PopSpot — Dominica

PopSpot

Dominica

OpenStreetMap · place= population tag analysis

▲ Upgrade candidates · 6 ▼ Downgrade candidates · 1 ◎ Monitor · 1 Generated · 2026-04-04 09:48 UTC

Disclaimer

Whether a settlement is a village, town, or city depends on a mix of factors — population, character, local identity, available services, and geographic context — none of which point to a single clear answer. Areas near larger settlements may function as suburbs or neighbourhoods rather than distinct places, while remote settlements often feel more significant than raw numbers suggest. Since reasonable mappers frequently disagree on these judgements, any classification changes should be made thoughtfully and with awareness of local context.

Population Distribution

Population distribution by place type for Dominica. Dots show individual tagged places, with the maximum, minimum, and median population values highlighted.

Population distribution plot for Dominica

Upgrade Candidates

Places whose population is more consistent with the next-higher type's distribution than their own. Ranked by signal strength, then percentile within current type.

#	Name	Type	Population	Own %ile	%ile among next higher	Signal	Edit	OSM
1	Salisbury	village	2,174	100%	83%	Weak	iD JOSM Level0	node/245509376
2	Marigot	village	1,277	98%	17%	Weak	iD JOSM Level0	node/245509766
3	La Plaine	village	1,128	95%	17%	Weak	iD JOSM Level0	node/245925926
4	Castle Bruce	village	1,087	92%	17%	Weak	iD JOSM Level0	node/245926011
5	Petite Savanne	village	1,000	90%	17%	Weak	iD JOSM Level0	node/246573529
6	Soufriere	village	973	88%	17%	Weak	iD JOSM Level0	node/242414184

Downgrade Candidates

Places whose population is more consistent with the next-lower type's distribution than their own. Ranked by signal strength, then percentile within current type.

#	Name	Type	Population	Own %ile	%ile among next lower	Signal	Edit	OSM
1	Colihaut	town	679	17%	78%	Weak	iD JOSM Level0	node/246579112

Monitor

Places that are consistent with their current type's distribution but sit at an extreme percentile within it. No reclassification signal yet, but worth revisiting as more population data is tagged.

#	Name	Type	Population	Own %ile	Edit	OSM
1	Roseau	town	15,000	100%	iD JOSM Level0	node/4950552057

Step 1 — Log-normal distribution fit

For each place type (city, town, village), a log-normal distribution is fitted to the population values tagged in OSM. Specifically, the mean and standard deviation of log₁₀(population) are computed, giving each type its own statistical fingerprint.

Step 2 — Log-likelihood ratio (LLR)

Each place is scored under its own type's distribution and under each adjacent type's distribution. A positive LLR means the place's population fits the neighbouring type's distribution better than its own — a statistical signal of potential misclassification. If both adjacent types fire simultaneously, the stronger LLR determines the flag direction.

Step 3 — Percentile corroboration

The LLR signal is cross-checked against percentile ranks: where does this place sit within its own type, and where would it land in the target type? Both signals must agree for a candidate to be rated Strong.

Confidence tiers

Badge	Direction	Criteria (all must hold)
Strong	Upgrade	LLR > 1.0 · own percentile ≥ 75th · target percentile 20–65th
Strong	Downgrade	LLR > 1.0 · own percentile ≤ 25th · target percentile 35–80th
Weak	Either	LLR > 0, but one or more Strong conditions not met
Monitor	—	No LLR signal, but own percentile ≥ 90th (top) or ≤ 10th (bottom)

Why extremes are excluded from Strong

The target-type percentile window deliberately excludes both tails. A place landing below the 20th percentile among the target type would be marginal even there — a weak fit. A place landing above the 65th percentile in the target type is unusually large even for that category, suggesting the reclassification destination may be wrong (should jump two levels) or the tag is erroneous. The middle range confirms the place would be a natural, unremarkable member of the target type — the hallmark of a genuine misclassification.

Table sort order

Within each section, Strong candidates always appear before Weak ones. Within each confidence tier, upgrade candidates are sorted by own-type percentile descending (most overqualified first) and downgrade candidates by own-type percentile ascending (most underqualified first). Monitor entries are sorted by own-type percentile descending.

Statistical flags are prompts for investigation, not conclusions. Place hierarchy in OSM reflects infrastructure, administrative status, and regional importance — population alone cannot capture that.