How does spatial heterogeneity in landscape structure influence de-

While initially the influence of spatial heterogeneity on the demographic performance and spread of the invasion was posed as two separate questions in the introduction, the inter- actions between the two render them so interlinked (as described in the previous section) that I found it more sensible to address these influences together. As discussed above, the spatial dynamics of a population can be extremely influential even in homogeneous settings; these dynamics are further complicated by applying them in heterogeneous land- scapes where both the suitability and spatial distribution of di↵erent habitats is variable. This landscape-level heterogeneity results from a combination of the underlying mech- anistic processes that determine the distribution of habitats within the landscape, and the di↵erential suitability of those habitats from the perspective of the specific organism. The interaction of these processes result in di↵erent patterns of fragmentation as perceived by the organism in question. This fragmentation can a↵ect the population dynamics of a species via three primary mechanism; either by reducing the total amount of suitable habitat available to a species through conversion to a di↵erent habitat type (Andren 1994; Fahrig 2002; Herrera & Garc´ıa 2010), reducing the connectivity of suitable habitat (Tall- mon et al. 2004; Noel et al. 2006; Dornier & Cheptou 2012), or by creating edge e↵ects, where the proximity to the edge of a habitat results in altered environmental conditions within a habitat (Murcia 1995; Aguilar et al. 2006; Watling & Orrock 2010). Currently there is no evidence to suggest if or how the performance of H. lepidulum varies in relation to distance to the habitat edge; I therefore made the simplifying assumption that abiotic conditions are equivalent throughout the entirety of the habitats. Instead, my analyses focused on comparing how the availability and connectivity of suitable habitat influences the dynamics of H. lepidulum populations.

I simulated the invasion of H. lepdiulum in homogeneous landscapes of each habitat in the study area, along with a range of di↵erent heterogeneous mixtures and configurations designed to represent the defining characteristics of the existing landscape. I utilised four di↵erent metrics of spread to assess the dynamics of these simulations; population size, occupancy rates, local densities, and spatial extent. In order to assess how these metrics were influenced by the specific landscape configurations, I developed a baseline of projections which estimated the metrics assuming that the spatial configuration of the landscape had no e↵ect. I accomplished this by assuming the invasion in a specific habitat would perform the same as in a homogeneous landscape, simply scaled down based on the proportion of landscape in each habitat (see Chapter 6 for specific details). Simi-

lar research examining spread in heterogeneous landscapes has concluded that when a minimum of 30 to 40% of the landscape is considered suitable to a species, population performance is likely to scale with the amount of available habitat, and not su↵er losses in performance due to connectivity constraints (Andren 1994; King & With 2002; Frater- rigo et al. 2009). Below this threshold, the lack of connectivity compounds the lack of available growing space, and the decrease in performance accelerates.The heterogeneous landscapes used in the simulations (designed to reflect the primary features of the real landscape) were all comprised of more than 40% suitable habitat (based on suitability estimations from Chapter 5), leading me to hypothesise that connectivity would not limit population performance, and that the projections based on the compositional makeup of the landscape would provide a good approximation of the simulated performance. How- ever, the agreement between the projections and the simulations was mixed; the spatial extent of the population (in terms of either the occupancy or extent) was well predicted by the projections, and was generally in agreement with the simulations, suggesting that the spread of the invasion was not compromised by the connectivity of the heterogeneous landscape configurations examined here. In comparison, the measures of population size at both the landscape (total population size) and local (mean density) scales were signifi- cantly lower in the simulations compared to the projections, revealing that the landscape configurations examined here had a much greater impact on the severity of the invasion compared to the invasion extent.

These results exemplify how the connectivity of the landscape can have a di↵erential e↵ect on the underlying mechanisms which determine the extent and severity of an invasion. Previous work has illustrated that connectivity is able to a↵ect the ability of invasions to reach and establish at a location (Jules et al. 2002; Florance et al. 2011). However, the variable nature of seed dispersal means that connectivity (in terms of the continued exchange of propagules) does not exist as an absolute yes or no attribute, but instead is better reflected as the potential level of exchange between locations. Even between areas with very little connectivity, there is usually a small chance that a single dispersal event could occur between the locations. Such a singular event is all that is needed for successful establishment, and subsequent increase in the absolute spatial extent of the population (i.e. spread) (Moody & Mack 1988; Clark et al. 1998). This is particularly applicable in the case of apomictic species such as H. lepidulum, which have the potential to establish a viable population from a single nascent individual. However, while a minimal level of connectivity between locations can allow for establishment events, the frequency of propagule exchange between the locations may be low enough to the point where the exchange is functionally non-existent (Ferreras 2001; Kadoya 2009). As evidenced by the results of Chapter 5, this exchange between populations can have a significant e↵ect on the viability and size of the local population, directly influencing the severity of the invasion at a given location. While extent and severity have already been recognised as being largely shaped by separate mechanisms (Coutts et al. 2010; Harris et al. 2011), this

example illustrates how the structure of the landscapes and its associated connectivity can di↵erentially a↵ect those mechanisms, and independently a↵ect the extent and severity of an invasion.

The only caveat is that it is possible that this underperformance in population growth may represent only a transient lag in population performance as opposed to a more enduring resistance of the landscape (Crooks et al. 1999; Frappier et al. 2003; Catford et al. 2009). The dynamics observed here could be indicative of a secondary lag, where long distance dispersal events from the initial introduction provide a rapid increase in extent, while local growth is delayed until the populations reach a reproductive state in which they can supply an increased the local density requires a more local and substantial seed input (Wangen & Webster 2006). It appears that the spatial expansion of the population of H. lepidulum is achieved by a relatively small number of long distance dispersal events, while local growth and metapopulation dynamics are a function of local dispersal. This period of rapid expansion of invasion extent would then potentially be followed by a more gradual ’filling in’ of occupied area as the local populations continue to grow and approach their full reproductive capacity.

Additionally, refinement of the dispersal component of the simulation could significantly a↵ect on the rate at which the invasion achieves the observed levels of extent. While variations between the dispersal kernels of di↵erent habitats appeared minimal, their di↵erences were statistically significant. The propagation of these di↵erences across an entire landscape could result in substantial di↵erences in spread, and could a↵ect the influence of specific landscape configurations (i.e. increasing dispersal distances in suitable habitats may increase the importance of that habitat to landscape level spread; Hoyle 2007; Andrew & Ustin 2010. In addition, this examination assumes that the seed are dispersed only via their standard means (wind); incorporating any additional dispersal vectors such as animals or humans that may be transporting seed could have a significant impact on the overall invasion dynamics (Higgins et al. 2003; Crespo-P´erez et al. 2011).