# Sihek (Guam kingfisher) phylogenetics — ND2 dataset and maximum-likelihood tree

**Focal taxon:** Sihek / Guam kingfisher, *Todiramphus cinnamominus* (Extinct in the Wild)
**Question:** How is the sihek related to other kingfishers?
**Data source:** NCBI GenBank, retrieved via Entrez E-utilities.

---

## 1. Dataset

Marker: **mitochondrial ND2** (NADH dehydrogenase subunit 2), the best-sampled locus
in Alcedinidae — 712 GenBank records at the family level versus 262 for COI and 99 for
cytochrome *b*.

| | count |
|---|---|
| Sequences in final alignment | 156 |
| Distinct taxa | 153 |
| Genera represented | 16 (all three kingfisher subfamilies) |
| Sihek individuals | 3 (MG008164, MG008165, KP291296) |
| Outgroups | 4 (*Merops*, *Meropogon*, *Momotus*, *Todus*) |
| Aligned length | 1,041 bp = 347 codons |
| Gap fraction | 0.17% |

Selection rule: one longest record per taxon, except the *T. cinnamominus* complex
(sihek, *T. albicilla*, *T. pelewensis*) where every individual was retained.
Four sequences were excluded for >25% ambiguous bases (*Pelargopsis amauroptera*,
*P. melanorhyncha*, *Tanysiptera carolinae*, *T. riedelii*).

## 2. Methods

1. **Retrieval** — Entrez esearch/esummary/efetch against `nuccore`.
2. **Alignment** — MAFFT v7 `--auto --adjustdirection`, then trimmed to columns with
   ≥50% occupancy.
3. **Frame check** — reading frame chosen by minimising internal stop codons under the
   vertebrate mitochondrial code (table 2). The selected frame gives **zero internal
   stop codons across all 156 sequences**, confirming a correct in-frame coding alignment.
4. **Tree inference** — IQ-TREE, partitioned by codon position (pos1/pos2/pos3), with
   ModelFinder selecting per-partition models by BIC:
   `TVM+F+I+G4` (pos1), `TIM3+F+I+R2` (pos2), `TIM3+F+ASC+R4` (pos3).
   Support: 1,000 ultrafast bootstrap replicates + 1,000 SH-aLRT replicates.
   Rooted on *Todus angustirostris*.
5. **Distances** — uncorrected p-distances and Jukes–Cantor corrected distances over
   pairwise-comparable sites.

Tree quality: 109 of 150 internal nodes (73%) have ultrafast bootstrap ≥95; 18 (12%) fall below 70.

## 3. Results

### 3.1 The sihek is a tree kingfisher (Halcyoninae)

It falls squarely inside Halcyoninae, sister to the river kingfishers (Alcedininae)
and water kingfishers (Cerylinae). Within Halcyoninae it sits among the *Todiramphus*
lineages — an Indo-Pacific island radiation.

### 3.2 It has no close living relative

Divergence from the sihek to the 59 other *Todiramphus* sequences spans **2.30–12.49%**
(uncorrected p-distance; median 3.09%). The load-bearing number is the **minimum**: the
single nearest congeneric sequence is **2.30%** away, while the sihek's own three
individuals differ by only **0.10–0.29%**. That is a **7.9-fold** gap between intraspecific
variation and the nearest relative, and it is a genuine discontinuity — **zero** of the 155
comparison sequences in the entire dataset (all genera, not just *Todiramphus*) fall
between 0.30% and 2.29%.

All three within-sihek comparisons (`within_sihek_distances.csv`):

| Pair | Differing sites / 1,041 | p-distance (%) |
|---|---|---|
| MG008164 × MG008165 | 2 | 0.19 |
| MG008164 × KP291296 | 3 | 0.29 |
| MG008165 × KP291296 | 1 | 0.10 |

(Note: `divergence_from_sihek.csv` is measured from MG008164 as reference, so it contains
only the first two of these; the 0.10% pair appears in `within_sihek_distances.csv`.)

The deeper end of the range reflects early-diverging *Todiramphus* lineages that are
distant from the whole core radiation, not just from the sihek: *T. winchelli* (12.49%),
*T. nigrocyaneus* (12.39%), *T. funebris* (8.02%) and *T. pyrrhopygius* (7.88%). These sit
at the base of the genus in Figure 1. The isolation claim therefore concerns the sihek's
*lower* bound — nothing is close to it — not a narrow overall range.

