Written and maintained by CASRAI Editorial Board
Published
On this page
The Forest Inventory and Analysis (FIA) DataMart is the official public data clearinghouse maintained by the USDA Forest Service for downloading nationwide forest inventory data across all ownerships in the United States. It provides standardized tabular extracts, state-level SQLite databases, and analytical tool inputs from the Forest Inventory and Analysis Database (FIADB).
Operating as the primary distribution hub of the national FIA program, the repository enables biometricians, ecological modelers, and resource analysts to evaluate forest extent, composition, volume, growth, mortality, and ownership across more than 700 million acres of forested land.
Definition, operational scope, and administrative foundation
The Forest Inventory and Analysis DataMart distributes data from the continuous forest census of the United States. Mandated by the Forest and Rangeland Renewable Resources Research Act of 1978 and subsequent Farm Bills, the FIA program conducts field inventories across all 50 states, Puerto Rico, the U.S. Virgin Islands, and Pacific territories. Administered through the Research and Development branch of the USDA Forest Service, the platform provides open public access to the FIADB.
In contemporary forestry-ecology research, the DataMart functions as empirical ground truth for landscape-scale ecological modeling. Unlike localized studies, the FIA program delivers an unbiased, probabilistic sample across federal lands, state reserves, tribal lands, corporate timberlands, and private family forests. The national protocol ensures uniform variable definitions and measurement standards across all ownership categories.
The operational scope of the repository encompasses five core measurement dimensions:
- Forest area dynamics: Tracking shifts between timberland, reserved forest land, unreserved non-timberland, agricultural margins, and urban boundaries.
- Growing stock volume: Volumetric calculations across species based on merchantable height standards and regional volume equations.
- Carbon stocks and biomass: Quantification of live aboveground biomass, belowground roots, standing dead trees, and forest floor litter.
- Stand structure: Diameter distributions, basal area densities, canopy cover proportions, stand age profiles, and regeneration stocking.
- Disturbance agents: Monitoring insect damage, disease epidemics, wildfire severity, weather events, and timber harvesting intensity.
How the FIA DataMart works in practice: data formats and access modalities
The FIA DataMart organizes holdings geographically by state and territory, alongside consolidated national tables for multi-state computational pipelines.
The clearinghouse provides inventory files through three primary distribution formats:
- Comma-separated value (CSV) archives: State directories contain compressed ZIP archives housing individual CSV files for every FIADB table (such as
AL_PLOT.csv,AL_COND.csv, andAL_TREE.csv). These tabular files support batch ingestion into R, Python, SAS, or database engines. - Pre-packaged SQLite databases: To eliminate manual relational assembly, the DataMart provides self-contained, fully indexed SQLite database files for individual states. A single SQLite container bundles the relational schema, foreign key links, and indexes, enabling immediate querying without database server setup.
- Forest Vegetation Simulator (FVS) input files: Specialized data files formatted for the Forest Vegetation Simulator, allowing biometricians to import stand structures directly into individual-tree growth projections.
To integrate FIA DataMart assets into research workflows, analysts follow a four-stage retrieval protocol:
- Geographic and temporal scoping: Identify target states and verify inventory cycles or annual panels across measurement years.
- Asset download and archive extraction: Download corresponding state-level ZIP packages or SQLite containers from the official portal.
- Relational integrity verification: Inspect primary keys (
CN) and foreign key linkages (PLT_CN) across downloaded tables to confirm parent-child record integrity. - Local environment indexing: For CSV pipelines, import tables into an analytical database engine, establishing indexes on state codes (
STATECD), plot numbers (PLOT), and measurement years (INVYR).
Relational architecture and core tables of the FIADB schema
The Forest Inventory and Analysis Database uses a normalized relational hierarchy where observations progress from landscape-level plot locations down to individual stems and subplots.
The core tables that form the foundation of any FIADB query include:
- The PLOT table: Represents the physical sampling location and survey event. Each record corresponds to a permanent plot visit, identified by control number (
CN). Attributes include inventory year (INVYR), state (STATECD), county (COUNTYCD), and plot status (PLOT_STATUS_CD). - The COND table (Condition Class): Captures stand characteristics within the plot. If a plot straddles distinct forest types, ownerships, or land use boundaries, crews split the plot into condition classes. Recorded variables include forest type (
FORTYPCD), stand age (STDAGE), and ownership group (OWNGRPCD), linking to plots viaPLT_CN. - The SUBPLOT table: Documents the four 24.0-foot fixed-radius sampling stations comprising the ground cluster, recording slope gradient, aspect, and condition proportion mappings.
