Naturally fragmented or rare ecosystems are important components of terrestrial plant biodiversity. Unfortunately, many are also threatened through anthropogenic activities leading to increased extinction risk for their flora. In Aotearoa / New Zealand, the limestone areas of the eastern South Island / Te Waipounamu are a naturally fragmented and rare ecosystem of particular concern as they contain many highly threatened plant taxa. To assist with the ongoing conservation management of limestone endemic plants in New Zealand, the overall aim of this thesis was to explore patterns of genetic diversity and connectivity within a group of threatened limestone gentians (subspecies of Gentianella calcis and G. astonii) as well as investigate the interspecific taxonomic delimitation of G. calcis and G. astonii and infraspecific taxonomic delimitation of G. calcis. In Chapter 2, using Single Nucleotide Polymorphisms (SNPs) from 174 G. calcis and G. astonii samples genotyped through genotyping-by-sequencing (GBS), I aimed to determine the amount of genetic diversity in each subspecies and sampled population of G. calcis and G. astonii, assess the extent of genetic connectivity among them, and understand the geographic structuring of genetic diversity and its relationship to the environment. I show that all taxa are characterised by high population structure and limited genetic connectivity, with the presence of three main genetic groups corresponding to the South Canterbury and North Otago, Waipara, and North Canterbury and Marlborough regions. Although Isolation-By-Distance appeared to explain the observed patterns of genetic connectivity, potential adaptation to local climate and habitat soils was also seen. Patterns of observed heterozygosity potentially reflect past demographic histories as well as the effects of polyploidy-induced paralogy in some SNPs. Based on these findings, I designate conservation Management Units to assist with current and future conservation of G. calcis.

DiscoSNP-RAD represents a novel SNP discovery approach that claims to not require the same parameter optimisation as other commonly used programs such as Stacks. There are very few published comparisons of its output to other SNP discovery programs, however, illustrating the need for empirical studies. Considering these factors, in Chapter 3 I aimed to assess the importance of using similarity-based parameters in SNP discovery by comparing both SNP discovery methods (i.e. Stacks and DiscoSNP-RAD) in terms of RAD loci assembled, data error, and population genetic inferences (e.g., estimates of population structure and genetic diversity). While both approaches provided similar patterns of population structure, estimates of genetic diversity and pairwise Fst differed between the two approaches. Using sample replicates, I show that this is likely due to increased SNP error in the DiscoSNP-RAD dataset, potentially reflecting a greater proportion of paralogy-induced SNPs caused by lower user control over the formation of RAD loci. Despite this, considering it has faster run-time and does not need extensive parameter optimisation, I suggest DiscoSNP-RAD is still a useful SNP discovery program.

In Chapter 4, synthesising the learnings from Chapter 2 and Chapter 3, I make populationspecific management recommendations for each Gentianella calcis population along with taxonomic delimitation recommendations using knowledge of the genetic patterns in G. calcis and G. astonii. Considering that to date no conservation genetics or genomics studies have considered limestone plants in New Zealand at the population level, this research represents an important step towards the integration of genomic data into their conservation management.