Why is public forestry data so hard to use?

Mostly because it is built to answer a different question from the one an operator asks. The main national forest inventory we look at here is designed to estimate whole populations: the timber volume, growth and removals of a region. To win access to private land and protect the owners who grant it, it deliberately blurs where its plots are. Most plot coordinates are moved within half a mile, the rest up to a mile, and some private plots are swapped with similar ones in the same county. Each plot is remeasured every 5 to 10 years. Below county scale the data are designed not to locate anything. At county scale, the inventory's own published sampling errors for growth and removals are often wide.

What the usual answer says

The usual answer is about formats and access. A 2024 paper by Vega-Gorgojo and colleagues puts it well. Governments, they write, "typically employ disparate data formats (sometimes proprietary ones) and published datasets are commonly disconnected from other sources, including previous versions of such datasets." Using one inventory's raw files, they note, "requires a suitable computing environment with a Microsoft Access license, non-trivial database skills, and good knowledge of the schema and ID codes". That is real. It suggests better portals would fix the problem, and for the inventory's plots, much of the difficulty is deliberate.

Blurred on purpose

The inventory's spatial data services page, last updated in August 2025, explains why. Plots are located by GPS, and ownership maps are public, so publishing exact plot locations would be "tantamount to revealing the owner's name and thus violating the law". The policy ensures that "data for any plot cannot be linked with certainty to the participating private landowner".

So the coordinates are altered. The "fuzzing" procedure relocates most plot coordinates "within one-half mile of their actual coordinates, with the remainder relocated up to 1 mile". That means "the actual plot location is generally masked within a 500-acre area". On private forest land, some plots are also swapped with similar plots in the same county, "for a total swapping of between 0 and 25 percent". The stated aim is to keep the "ecological signal" while introducing "enough uncertainty to decouple the plot-landowner relationship". Known locations could also tempt people to alter conditions on the plot, harming "the integrity of data that are collected the next time the plot is measured (in 5 to 10 years)".

Access to the true locations is by agreement, and the page warns that approval "will take several months and may not be approved". At the time of writing it also says: "Due to the recent reduction of the federal workforce, we cannot process requests for confidential information at this time."

What that means in practice

A 2025 study by Cao and colleagues used the inventory to validate a commercial satellite biomass product. At county level, the two agreed closely, with an R² of 0.90. But the authors note that "the spatial resolution of the validation was limited to aggregated hexagon and county scales, owing to the location perturbation of FIA plots". Their hexagons were 64,000 hectares each. If researchers validating a national product must work at that scale, an operator checking one property can't use the plots directly.

County scale has its own limits

County estimates come with a published sampling error, and it is worth reading. The inventory's county estimates map service holds a 2019 edition, the latest it carries, with each estimate's "Percent Sampling Error (68% confidence level)". We downloaded it and counted. For net volume of live trees, the median county error was 24% across the 1,672 counties with an estimate. For annual growth it was 34%, across 1,542 counties. For annual removals it was 62%, across 1,324 counties, and 859 of those had an error of 50% or more.

A removals figure with a 62% error at 68% confidence is a direction, not a number. Supply studies built on single counties inherit that. Aggregating counties into a supply area narrows the error, which is what the inventory is designed for.

Other public series have their own limits, such as late releases and unpublished estimates. Eurostat's forestry metadata, for example, warns of delays of "up to 4 months late for wood products". Each is manageable once it is known.

Using it at the right scale

Use inventory estimates for supply areas large enough that the published error is acceptable, and quote the error with the figure. Keep the release each figure came from, since later releases revise earlier ones. For a single tract, public data adds context the tract's own records can't: whether the surrounding area is being cut faster than it grows, within an error you can state.

Where AI helps

AI helps with the mechanics: pulling tables, reading method notes, keeping each figure with its sampling error and release date, and aggregating counties until the error falls below a threshold you set. It shouldn't be pointed at recovering true plot locations, because the fuzzing and swapping exist to keep the inventory within the law that protects its landowners.

When it doesn't apply

Analysts with an approved agreement for the confidential data can work at finer scales, when such requests are being processed. Other countries' inventories follow their own rules, which we haven't checked here. For regional supply analysis, the public data is already at about the right scale.

Quarri for forest management is built around how a forest operation runs, from the cruise to the settled account.

Quarri is an AI-native data platform for the timber supply chain. It connects buying, production, sales and inventory for forest management, sawmill, wood products and pulp, paper and packaging operators.