As far as I understand, URIEL+ contains information on family relationships from Glottolog and uses this for value imputation. The trees in Glottolog are derived from classical historical linguistics, filtered through the Glottolog editors. One drawback with the trees in Glottolog is that they lack branch lengths, all branches are the same length (1). This means that contemporary languages do not necessarily have the same distance to the root, below is an example with Polynesian where Austral is further way from the root than Niuean.
Depending on (i) what you use the trees for, (ii) what you take branch lengths to represent (e.g. time since split, lexical change) and (iii) the precise structure of the trees (if there are more splits closer to the tips or if they are more evenly spread out), lacking branch lengths can be an issue. I think that for value imputation in URIEL+, this may be a relevant issue for you to consider.
There are language trees with branch lengths you can use, many of them are catalogued here in phlorest
There is also the as of yet unpublished "global" tree that is used in several evolutionary linguistics papers:
Bouckaert, R., Redding, D., Trebski, A., Sheehan, O., Gray, R., Jones, K. E., & Atkinson, Q. (2026, March 6). Global language diversification is linked to socio-ecology and threat status.
Paper: https://osf.io/preprints/socarxiv/f8tr6_v2
Tree files: https://github.com/rbouckaert/global-language-tree-pipeline/releases/tag/v2.0.0
As far as I understand, URIEL+ contains information on family relationships from Glottolog and uses this for value imputation. The trees in Glottolog are derived from classical historical linguistics, filtered through the Glottolog editors. One drawback with the trees in Glottolog is that they lack branch lengths, all branches are the same length (1). This means that contemporary languages do not necessarily have the same distance to the root, below is an example with Polynesian where Austral is further way from the root than Niuean.
Depending on (i) what you use the trees for, (ii) what you take branch lengths to represent (e.g. time since split, lexical change) and (iii) the precise structure of the trees (if there are more splits closer to the tips or if they are more evenly spread out), lacking branch lengths can be an issue. I think that for value imputation in URIEL+, this may be a relevant issue for you to consider.
There are language trees with branch lengths you can use, many of them are catalogued here in phlorest
There is also the as of yet unpublished "global" tree that is used in several evolutionary linguistics papers: