TWIG Data Resources

Quick answer: The TWIG Viewer maps TWIG's treatments alongside wildfire perimeters and summarizes them by state, county, congressional district or federal land unit, with no GIS software needed.

• TWIG Viewer. Downloads from the Viewer are capped at 10,000 records. Video tutorials are on the Program Resources page.
• Treatment index REST API: an ArcGIS feature service of the treatment index. It returns up to 1,000 records per request and has no published rate limit, so larger queries need paging; ArcGIS Pro, QGIS and most R and Python libraries page automatically.
• Layer details on ArcGIS.

Quick answer: Download up to 10,000 records from the TWIG Viewer, pull up to about 100,000 through the REST API, or download the full treatment index as one zipped Esri file geodatabase (about 1.3 GB, updated weekly).

Pick a route by how many records you need:

• Up to 10,000 records: select and download them in the TWIG Viewer. Viewer downloads are capped at 10,000 records.
• 10,000 to about 100,000 records: use the REST API (see Explore TWIG online).
• More than about 100,000 records: download the full treatment index (treatment_index.zip), about 1.46 million records as of November 2026.

Why these break points? The API returns up to 1,000 records per request and has no published rate limit (as of November 2026, this may change in the future). In tests, each request for 1,000 records with their shapes took about 2 seconds and returned about 10 MB of JSON. At 100,000 records that comes to about 100 requests and roughly 1 GB of data, close to the size of the whole compressed download (about 1.3 GB). If you don’t need the shapes, request attributes only; the responses are much smaller.

How many records to expect (as of Nov 4, 2026):

“All years” means 1984 to the present, plus planned treatments; date ranges use treatment_date. State counts use the state field. County counts include every polygon that touches the county.

State and NGO records are Viewer-only for now. About 100,000 records from state agencies and NGOs, shared through the National Fuels Treatments Initiative (NFT), appear in the Viewer but are not included in downloads. Our current data-sharing agreements don’t allow us to redistribute them. As of November 2026, we are currently working on a new data-sharing agreement and expect most states’ records to become downloadable through TWIG in the coming months. You can also explore these records in the NFT Explorer.

Quick answer: All code that builds TWIG is open source on GitHub under the GNU GPL v3. Each source has its own download and insert steps. FACTS Common Attributes gets by far the most filtering, and its inclusion rules are in treatment_index/processing.py.

To build your own version of TWIG from the code, start from the schema and lookup tables: an empty file geodatabase that defines the treatment index fields and the lookup tables the code uses. It contains no treatment records.

Quick answer: Please cite our data descriptor paper in any publication that uses TWIG data, and give the date you downloaded the data.

Call, A., Tomczyk, N., Withnall, K.A., Dappen, P., Heusinkveld, D., Mueller, S.E., Holloway, B., Shennan, K., Herring, J.A., Franko, A., Colavito, M.M., Kimple, A.D., Sánchez Meador, A.J. & Stevens-Rumann, C.S. A new geodatabase of fuel treatments across federal lands in the USA. Sci Data 12, 1485 (2025). https://doi.org/10.1038/s41597-025-05859-z

The paper (published Aug 26, 2025) explains why and how TWIG was built. It describes the July 2025 version, which held FACTS and NFPORS records only. To reproduce the paper’s results, use the archived snapshot on Zenodo, TWIG_treatment_index_20250709.zip (doi:10.5281/zenodo.15851622). For current data and documentation, use this page.

License: TWIG data are licensed under CC BY 4.0; TWIG code is licensed under the GNU GPL v3.

Quick answer: Each week, TWIG downloads records from each source system, keeps the ones that describe fuel-related work, maps them onto one shared set of fields, flags likely errors, and publishes the result.

The treatment index is the heart of TWIG, and the place where fuel treatment data is kept. It is a single table where each record is one management activity with one polygon (or multipart polygon) and its attributes: what was done, when, by whom, and, where the source records them, funding and estimated cost. TWIG keeps each record as its source system has it. It doesn’t correct errors; it flags them (see Error flags).

The weekly build runs in six steps:

1. Download. FACTS Hazardous Fuels (one Esri file geodatabase) and FACTS Common Attributes (nine Esri file geodatabases, one per Forest Service region), IFPRS from its feature service, and state and NGO records from the NFT. The NFPORS service is offline, so TWIG reuses the latest available copy (records last modified Feb 20, 2025).
2. Filter. Keep records that have a polygon with some area. For federal sources, keep records dated 1984 or later. From FACTS Hazardous Fuels, drop records whose category is “Wildfire Non-Treatment.” From FACTS Common Attributes, keep only records that pass TWIG’s inclusion rules and aren’t already in FACTS Hazardous Fuels. From IFPRS, keep fuels records classed as actual or estimated treatments. From NFT, TWIG receives state and NGO records only (federal records come straight from FACTS and IFPRS) and keeps those with a completion date.
3. Crosswalk. Map each source’s fields onto the treatment index fields, and record the source in the identifier_database field.
4. Clean and classify. Trim stray spaces, fix known typos (e.g. “Biomass Removall”), convert state names to two-letter codes, group fund codes into fund_source, calculate total_cost, and assign twig_category and fallon_category.
5. Flag. Mark duplicates, extreme costs, unusual units, very large polygons, polygons outside their stated state, and simplified or buffered shapes in the error field.
6. Publish. Update the Viewer and API, then write the downloadable file geodatabase (the file geodatabase contains data from federal sources only, for now).

