USGS Bedrock Geology & Lithology Screener

Point-in-polygon screening of any US coordinate against the USGS State Geologic Map Compilation (SGMC) — a seamless compilation of the 48 conterminous state geologic maps, 313,732 polygons, published scales 1:50,000 to 1:1,000,000 (doi:10.5066/F7WH2N65).

You get the mapped bedrock unit, its formal stratigraphy, its chronostratigraphic age, its full lithologic composition, and eight derived engineering and environmental hazard flags — each with the exact keywords that triggered it, so every call is auditable.

Batch a whole portfolio in one run. No API key. No scraping.

---

Who this is for

Geotechnical engineers scoping an investigation. Before you site a single boring you want to know whether you are on carbonate (grout takes, voids), on smectitic shale (shrink-swell), on loess (hydrocompaction on wetting), on a mapped landslide deposit, or on hard crystalline rock that will need blasting rather than ripping. rippability_class, major_lithology_percent and unit_comment — which frequently carries the published unit thickness — turn a coordinate into a scoping memo.

Phase I ESA and environmental due diligence. ASTM E1527-21 asks you to describe the geologic setting. reference_citation is the full bibliographic citation of the state geologic map the answer came from, dataset_doi cites the compilation itself, and source_scale / positional_uncertainty_m state honestly how precise the answer is. Drop them straight into the report. acid_rock_drainage_risk and radon_parent_rock flag the two geologic conditions that most often become a finding.

Solar, BESS and data-center siting. Pile-driving feasibility is a lithology question. Karst wrecks a fixed-tilt racking budget; hard rock forces pre-drilling; soft compressible ground forces deep foundations. Screen a thousand candidate parcels in one run and rank them before anyone drives out.

Naturally-occurring asbestos compliance. California's CARB Asbestos ATCM, and equivalent rules in North Carolina and Virginia, are triggered by ultramafic and serpentinite rock. This actor scans the union of every lithology field and the SGMC lithology hierarchy — the token Ultramafic appears in the hierarchy on 153 units and never in the primary-lithology column, so a screen that reads only the obvious field misses them.

AI agents. See "Use as an MCP tool" below.

---

Why the flags read more than the obvious column

SGMC has a GENERALIZED_LITH column that looks like the natural thing to key a hazard flag on. It is not sufficient. 11,790 polygons carry MAJOR1 = 'Limestone' while GENERALIZED_LITH says only 'Sedimentary, undifferentiated' — 35% of every limestone polygon in the country, invisible to a generalized-lithology-only karst screen. Mammoth Cave, Kentucky is one of them.

So every flag scans the union of GENERALIZED_LITH, MAJOR1-3, MINOR1-5, INCIDENTAL, INDETERMINATE, and the SGMC Lithology table's full classification hierarchy (Igneous - Plutonic - Ultramafic - Peridotite) plus its lowest-level rock name. Matching is word-boundary based on normalized text, so sand never matches sandstone and clay never matches claystone.

The depositional-form column (LITH_FORM) is scanned separately, because its controlled vocabulary mixes real depositional settings (Alluvial, Playa, Eolian-loess, Landslide) with metamorphic facies labels (Greenschist, Amphibolite epidote-amphibolite) that describe pressure and temperature grade rather than rock type. Merging them would flag every greenschist-facies terrane in the country as asbestos-bearing.

---

Example input

{
  "assets": [
    { "lat": 37.187, "lon": -86.1,   "label": "Mammoth Cave, KY" },
    { "lat": 36.38,  "lon": -120.75, "label": "Clear Creek, CA" },
    { "lat": 39.9,   "lon": -102.5,  "label": "Eastern Colorado plains" },
    { "lat": 61.2,   "lon": -149.9,  "label": "Anchorage, AK" }
  ],
  "bufferMeters": 500,
  "maxResults": 1000
}

Input	Required	Default	Notes
assets	yes	—	Array of {lat, lon, label}. latitude / longitude / lng aliases accepted. Omitting it is an HTTP 400 before the run starts, so nothing is billed.
bufferMeters	no	500	Radius for the nearby-unit search, 0–50,000 m. The query is paginated, so a large radius returns every polygon rather than silently stopping at 2,000. Set 0 to disable.
maxResults	no	1000	Row cap. One row per asset.

---

Output fields

The site and the mapped unit

Field	Description
asset_label, lat, lon	Echoed from your input.
state	Two-letter code of the state compilation the unit belongs to.
unit_name	Name of the bedrock map unit at the coordinate.
sgmc_label, unit_link	SGMC harmonised label, and the key joining the polygon to the lookup tables.
orig_map_label	The label as printed on the published state sheet — often different from sgmc_label.
unit_description	Full descriptive text for the unit.
stratigraphic_units	Formal formation and member names the unit contains. Populated on 612 units.
unit_comment	Compiler note; frequently carries unit thickness ("thickness ranges from 97-282 m").
province	Physiographic / geologic province. Populated on 1,754 of 6,627 units, in PA NH MD SC NY NJ VT AL GA TN NC UT.

