UC Davis

Scoliid wasps comprise a clade of aculeate insects whose larvae are parasitoids of scarabaeid beetle grubs. While scoliids have been studied and used as biological control agents, research into the group's evolution, as well as the stability of scoliid taxonomy, has been limited by a lack of reliable phylogenies. In Chapter 1, ultraconserved element (UCE) data are used under concatenation and the multispecies coalescent to infer a phylogeny of the Scoliidae. Data filtering experiments using posterior predictive checks and matched-pairs tests of symmetry are performed in order to mitigate potential issues arising from model misspecification. Analyses confirm the position of Proscolia as sister to all other extant scoliids. There is also strong support for a sister group relationship between the campsomerine genus Colpa and the Scoliini, rendering the Campsomerini non-monophyletic. Campsomerini excluding Colpa (hereafter Campsomerini sensu stricto) is inferred to be monophyletic, with the Australasian genus Trisciloa recovered as sister to the remaining members of the group. Out of nine genera in which more than one species was sampled, Campsomeriella, Dielis, Megascolia, and Scolia are inferred to be non-monophyletic. Analyses incorporating fossil data indicate an Early Cretaceous origin of the crown Scoliidae, with the split between Scoliini + Colpa and Campsomerini s.s. most probably occurring in the Late Cretaceous. Posterior means of Scoliini + Colpa and Campsomerini s.s. crown ages are estimated to be in the Paleogene, though age 95% HPD intervals extend slightly back past the K-Pg boundary, and analyses including fossils of less certain placement result in more posterior mass on older ages. Estimates of the stem ages of Nearctic scoliid clades are consistent with dispersal across Beringia during the Oligocene or later Eocene. This study provides a foundation for future research into scoliid wasp evolution and biogeography by being the first to leverage genome-scale data and model-based methods. However, the precision of dating analyses performed here is constrained by the paucity of well-preserved fossils reliably attributable to the scoliid crown group. Despite concluding that the higher-level taxonomy of the Scoliidae is in dire need of revision, the chapter ends with the recommendation that taxonomic changes be predicated on datasets that extend the geographic and taxonomic sampling of the current study.When used for phylogenetic inference, exonic DNA sequences can be coded in multiple ways, including as nucleotides, amino acids, and codons. In empirical studies, the choice of data type and associated model is often predicated on which model is less expected to be violated in ways that lead to inaccurate inference. Posterior predictive checks are one method for assessing the adequacy of phylogenetic models and potentially providing an indication of inference reliability. In Chapter 2, a simulation-based approach is used to explore how the ability to detect model inadequacy using phylogenetic posterior prediction, as well as the associated inference errors, may vary with data coding. Specifically, data were simulated under multiple models, including codon models featuring process heterogeneity across lineages, selection heterogeneity across sites, and selection for codon usage. Inference and posterior predictive checks were then performed under nucleotide and amino acid models from the GTR family. Some simulation conditions resulted in large differences, between amino acid and nucleotide treatments, in the ability to detect model violation, even when the magnitude of error in an estimate of interest was similar. Moreover, the results of other studies indicating that error in tree length estimation is not always correlated with error in topology reconstruction are corroborated. Although the use of amino acid models generally resulted in more accurate topologies, tree length errors were often greater than for nucleotide models when the data being analyzed were generated using branch-heterogeneous codon models. The results demonstrate that the magnitude and direction of tree length estimation error can depend on both data coding and properties of the data-generating process. Chapter 2 ends with the conclusion that if posterior predictive checks are to be used for purposes such as data filtering, practical effect size thresholds indicative of low inference reliability must be established separately for amino acid and nucleotide data. Caution and careful selection of models and data coding are recommended when performing analyses where accurate inference of tree length is important. Existing resources for the identification of Nearctic scoliid wasps have multiple shortcomings including limited geographic and taxonomic coverage, the use of outdated taxon names, and factual errors. Chapter 3 seeks to remedy the situation by providing a new key the Nearctic species. Additionally, molecular phylogenetic analysis and examination of morphological characters are used to demonstrate that specimens identified as Scolia bicincta using existing keys and commonly labeled as such in collections belong to two different species. One of these groups is sister to Scolia dubia. The other is sister to or conspecific with Scolia mexicana. Until the identity of the Scolia bicincta type is definitively established, the specimens related to S. dubia are treated as S. bicincta, and the specimens related to S. mexicana are treated as a geographic variant of S. mexicana.