Neotropical forests have the highest tree diversity on earth, with an estimated 22,000 species. In contrast temperate North America, Europe and Asia combined support only 1,166 tree species. In the western Amazon rainforest, complex patterns of edaphic heterogeneity have been invoked as potential drivers of plant diversity through local. Numerous studies have demonstrated that physiological trade offs associated with adaptation to different habitat types leads to ecological sorting which in turn drives the spatial distribution of tropical trees along environmental gradients. If populations that are locally adapted to different environments experience a reduction in gene flow they may continue to differentiate potentially leading to new species formation.

Ecological speciation, whereby divergent natural selection results in reproductive isolation occurs as a direct result of adaptation to ecological conditions, is thought to be an important driver of plant diversification. Ecological speciation is known to occur in allopatry and a number of studies have demonstrated that it may be possible in parapatry and sympatry. However, the latter scenario, in which divergence and speciation occurs in the face of gene flow remains a point of contention. In the absence of geographic barriers to gene flow, disruptive selection must be strong enough to overcome genetic `dilution' from neighboring populations. Environmental factors commonly associated with ecologically based disruptive selection include variation in topography, microenvironmental differences, soil heterogeneity, herbivore pressure, and pollinator differentiation, all of which can reduce gene flow, potentially leading to the evolution of reproductive barriers.

Reproductive barriers can act to isolate diverging lineages before or after fertilization. Prezygotic isolating mechanisms in plants can be intrinsic, usually entailing pollen incompatibility with the stigma or ovule, or extrinsic, where isolation is ecologically driven, as with pollinator-mediated barriers or changes in flowering phenology. Postzygotic isolating mechanisms can be intrinsic, in the case of hybrid sterility and inviability, or extrinsic, where isolation is enforced through ecologically based natural selection where hybrid offspring are less suited for survival in either parental habitat type. To date, only a handful of studies have comprehensively quantified the strength of individual reproductive isolating mechanisms among closely related plant species, the majority of which are herbaceous and found in temperate zones. Even fewer studies have focused on tropical plants and those that have focused on herbaceous lineages.

This dissertation aimed to investigate the broader role of divergent natural selection as an agent in generating and maintaining Amazonian tree diversity using edaphically divergent populations of Protium subserratum. Using a combination of molecular population genetics field observations and hand-pollination experiments I: (1) Developed nuclear microsatellite markers in order to evaluate population level differentiation and gene flow between populations of P. subserratum associated with different soil types. (2) Inferred the role of divergent natural selection in driving the genetic structure of P. subserratum populations found on clay, brown-sand, and white-sand soils distributed more than 100km across the Peruvian Amazon. (3) Investigated the role of habitat and distance in driving turnover in stingless bee communities, the putative effective pollinators for P. subserratum. (4) Systematically evaluated the role of multiple pre- and post-zygotic barriers to reproduction in maintaining population integrity between parapatric populations of white-sand and brown-sand P. subserratum.

I successfully developed 17 polymorphic nuclear microsattelite markers, which were subsequently used to assess population variation across populations of P. subserratum, found on different soil types. I found evidence that suggests that edaphic specialization has occurred multiple times in P. subserratum and that natural selection may be driving divergence across edaphic boundaries. I found that location and soil type play a significant role in structuring stingless bee communities in white-sand and non-white sand forest and that community turnover may be more strongly influenced by distance in white-sand habitats than non-white sand habitats. I was able to identify four active barriers to reproduction between parapatric white-sand and brown-sand populations of P. subserratum including ecogeographic isolation, differential pollen adhesion, differences in pollinator assemblages, and low levels of hybrid seed development. I demonstrated that a combination of pre-zygotic and post-zygotic barriers to reproduction act to maintain near complete reproductive isolation between edaphically divergent populations of the tropical tree, P. subserratum.

Adaptive radiations are thought to be one of the most important processes generating biological diversity on Earth. Although the existence of adaptive radiations is not in doubt, the exact mechanisms via which adaptation to different habitats translates into lineage splitting has been debated for over a century. Since the time of Darwin, biologists have invoked trade-offs during adaptation to environmental gradients as being the key to linking adaptation to species formation. However, identification of the causal gradients, trade-offs and extrinsic selective regimes involved in adaptive or ecological speciation requires detailed fieldwork and experimentation and cannot be inferred using genetic or observational data alone.

This series of studies provides an empirical, experimental basis for the conclusion that adaptation to different habitats is driving divergence in Encelia (Asteraceae) a genus of perennial desert shrubs that has radiated extensively throughout the Mojave and Sonoran deserts. Encelia is an ideal system for studying the mechanisms of adaptation and speciation because all taxa are completely interfertile, and many are considered habitat specialists and form hybrid zones wherever their distributions abut. Reciprocal transplant field experiments between two taxa, Encelia palmeri and E. ventorum, showed that extremely strong postzygotic divergent natural selection is primarily responsible for preventing species fusion. A resource manipulation experiment between the same two species further showed that this was due to trade-offs caused by a gradient in water availability between dune and desert habitats. Patterns of seed germination, herbivory, and burial by sand were also important and showed interactions with taxa and habitat indicating that divergent selection at this site is complex and multifaceted.

A second study involving the same two species asked whether a signature of postzygotic selection can be seen in the distribution of phenotypes through time and along a gradient of habitat disturbance. I found that novel phenotypes suggestive of recombination are produced at a high frequency but are not present in the adult population indicating a role for postzygotic natural selection in removing the products of recombination from the population. This conclusion was further strengthened by looking at disturbed versus undisturbed habitats. Disturbed habitats contained novel phenotypes suggestive of recombination that were absent in the undisturbed sites indicating that the hybrid swarms that frequently follow disturbance are likely caused by an alteration of postzygotic selective pressures. The higher resource availability of the disturbed sites suggests that a relaxation of selection is likely responsible.

These results are elaborated on further by conducting a literature review of cases where hybridization, species fusion or hybrid swarm formation are associated with disturbances or changes in environmental forcing factors. I found that there are many cases in the literature that describe taxa maintained primarily if not solely by extrinsic postzygotic selection, although this appeared to be more true for plants than animals. Animals, in contrast, were isolated primarily by prezygotic barriers including allopatry and sexual selection. This discrepancy may help explain the disagreement between zoologists and botanists for the past century about species definitions; zoologists have typically favored prezygotic criteria while many botanists have pointed out that definitions based on reproductive compatibility fail to capture much of the variation observed among plants.

Finally, I conducted a common garden experiment and an analysis of climatic niches with eight species in Encelia in order to determine whether there is evidence that selection is driving divergence among other taxa in the genus in addition to E. palmeri and E. ventorum. All taxa studied showed strong climatic differentiation according to temperature and precipitation, and trait divergence in the group was high with some taxa showing evidence of the evolution of key traits allowing colonization of high altitude and hyper-saline habitats. Despite this, trait variation did not reliably follow the predictions of leaf economic theory either within or among taxa. This may be due to the existence of multiple alternative ecologically equivalent strategies that may introduce noise into low-dimensional analyses of functional traits and climate. There were, however, exceptions to this pattern. Specific leaf area showed coherent variation within taxa but not among taxa, and ecotypes E. farinosa also varied in the directions predicted by leaf economic theory for nearly all of the traits examined.

These and previous studies establish Encelia as a classic case of adaptive radiation and underscore the importance of empirical, field-based studies for disentangling the complex mechanisms driving adaptation and the formation of new species.