public-usgs/kaiparowits-coal-resources: USGS OFR 97-709 Kaiparowits Plateau coal coverages

[cassiebuhler]

Source

Landing page: https://pubs.usgs.gov/of/1997/ofr-97-0709/
Citation: Biewick, L.R.H., Hettinger, R.D., and Roberts, L.N.R., 1997, Selected ARC/INFO coverages
created for investigations of the distribution and resources of coal in the Kaiparowits Plateau,
southern Utah: an accompaniment to Hettinger and others, 1996, USGS Open-File Report 97-709.
Edition: version 1 (upstream's own label), files dated 1997-11-07. Accessed 2026-09-22.

Single archive: kaip.tar.gz, 5,368,079 bytes, md5 d680d23c47172d5ee1061728d7cd1632.

Verified 2026-09-22:

curl -I .../kaip.tar.gz                    -> HTTP/2 200, application/gzip, 5368079
ogrinfo -so kaip/coverages/scale125k/allcoal PAL  -> AVCBin, 5222 polygons

What the data is, and what it is for

The GIS layers behind the USGS National Coal Assessment of the John Henry Member of the Straight
Cliffs Formation, Kaiparowits Plateau. 20 ARC/INFO coverages at three scales. Coverage
descriptions: https://pubs.usgs.gov/of/1997/ofr-97-0709/Appendix1.html

The one that matters is allcoal (1:125,000, 5,222 polygons) — the union of net-coal thickness,
overburden, reliability, dip and ownership, carrying a per-polygon tonnage:

field	meaning
SHTONS_MIL	coal resource, million short tons
THICK	net coal thickness, ft
OVERB	overburden class, thousands of ft — 0-1, 1-2, 2-3, 3-6, >6
REL	reliability — iden (identified) / hypo (hypothetical)
COAL_OWN	fed / nonfed
SURF	surface owner — BLM, FS, STATE, PRIVATE, NPS, NREC, WATER
DIP	inclination class, degrees — 0-6, 6-12, 12-25, >25
INT	T total / P partially eroded coal-bearing interval
QUAD, CNTY, TR	7.5' quad, county, township-range

Totals as read from the source: 72,129.6 million short tons across 5,222 polygons, of which
COAL_OWN = 'fed' is 65,997.6 and REL = 'iden' is 54,878.8.

This is the only published source that carries Kaiparowits coal tonnage as geometry. The existing
coal-deposits layer (UGS 1988, in public-utah) is deposit-area outlines with no tonnage.

Consumer: utah-public-lands.

Companion coverages in the same archive

coverage	scale	geom	n	contents
allcoal	125k	polygon	5,222	the tonnage layer, above
fig22	125k	polygon	815	areas geologically favourable for current underground mining technology; same attributes as allcoal minus THICK/SHTONS_MIL
kaipcoal	24k	point	209	drill-hole/measured-section point source data — per-bed thickness and counts by thickness class (THK1-2.4 … THK>40, NUM*), DEPTHCOALTOP, DEPTHCOALBOT, TOTALCOAL, SURFELEV
kjh_thk	125k	polygon	124	Calico + A-sequence thickness contours
csb_struct	125k	polygon	239	structure contours, Calico sequence boundary
dip	125k	polygon	57	inclination of strata
cb_int, csb, kdak, kjh_bkw, kjh_kdt	125k	polygon	6–26	stratigraphic outcrop/extent outlines
kdak_dd	125k	polygon	26	kdak in NAD27 geographic decimal degrees (EPSG:4267), not distinct content — out of scope
structure, m_adit	125k	line	—	structural features; coal mine adits
twnshp	24k	polygon	190	township and range
bb_surf, county, roads, ut_st_dd, ut_fedl_dd	100k / 5m	—	—	generic basemap, superseded by catalog layers — out of scope

Scope the ingest to the assessment coverages. The basemap ones duplicate data already in the
catalog.

In scope, 14 coverages: allcoal, fig22, kaipcoal, kjh_thk, csb_struct, dip, cb_int,
csb, kdak, kjh_bkw, kjh_kdt, structure, m_adit, twnshp.

Two source characteristics that will bite

1. The exports/*.e00 files are compressed E00 and GDAL cannot read them. ogrinfo on
   allcoal.e00 returns ERROR 4: This looks like a compressed E00 file and cannot be processed directly. The archive also ships the native binary coverages under kaip/coverages/<scale>/,
   which GDAL's AVCBin driver opens without preprocessing. Polygons are the PAL sublayer.
2. The CRS is NAD27, and the coverages do not say so. GDAL reports UTM zone 12N on a Clarke 1866
   ellipsoid with DATUM["Not specified"]. That is EPSG:26712 (NAD27 / UTM 12N), not 26912.
   Assigning NAD83 instead displaces the data by roughly 100 m.

