Predicting patterns in vegetation across landscapes: from structure and composition to quality and quantity

A fundamental principle underpinning conservation is to evaluate the condition of native vegetation relative to a reference state. Many ecological applications require vegetation condition as a spatially-explicit continuous map layer that covers broad landscape-scaled extents. Typically, vegetation condition is assessed by comparing contemporary vegetation to a historical, pre-colonial or pre-industrial reference state. Benchmarks used to define historical reference states have a high degree of uncertainty. Furthermore, their relevance for contemporary ecosystem management is questionable given changes that have occurred to ecosystems through time (e.g. due to human disturbance or climate change). I outline an alternative conceptual framework for defining the reference state. I propose the contemporary reference state offers a quantifiable approach to assessing vegetation condition relative to empirically-defined, Best-on-Offer benchmarks. In the first section of my thesis, I overcome several challenges associated with synthesising data from plot-based floristic inventories. I addressed the critical and universal problem of transforming estimates of plant foliage cover recorded on an ordinal scale (e.g. Braun-Blanquet cover-abundance scale) to continuous quantitative estimates. I demonstrate that by accounting for the underlying skew of these data, more robust transformations of ordinal cover-abundance were generated. Unlike previous approaches to transforming ordinal data, this method does not overinflate summed cover. Moreover, I show that different growth forms (such as trees, shrubs, forbs and grasses) require different transformation values. Another issue associated with floristic plot data that profoundly affects spatial modelling are errors in the location of on-ground, geo-referenced location of floristic plots. Datum errors are prevalent in archival data but difficult to detect. The method I outline to identify enigmatic sources of error is crucial for building accurate spatially-explicit models. After solving these data challenges, I aggregate floristic observations into plant functional groups. I then calculate the summed foliage cover and summed species richness of four growth forms as well as the total foliage cover and total native species richness. I further develop this modelling approach to score the structural and compositional characteristics for plant functional groups by comparing the observed values (either summed cover and richness) to their empirical benchmark values. I used these models to generate spatially-explicit surfaces of the status of plant functional groups relative to their benchmarks, thus mapping the condition of vegetation consistent with my conceptual model to help achieve better biodiversity outcomes in contemporary ecosystems. In completing this research, I demonstrate how a large repository of existing floristic plot data can be value-added to predict vegetation relative to a reference state and therefore offer an estimate of condition. These new maps complement existing maps of vegetation communities but deliver more specific detail on the structure, composition or relative quality of vegetation. Notably, with the ever-growing quantity of plot-based floristic inventories, this research has global applications to strengthen, evidence-based decisions for conservation planning and landscape management using pan-continental datasets