In one paragraph
What the mine pages claim, exactly
Two public United States Geological Survey files are read around any address, ZIP code or town, and every record in them is placed inside the Census Bureau town and county boundaries by its own coordinates. Each page then names the office in that state that holds the underground mine maps, because those maps, not these files, are the record that shows what runs under a parcel.
No page says that a mine is under a given house. A point in either file is a location, often the opening a surveyor saw at the surface, and the workings it leads to are drawn only on the mine map. Finding nothing in the files is not proof of nothing in the ground: a working that was never cataloged and never drawn on a survey sheet is in neither file, and on coal ground that describes a great deal of what was mined.
Downloads
The files, each one public and each one downloadable
| File | What it holds and what it is used for | Download |
|---|---|---|
| USGS Mineral Resources Data System | Named mines, prospects and mineral occurrences with the commodity, the development status, the operation type and a position. 304,632 records, 274,114 with a position. The USGS says it has had no systematic update since 2011. It holds 157 coal records in the whole country and none in Pennsylvania, West Virginia or Illinois. | 25 MB, mrds-csv.zip |
| USGS USMIN topographic mine features, version 10 | Shafts, adits, mine entrances, prospect pits, quarries and tailings digitized from the historical topographic sheets, each point carrying the sheet name and the survey year. 466,747 points, 97,564 of them underground openings before the cave rule below. | 253 MB as the national shapefile from the ScienceBase item |
| USGS Open File Report 96-92, coal fields of the conterminous United States | The National Atlas coal fields layer, 993 polygons drawn at 1:5,000,000, a digital version of Trumbull's 1960 map with the Gulf Province updated from state geological maps, as the report documentation explains. The layer carries the field and the coal rank, never a single working. | 5 MB as a shapefile from the Stanford Earthworks copy |
| Census Bureau TIGER 2024 county boundaries | The county, parish, borough and census area polygons every point is tested against. For Connecticut the 2024 file carries the nine planning regions, so the eight counties come from the TIGER 2020 county file instead. | Shapefile, national |
| Census Bureau TIGER 2024 place boundaries | The incorporated places and census designated places, one file per state. A town page counts only what falls inside its own limit. | Shapefile, one per state |
| Census Bureau national county file | The official county names, read per state, so a page prints Parish, Borough, Municipio and the spelling the Census uses: St. Clair, Prince George's, McKean. | Text, 3,235 rows |
| Census Bureau geocoder | Turns a street address typed into the search into a coordinate. It is called from the browser with a JSONP callback because the service sends no CORS header; the API reference documents the call. | Free web service |
Point in polygon
How every record was placed
Every point of both files is tested against every county and every incorporated place by its own coordinates, with a ray casting point in polygon test and a one degree grid index, written without a geographic library. The county name written in the record is never used, because it is blank or wrong often enough to matter: all 12,958 Alaska rows of the mineral file carry no county name at all. A point that falls outside every boundary, offshore or in a gap, is counted in no county and no town.
Connecticut needed its own pass. The 2024 county file carries the state's nine planning regions, which no page is named after, so every Connecticut point fell in a region and the eight county pages read zero while 61 town pages carried records. The 2020 county file places them: 1,267 mineral records and 1,289 sheet features, seven of the mineral records falling outside every county, offshore.
Puerto Rico has no town or county pages on the site, so its 721 mineral records and 357 sheet features are placed inside the 78 municipios and named on the island's page rather than linked. The District of Columbia has no county, so its one town page is the District.
