Table 1. Ecological predictors of forest outcomes.

Deforestation				Forest Replacement
Predictor	RVI	Beta	95% C.I.	Predictor	RVI	Beta	95% C.I.
Log(Rainfall)	1	-1.599	-2.086, -1.111	Log(Area)	0.998	-0.124	-0.184, -0.063
Log(Elevation)	0.911	-0.689	-1.083, -0.294	Tephra = 3	0.987	-1.215	-2.221, -0.208
Log(Isolation)	0.895	0.263	0.076, 0.45	Log(Isolation)	0.6	0.057	0.001, 0.113
Abs. Latitude	0.642	0.019	0.002, 0.037	Abs. Latitude	0.547	-0.028	-0.056, 0.001
Tephra = 2	0.626	-0.624	-1.218, -0.03	Tephra = 2	0.543	-0.736	-1.543, 0.072
Log(Area)	0.373	-0.160	-0.375, 0.056	Age	0.257	-0.038	-0.106, 0.029
Tephra = 3	0.337	-0.226	-0.624, 0.172	Dust	0.207	-0.001	-0.002, 0.001
% Makatea	0.235	-0.259	-1.68, 1.162	Log(Elevation)	0.157	0.019	-0.124, 0.163
Dust	0.198	-0.0003	-0.001, 0.001	Log(Rainfall)	0.152	0.008	-0.062, 0.077
Age	0.163	0.059	-0.223, 0.342	% Makatea	0.15	0.022	-0.352, 0.395
Dependency		Mean	p-value	Dependency		Mean	p-value
Cultural (λ')		0.017	0.957	Cultural (λ')		0	1.0
Geographic (ϕ)		0.082	0.932	Geographic (ϕ)		1.0	<0.001
Independent (γ)		0.901	-	Independent (γ)		0	-

Table shows relative variable importance, Akaike weighted beta estimate and 95% confidence interval for PGLS analysis (Materials and Methods) of the effects of putative ecological predictors, phylogeny and geographic proximity on deforestation (n = 76) and forest replacement (n = 72). Previously identified significant predictors of deforestation and forest replacement are shown in bold. All values integrate over phylogenetic and sampling uncertainty across 100 replicates from our posterior distribution of language trees.