- The TREE table: Contains measurements for live and dead trees (DBH 5.0 inches or greater) tallied on subplots. Attributes include botanical species (
SPCD), DBH, height, tree status, merchantable volume, and dry biomass. Records link to plots viaPLT_CNand to conditions viaCONDID. - The SEEDLING table: Tracks regeneration on nested 6.8-foot radius microplots, sampling saplings (1.0 to 4.9 inches DBH) and seedlings (stems under 1.0 inch DBH) by species and condition.
The following table compares the operational trade-offs across access tools and analytical platforms supporting the FIA DataMart:
| Access Tool or Platform | Delivery Interface | Computational Skill Level | Primary Strength | Analytical Limitation |
|---|---|---|---|---|
| FIA DataMart Web Directory | Direct HTTP and Browser download | Beginner to Intermediate | Provides direct access to complete state CSV archives and SQLite files | Requires manual file management and local relational database assembly |
| Pre-Packaged SQLite Databases | Local relational database file | Intermediate (SQL and Python) | Pre-indexed relational structure with zero server configuration | Large file sizes for multi-state regions; requires local disk allocation |
| USFS EVALIDator (Web and API) | Browser interface and REST API | Beginner (Web) to Advanced (API) | Automatically computes official design-based population estimates | Pre-defined aggregation parameters; limited flexibility for custom models |
| rFIA Package (R CRAN) | Programmatic R library interface | Intermediate to Advanced | Automates design-based estimation and spatial analysis using FIADB | High memory consumption when executing multi-state national queries |
| Forest Vegetation Simulator (FVS) | Command-line and GUI simulation tool | Advanced (Biometrics) | Projects individual-tree growth, mortality, and harvest management | Focuses on future growth simulations rather than historical synthesis |
For institutions maintaining an enterprise data warehouse, ingesting these core tables establishes continuous ecological baseline tracking. Aligning tabular structures with a standardized metadata schema ensures cross-decadal comparability across research teams.
Spatial coordinate fuzzing and swapping: rules and analytical boundaries
Under Section 1770 of the Food Security Act of 1985 (7 U.S.C. 2276) and subsequent privacy provisions in the 1998 Farm Bill amendments, the USDA Forest Service is legally prohibited from disclosing exact coordinates of field plots on private lands. Confidentiality protections are enforced to safeguard landowner privacy and prevent commercial exploitation.
To comply with federal privacy mandates while preserving data utility for regional ecological research, the Forest Service applies a dual-layer perturbation algorithm to private plot locations:
- Coordinate fuzzing: Coordinates published in the
PLOTtable (fieldsLATandLON) are perturbed using a random directional displacement. The published coordinate is shifted within a defined radius, typically up to 0.5 miles for private plots, and up to 1.0 mile for sensitive locations, remaining within the actual ground plot county. - Coordinate swapping: To prevent reverse-identification of landowners owning prominent forest tracts, up to 20 percent of private plot coordinates are swapped with coordinates of another private plot in the same county sharing similar forest types and stand conditions.
While public plots on National Forest System lands generally retain true field coordinates, fuzzed and swapped records on private lands define distinct analytical boundaries:
- Valid analytical applications: County-level aggregations, regional timber supply analyses, statewide ecological assessments, ecoregion climate vulnerability evaluations, and broad-scale monitoring remain statistically sound because coordinates remain confined within their respective counties and strata.
- Invalid spatial procedures: Extracting pixel values from high-resolution remote sensing imagery (such as 10-meter Sentinel-2 or 30-meter Landsat rasters) using published public coordinates introduces spatial mismatch errors. Overlaying fuzzed coordinates against parcel property boundaries or municipal zoning lines yields erroneous conclusions. Researchers requiring true coordinates must submit a formal research proposal to the Forest Service to work in a secure Spatial Data Services (SDS) environment under a non-disclosure agreement.
Design-based population estimation and sampling weight calculations
Calculating unweighted arithmetic means across FIA plot or tree records introduces severe sampling bias because the sampling frame is not a simple random sample across uniform terrain. The national design operates on an enhanced three-phase probabilistic framework formalized by Bechtold and Patterson (2005) in General Technical Report SRS-80.
The three operational phases of the FIA sampling design include:
- Phase 1 (Stratification): Utilizes remote sensing imagery and land cover classifications to stratify the landscape into homogeneous strata, reducing sampling variance.
- Phase 2 (Ground Inventory): Consists of the standard permanent ground plot grid, spaced systematically at approximately one plot per 6,000 acres (~125,000 active plots nationally). Plots are measured on rotating annual panels: typically a 5-year cycle (20 percent per year) in the East and a 10-year cycle (10 percent per year) in the West.
- Phase 3 (Forest Health Indicators): A subset of Phase 2 plots (~1 in 16 plots) measured for intensive biophysical health indicators, encompassing soil chemistry, lichen diversity, crown condition, and down woody material transects.