What counts as what
What each number on a page means
| Word on the page | What it is |
|---|---|
| Old mine site | A mineral file record whose development status is past producer, prospect or occurrence. A producer is counted separately and shown with its status. |
| Site with no status | A mineral file record with the status field blank, common on sand and gravel. 555 pages had records and no section until these were counted, because the page list keyed on the status. |
| Hole drawn on the old survey maps | A point from the topographic features file: a shaft, an adit, a mine entrance, a prospect pit, a quarry or a tailings pile, with the sheet and the year. |
| Underground opening, or shaft and tunnel | A sheet feature whose type is an adit, a shaft, a tunnel or a mine entrance. A feature the old sheets drew with the adit symbol but named a cave, a cavern, a grotto or a rock shelter is a mapped feature and not an opening: 278 of them. A name that also says mine, shaft, adit, prospect or quarry stays a working. |
| Working type | Read from the mineral file's operation type field, which the first build missed because it read a different column that is empty on 279,032 of the 304,632 rows. Across the search tiles: 46,396 underground, 102,233 surface, 12,898 worked both ways, 112,587 not recorded. |
| Records behind the sites | The raw record count printed beside the site count on every page, so the merge below is visible and can be checked. |
The merge
How the same working is counted once
Both files record the same working more than once, and the mineral file does it to itself: the Halifax shaft at Tonopah is three records, a survey map shaft and two mineral file entries 25 meters apart. Records inside the same county are clustered when their names match once the filler words are stripped (mine, company, shaft, number, claim and the like), common abbreviations are expanded (Consolid. to Consolidated) and the town's own name is removed, and when they sit within 150 meters of each other. Two records whose names match exactly may sit up to 700 meters apart, because the mineral file itself says on the Roxbury Iron Mine record that it combined the record with the other Roxbury Iron Mine entry. A pair the file cross references that way is one site at any distance, and 164 pairs do. Records named for a mining district rather than a working are joined within 60 kilometers, since the file holds three rows called Tonopah District two kilometers apart.
What is never merged: an unnamed sheet feature, and a sheet feature whose name is only its own type, because a surveyor labels an unnamed hole with the word for what it is. That rule came from Litchfield, where 27 separate gravel pits had been folded into 11. The row that is kept takes the fullest written out name, the survey map position where there is one, the status of any record that has one (a producer's status code is zero, which an earlier test read as empty and threw away), the minerals of every record named once, and every record number, so nothing is lost and every merge can be undone by hand.
The result at county level today: 728,320 records became 681,415 sites, 46,905 of them folded into another. Inside town limits: 48,722 records became 45,424 sites. In the search tiles, 48,458 rows were merged so that one mine is one dot.
Today
The totals as the pages print them today
| Count | Number |
|---|---|
| Counties with a record | 3,056 of 3,062 |
| Records inside counties | 728,320 |
| Sites inside counties after the merge | 681,415 |
| Old mine sites | 171,273 |
| Producers | 28,813 |
| Sites with no status in the records | 22,962 |
| Holes drawn on the old survey maps | 458,367 |
| Underground openings, caves excluded | 92,654 |
| Towns and cities with a record inside their own limit | 6,356 |
| Records inside town limits, and sites after the merge | 48,722 and 45,424 |
| Counties over a mapped coal field | 1,148 |
| Towns over a mapped coal field | 4,964 |
| Search tiles, one degree cells | 1,809 |
The gap
The coal field test, and why it exists
The mineral file holds 157 coal records in the country and the sheets only show a hole where a surveyor drew one, so Scranton showed five records while sitting on the anthracite. The coal fields layer closes that gap. Every town and county boundary is tested against the 993 polygons by its center and by up to 48 points of its own outline, because the layer is drawn at 1:5,000,000 and a town center can fall in a gap the small scale leaves. A place caught only at the edge of a field is told it lies on the edge, not over it.
1,148 counties and 4,964 towns sit over a mapped field. The 3,454 that sit over coal with not one record in either file carry a section of their own saying exactly that, because that is the purest form of the problem: a town over coal with nothing in the files.
The hub tool
How the address search works
The records are cut into 1,809 one degree tiles, the largest 1.45 MB, and the browser loads only the tiles the chosen radius touches. The point comes from a ZIP code centroid, a Census place, the Census geocoder for a street address, or the browser's own location, and the radius runs from half a mile to 25 miles. Filters by source, status, working type and mineral act on the rows after they load, so unticking every status box leaves no mineral file row, not all of them.
