Global Database of Cement Production Assets and Upstream Suppliers — Asset-Level Dataset

Most cement statistics arrive as country totals; this database works one asset at a time. Compiled under the GeoAsset programme of the Spatial Finance Initiative at the University of Oxford with Astraea Inc., it lays out the world's cement industry as two linked tables. The assets table describes each plant: city, state/province and country, ISO 3166 country codes, region and subregion, WGS84 latitude and longitude carrying an accuracy flag, operating status, plant type (Integrated or Grinding), wet-or-dry clinker process, capacity in millions of tons with a confidence rating and its reporting source, the year production started, and an ownership chain resolved through Refinitiv PermIDs, Legal Entity Identifiers, private-or-public holding status and exchange tickers for up to two ultimate parents.

The second table does something almost no other cement resource attempts: it attributes each plant to its upstream raw-material mines by country, with supplier coordinates attached, and classifies each plant's input strategy as locally sourced, imported or hybrid across limestone, clay, gypsum, sand and coal. Established reporting from development banks and governments sits alongside geospatial computer-vision and language-model extraction, and the whole construct is documented in a peer-reviewed Scientific Data article. Get a sample of this dataset and inspect both tables before you commit to anything.

Two linked tables land in one delivery: an assets table with one row per cement plant, and a suppliers table attributing each plant to upstream raw-material mines. Illustrative rows in the delivered column order:

# assets table -- one row per cement plant
uid         : plant_0417
location    : Greyrock, Alberta, Canada      iso3 : CAN      region : Americas / North America
lat_lon     : 51.05, -114.07                 geom_accuracy_flag : present per record
status_code : operating                      plant_type : Integrated      production_type : Dry
capacity    : 1.80 Mt                        confdnc : confidence-rated   cap_source : company disclosure report
year        : 1978
owner       : Northbridge Cement Ltd         owner_permid : 5078001234    owner_source : company disclosure report
parent      : Northbridge Materials Group    parent_permid : 5063009871   parent_lei : 5493001EXAMPLELEI22X0KX34
holding     : Public   ticker : NBM   exchange : TSX    ownership_stake : n/a (wholly owned)
sourcing    : hybrid                         raw_mtrl : limestone, clay, gypsum, coal    clinker : yes

# assets table -- a grinding station with a joint-venture parent
uid         : plant_1129
location    : Puerto Litoral, Colombia       iso3 : COL      region : Americas / South America
lat_lon     : 10.62, -75.51
status_code : operating                      plant_type : Grinding        production_type : n/a
capacity    : 0.65 Mt                        confdnc : confidence-rated   cap_source : news media
year        : 2011
owner       : Andina Cementos S.A.S.
parent      : Meridian Building Materials    ownership_stake : 60% (joint venture; second parent rides in the _2 columns)
holding     : Private   ticker : --   exchange : --
sourcing    : imported                       raw_mtrl : clinker, gypsum   clinker : yes

# suppliers table -- one row per plant-to-mine attribution
facility.country : Canada                    supplier.country : Canada
supplier.lat_lon : 52.11, -113.42

Plant names, identifiers and figures above illustrate the shape of a row rather than quote certified extracts; your sample pulls live records. The ownership block is the payload: PermIDs and LEIs resolve every operator and ultimate parent to a capital-markets entity, so plant capacity rolls up into issuer-level exposure in one aggregation.

Forty-two documented columns span the two tables, verified against the deposit's README documentation and the accompanying peer-reviewed article. Missing values are stored as an explicit 'n/a' token rather than blanks, so downstream filters and parsers behave predictably. The columns below cover the load-bearing fields; controlled vocabularies and the joint-venture column block fold into the final row.

Geography - global scope, with every plant coded to ISO 3166 in both three-letter and numeric form, grouped by region and subregion, and pinned to a WGS84 point coordinate that carries an accuracy flag. Mapping, distance-to-market and clustering work straight off the table.

Temporal - a single versioned snapshot, October 2023 vintage. Each plant carries its production start year, so fleet-age and vintaging analyses run without any extra joins, but there is no year-over-year panel inside the deposit itself. When you need the flow side, pair it with monthly global cement production data.

Granularity - asset-level throughout: one row per cement plant in the assets table, plus plant-to-supplier attribution rows linking each facility to upstream mines by country with mine coordinates. Nothing above plant level unless you aggregate it yourself.

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

• ESG and emissions analysis - start year, wet-or-dry process, capacity and confidence ratings are exactly the variables utilization-rate and greenhouse-gas models need per kiln; the deposit was built as a foundation for spatial finance work of this kind. More on our ESG and emissions analysis use case.
• Supply chain mapping - the supplier table ties each plant to upstream mines by country across limestone, clay, gypsum, sand and coal, so input-side dependency maps build in one hop; see supply chain mapping.
• Market sizing and competitive mapping - capacity by plant rolled up through PermID- and LEI-linked parent chains puts hard numbers under market-share claims; see market sizing.
• Investment and credit screening - tickers, exchanges and public/private holding status join plant-level capacity to securities master files, turning asset registers into issuer exposure views.
• ML model training - clean, labeled, globally consistent point geometry with categorical targets (integrated versus grinding, wet versus dry, local versus imported inputs) is rare training material for industrial-classification and remote-sensing models; more for data scientists.

ESG and sustainability analysts get plant age, process and capacity in one table - the difference between a corporate net-zero pledge and an auditable list of which kilns exist and how they fire. Data scientists and geospatial engineers get globally consistent point coordinates with accuracy flags and identifier-clean labels, ready for feature pipelines. Competitive intelligence product teams get ownership resolved to PermIDs and LEIs, so corporate-family trees assemble themselves; more for competitive intel product teams. Developers and builders get two flat CSV-shaped tables that load without transformation gymnastics - see developers and builders.

Against the Global Cement and Concrete Tracker, the division of labor is clear: the tracker is the broader, maintained census with pipeline and retirement status; this database contributes the harder-to-source columns - PermIDs, LEIs, tickers, production start years, capacity confidence ratings and upstream mine attribution. It is in no way a rival: Global Energy Monitor cites it as one of the seed sources for its own cement tracker.

Against emissions ledgers such as Global Carbon Project's national cement CO2 series, the contrast is flows versus assets: those count tonnes of CO2 by country and year, while this supplies the asset base needed to attribute emissions to specific plants. A direct methodology comparison sits on our GCP-CEM vs this database page, and the wider slate lives on our best construction materials datasets ranking.

Three things. First, this is a versioned snapshot, not a rolling series: the October 2023 vintage carries production start years but no year-by-year panel, so trend work needs a companion flow dataset rather than wishful thinking.

Second, a handful of columns - operating status, geolocation accuracy and the capacity confidence rating - ship as compact codes whose full vocabularies live in the file documentation; the field meanings below are verified, and the code lists arrive with your delivery.

Third, nobody publishes the plant count anywhere on the record, which we regard as mildly absurd - so the row count is the first thing your sample tells you. Get a sample of this dataset and count for yourself.