Todiramphus cinnamominus · Extinct in the Wild · reintroduced to Palmyra Atoll, 2024

What 156 kingfisher sequences say about the sihek.

A mitochondrial ND2 phylogeny of the Guam kingfisher against every other kingfisher sequence in GenBank, and the genomic resources that exist for following it up. The short version: nothing alive is close to this bird.

156
ND2 sequences
0
sequences in the moat
2.02 pp
width of the void
2.30 %
nearest other kingfisher
7
empty annuli in the family
The figure

The Moat

Every sequence plotted by its ND2 distance from the sihek, with the empty intervals inked instead of the data. The gap around the bird is not sparse — it is empty, and it is 2.02 percentage points wide. Hover any mark for its identity, or click through to that bird on Wikipedia.

Open the figure →

The three static figures

Maximum-likelihood tree of Todiramphus with the sihek highlighted Figure 1. The sihek's closest kin are not the birds it was once classified with. Family-level kingfisher tree with subfamilies marked Figure 2. The sihek is a tree kingfisher, nested among Todiramphus lineages. Histogram and dot plot of ND2 divergence from the sihek Figure 3. An empty gap isolates the sihek; its former subspecies are not its nearest kin.

What is here

REPORT.md
Dataset, methods, results and caveats in full. Renders on GitHub.
data/
The 156 × 1,041 bp alignment, ML tree, IQ-TREE log, distance tables, accessions and citations.
figures/
The three figures above, moat.html, and the JSON inputs it is built from.
scripts/
Deterministic figure build, the palette validator, the Wikipedia resolver, and the genome fetcher.
papers/
The three open-access (CC-BY) sources cited in data/citations.csv.
genome/
NCBI assembly GCA_033439825.1 — not in git, see below.

The genome

GCA_033439825.1 (ASM3343982v1) is the only Guam kingfisher assembly in existence — Iridian Genomes, released 2023-11-08, Illumina/SPAdes. It is 1.16 Gb across 138,582 scaffolds, scaffold N50 60,279 bp, longest scaffold 590 kb, and zero annotated genes.

It is excluded from git: genomic.fna alone is 1.1 GB, past GitHub's 100 MB per-file limit, and every byte is reproducible from NCBI. Only its checksums and assembly stats are tracked, so the assembly can be identified and verified without downloading a gigabyte. 

scripts/fetch_genome.sh   # ~353 MB, MD5-verified against NCBI's manifest

Because the assembly is fragmented and unannotated, a genome browser will show sequence and nothing else. For anything interpretive, a synteny dot plot against a chromosome-scale relative (T. chloris collaris, GCA_025345505.1, N50 17.5 Mb) is the more informative view.

Rebuilding the figure

figures/moat.html is self-contained and needs no build step to view. To regenerate it from the underlying data: 

python3 scripts/prep_data.py     # tree order + distances + taxonomy
python3 scripts/wiki_lookup.py   # verify taxa against the Wikipedia API
python3 scripts/build_moat.py    # render the SVG server-side

The build is deterministic. scripts/tokens.css holds the design tokens shared by this page and the figure, so the two cannot drift.

Caveats that travel with the results

  • Single, maternally inherited locus. ND2 traces the matriline only. A genome-scale study of the genus found nuclear and mitochondrial data give no concordant topology.
  • The sister lineage is unresolved. T. recurvirostris carries only 66% bootstrap support. The isolation is solid; the ancestry is not.
  • The nearest neighbour is not identified. T. sacer pealei (2.3041%) and T. reichenbachii (2.3055%) differ by 0.0014 points over different site counts (868 vs 1,041). Treat 2.30% as a bound, not an identification.