FIA forest inventory plot data: the national forest census, measured tree by tree

It is the row-level record of the United States counting its trees the way a central bank counts money: continuously, uniformly, with a mandate behind it. Forest Inventory and Analysis is a congressionally mandated USDA Forest Service program whose permanent plot network blankets all 50 states, each plot revisited on a five-to-ten-year cycle, every live tree inside a plot measured for species, diameter, height and merchantable volume. Annualized statewide panels have run since roughly 2004 under the 1999 and 2014 Farm Bills, with Interior Alaska explicitly outside the annualized design. Around the core inventory sit three siblings: a woodland-owner survey, an urban variant, and National Resource Use Monitoring - the home of Timber Products Output, which surveys the mills and turns harvested roundwood into a classified ledger.

In the catalog it ships as the USDA Forest Service FIA DataMart and Timber Products Output (TPO) record - one entry, two instruments pointed at the same wood.

The inventory half measures forests from the ground. Plot rows fix where and when a field visit happened; condition rows describe the land underneath - ownership group, forest type, stand age, site productivity and site index; tree rows are the atomic unit, roughly 105 documented columns spanning species and species-group codes, status, diameter at breast height, height, crown ratio and both gross and net cubic-foot volume. About 59 distinct table types ride alongside per state, including seedling counts, soils, vegetation subplots and the growth-and-mortality family, and population estimates with sampling errors sit above the raw rows so any published total arrives with its uncertainty attached.

The TPO half flips the camera. Wood-product manufacturers report intake and output, the results reconcile against the inventory, and every harvested unit lands in a coded cell: roundwood product, logging residue, other removals, mill residue, or the separately estimated residential-firewood bucket that never passes a mill door.

Rows exactly as they arrive - verified during cataloging, August 2026:

table : PLOT   (one field-visit row)
STATECD : 1   UNITCD : 3   COUNTYCD : 123   PLOT : 90042
INVYR  : 1990
LAT/LON: 32.850131 / -85.835005 (fuzzed)   ELEV : 980 ft

table : SURVEY  (inventory-cycle row)
STATECD : 56 (Wyoming)   INVYR : 2011
ANN_INVENTORY : Y   NOTES : "Annual 01 of 10 subcycles"

table : TREE   (the atomic unit)
PLT_CN  : 157499420010854
STATUSCD: 1 (live)   SPCD : 693
DIA     : 10.2 in   HT : 68 ft   VOLCFNET : 6.573 cu ft

Read the tree row as a sentence: loblolly pine, ten inches through, sixty-eight feet up, 6.573 net cubic feet in the saw-log portion and above. Three structural facts make millions of such rows joinable. A sequence-generated identifier (CN) is the spine - TREE.PLT_CN resolves to its parent PLOT.CN, so a county-level volume roll-up is a group-by rather than a spatial guess. Location rides on four integer keys - state, survey unit, county, plot number - which survive even where coordinates deliberately do not. And the coordinates themselves are blurred by design: exact plot positions are confidential under federal law, so geography work keys on the location codes and stays honest at county scale and above.

Every field above carries a verified definition with an example in the delivered dictionary. Request a sample cut to your states, counties and species groups and the dictionary travels with the rows.

Deeper than the headline design suggests. Annualized statewide panels date to roughly 2004, but the rotation beneath reaches back decades - observed collections begin in the mid-1980s depending on when a state entered - and because every plot returns every five to ten years, a single cross-section quietly contains a decade of staggered field visits. Growth and drain therefore get modeled from remeasurement pairs, not asserted from two snapshots.

On the mill side, TPO survey years accumulate from 1997 through 2024, with per-state participation documented rather than implied. That documentation matters: a state can hold current-cycle plots while its latest mill-side survey dates several cycles back, and treating the two clocks as identical is the most common way analyses on this record go wrong. Scope by survey year explicitly, and the gaps become known quantities instead of surprises.

