Ancestral niche reconstruction using bins
Our new package, nichevol, contains tools for performing
ancestral state reconstructions on binned datasets using two methods:
maximum parsimony (as implemented in castor v.1.4.3; Louca and
Doebeli 2018) and maximum likelihood (as implemented in apev.5.3; Paradis and Schliep 2018). For each algorithm, ancestral state
reconstructions are first performed for each bin separately, treating
bin scores as discrete characters under an equal transition rate model.
The results are then smoothed such that the reconstructed ancestral
niche at each node is uninterrupted by “unsuitable” scores, again
following the assumption of a unimodal response to environmental
conditions (Maguire, 1973). This smoothing algorithm also adjusts
reconstructions to account for the evolutionary non-independence of
single bins. As a simple illustration of how this analysis works, we
simulated data for a scenario in which there was a shift from the
ancestral warm niche to the cool niche took place, to demonstrate the
ability of our new methods to identify instances of niche evolution. To
this end, we generated a single, 6-taxon stochastic birth-death tree
(birth rate = 1, death rate = 0), and assigned simulated niche
characterizations from the warm niche group to tips in the tree (“t1”,
“t2”, “t3”, and “t6”). The remaining monophyletic clade of tips in
the tree (“t4” and “t5”) were assigned simulated niche
characterizations from the cool niche simulation group. We then
performed parsimony and likelihood ancestral state reconstructions on
binned niche characterizations as described above.
We note that we have kept these initial explorations simple for the
purpose of illustration—many improvements could be made to this
methodology, such as implementing different evolutionary models (e.g.
Ornstein-Uhlenbeck), Bayesian approaches in inferring ancestral
character states, and consideration of joint effects of environmental
dimensions (e.g. temperature, precipitation) that are here considered
independently. Furthermore, phylogenetic comparative methods are
notoriously “data-hungry”, and our proposed ancestral character
inference strategy will benefit from further detailed simulation-based
examinations in the future. Our purpose overall was to illustrate the
crucial importance of explicit incorporation of uncertainty in these
analyses of climate niche evolution. Code for analyses can be found in
Supplement 5; further explanation and illustration can be found in
annotated code with tables and figures provided as an HTML document in
Supplement 6.