Sentinels: OVERB = '-99', REL = 'hole' and empty-string ownership mark the 5 polygons that are
map holes, all with SHTONS_MIL = 0.

The same sentinel convention runs through the companion coverages, with values the allcoal list does
not cover: kjh_thk.THICK = -9999 (5 of 124), csb_struct.ELEV = -9999 (1 of 239), dip.DIP = ''
(5 of 57), and kaipcoal, whose metadata states -99 in field means NO DATA across its numeric fields.
None of these is a zero.

Spatial extent

Full upstream extent: the Kaiparowits Plateau study area, southern Utah, east of 112°W.
allcoal bounds (EPSG:26712): 411,784–499,408 E, 4,102,866–4,199,889 N.

H3 native and parent resolutions

native 8, parents 0 — mean allcoal polygon is ~1.6 km² against res 8's 0.74 km², and the source is
compiled at 1:125,000, so res 9+ would assert precision the mapping does not have. h8 is the
catalog join key and is carried natively.

Reducer (rasters only)

n/a — vector.

Target bucket and dataset path

s3://public-usgs/, dataset kaiparowits-coal-resources, one dataset per coverage beneath it:

public-usgs/kaiparowits-coal-resources/stac-collection.json   <- grouping collection, child links only
public-usgs/kaiparowits-coal-resources/<coverage>.parquet
public-usgs/kaiparowits-coal-resources/<coverage>.pmtiles
public-usgs/kaiparowits-coal-resources/<coverage>/hex/h0=*/data_0.parquet
public-usgs/kaiparowits-coal-resources/<coverage>/stac-collection.json

One collection per coverage, not one for all 14: verify-stac.py keys assets {last-segment}-{format}
and pairs each flat GeoParquet with its hex, so the coverages cannot share a collection. Layout follows
padus-4-1.

Licence, with evidence

public-domain (SPDX). USGS Open-File Report, US federal agency product, no access or use
constraint stated in the report. https://www.usgs.gov/information-policies-and-instructions/copyrights-and-credits

Acceptance criteria

• Raw kaip.tar.gz staged before processing, with the size and md5 above recomputed from the stored object.
• allcoal present with all 11 attributes including SHTONS_MIL and THICK as numerics.
• SUM(SHTONS_MIL) over the published allcoal = 72,129.6 (± float noise) across 5,222 features. A mismatch means geometry or rows were dropped.
• Source CRS recorded as EPSG:26712 (NAD27) and reprojected to EPSG:4326 through a NAD27→NAD83 transformation, not a null transform.
• SHTONS_MIL units (million short tons) stated in table:columns.
• STAC records that these are in-place resource figures at the stated thickness/overburden cutoffs, not recoverable reserves, and not economically recoverable tonnage. The 72 Bt total spans every overburden class including >6 (thousand ft).
• REL (identified vs hypothetical) documented — the two are not interchangeable and hypo should not be summed with iden without saying so.
• -99 / hole / empty sentinels documented as no-data, never as 0.
• Basemap coverages (bb_surf, county, roads, ut_st_dd, ut_fedl_dd) excluded.
• h3:native_resolution = 8, h3:parent_resolutions = [0].
• SPDX public-domain + {"rel": "license"} link.
• table:columns populated; scripts/verify-stac.py --bucket public-usgs --dataset kaiparowits-coal-resources
  exits 0, and so does a run against each of the 14 per-coverage collections.
• The NAD27 to NAD83 step uses the NADCON grid us_noaa_conus.tif (NAD27 to NAD83 (1), 0.15 m), not
  PROJ's grid-free ballpark Helmert fallback (NAD27 to WGS 84 (4), 10 m).
• STAC records that structure (327 arcs) and m_adit (50 arcs) carry no thematic attributes in the
  source, so their attribute-free parquet is not read as a dropped column.
• STAC records that kaipcoal ships fewer fields than its own FGDC metadata documents: THK1-2.4,
  NUM1-2.4, THK2.5-3.4, NUM2.5-3.4, THK3.5-7, NUM3.5-7, THK>40 and NUM>40 are defined in
  the metadata and absent from KAIPCOAL.PAT.
• STAC records that SHTONS_MIL is computed, not measured: AREA * 0.0002471 * THICK * 1800 / 1e6,
  assuming bituminous rank at 1,800 short tons per acre-foot.

