Proposal:Surface=laterite/Data

From OpenStreetMap Wiki
Jump to navigation Jump to search

Supporting subpage for Proposal:Surface=laterite. See also: Proposal:Surface=laterite/Science and Proposal:Surface=laterite/Community.

This page documents the usage data, snapshot analysis, and bot migration history that underpin the proposal. The key argument on the main page (that surface=laterite was erased by a bot and needs to be reinstated) is evidenced here in full.

Current snapshot data (full table)

The table below summarises Overpass snapshot counts across related surface and material tags. "% highway" is the share of each tag's total features that carry a highway=* key.

Tag Total features Highway features % highway Notes
surface=laterite 1 0 0% Only 1 feature globally; a tennis court in China, not a road. Absence is explained by the bot migration that converted existing surface=laterite highway uses to material=laterite without adding a surface fallback; see row below.
material=laterite 845 725 86% Primary displacement tag. Dominated by roads (residential, unclassified, secondary, tertiary, trunk). Highway features are concentrated in Cameroon (55%) and Cambodia (43%), with minor presence in Gabon and Laos; this geographic distribution reflects the bot migration that introduced the tag. Non-highway uses include natural=bare_rock (80), barrier=wall (17), and aeroway=runway (4); those are out of scope for this migration.
surface=murram (lowercase) 17 6 35% All in East Africa. Lowercase variant skews toward aeroway (65%).
surface=Murram (capital-M) 33 30 91% All from a single Kenya survey cluster. Capital-M variant is 91% highway. Combined with lowercase: 50 features, 36 highway retagging candidates.
surface=red earth 24 24 100% All in Auroville, Tamil Nadu, India (lat ~12°N, lon ~79.8°E). 100% highway: tracks, paths, residential, unclassified. A local English name for laterite in this area of South India; a retagging candidate.
surface=clay 41,793 2,363 6% 87% leisure/sport (dominated by tennis: 29,786 features). Highway uses are candidate migration targets where laterite can be confirmed. Current highway breakdown: track 670, footway 647, path 339, residential 230, service 110, unclassified 100; higher-classification road uses (secondary–trunk) are small in absolute terms. The "rare edge case" framing in this proposal refers to the expected post-migration state, not the current count.
material=clay 263 1 0.4% Used as a construction material tag (embankments 84, boundary stones 79, artwork 28). Semantically distinct from the surface tags; not a road-surface value.
surface=caliche 14 14 100% All private oil-field service roads in a small area of west Texas (USA). Caliche is calcium-carbonate hardpan, distinct from laterite, but the pattern is directly analogous: a region-specific natural soil term used by local mappers exclusively for road tagging. Confirms that region-specific natural soil terms for roads exist and are used consistently.
Combination forms: surface:material=*, dirt=* 0 0 N/A No usage found. Mappers did not organically adopt any sub-key or combination approach; displacement went entirely to material=laterite alone, stripping the surface signal.

Key readings

  • material=laterite is overwhelmingly a highway tag (86%), confirming that the bot migration displaced road surface data.
  • OSH history on Cambodia and Cameroon directly confirms 711 bot-driven surface=laterite to material=laterite transitions with zero fallback surface=* tag; see OSH history audit below.
  • surface=clay is overwhelmingly not a highway tag (6%); its primary use is sports courts. The highway uses represent the laterite/clay conflation this proposal resolves.
  • surface=murram combined with its capital-M variant totals 50 features, 36 highway (72%). The lowercase variant alone skews toward aeroway; the capital-M variant (from a single Kenya survey cluster) is 91% highway. Combined, the murram migration pool is 36 highway retagging candidates.
  • surface=caliche confirms that region-specific natural soil terms for roads exist and are used consistently by local mappers, providing a structural parallel to murram and to the proposed laterite tag.
  • Combination forms confirm that no organic alternative emerged from the bot migration; the surface signal was simply lost.

OSH history audit

Methodology

The OSH history audit extracted full change history for highway=* ways in two confirmed laterite-region areas: Cambodia and Cameroon. The audit filtered for ways where at any historical version surface=laterite was present. For each such way, all historical versions were scanned to identify:

  • Whether surface=laterite was replaced by material=laterite (bot-migration pattern)
  • Whether the transition was authored by the bot account or a human mapper
  • Whether any fallback surface=* tag was added at the same time or in subsequent versions before the next human edit

Findings

  • 711 transitions identified where surface=laterite was replaced with material=laterite
  • All 711 were bot-authored (Mateusz Konieczny bot account)
  • 0 of 711 added any fallback surface=* tag at time of transition
  • Among 84 ways in the same dataset with no prior surface=laterite state on the same way history: 80/84 are version-1 entries (likely split/copy inheritance from a nearby way); 4/84 are non-version-1 stronger manual-add candidates
  • This separates two practical review pools: 711 priority bot-repair candidates and 84 residual non-bot review candidates

Snapshot counts remain useful for scope sizing (725 material=laterite highway features globally at snapshot time), while OSH history provides causal evidence for the displacement.

