USGS Mineral Resources Data System (MRDS): Every Known Mineral Site as a Row

One federal inventory, two merged lineages, one row per site. MRDS combines the original Mineral Resources Data System with the US Bureau of Mines' MAS/MILS archive into a worldwide point database of metallic and nonmetallic mineral resources - 304,632 site records spanning mines, prospects, occurrences and plants, deepest inside the United States.

What makes it usable at scale is discipline rather than size. Every site resolves against 46 flattened fields: a 12-digit dep_id tying all records for a property together, preferred and previous names, WGS84 coordinates, administrative geography down to county and district, a metallic/nonmetallic flag, commodities in three tiers with hyphenated qualifiers, operation type from placer to leach to processing plant, development status, ore and gangue minerals, deposit model names such as {Climax Mo}, host-rock codes drawn from the LithClass 6.2 lithology scheme, discovery and production years, and the bibliographic references underneath. USGS then graded each record A through E on information richness so users could find the detailed reports first.

Scale, concretely: counting the primary commodity field alone, about 69,940 records list Gold and 24,443 Silver, alongside 1,002 Platinum and 180 Palladium. Within Datadory's catalog of 1,744 datasets across 159 viable industries, this record scores 9/10 for quality, with field definitions verified.

Get a sample of this dataset

Three real records chosen to span the grade range, flat exactly as their core fields arrive:

dep_id     : 10028178            site_name : E. A. Culver
coords     : 38.31935, -120.72045          (California, United States)
commod1    : Gold                          oper_type : Unknown
dev_stat   : Occurrence                    score     : D

dep_id     : 10056347            site_name : Big Bell Mine
geography  : Western Australia, Australia  (no coordinates in this record)
commod1    : Gold,Silver                   dev_stat  : Producer
prod_yrs   : 1916-1923, 1937-1955, 1989-   score     : A

dep_id     : 10012410            site_name : Henderson Mine
coords     : 39.76916, -105.84199          (Colorado)
disc_yr    : 1965    yr_fst_prd : 1976     oper_type : Underground
commod1    : Molybdenum,Lead,Tungsten,Tin,Zinc,Iron    dev_stat : Producer
model      : {Climax Mo}                   score     : A

Four buyer questions settle themselves here. First, shape: one flat row per site, keyed on dep_id, so a filtered pull lands in a dataframe without a parsing pass. Second, tiers: commodities split into primary, secondary and tertiary lists - Big Bell's Gold,Silver reads as a gold operation with silver credit, which is exactly the distinction commodity screens need. Third, time: production spans arrive as interval text like 1916-1923, 1937-1955, 1989-, gaps included, because districts reopen. Fourth, trust: the E. A. Culver occurrence carries a D grade and Henderson an A, so weighting records by documentation richness is a filter, not a judgment call.

Every other record repeats the same shape - the table simply fills in different sites, statuses and commodity mixes.

46 flattened fields in the full digest, definitions verified during research - twenty-one form the working core in the dictionary below, grouped into five families: identification (dep_id, site name), geography (latitude, longitude, country, state), commodities (com_type, the commod1-3 tiered lists), operation and status (oper_type, dev_stat), and description (ore, gangue, host-rock code, deposit model, discovery and production years, references) plus USGS's own quality grade (score).

The remaining twenty-five are the deep-geology apparatus - deposit type, orebody form, workings type, alteration, concentration processes, ore controls, associated rock units, structure and tectonic setting among them. They ship under additional fields on request rather than cluttering the default cut, mapped to your use case at sampling.

One convention deserves a flag before any join: multi-value fields are comma-delimited within a field, and qualifiers follow each commodity after a hyphen, so Gold-refinery and bare Gold mean different metal-accounting bases. Parse against the commodity-codes list rather than assuming a flat string match.

Six groups sit beyond the core dictionary and get mapped to your use case at sampling:

• Deposit geology block - deposit type, orebody form, workings type, alteration, concentration processes and ore controls, the descriptors that explain how a site formed and was worked.
• Host and wall rock detail - host rock and associated rock units with their LithClass 6.2 codes, plus structure and tectonic setting where reported.
• Commodity qualifier strings - the hyphenated suffixes distinguishing refinery product from contained-metal figures, decoded against the controlled commodity-codes list.
• Administrative geography - region, county, district and land status extending the country-and-state pair shown in the samples.
• Workings history text - free-text production-history notes, discoverer attribution and previous site names preserved beside the current preferred form.
• Legacy identifiers - mrds_id and mas_id keys for reconciling against older Bureau of Mines extracts still sitting in internal systems.

