Next generation sequencing (NGS) technologies have seen successful applications in ecology and evolutionary biology. The ever expanded applications of NGS to non-model organisms in the wild populations make it possible to address important ecological and evolutionary questions at larger and more precise scales. These questions vary from deciphering the genetic basis of traits underlying rapid ecological adaptation and speciation to identifying the genome wide differentiation patterns of populations at different divergence stages. The later question can only be answered when the whole genome of the species is available. The Australian groundsel, i.e., Senecio lautus is a good system to study ecological speciation. This species complex consists of multiple ecotypes adapting to different environments, among which the Dune and Headland ecotypes occur proximately to each other in several coastal localities of Australia, displaying very contrasting morphologies despite being interfertile. Phylogenetic analyses supported independent origination of each Dune and Headland pairs and ecological studies have correlated phenotypic difference with environmental adaptation. Here I used methylation filtration to develop gene space and transcriptome sequencing (RNAseq) to construct reference transcriptome for this species. I explored multiple strategies to optimize the final results in terms of assembly completeness, accuracy and contiguity. For gene space sequence assembly, I found hybridizing results from different assemblers used for assembling specific data type is better than combining results from hybrid assemblers (i.e., assemblers assembling all types of data at once). In the case of transcriptome assembly, I found the multiple spectrum assembly strategy (i.e., multiple assembly parameters and multiple assemblers) reconstructed more genes, but introduced more redundancy, complicating downstream analyses such as gene family reconstruction, phylogenetic and population genetics analyses. I also found that redundancy reduction based on expression level is better than the other methods. I used gene space assembly as reference sequences to detect the genome wide divergence patterns of multiple Dune and Headland ecotypes and found that geographic distance affects the magnitude of genetic differentiation between populations, but not the genomic divergence patterns. Applying pooled RNAseq to four S. lautus ecotypes allows population genomics and molecular evolution analyses leading to the finding of short divergence time among these ecotypes and candidate genes that have potentially contributed to adaptive divergence of S. lautus. Moreover, I also found decoupled differential gene expression and coding sequence divergence patterns, suggesting the rapid divergence of S. lautus has been achieved through evolution at both levels. These works are essential to shape S. lautus up as an excellent system to study ecological speciation with gene flow.