Age

Field	Description
age_min, age_max	Chronostratigraphic paths from the polygon attributes.
age_min_ma, age_max_ma	Numeric bounds in millions of years.
age_min_era / _period / _epoch and age_max_era / _period / _epoch	The full hierarchy, split for filtering.
unit_age_text	The age in the state survey's own words, e.g. "Upper Mississippian".
age_dating_type	Absolute (isotopic) or Relative (stratigraphic correlation).
age_note	Compiler note, including isotopic dates where cited. Populated on 1,807 units.

Lithology

Field	Description
generalized_lith	SGMC generalized class. Coarse by design — see above.
major_lith_1/2/3, minor_liths, incidental_lith, indeterminate_lith	The polygon's own lithology attributes. indeterminate_lith is sparse (2,709 of 313,732 polygons, 19 states); a recorded input that populates it is {"lat":33.85734,"lon":-111.97056}.
lithology_hierarchy	Metamorphic - Metaigneous - Serpentinite (Major) | Igneous - Plutonic - Ultramafic - Peridotite (Minor).
lithology_detail	Structured array, one entry per lithology row: rank, the five classification levels, form, comment, percent_of_unit.
lithology_forms	Distinct depositional forms and metamorphic facies (Landslide; Colluvial, Alluvial; Eolian; Playa).
lithology_notes	Compiler notes from the lithology table. Populated on 7,568 of 18,418 rows.
major_lithology_percent	Composition estimate: Limestone 40%; Sandstone 40%; Siltstone 15%; Shale 5%. Only unambiguous single-percentage notes are parsed, so an absent value means "not stated", never zero.

Hazard flags

Every flag is true, false or null. null means not checked — the source backing it did not answer. It never means "no hazard".

Field	Fires on
noa_risk	Serpentinite, ultramafic, peridotite, dunite, pyroxenite, amphibolite, talc and soapstone, tremolite / actinolite / anthophyllite, greenstone, ophiolite.
karst_bedrock	Limestone, dolostone, dolomite, marble, chalk, coquina, travertine, marl — plus the far more soluble gypsum, anhydrite, halite and other evaporites.
acid_rock_drainage_risk	Black, oil and alum shale, pyrite, sulfides, marcasite, gossan, coal, lignite, anthracite, graphite.
expansive_soil_parent	Smectite, montmorillonite, bentonite, claystone, clay, mudstone, shale, marl, argillite, saprolite.
radon_parent_rock	Uranium-bearing black / alum / oil shale, marine phosphorite, and felsic plutonic rock (granite, alkali granite, syenite, pegmatite). Screening only — not a substitute for a measured test.
collapsible_loess_risk	Loess recorded as the depositional form: hydrocompaction on wetting. This signal exists only in the lithology table.
soft_compressible_ground	Peat, muck and organic-rich lithologies, or playa / lacustrine / estuarine / deltaic / swamp deposition.
landslide_colluvium_parent	The map unit is itself a landslide, mass-wasting, colluvial, talus or melange deposit.
hazard_flag_count	How many fired. null when nothing could be checked; 0 means checked and clean.
risk_basis	The exact keywords that triggered each flag: karst: limestone | expansive: shale. Audit every call.
rippability_class, rippability_note	unconsolidated_rippable, weak_rock_rippable, intermediate_marginal, hard_rock_blasting_likely, not_applicable, unknown — plus prose.

Nearby units, because the point is not the whole answer

Field	Description
units_within_buffer	Distinct units inside the buffer, each with its own hazard flags, computed through the same union.
units_within_buffer_count, multi_unit_ambiguity, buffer_meters	More than one unit means the site sits near a mapped contact.
nearby_noa_risk, nearby_karst_bedrock, nearby_acid_rock_drainage_risk, nearby_expansive_soil_parent	True if any nearby unit is; false only when every nearby unit was checked and negative.
nearby_hazard_summary	Which hazards are nearby, and exactly which unit links carry them.
nearby_flags_status	ok, partial or not_applicable.

The point answer is only as good as the map. At 1:500,000 a drafted contact is worth ±250 m on the ground, so a site reading "granite, no asbestos" with a serpentinite body 300 m away has not been cleared.

Provenance and precision

Field	Description
source_scale, source_scale_denominator	Publication scale of the source sheet — the coarsest, when the citation lists several.
positional_uncertainty_m	Ground uncertainty from the 0.5 mm drafted-line convention: denominator × 0.0005 m. 250 m at 1:500,000.
map_polygon_area_km2	Approximate area of the specific polygon, de-scaled from Web Mercator by cos²(latitude).
map_precision_class	site_scale, reconnaissance or regional_generalization.
point_features_returned	Polygons intersecting the coordinate. Normally 1.
reference_citation	Full bibliographic citation of the source state geologic map.
ngmdb_url, digital_data_url	NGMDB product page, and the publisher's GIS download.
dataset_name, dataset_doi	The SGMC data release itself, for a references section.