No point or line records come from federal sources. Some state records began as points or lines, and NFT converted them to polygons before TWIG received them (see Error flags).

TWIG is a weekly snapshot, not a real-time feed. Each source system publishes on its own schedule, and TWIG rebuilds once a week. A record edited in FACTS today may take a week or more to reach TWIG, because TWIG reads the copies each source publishes, not the live systems. For example, at the Oct 27, 2026 sync, IFPRS was current to that day, but the newest FACTS edits were from Oct 15–16; the Forest Service republishes its public FACTS files on its own schedule. NFT updates each state’s data on schedules that range from near real-time to once a year. If a weekly build fails, changes wait for the next one. The date_current field shows when each record last changed in its source, as of TWIG’s latest sync.

Quick answer: TWIG combines six record sets from four systems. The Forest Service’s FACTS supplies most records, IFPRS and an archived copy of NFPORS cover Interior agencies, and the NFT supplies state and NGO records. The identifier_database field says where each record came from.

Programming Tip: To keep only the records each agency itself reports as fuel treatments, drop TWIG’s extra FACTS records: identifier_database <> 'FACTS Common Attributes'

Quick answer: Use twig_category to group any TWIG record by how the work was done: fire, machines or hand tools, chemicals, or animals. Use fallon_category (new, in beta) to group records by what the work did to fuels. The other classification fields come straight from the source systems and only mean the same thing within one source.

Every source system describes treatments its own way. FACTS uses hundreds of activity names, IFPRS and NFPORS use their own category and type lists, and each state uses its own terms. TWIG keeps every source’s labels and adds two fields that work across sources.

Which classification field should I use?

The Treatment Index includes some wildfires, but it is not a wildfire database. About 66,000 records are wildfires (twig_category = 'Unplanned Ignition'), almost all from FACTS Common Attributes. TWIG includes a wildfire only when the reporting agency recorded it as a management activity, in effect counting it as a treatment, such as a fire managed for resource benefit or a planned treatment that burned in a wildfire. FACTS also records some wildfires by ignition source, natural or human. Agencies count wildfire acres differently, and their practices have changed over time. If you’re counting treated acres, decide whether these wildfires belong, and filter on twig_category. For a complete record of wildfire occurrence, use the wildfire perimeter layers in the TWIG Viewer, which come from the National Interagency Fire Center (NIFC) and other sources.

Programming Tips

• No wildfires: twig_category <> 'Unplanned Ignition'
• Prescribed and pile burning: twig_category = 'Planned Ignition'
• Drop activities that likely didn’t change fuels: fallon_category IS NULL OR fallon_category <> 'Non Fuel Treatment' (keeps state and NGO records, which have no fallon_category yet)

Quick answer: treatment_date is the completion date when the work is done; otherwise it’s the best “planned” or “general” record date the source has. status says which kind of date it is: Completed, Planned, Started or Other. For trends over time, filter to status = 'Completed' and summarize by calendar year.

For planned FACTS Common Attributes records, treatment_date is the date the record was created, not a planned work date.

How accurate are the dates? People enter completion dates by hand without any additional checking mechanisms. A completion date usually marks the last day of work, which may come weeks or months after most of the work was done. When the exact date is unknown, some managers may enter the end of the fiscal year (Sept 30). Planned dates can sit years or decades in the future. Summarizing by year rather than by day reduces these problems.

Quick answer: TWIG never edits or deletes source records. It writes warnings into the error field instead, separated by semicolons when a record has more than one. About 87,000 records (5.6%) carry at least one flag. Flag names use underscores, such as DUPLICATE_DROP.

Programming Tips

The following examples show code snippets that can be used for filtering error flags in SQL, R (dplyr) and Python (pandas or geopandas). Flag names must match exactly, underscores included.

SQL. A WHERE clause. Paste it into a definition query in ArcGIS Pro or QGIS, the REST API’s where parameter, or a query against a SQL database.

-- Drop duplicate copies
error IS NULL OR error NOT LIKE '%DUPLICATE_DROP%'
-- Records safe for cost totals
error IS NULL OR (error NOT LIKE '%HIGH_COST%' AND error NOT LIKE '%CHECK_UOM%')

R (dplyr). Assumes the treatment index is loaded as a data frame or sf object named twig.

library(dplyr)
# twig <- sf::st_read("treatment_index.gdb", layer = "treatment_index")
# Drop duplicate copies
twig_nodup <- twig %>%
  filter(is.na(error) | !grepl("DUPLICATE_DROP", error, fixed = TRUE))
# Records safe for cost totals
twig_cost <- twig %>%
  filter(is.na(error) | !grepl("HIGH_COST|CHECK_UOM", error))

Python (pandas or geopandas). Assumes the treatment index is loaded as a DataFrame or GeoDataFrame named twig.

