modelling of plant species distribution in recent years. In the context of plant invasions, "these are paving the way for very detailed and novel studies of species patterns and processes" (Richardson,2004). Integrated GIS and RS have been successfully applied to detect and map the distribution of several alien plant species (e.g.Dark,2004;Rewet al.,2005;Higginset al.,1999;Evangelistaet al.,
Bradley and Mustard,2006;Heet al.,2011), their ecosystems (e.g.Deutschewitz
et al.,2003), their bio-climatic conditions (e.g.Rougetet al.,2004) and the drivers
affecting invasions (e.g.Rouget and Richardson,2003;Foxcroft et al.,2004;Pino
et al.,2005). For instance, climate, environmental (e.g. topography) and potential
human-related disturbance (e.g. land-use and propagule pressure) have been shown to influence alien species (e.g.Rouget and Richardson,2003; Pinoet al.,2005) and native species distribution (Deutschewitzet al.,2003;Dark,2004).
The ability to analyse, map and model plant species are just a few of the many advantages of using GIS and RS for this work. These technologies contribute to our understanding of the width dynamics of species patterns, and integrated with spatial analysis allow us to design objective, efficient sampling methods (e.g. random, systematic or stratified) and/or to evaluate their adequacy in capturing species and environment (Neldneret al.,1995). We tend to evaluate sampling methods especially when the study area is large and cannot be entirely sampled (Rewet al.,2006). For instance, opportunistic sampling method (i.e. samples collected from known localities or in easily accessible areas) may be preferred to unbiased methodssensuRewet al.
(2006), such as random or stratified random sampling, but the latter may fail to capture the spatial variation or spatial dependency of the environment, leading to difficulties in detecting spatial relationships or incorrectly inference statistical models (Legendreet al.,2002;Fortin and Dale,2005).
To achieve a representative sample of the population (e.g. the Banks Peninsula vegetation), the appropriate theories to guide sample design need to be based on an understanding of the geographical structure and ecological organization of plant communities with regard to the heterogeneity of the sampled vegetation type (Huebner,2007) or the environments associated with species occurrence (Stohlgren
et al.,2003b). For instance, an opportunistic sampling method is useful in defining
the breadth of environmental conditions that characterize each species location (i.e. environmental envelope;Jarnevichet al.,2007;Evangelistaet al.,2008). However, this non-objective sampling method has some limitations such as over-estimating species presence or missing species presence in localized random samples (Stohlgren and Schnase,2006;Fitzpatricket al.,2009).
In this chapter, I describe the GIS, RS operations and tools, and spatial analysis techniques I used to analyse the range of environments on Banks Peninsula associated with the floristic survey and to ask:
1. In this highly variable landscape, how can plant species distribution be best described using available spatial data?
2. Is a systematic sampling scheme suitable for providing a truly representative picture of the Banks Peninsula environments?
3. Do the environmental variables extracted using this sampling approach violate any statistical assumptions?