Overpass queries

Bot migration analysis

The "hopefully confused" annotation

The bot migration source code is public under GPL 3.0 at the bot migration wiki page. The relevant line is: 

'laterite': 'laterite', # hopefully confused, may refer to type of soil or bricks or stone

This annotation is the critical evidence that the bot author recognised laterite as potentially misclassified but migrated it anyway. All other entries in the same dictionary (granite, marble, sandstone, limestone, basalt, etc.) are unambiguously rock types with no annotation. The comment "hopefully confused" indicates the bot author's own uncertainty about whether surface=laterite was a rock-type surface value or a soil surface value. The migration proceeded regardless.

Why the bot migration was incorrect for laterite

The bot migration's logic was correct for rock types: surface=granite is ambiguous between paving_stones, sett, cobblestone, and rock. Removing it and replacing with material=granite correctly strips the surface signal while preserving the material information.

For laterite, this logic does not apply:

  • Laterite as a road surface takes only one form: naturally formed cohesive tropical soil. There is no ambiguity between multiple surface sub-types the way there is for granite.
  • surface=laterite on a highway way has a single, unambiguous reading: an unpaved soil road with seasonal plasticity.
  • material=laterite alone on a highway way provides no speed or risk signal to routers; it says only that the material is laterite, with no indication of surface condition, seasonal behaviour, or passability.

The bot migration discussion explicitly noted that some non-rock values were out of scope; surface=mud was the example given. The same reasoning applies to laterite: it is a soil surface value, not a rock type.

Geographic distribution of material=laterite highway features

The geographic distribution of material=laterite highway features reflects the bot migration, not organic mapper behaviour:

  • Cameroon: approximately 55% of 725 highway features
  • Cambodia: approximately 43% of 725 highway features
  • Minor presence: Gabon, Laos

This concentration in two countries reflects where laterite roads were mapped before the bot ran, not where laterite is globally distributed (which spans the tropics broadly). Post-proposal, organic mapper adoption is expected across Africa, Southeast Asia, Latin America, South Asia, and northern Australia.

surface=murram analysis

Lowercase vs capital-M

Two variants exist in the snapshot:

  • Lowercase (surface=murram): 17 features, 35% highway, 65% aeroway
  • Capital-M (surface=Murram): 33 features, 91% highway; from a single Kenya survey cluster

The capital-M variant (from a single Kenya survey effort) is 91% highway, confirming that it represents systematic road surveying where the mapper used the local term with a capitalised first letter. The lowercase variant is less certain due to the high aeroway share.

Combined migration pool: 36 highway features (72% of 50 total). Both variants are migration candidates to surface=laterite after quick local verification.

The aeroway features (14 total across both variants) are not migration candidates; airstrips tagged with murram are out of scope for this proposal.

surface=caliche structural parallel

surface=caliche (14 features, all highway, all in a small area of west Texas) is a useful structural parallel:

  • A region-specific natural soil term (calcium-carbonate hardpan)
  • Used exclusively for road tagging by local mappers in a specific geographic cluster
  • Near-zero usage elsewhere because the material is regionally specific
  • The pattern confirms that region-specific natural soil terms for roads exist and are used consistently

The same pattern applies to murram, red earth, din daeng, and the proposed surface=laterite: local mappers use the local term; the proposal establishes a canonical global value that these local terms map to.

surface=red earth (Auroville) analysis

All 24 features using surface=red earth are in Auroville, Tamil Nadu, India (approximately lat 12°N, lon 79.8°E). All 24 are highway features (tracks, paths, residential roads, unclassified roads). "Red earth" is a local English name for laterite in this part of South India. These 24 features are straightforward migration candidates to surface=laterite after local confirmation.

Combination forms: zero usage

No usage was found for any combination approach:

Mappers did not organically adopt any combination approach after the bot migration. Displacement went entirely to material=laterite alone, stripping the surface signal completely. This confirms that a dedicated surface=laterite value is the correct restoration, not a combination tagging scheme.

Migration scope summary

Source tag Features Highway Priority Action
material=laterite (bot-converted) 711 confirmed 711 High Revert to surface=laterite; bot-repair candidates
material=laterite (non-bot) ~14 remaining highway 14 Medium Manual review before conversion
surface=murram + surface=Murram 50 total 36 High Retag to surface=laterite after quick local verification
surface=red earth 24 24 Medium Retag to surface=laterite after local confirmation
surface=clay on highway in tropical regions 2,363 currently 2,363 Low Review only; no mass migration; targeted MapRoulette challenge recommended

Non-highway material=laterite uses (bare_rock, barrier, aeroway) are not affected by this proposal.

See also

References

Retrieved from "https://wiki.openstreetmap.org/w/index.php?title=Proposal:Surface%3Dlaterite/Data&oldid=3040634"