Name the ones you want when you request the sample; they arrive as ordinary columns in the same rows, not as a side file.

Geography - worldwide point coverage with the United States carrying the deepest detail; roughly 19% of rows lack country or state values, so coordinate-based filters outperform name-based ones. Alaska's fine-grained detail lives in its companion Alaska Resource Data File rather than here.

Temporal - treat the table as a research snapshot: extraction dated 15 March 2016, file regenerated August 2022, and USGS systematic updates having ceased in 2011. Human-activity fields - operating status, ownership, production - reflect only the date of each record's underlying source report, which is why a row can say Producer while the site sits idle. Open-ended bounds carry < and > modifiers on first and last production years, so "began after 1905" stays distinguishable from "began in 1905".

Granularity - one row per mineral site, prospect, occurrence or plant across 304,632 features, keyed on the 12-digit dep_id. There is no annual time series hiding inside; year fields describe a site's life, they do not build a production panel. Pair with a statistics series when the question is tonnage over time rather than where sites sit.

API, files, or your warehouse. Daily, weekly, or hourly.

Name the commodity set, the geography and the status band when you request the sample - every gold-bearing prospect in one state, all Past Producers within a bounding box, the full A-graded subset worldwide. The sample ships first either way; the ongoing feed lands on whatever cadence your workflow needs, shaped to the scope you named rather than as an undifferentiated dump.

A national-scale site inventory earns its keep in five jobs:

• Exploration targeting and ground screening - Filter the 304,632 rows to a commodity, status band and bounding box - say gold occurrences in a Nevada county - and rank acreage before anyone files a claim.
• District dating and mining-history research - Discovery year, first and last production years and per-record references turn 'this district is old' into dated evidence of when ground was found, worked and abandoned.
• M&A and land-package comps - Past Producer and Producer rows around a target give deal teams the neighborhood comparables - what operated nearby, under what deposit model, until when.
• Supply-chain and critical-minerals mapping - Commodity tiers across every site map where antimony, tungsten or platinum-group exposure actually sits, country by country, without waiting on company disclosure.
• Prospectivity modeling and geospatial ML - Labeled positive points with grades, deposit models and host-rock codes assemble straight into training sets - the A-E letter doubles as a confidence weight.

Investors & Quant Researchers get deposit-level evidence behind a junior's claims, with status and production years to test the story. Data Scientists & ML Engineers get one flat table keyed on dep_id where features join without surprises. Developers & Builders get stable coordinates and names their map products can plot against, with the grade letter built in for demotion logic. Market Researchers & Consultants get citable federal site counts for client decks. Competitive Intelligence & Product Teams get land-package context grounded in the public record. Start from the best precious metals & minerals datasets shortlist, then pull a sample.

Provenance note - compiled by the U.S. Geological Survey's Mineral Resources Program, merging the original MRDS with the Bureau of Mines' MAS/MILS archive; hundreds of reporters contributed over decades, which is why terminology varies record to record and the A-E grade exists in the first place.

Methodology note - screen with the grade before treating a site as firm ground: A records carry the fullest descriptions and reference lists, E records generally lack bibliographic backing. Commodity strings need parsing against the codes list, multi-value fields are delimiter-split rather than atomic, and absence differs from zero - a blank commodity tier means unreported, not none.

Completeness note - Datadory scores this record 9/10, and the twenty-one-field core above maps fully to the verified dictionary. The snapshot vintage stated in coverage is asserted by the publisher itself, caveats included, which is rarer than it should be and worth trusting.

Where to go next - keep the site ledger and add the layers around it: USGS Mineral Resources Online Spatial Data (MRDATA) wraps this register inside a catalog of assessments, geochemistry and geophysics; Mining Data Online: mining projects and companies database adds the corporate and cost side; MSHA Mine Data Retrieval System & Reports covers the operating present for US mines. The cards below collect the near neighbors.