A tile that answers 404 is empty ground. Any other failure is said in words, not reported as nothing nearby, because on a page about safety an empty result after a failed load is a false all clear. Every run carries a token so a late answer from an earlier search is dropped, the search is written into the address bar so a link reopens the same result, and the tile addresses carry a version so a cached tile cannot answer with old data.
Where a place name is used twice in one state, Cottonwood in Arizona 171 miles apart for one, the search asks which. St. Louis, Saint Louis and St Louis all find the same place.
Stated on every page
The limits
| Limit | What it means |
|---|---|
| A sheet shows the surface | A topographic sheet shows an opening at the surface, not the rooms under it. A working can run for miles from the point. |
| A mineral record is a deposit | Its position is of varying precision and is not a property line. |
| Nothing found is not a clearance | A working never drawn on a sheet and never entered in the mineral file is in neither count, and on coal ground that is common. |
| The files are dated | The mineral file has had no systematic update since 2011; the sheets were surveyed between the 1900s and the 2000s, and every state page prints its own year range. |
| The boundaries are the Census's | 2024 limits, which may not match a municipal record or an older annexation. |
| The geocoder can miss | A street address can be placed on the wrong block, and a rural address on the road rather than the house. |
| The merge is a rule, not a judgment | Two different workings of one name within 700 meters read as one site; the record numbers on the row let anyone check. |
The log
What was wrong and when it was fixed
| When | What |
|---|---|
| October 22, 2026 | West Virginia was missing from the town data: the place to page key tried the state names in dictionary order and every West Virginia page matched Virginia first. The key tries the longest name first and asserts it. |
| October 22 | 201 descriptions printed a count of zero as a fact; Louisiana pages said County for a Parish; Saint Clair and Prince Georges were spelled as the file name spells them. |
| October 23 | The working type was read from an empty column, so the Underground filter hid the whole mineral file. Connecticut had nothing at county level. The District said it had no mines when the files hold 48 old pits and quarries inside it. Seventeen faults in the search, the worst a failed data load reported as nothing nearby. |
| October 23 | The coal fields layer was added, and Puerto Rico became the 52nd page. |
| October 24 | Caves drawn with the adit symbol were counted as mine openings: 278 removed. The same working was counted once per record: the merge above. 555 pages with records and no section, because the status field was blank. The named list was capped at 40 and 30 rows and said one number for two different things. |
| October 26 | Every producing mine had lost its status, because a status code of zero was read as empty. The merge had never stripped the town's own name, so Belmont and Tonopah Belmont Development Co. never met; an exact name match now reaches 700 meters; and a sheet label that is only a type is never merged, after 27 Litchfield gravel pits became 11. Washington DC had no section at all because its page wrote the FAQ hard against the section before it. |
For an auditor
How to check any number on any page
Download the two files and the two TIGER layers above, pick any county, place every point by its coordinates and apply the merge rules in the order given. The page prints the raw record count beside the site count so the merge can be checked on its own. Every row of the named list on a county or town page opens the USGS record it came from, or the USGS records at its own coordinates for a sheet feature that has no record page. To see what the surveyor drew, the USGS topoView shows every edition of the sheet, free.
Every external link on the 52 state pages was opened from this server when the pages were built, and where a host shields itself from a server the exact address was confirmed in a search index instead. The same was done for every link on these guides on October 29, 2026.
Your own state
Abandoned mine maps, state by state
Each page names that state's own offices, its own counties and its own numbers.
Please read
Disclaimer
The mine information here is for general information only. Old mine records can be incomplete, outdated or wrong, and finding nothing here does not mean there is nothing there. Check with the mine map office named here before you rely on it, and have a qualified engineer look at any ground you are worried about. PublicRecordCenter.com assumes no responsibility or liability for any errors, omissions, or consequences arising from the use of this information.