To produce unbiased population totals, analysts utilize the population evaluation tables embedded within the FIADB schema:
POP_EVAL: Defines an evaluation group, representing a specific analytical inquiry (such as total forest volume in a state for a reporting year) and identifying valid annual panels.POP_ESTN_UNIT: Delineates estimation units (frequently counties or county groups) within the evaluation group.POP_STRATUM: Contains stratification weights (STRATUM_WEIGHT), stratum areas, and expansion factors used to scale individual plot measurements to landscape-level totals.
These design-based estimation formulas guarantee that empirical inventory conclusions support defensible forest management strategies and regional timber availability assessments.
Programmatic querying with rFIA, Python, and SQL pipelines
Modern analytical pipelines frequently automate data retrieval directly from the FIA DataMart to ensure version control, reproducible research, and scalable multi-state analysis.
In the R ecosystem, the rFIA package (developed by Hunter Stanke, Jeffrey Doser, and collaborators, available on CRAN) serves as the primary open-source toolkit for processing FIADB assets. The package automates database downloading, schema construction, and the implementation of Bechtold and Patterson design-based estimation algorithms. Core functions handle complex sample weighting automatically:
getFIA(): Connects directly to the FIA DataMart, downloading specified state archives and saving them as local CSV or SQLite databases.readFIA(): Loads local database files into a structured in-memory representation, optimizing data types and establishing table relationships.tpa(): Calculates trees-per-acre and basal area per acre estimates across user-defined domain groupings.biomass(): Estimates aboveground and component biomass pools and carbon stocks along with corresponding sampling errors.volume(): Generates merchantable net volume, gross volume, and sawtimber volume estimates across geographic boundaries.
In Python environments, data engineers utilize libraries such as sqlite3 and pandas to query DataMart SQLite databases directly. For example, an analytical join linking plot metadata, condition characteristics, and individual tree biomass can be executed using concise SQL syntax:
SELECT
p.STATECD,
p.INVYR,
c.FORTYPCD,
c.STDAGE,
t.SPCD,
t.DIA,
t.DRYBIO_AG
FROM PLOT p
JOIN COND c ON p.CN = c.PLT_CN
JOIN TREE t ON c.PLT_CN = t.PLT_CN AND c.CONDID = t.CONDID
WHERE p.PLOT_STATUS_CD = 1
AND t.STATUSCD = 1
AND p.INVYR = 2024;
This direct querying approach integrates smoothly into professional timber cruise compilation tools and institutional forest carbon accounting platforms, enabling rapid synthesis across millions of individual tree measurement records.
Where forest inventory data matters for research facilities and environmental monitoring
The applications of FIA DataMart holdings extend far beyond classical timber harvest projections. Modern environmental monitoring facilities, academic laboratories, and sustainability groups rely on FIADB data as a fundamental operational benchmark:
- Carbon offset verification: Carbon markets require verifiable baseline estimates of forest carbon sequestration. FIA DataMart records establish historical carbon densities and growth-to-drain ratios, serving as independent baselines against which carbon offset projects are audited.
- Remote sensing calibration: Earth observation programs, including spaceborne LiDAR and radar missions, utilize ground-measured FIA tree heights, canopy cover estimates, and biomass calculations to train and validate spaceborne predictive algorithms.
- Wildfire risk characterization: Fire behavior modeling systems integrate FIA Phase 3 down woody materials data, canopy base heights, and stand density indices to simulate fire spread, crown fire potential, and smoke emissions under changing climatic scenarios.
- Biodiversity assessments: Long-term remeasurement records track native species migrations, regeneration failures, mortality pulses associated with drought, and the geographic spread of invasive forest pathogens across broad ecological gradients.
The structured tabular data provided by the DataMart is summarized in the reference dataset below, outlining the primary tables, relational keys, and analytical functions that underpin the national database architecture.
Common misunderstandings and analytical pitfalls
Working effectively with the FIA DataMart requires avoiding several common conceptual errors:
- Conflating forest land with timberland: Timberland is a commercial subset of forest land producing in excess of 20 cubic feet per acre annually of industrial wood, excluding reserved areas like Wilderness. Treating all forest land as timberland severely inflates harvestable resource projections.
- Mixing annual inventory panels across incompatible cycles: Because annual inventories measure rotating panels, pooling plots from separate cycles without evaluation weights (
POP_EVAL) creates temporal inconsistency. Queries should always group by designated survey evaluations. - Treating fuzzed coordinates as true plot centers: Attempting to link high-resolution drone or pixel data directly to published latitude and longitude produces false correlations due to intentional coordinate displacement. Spatial analyses must remain aggregated at landscape scales.