Per-source status — read this before you trust a false

Field	Description
geology_layer_status	ok, no_coverage, unavailable.
units_lookup_status, age_lookup_status, lithology_lookup_status	ok, no_match, unavailable, not_applicable.
buffer_status, buffer_truncated	Whether the nearby-unit list is complete.
flags_complete	true when the lithology hierarchy was actually read, so a false flag genuinely means checked-and-negative.
sources_unavailable, data_completeness	Which sources did not answer, and complete / partial.
out_of_coverage, coverage_note	See below.
error, source_url, retrieved_at	Per-asset error, the exact reproducible query, and the fetch timestamp.

---

Coverage and failure behaviour

SGMC maps the conterminous 48 states only. Alaska, Hawaii, Puerto Rico and the territories are not in it. A coordinate there gets an explicit out_of_coverage: true row with an explanatory coverage_note and every hazard flag set to null — never false. An unmapped coordinate is not checked, and a screener that reported "no asbestos" for Anchorage would be lying to a buyer about to break ground. The same applies offshore and in gaps between state compilations.

The actor fails loudly rather than returning a plausible empty answer:

• Omitting assets is an HTTP 400 at run creation. The run never starts and nothing is billed.
• Four live drift assertions run before any screening: the geology layer's polygon count and the lithology table's row count must fall within their calibrated bands, the token Ultramafic must still exist in the lithology hierarchy, and the limestone-behind-a-generic-generalized-lithology population must still be non-zero. If the upstream has been silently truncated or reshaped, the run fails and nothing is billed — because the flags would no longer mean what they say.
• If every in-coverage asset fails against the service, the run fails rather than billing you for a page of error rows.
• If all three lookup tables are down at once, the run fails, because every hazard flag would have been null.
• If some sources answered, you get your rows, and each one states on its face which sources did not answer and which flags are therefore unverified.

---

Use as an MCP tool

Available to AI agents through mcp.apify.com. Every input and output field carries a description, so an agent can call it unassisted and reason about the result. It chains cleanly: resolve an address to a coordinate, screen the geology, then hand karst_bedrock, noa_risk and rippability_class to a downstream siting or cost model. risk_basis gives the agent evidence to cite, and flags_complete tells it when not to.

---

Pricing

Pay per result: one event per row, one row per asset. A 500-site portfolio screen is 500 results. Graduated discounts apply on paid Apify plans. Failed runs emit no rows and are never billed.

---

Related actors

Bedrock lithology is one layer of a site screen. These cover the rest:

• Karst & Sinkhole Risk Screener — the USGS karst-susceptibility polygons. Complementary rather than duplicative: this actor reports the mapped surface unit's lithology, while the karst screener includes buried carbonate and evaporite units at depth. Denver reads Gravels and alluviums (COQg;0, no karst) here and is a genuine buried-evaporite hit there. Both are correct; they answer different questions. Run both.
• USGS Landslide Proximity Screener — the mapped landslide inventory. landslide_colluvium_parent here tells you the bedrock map unit is a slide deposit; that actor tells you which inventoried slides are nearby.
• USGS Seismic Design Screener — ASCE 7 design values. Site class is a function of the material this actor identifies.
• Site Due Diligence Bundle — 20 layers for one coordinate in one call, with a composite score and a go / caution / no-go verdict.

---

FAQ

How current is the data? SGMC is a compilation of published state geologic maps, so its currency is the currency of each state's sheet — anywhere from the 1980s to the 2010s. That is the nature of bedrock geology: the rocks have not moved. What changes is mapping detail, which is why every row carries source_scale, reference_citation and positional_uncertainty_m.

Is this a substitute for a site investigation? No, and nothing costing cents per site is. It is a screening tool: it tells you where to spend investigation budget and what to expect, at a scale of 1:50,000 to 1:1,000,000. Read map_precision_class before leaning on any single answer.

Why is a flag null instead of false? Because the source backing it did not answer. null means not checked; false means checked and negative. Look at flags_complete, lithology_lookup_status and sources_unavailable on the row.

The point says no karst but a nearby unit is limestone. Which do I believe? Both. karst_bedrock describes the polygon your coordinate fell in; nearby_karst_bedrock describes the buffer. Compare the distance you care about against positional_uncertainty_m — inside that distance the map genuinely cannot distinguish them.

Why does my coordinate return out_of_coverage inside the US? Either it is offshore or in a large water body, or it falls in a gap between state compilations. coverage_note says which.

Can I get Alaska or Hawaii? Not from SGMC — they are not in the compilation. Those rows come back honestly flagged rather than silently blank.

How many requests does a large run make? The lookup tables are fetched once for the whole run, batched by the distinct set of units across every asset — three requests total for a typical multi-state portfolio, never one per asset. Spatial queries are two per in-coverage asset: the point, plus a paginated buffer.

---

Data source: U.S. Geological Survey, State Geologic Map Compilation (SGMC), conterminous United States, doi:10.5066/F7WH2N65. Public domain. This actor is not affiliated with or endorsed by the USGS.