Forecasting spatial changes in mangrove species' patterning under future sea level rise for Darwin Harbour, Australia

© 2013 Dr. Beth CraseMangroves forests provide critically important ecosystem services and products for human wellbeing and support highly diverse fauna. In addition to pressure from direct clearing, sea level rise over the past 30 years is altering their distribution and forecasts of their future spatial extent are required to manage these valuable ecosystems. Species distribution models (SDMs), used extensively to forecast spatial shifts in response to climate change for terrestrial species and to a limited extent in marine and freshwater ecosytems, are used here to forecast changes in spatial patterning in response to higher sea levels in an extensive mangrove forest in northern Australia. Key environmental variables influencing the distribution of habitat suitable for the dominance of three mangrove species, Sonneratia alba, Rhizophora stylosa and Ceriops tagal, are identified (chapter 2). The relationships between species’ dominance and the hydroperiod (defined as the duration of inundation in a year), soil salinity and the salinity of inundating water were determined using Generalized Linear Models (GLMs) and Boosted Regression Tree models (BRTs). In the models of species dominance, the hydroperiod was identified as the key variable influencing distribution of dominance, followed by soil salinity and the salinity of inundating water. Changes in the distribution of the mangrove forest in response to a sea level rise of 0.5 and 1 m were forecast in chapter 3. A reduction in the total extent of forest was forecast to be 3.4 and 8.1% under sea level rises of 0.5 and 1 m respectively. I developed forecasts of the distribution of the three focal species and of three plant community types, in addition to changes in the spatial pattering of dominance. Mangroves forests are structurally simple forests, dominated by few species that occur across a clearly partitioned environmental gradient, thus providing a model system within which to test whether these different components (species, dominants and communities) respond differently to sea level rise. The response variables (fitted in a total of nine separate models) were the presence or absence of S. alba, R. stylosa or C. tagal (as recorded at field sites); the dominance or non-dominance of S. alba, R. stylosa or C. tagal (as recorded in at field sites); and the presence or absence of the S. alba community, R. stylosa community or C. tagal community (as shown in maps of community distribution). Two seedling establishment scenarios were implemented. Under a scenario of limited seedling establishment, the area occupied by all the three focal taxa declined substantially. However, under a scenario of unlimited seedling establishment, substantial declines were forecast only for C. tagal (species, areas suitable for dominance, and the community type) while the habitat suitable for S. alba and R. stylosa expanded. As SDMs are widely used to forecast changes in the spatial distribution of species and communities it is crucial that the assumptions of these techniques are not violated and that SAC is accounted for in order to reduce bias and inaccuracies. As SDMs do not usually incorporate SAC (environment-only models), in Chapter 5 I quantified the spatial difference in forecasts of habitat suitability produced by environment-only models and two modelling approaches that can incorporate SAC, the RAC approach developed in Chapter 4, and the widely used autologistic approach. BRT models were fitted to simulated data and observed data for three mangrove communities, and forecasts produced for a series of climate change scenarios. Although the fit of all three modelling approaches to the initial distribution of organisms was excellent in most cases, the environment-only models forecast substantially greater range shifts under simulated climate change than did the models incorporating SAC. Furthermore, the magnitude of these differences increased with increasing climate change. The RAC approach forecasts were intermediate between environment-only and autologistic approaches due to the incorporation of both extrinsic and intrinsic SAC. As published forecasts rarely incorporate SAC, changes in the spatial distribution of suitable habitat and total available habitat under climate change may have been overestimated for many of these species. The findings of this thesis have important implications for modelling future distributions of species distributions in general, and of mangrove species in particular. Furthermore, incorporating spatial autocorrelation in species distribution modelling, will lead to improvements in species distribution modelling practice, and increased confidence in the future predictions of species distributions