Known blockers

• None on access. The compressed-E00 and NAD27 items above are source characteristics, documented so they are not rediscovered mid-build.
• Fixed vintage: 1997, superseded by nothing. This dataset will never update.

Related

#478 USGS MRDS · #700 USGS NOGA assessment units

[cassiebuhler]

Scope decisions, fixed before build (2026-09-22)

Source verified against the staged archive: kaip.tar.gz 5,368,079 bytes, md5 d680d23c47172d5ee1061728d7cd1632,
matching the body. allcoal reproduces every number in the body exactly: 5,222 features, SUM(SHTONS_MIL)
72,129.6, COAL_OWN='fed' 65,997.6, REL='iden' 54,878.8, 5 map-hole polygons.

1. Catalog layout: one collection per coverage, under a grouping collection

verify-stac.py keys assets {last-segment}-{format} and pairs each flat GeoParquet with its hex, so a
single collection cannot hold 14 coverages. Layout follows padus-4-1:

• s3://public-usgs/kaiparowits-coal-resources/<coverage>.parquet / .pmtiles, hex under
  kaiparowits-coal-resources/<coverage>/hex/h0=*/data_0.parquet
• one collection per coverage at kaiparowits-coal-resources/<coverage>/stac-collection.json
• a grouping collection at kaiparowits-coal-resources/stac-collection.json holding child links, itself
  a child of public-usgs/stac-collection.json

The acceptance criterion verify-stac.py --bucket public-usgs --dataset kaiparowits-coal-resources targets
the grouping collection; every child is verified as well.

2. kdak_dd dropped: it is kdak in a different projection

kdak_dd is the same 26-polygon plateau outline as kdak, stored in NAD27 geographic decimal degrees
(EPSG:4267, Projection GEOGRAPHIC / Units DD in its prj.adf) instead of UTM. Both publish to EPSG:4326,
so ingesting both yields two copies of one layer. Same reasoning the body already applies to
ut_st_dd / ut_fedl_dd: _dd is this archive's decimal-degree naming convention, not distinct content.

In scope, 14 coverages: allcoal, fig22, kaipcoal, kjh_thk, csb_struct, dip, cb_int, csb,
kdak, kjh_bkw, kjh_kdt, structure, m_adit, twnshp.

3. Datum transformation is pinned to NADCON, not left to PROJ's default

Every coverage but kdak_dd is EPSG:26712, confirmed by prj.adf (UTM / Zone 12 / CLARKE1866) and by
allcoal.meta, which states Horizontal_Datum_Name: North American Datum of 1927 outright.

With no grid installed PROJ silently falls back to a ballpark Helmert (NAD27 to WGS 84 (4), 10 m). The build
downloads us_noaa_conus.tif from cdn.proj.org and sets PROJ_DATA, which selects NAD27 to NAD83 (1) at
0.15 m. Measured at the plateau centre (455000, 4150000), mis-assigning NAD83 displaces the data 212 m
(66 m E, 201 m N), larger than the ~100 m in the body and well beyond the source's own 1:125,000 accuracy.

4. Attribute policy

Dropped everywhere: ARC/INFO bookkeeping only, AREA, PERIMETER, <COV>#, <COV>-ID, ArcIds, UserId,
ValueId, PolyId, FNODE_, TNODE_, LPOLY_, RPOLY_, LENGTH. kaipcoal.X / .Y are dropped as well:
they are NAD27 UTM coordinates of the point, redundant with the published WGS84 geometry and misleading next
to it. Every thematic attribute is kept. allcoal keeps exactly the 11 the body lists.

kaipcoal column names are made parquet-safe: THKC-D to THKC_D, THK7-14 to THK7_14, NUM7-14 to
NUM7_14, THK14-20 to THK14_20, NUM14-20 to NUM14_20, THK20-40 to THK20_40, NUM20-40 to
NUM20_40, MAP# to MAP_NO. Source names are recorded in table:columns.

5. Source facts found during verification, to be documented in STAC

Unrecoverable from the published parquet, so they go in the collections:

• More no-data sentinels than the body lists. kjh_thk.THICK = -9999 (5 of 124), csb_struct.ELEV = -9999
  (1 of 239), dip.DIP = '' (5 of 57). kaipcoal.meta states -99 in field means NO DATA across its numeric
  fields. None of these is a zero.
• structure and m_adit carry no thematic attributes at all. 327 and 50 arcs, geometry only. The
  syncline / anticline / fold / fault distinction structure.meta describes exists only in the printed figure 9
  of Hettinger and others (1996), never in the coverage. Nothing was dropped in ingest.
• kaipcoal ships fewer fields than its own metadata documents. The FGDC record defines THK1-2.4,
  NUM1-2.4, THK2.5-3.4, NUM2.5-3.4, THK3.5-7, NUM3.5-7, THK>40, NUM>40; the shipped KAIPCOAL.PAT
  has none of them, only THK7-14 / THK14-20 / THK20-40 and their NUM* pairs. The field list in the body
  came from the metadata. There is no second INFO table: the coverage holds only PAT, AAT, TIC and BND.
• DIP is coded >25 in the data, gt25 in the metadata. The data spelling is what is published.
• kdak.LINK is an ArcView hotlink field (25 of 26 empty, one K2.APR), described as such in the FGDC
  record. Kept, and labelled an authoring artifact rather than data.
• allcoal.SHTONS_MIL is derived, not measured: AREA * 0.0002471 * THICK * 1800 / 1e6, assuming
  bituminous rank at 1,800 short tons per acre-foot. THICK is itself interpolated from 209 data points, each
  polygon taking the midpoint of its bounding contour interval.
• Upstream states Access_Constraints: None, which is the evidence behind public-domain.

6. Hex

Native 8, parents [0] for every coverage, as the body sets. structure and m_adit are lines and hex by
circumradius buffering at res 8; kaipcoal is 209 points, one cell each.

[cassiebuhler]

Built and published

All 14 coverages are live under s3://public-usgs/kaiparowits-coal-resources/, each with
GeoParquet, PMTiles and H3 resolution 8 hex, its own STAC collection, and a grouping collection
registered as a child of public-usgs/stac-collection.json. README at
kaiparowits-coal-resources/README.md. PR #712.

Acceptance criteria, as measured against the published data

criterion	result
raw staged with size and md5 recomputed from the object	5368079 bytes, md5 d680d23c47172d5ee1061728d7cd1632; the job fails the build on a mismatch
allcoal with all 11 attributes, SHTONS_MIL and THICK numeric	present, both DOUBLE
SUM(SHTONS_MIL) = 72,129.6 over 5,222 features	72,129.6 over 5,222 rows, 5,222 distinct _cng_fid; fed 65,997.6, iden 54,878.8
EPSG:26712 reprojected through a real NAD27 to NAD83 transformation	NADCON NAD27 to NAD83 (1), asserted numerically in the build
SHTONS_MIL units stated in table:columns	million short tons, with the formula that computed it
in-place resource, not reserves, recorded in STAC	on every tonnage-bearing collection and the grouping collection
REL identified vs hypothetical documented	with a SQL example rather than a prose imperative
sentinels documented as no-data, never 0	see the correction below
basemap coverages excluded	excluded, and kdak_dd with them
h3:native_resolution 8, h3:parent_resolutions [0]	on all 14 hex assets
SPDX public-domain + rel: license link	on all 15 collections
verify-stac.py --bucket public-usgs --dataset kaiparowits-coal-resources exits 0	exits 0, and so does each of the 14 per-coverage collections

Hex integrity: allcoal hex holds 7,424 rows over 4,344 cells with all 5,222 features present and
no null h8. Every declared values array was checked against the ingested DISTINCT set.

One correction to the issue's own field notes

The body records -99 as kaipcoal's no-data marker, which is what the FGDC record states for the
whole point attribute table. Measured against the published parquet it occurs in 8 of the 17
numeric columns and never in the other 9:

• carries -99: DEPTHCOALTOP (51 of 209), DEPTHCOALBOT (51), ELEVCSB (51), SURFELEV (45),
  DEPTHCSB (154), ESTDEPTHCSB (106), THKC_D (174), THKDT (196)
• never does: TOTALCOAL, NUMBEDSTOTALC, the six per-class thickness and bed-count columns, and
  MAP_NO

In the second group a 0 is a real measurement, no coal in that thickness class, so publishing the
blanket claim would have told consumers to filter out data. Each column now carries its own measured
count and the collection says which columns have the marker and which do not.

Two advisories left standing deliberately

• values-extra on the columns whose only unmatched value is "". The empty strings are in the
  data (CNTY = '' matches exactly the 5 map-hole polygons); they do not survive the round trip
  through the verifier's query transport. The declared arrays are correct.
• recall-candidate-categorical on fig22.QUAD. It is a list of 7.5-minute quadrangle names, not a
  coded domain. Declaring the 31 names satisfies the recall check but then trips the inline
  CODE=Definition requirement, since a few names are single words that the self-describing
  heuristic does not recognise, and the only way to clear that is 31 entries of the form
  HENRIEVILLE=Henrieville quadrangle. Left as an advisory rather than adding that.