# import geopandas as gpd
# twig = gpd.read_file("treatment_index.gdb", layer="treatment_index")
# Drop duplicate copies
twig_nodup = twig[~twig["error"].fillna("").str.contains("DUPLICATE_DROP", regex=False)]
# Records safe for cost totals
twig_cost = twig[~twig["error"].fillna("").str.contains("HIGH_COST|CHECK_UOM")]

Quick answer: TWIG mirrors its sources, so errors in those sources are mirrored in TWIG. Automated flags catch some of them. Outside studies suggest that many FACTS polygons don’t match visible change on the ground and that some treatments never get recorded. As of November 2026, we are currently developing methods of checking TWIG against satellite imagery to provide independent validation of polygon, treatment date, and treatment type accuracy.

What TWIG checks automatically. As of November 2026, 65,615 records (4.2%) carry a duplicate flag, 34,899 of them DUPLICATE_DROP. Another 636 carry HIGH_COST and 845 carry SPATIAL. The duplicate check compares FACTS records only, so the same treatment recorded in two different systems isn’t caught. See Error flags for every flag and its limits.

Spatial accuracy and completeness. In a study of FACTS Hazardous Fuels data in California, trained observers compared FACTS polygons with changes detected in Landsat imagery. Only 54% of FACTS polygons matched a detected change, and only 79% of those matched closely (within 25% of total area). In the same analysis, 80% of CAL FIRE treatments were detectable. Not every treatment should be visible from space, but the gap points to both misplaced polygons and records of work that may not have happened (errors of commission).

Missing records (errors of omission) are harder to measure. A study of forest management in New Mexico compared regional and state records with records gathered from local offices. Missing treated area ranged from 0% to 57.4% across five watersheds and averaged 7.4%. Those figures cover federal, state and private records, so treat them as a benchmark rather than a TWIG estimate. Areas dominated by federal land are likely better covered in TWIG than areas that are mostly private or state-owned.

Polygon quality also varies with age and method. Some older records, especially before 2005, have poor spatial records. Newer ones are more often mapped by GPS or from imagery. Even when accurately mapped, activities like thinning, sanitation cuts and prescribed fires often change fuels on only part of the area drawn. TWIG treats every polygon the same, however it was drawn.

Costs. Cost accuracy is hard to judge: accounting practices differ among agencies and regions, and cost estimates aren’t reconciled with actual spending. Extreme values are the easiest errors to spot and the most damaging. On the Dixie National Forest in Utah, the record “Gum Hill Chaining Maintenance 22” lists $405,451 per acre across 1,729 acres, a total of about $701 million. One record like that can outweigh every other cost in a national forest’s summary. This record carries a HIGH_COST flag, but (as of November 2026) IFPRS and state records are not screened in the same way, and similar errors will not be flagged in those records. We strongly urge users to account for possible errors when calculating cost totals or averages.

New validation work. We are expanding TWIG’s validation in two directions. Remote sensing: we are comparing TWIG polygons and dates with disturbance detected in satellite imagery, to estimate how often recorded treatments show up on the ground and how well their outlines match. Classification accuracy: we are checking how well twig_category and fallon_category agree with expert review of records and imagery. We’ll post methods and results here as they’re ready.

Quick answer: TWIG works best for regional and national questions that span several years. It isn’t built for real-time decisions and isn’t usually the best tool for mapping single projects or stands.

Good uses

• Tracking where and when treatments happened across agencies and land ownerships
• Planning and prioritizing at landscape scale
• Research on treatment effects and on how treatments and wildfires interact
• Policy analysis of wildfire risk reduction

Limits

• Not real-time. TWIG is a weekly snapshot of systems that update on their own schedules. Don’t use it for emergency management or real-time decisions.
• Not fine-scale. Without checking against other data, don’t use TWIG for fine-scale mapping, planning or analysis, such as single treatment projects, stands or trees.
• Reliability grows with scale. Analyses of small areas or single years should be read with care.
• Errors are flagged, not fixed. Filter and check the data for your purpose (see Error flags).
• Polygons vary. Spatial accuracy varies, especially for older records.
• Costs are rough. Cost fields are estimates, mean different things by source, and are often missing.
• Coverage is uneven. Federal coverage is the most complete. State coverage depends on which states share data with NFT, and work on private land is mostly missing.

TWIG isn’t the right tool for every job; local or specialized tools may serve better. Similar resources are listed on our Other Resources page. TWIG grows with user feedback, so please tell us what you need.

Quick answer: Every TWIG record has the 26 fields below. The first table says what each field holds; the second shows which source field fills it; the panels underneath give details and cautions.

Download this reference: Excel workbook (fields, value lists and changes since October 2025) or CSV (fields only).

Where each field comes from

Quick answer: The most recent changes to TWIG’s data and documentation, newest first.

Quick answer: The TWIG Viewer lets you search, map and summarize TWIG treatments and wildfire perimeters in a web browser, with no GIS software needed.

The Viewer shows every TWIG record, including state and NGO records that aren’t yet in the downloadable file. Short video tutorials are on the Program Resources page. A full guide to the Viewer is coming soon.