Probabilistic Modeling of Bursts and Gradual Changes of Diversification Rates in Bird Evolution

In our previous work, the PhyPPL concept paper [1], we used the sequential Monte-Carlo algorithm in a
probabilistic programming language framework to do a Bayes factor comparison for the diversification models
CRBD, TDBD, BAMM, LSBDS (not shown, a weaker variation of BAMM), and the novel ClaDS on the 40 bird
clades having more than 50 species, taken from https://birdtree.org/. We identified four patterns:
● Group 1 (small trees such as Alcedinidae): simple models are adequate (CRBD as good as the others);
● Group 2 (e.g. Muscicapidae-+): evidence for slowing down of diversification (TDBD outperforms others);
● Group 3 (e.g. Accipitridae): cladogenetic changes in diversification (ClaDS outperforms others);
● Group 4 (e.g. Anatinae): cladogenetic and punctuated anagenetic lineage-specific changes. (ClaDS and BAMM/
LSBDS best).
We introduce two new models (AnaDS-GBM and AnaDS-GBM+LS) to investigate the effect of gradual and
punctuated anagenetic changes in speciation rates. For this poster we tested AnaDS-GBM on one tree per group.