- Omitting status code and cull deductions: The
TREEtable distinguishes between live trees (STATUSCD = 1), dead stems (STATUSCD = 2), and removals (STATUSCD = 3). Querying diameters without status filtering or sound volume equations inflates commercial wood estimates. - Overlooking historical regional protocol differences: Prior to national standardization in the early 2000s, regional stations utilized varying sampling designs. Long-term temporal analyses spanning pre-2000 data must account for these variations using regional conversion documentation.
Frequently asked questions
What data can you download from the FIA DataMart?
The FIA DataMart provides public access to state and national tables of the Forest Inventory and Analysis Database. Available assets include CSV files for all tables (PLOT, COND, TREE, SEEDLING), pre-packaged state SQLite database containers, Forest Vegetation Simulator input files, desktop EVALIDator tools, and state analytical reports.
How do you query and download FIA DataMart tables?
Users download data directly via the USDA Forest Service web interface by selecting specific states and table packages, or programmatically in R and Python. The rFIA package automates downloads via getFIA(), saving state datasets directly to local working directories for analysis.
What is the difference between FIA DataMart and EVALIDator?
The FIA DataMart distributes raw relational tables and database containers for local custom queries. EVALIDator is an analytical calculation tool and API that queries the FIADB database to produce official design-based population estimates and standard errors without requiring manual SQL joins.
Why are FIA plot coordinates fuzzed and swapped?
Under Section 1770 of the Food Security Act of 1985 and Farm Bill amendments, the Forest Service must protect private landowner confidentiality. Coordinates for private plots are fuzzed by up to 0.5 to 1.0 miles, and up to 20 percent of private coordinates are swapped within the same county.
How does rFIA simplify working with the FIADB database?
The rFIA package handles design-based sampling weights and post-stratified estimators automatically. Instead of writing complex SQL joins across PLOT, COND, TREE, POP_EVAL, and POP_STRATUM, users execute functions like tpa(), biomass(), and volume() to obtain population totals.
What is the difference between timberland and forest land in FIA data?
Forest land is land at least 10 percent stocked by trees of any size spanning at least 1.0 acre with 120-foot width. Timberland is unreserved forest land capable of producing at least 20 cubic feet per acre per year of industrial roundwood.
Sources
- USDA Forest Service Research and Development: FIA DataMart, checked 2026-11-20
- USDA Forest Service Research and Development: The Forest Inventory and Analysis Database User Guide (NFI), checked 2026-11-20
- USDA Forest Service Research and Development: Nationwide Forest Inventory Field Guide, checked 2026-11-20
- USDA Forest Service Research and Development: The enhanced forest inventory and analysis program – national sampling design and estimation procedures, checked 2026-11-20
- Comprehensive R Archive Network: CRAN: Package rFIA, checked 2026-11-20
- GitHub: doserjef/rFIA: rFIA · GitHub, checked 2026-11-20
Data: Core Tables and Architecture of the Forest Inventory and Analysis Database
A structured compilation of primary tables within the USDA Forest Service FIADB schema (v9.4), specifying record granularity, relational hierarchy, key identifiers, and principal ecological metrics compiled from official USFS documentation.
Period covered: 2000-2026. Coverage: United States and Territories.
| Table Name | Hierarchy Level | Primary Key | Foreign Key Links | Sampling Unit | Core Variables Measured |
|---|---|---|---|---|---|
| PLOT | Plot level | CN | None (Top-level table) | 1/6-acre cluster (4 subplots) | Latitude, longitude, cycle, measurement year, plot status, county code |
| COND | Condition level | CN | PLT_CN -> PLOT.CN | Condition within plot (min 1.0 acre) | Forest type, stand age, ownership class, stand size class, disturbance code |
| SUBPLOT | Subplot level | CN | PLT_CN -> PLOT.CN | 24.0-foot fixed-radius subplot | Slope, aspect, condition proportion, snow/water depth |
| TREE | Individual tree level | CN | PLT_CN -> PLOT.CN, CONDID -> COND.CONDID | Tally tree (>= 5.0 in DBH) | Species code, DBH, height, tree status, net volume, biomass, defect, damage |
| SEEDLING | Regeneration level | CN | PLT_CN -> PLOT.CN, CONDID -> COND.CONDID | 6.8-foot radius microplot | Species code, seedling count, condition proportion |
| POP_EVAL | Population evaluation | CN | None (Survey metadata) | State/Cycle evaluation group | Evaluation type (volume, area, growth), inventory year, evaluation sequence |
| POP_STRATUM | Stratification level | CN | EVAL_CN -> POP_EVAL.CN | Remote sensing stratum polygon | Stratum weight, pixel count, area expansion factor |
Sources
- The Forest Inventory and Analysis Database User Guide (NFI) (accessed 2026-11-20)
- The enhanced forest inventory and analysis program - national sampling design and estimation procedures (accessed 2026-11-20)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