Nearest relatives by ND2 distance (the five smallest of 59 comparisons):

| Taxon | Accession | p-distance (%) |
|---|---|---|
| *T. sacer pealei* | KP291354 | 2.30 |
| *T. reichenbachii* | MG008179 | 2.31 |
| *T. chloris santoensis* | KP291239 | 2.50 |
| *T. sacer manuae* | KP291342 | 2.50 |
| *T. sacer* | KP291327 | 2.59 |

### 3.3 Its former subspecies are NOT its closest kin

The sihek was historically treated as one species with the Micronesian kingfisher
of Pohnpei (*T. reichenbachii*), Palau (*T. pelewensis*) and the Marianas
(*T. albicilla*). The tree does not support that grouping:

- *T. pelewensis* — 2.88–2.98% divergent, placed with *T. reichenbachii* in a separate clade
- *T. albicilla* — 3.07–3.09% divergent, nested deep inside the *T. chloris* / *T. sanctus* complex

Both are **more** distant from the sihek than eleven other *Todiramphus* sequences: the
nearest *T. pelewensis* ranks 12th of the 59 congeneric comparisons, and *T. albicilla*
23rd. This is
concordant with the modern taxonomic split of the old "Micronesian kingfisher" into
separate species, and it means the sihek is an isolated lineage, not one member of a
closely related trio.

### 3.4 Its exact sister lineage is unresolved

| Step outward from sihek | Sister group | SH-aLRT | UFboot |
|---|---|---|---|
| 1 | *T. recurvirostris* | 0.0 | **66** |
| 2 | 37-taxon *chloris*/*sanctus* clade | 92.5 | 95 |
| 3 | *T. farquhari* | 82.9 | 99 |
| 4 | *T. australasia* | 94.0 | 99 |
| 5 | *albonotatus* group (5 taxa) | 84.6 | 98 |

The sihek's immediate sister (*T. recurvirostris*, Marquesas) has only **66% bootstrap**
and 0 SH-aLRT — not supported. Its placement *within* the broader *Todiramphus* radiation
is well supported (95%), but the single-locus data cannot resolve which lineage it split
from. This is the expected signature of a rapid island radiation: short internal branches
that one mitochondrial gene cannot separate.

## 4. Caveats

- **Single locus, maternally inherited.** ND2 traces the matriline only. Mitochondrial
  trees can be misled by introgression and incomplete lineage sorting — both plausible in
  an island archipelago radiation.
- **GenBank nomenclature lags.** DQ111853 is filed as *Halcyon ruficollaris* but is the
  same species as *Todiramphus ruficollaris*; many *T. chloris* subspecies records reflect
  the pre-split taxonomy. Names in the figures were reconciled where the synonymy is
  unambiguous.
- **Genus-level non-monophyly.** *Halcyon*, *Dacelo* (paraphyletic w.r.t. *Clytoceyx*) and
  *Actenoides* (w.r.t. *Caridonax*) are not monophyletic here. This is consistent with the
  published literature on Alcedinidae and is not an artefact specific to this dataset.
- Support values are ultrafast bootstrap, which is less conservative than standard
  nonparametric bootstrap; ≥95 is the appropriate threshold.

## 5. Genomic resources available for follow-up

| Assembly | Species | Submitter |
|---|---|---|
| GCA_033439825.1 | *Todiramphus cinnamominus* (sihek) | Iridian Genomes |
| GCA_030710405.1 | *T. albicilla* | University of Kansas |
| GCA_025345505.1 | *T. chloris* | Field Museum |
| GCA_024582635.1 | *T. saurophagus* | Iridian Genomes |

Also available: 4 SRA runs for the sihek, 124 across *Todiramphus*; UCE datasets
(KDIZ00000000 for the sihek, KDJD00000000 for *T. albicilla*) and nuclear loci
(RAG1, TGFb2, CCDC, HMG, MUSK) from the same vouchers used here.

## 6. Files

- `nd2_alignment.fasta` — final 156 × 1,041 bp codon-aware alignment
- `nd2_ml.treefile` — ML tree, Newick, with SH-aLRT/UFboot node labels
- `taxa_accessions.csv` — every sequence used, with accession and subfamily
- `divergence_from_sihek.csv` — distances from sihek MG008164 to all 155 other sequences
- `within_sihek_distances.csv` — all three pairwise comparisons among the sihek individuals
- `sihek_placement_support.csv` — nested sister groups and their support values