Two coverage caveats are printed rather than buried. Interior Alaska sits outside the annualized design, so pan-national comparisons should exclude or separately model it. And the fuzzed-coordinate policy means the finest honest mapping unit is the county, not the tree - which is precisely why every delivered extract carries the location keys forward intact.

This is the least famous, most commercially interesting part of the record. Each cell is keyed by survey year, harvest state, mill state and county, then classified along four code families that decode into clean columns once the dictionary is attached. Removal class splits every harvested unit into roundwood product, logging residue, other removals, a residential-fuelwood estimate, or mill residue. Ownership says whose land the cut came off - national forest, other public, forest industry, other private. Forty-one species groups identify the trees, from group 8 loblolly-shortleaf pines to group 33 select white oaks. Product and source codes say what the mill made - sawlogs, veneer logs, pulpwood, composite panels, bioenergy fuelwood - and what it discarded, down to bark, coarse residues and fine residues.

Volumes arrive stated twice: MCFVOL in cubic feet, and RPA_STD_AMOUNT re-expressed the way the trade quotes it - board feet International 1/4-inch for sawlogs and veneer, cords for pulpwood and fuelwood. Having both on one row deletes the unit-conversion step where homegrown analyses quietly go wrong. For anyone siting a pellet plant or pricing a fiber contract, the mill-residue bucket plus the fuelwood product code usually decide the economics before roundwood prices do; see roundwood removals and drain and roundwood for the vocabulary.

Nothing else in the catalog records individual trees; the comparators trade granularity for reach.

FAO's Global Forest Resources Assessment is breadth itself: country-reported values for 236 countries and areas at benchmark years 1990 through 2025, one figure per indicator table - the frame for global comparisons, not for growth modeling.

Global Nature Watch measures from orbit: umd_tree_cover_loss at 30 m annually for 2001-2025 with country and admin-2 summary tables attached. It evidences change, not standing volume - the two records agree on where forests are and disagree, productively, on what they are worth; the pairing is worked through on our tree cover loss by country page.

UK Forestry Statistics and Facts & Figures is the yearbook counterweight: planting series from 1971, removals from 1976, sawmill deliveries, prices and trade with England/Wales/Scotland/NI breakdowns. Everything money touches lives on the British side; everything biology touches lives on the American side. The scored head-to-head sits on the comparison page.

Luke in Finland shows what a national statistical database does instead: roundwood trade from 1949 across eight price regions - seventy-five years of stumpage structure, but no trees.

Route by the job: physical supply physics wants the American plot record; global benchmarks want FRA; harvest footprints want satellites; market context and prices want the yearbooks.

Ranked by how directly tree-level rows answer their day job:

1. Timberland investors and REIT analysts. Site index, stand age, ownership group and per-acre volume by county separate cheap acres from productive ones - the inputs behind every valuation memo. The workflow pattern sits on investors and quants x forest products.
2. Mill procurement and fiber-supply teams. TPO removals and mill residues by county show where fiber already moves before bidding on supply contracts, and flag regions where harvest outran regrowth.
3. Bioenergy and pellet-market analysts. The fuelwood product code read against the mill-residue removal class decides whether a plant pencils; residue markets usually beat roundwood prices to that answer.
4. Data scientists. A national repeated-measures plot panel is rare ground truth for spatial models and calibration sets - see data scientists x forest products.
5. Academics, journalists and consultants. Citation-grade numbers from a named federal program with a documented protocol outrank trade-association estimates; patterns live on our market sizing and citation-grade research pages.

The USDA Forest Service FIA DataMart and Timber Products Output (TPO) product page carries the full field dictionary, captured sample rows and coverage detail. For the rest of the shelf - production and trade cubes, lumber prices, satellite monitoring, certificate registries - start with the forest products data guide and the forest-products data hub.

Every record here ships from Datadory as typed rows with dictionaries attached, codes resolved to labels and stable join keys - delivered daily, weekly, or hourly, your call. Name the states, species groups, removal classes and years you need and get a sample; real rows come back first.