# Krummholz Forestry datasets - PRELIMINARY

Release date: 2026-10-22. Derived from the previously acquired release-candidate snapshot.

Maine forest inventory measurements for exploratory models of tree dimensions, biomass, carbon and forest condition. The two tree source parts are combined before deduplication. All available years are retained.

## Contents and audit

Table | Input observations | Excluded | Retained observations | Unique published feature rows
--- | ---: | ---: | ---: | ---:
trees | 1,066,726 | 0 | 1,066,726 | 880,126
conditions | 28,529 | 0 | 28,529 | 23,128

A CSV row is a unique normalized feature vector. `record_count` records how many retained observations it represents. Duplicate vectors are stored once; legitimate repeated measurements are not silently discarded. Do not sum counts from different tables as if they were independent people, fish or trees.

All parts have their own header. Concatenate the data rows of all tree parts to reconstruct the deduplicated tree table. Keep conditions, trips, catch and size tables separate. `schema.json` defines exported columns; `audit.json` contains exact removal lists, missingness, row reconciliation and source/output checksums.

## Cleaning and privacy

Coordinates (including publicly fuzzed coordinates), county/unit/plot/tree/condition identifiers and link keys, ownership, exact visit or administrative timestamps, and site-location details are omitted. Plot and population-estimation tables are excluded because their primary role is location, linkage or survey administration. Biological condition and tree records remain separate feature tables.

Whitespace and equivalent numeric spellings are standardized without rounding. Blank and documented missing-token spellings become empty CSV cells. No synthetic values or fabricated labels are introduced. Source files remain unchanged in private staging; original row-level identifiers and rejected records are not bundled.

Columns with no observed values are omitted: trees: BHAGE, TOTAGE. The species codebook appears in both download packages. Short source-definition excerpts identify original manual sections; consult the linked guide for full code tables and footnotes.

## PRELIMINARY limitations

FIA is a repeated multi-year sample, not a current tree census. Original plot/tree relationships are deliberately removed, so the release cannot reproduce design-based estimates or support longitudinal record joins. Repeated measurements can remain correlated. For model evaluation, use temporal holdouts and domain review rather than claiming independent random-row performance. Legacy definitions, derived biomass/carbon targets and missingness vary across inventory years.

The release removes direct identifiers and fine location fields, but is not a formal anonymity guarantee, domain-expert quality certification, or validated production model dataset. Numeric checks do not establish that every measurement or code is scientifically correct. Choose task-specific targets and predictors carefully: calculated volumes, biomass and carbon can leak information when used together.

## Attribution and source terms

USDA public-use data - CC0 identified in USDA’s dataset record

The USDA Ag Data Commons Forest Inventory and Analysis database record identifies its license as CC0 and links the dataset DOI to the FIA DataMart, the source of the raw Maine CSV downloads used here.
USDA dataset record: https://agdatacommons.nal.usda.gov/articles/dataset/Forest_Inventory_and_Analysis_database/27011560
CC0 deed: https://creativecommons.org/publicdomain/zero/1.0/
Dataset DOI: https://doi.org/10.2737/RDS-2001-FIADB

Metadata conflict (verified 2026-10-22): the separate Data.gov FIADB catalog lists CC BY 4.0, including for its DataMart distribution. This release follows the CC0 identification in USDA's dataset record and records the conflicting catalog statement rather than presenting CC BY 4.0 as the sole source license. This is source-license metadata, not a new license assigned to the upstream data or a separate CC BY license for Krummholz's additions.
Data.gov catalog: https://catalog.data.gov/dataset/forest-inventory-and-analysis-database
CC BY 4.0 deed cited by that catalog: https://creativecommons.org/licenses/by/4.0/

Source attribution: U.S. Department of Agriculture, Forest Service, Forest Inventory and Analysis program.
Suggested source citation: USDA Forest Service. 2019. Forest Inventory and Analysis database. St. Paul, MN: USDA Forest Service, Northern Research Station. https://doi.org/10.2737/RDS-2001-FIADB

Krummholz modified the source by selecting features, removing location/linkage fields and entirely missing columns, normalizing missing/numeric values, and compacting repeated feature vectors with observation counts. The accompanying schema and audit describe these changes. No agency endorsement is implied.


Verified official sources (2026-10-22):

- USDA Forest Service Forest Inventory and Analysis: https://research.fs.usda.gov/programs/fia
- FIA DataMart: https://apps.fs.usda.gov/fia/datamart/datamart.html
- USDA Ag Data Commons FIADB dataset record identifying CC0: https://agdatacommons.nal.usda.gov/articles/dataset/Forest_Inventory_and_Analysis_database/27011560
- FIADB dataset DOI resolving to FIA DataMart: https://doi.org/10.2737/RDS-2001-FIADB
- Data.gov FIADB catalog with conflicting CC BY 4.0 metadata: https://catalog.data.gov/dataset/forest-inventory-and-analysis-database
- FIA plot privacy and public coordinates: https://research.fs.usda.gov/programs/fia/sds
- FIADB User Guide version 9.4: https://research.fs.usda.gov/sites/default/files/2025-08/wo-v9-4_Aug2025_UG_FIADB_database_description_NFI.pdf