itrdbMeasurementsClone

A copy of the tree-ring measurement data from the International Tree-Ring Data Bank (ITRDB), parsed into R objects you can load in one line.

Part of the openDendro project.

Crossdating reports for the whole archive. Every ring-width file here has a COFECHA-style report made by dplR's xdate.report(): 6,807 reports, 1.5 million segments tested. Start with the overview, or jump to an example.

Why this exists

The ITRDB is the archive of record for tree-ring measurements. It holds several thousand studies contributed over decades. Getting all of it into R, though, means downloading thousands of files, working out what each one measures, matching it to its study's metadata, and parsing a fixed-width format that has drifted over forty years of use.

Every project that wants to work across the whole database repeats that work. This repo does it once and publishes the result: two files that load in a second and give you every readable measurement file with its metadata attached.

That makes questions across the whole archive tractable — how sample depth varies by species, what the age structure of the network looks like, how many series cover a given period — without each researcher rebuilding the same pipeline.

What you get

File	What it holds
Rdatafiles/rwls.rds	A list of rwl objects, one per .rwl file in the ITRDB. Each is a data frame of measurements with years as row names and one column per series.
Rdatafiles/rwls_meta.rds	A data frame with one row per file: what the file measures, its study, location, elevation, and taxonomy.

The two line up by position: rwls[[i]] is the data for rwls_meta[i, ]. Both are keyed by file name without .rwl, so you can pull a file by name:

rwls[["ca671"]]
rwls_meta["ca671", ]

Not every file is ring width. The ITRDB also holds earlywood and latewood width, density, blue intensity and other measurements, often several files per study. variableShort says what each file holds, as NOAA records it. To get ring width:

rw <- rwls[rwls_meta$variableShort %in% "total ring width"]

Main rwls_meta columns: file, code (the ITRDB study code), variableShort, unit, studyName, Lat, Long, Altitude, AltitudeSource, scientificName, GenusSpp, Genus, Family, Order, Group, doi, landingPage.

Quick start

library(dplR)

rwls <- readRDS("Rdatafiles/rwls.rds")
rwls_meta <- readRDS("Rdatafiles/rwls_meta.rds")

# ring-width files only
rw <- rwls_meta$variableShort %in% "total ring width"
rwls_meta[rw, ][1, ]
aRWL <- rwls[rw][[1]]

rwl.report(aRWL)
summary(aRWL)
plot(aRWL)

Everything after that is ordinary dplR. To pull every ring-width file of a given species:

piab <- rwls[rw & rwls_meta$GenusSpp %in% "Picea abies"]
length(piab)

Crossdating reports

reports/correlation-stats/ holds a crossdating report for every ring-width file, made by dplR's xdate.report() in the layout of the COFECHA output NOAA publishes with each study. It shows the function running across the whole archive. It is not NOAA's published statistics.

• The overview summarises the run and lists the files that need a look: 1,278 have 10% or more of their segments flagged. Each continent's folder lists its own reports, most flagged first.
• Every report comes twice, as fixed-width text and as a self-contained HTML page, in txt/ and html/ under the same folder its .rwl has. For ca671.rwl: txt/northamerica/usa/ca671.txt and html/northamerica/usa/ca671.html. GitHub shows .html as source; download it to read it as a page.
• Every number is in reports/xdate-summary.csv, one row per file: series, segments tested, A and B flags, percent flagged, series intercorrelation, and any notes.

From a study code to its report in R:

s <- read.csv("reports/xdate-summary.csv")
x <- s[s$code == "CA671", ]
x[, c("file", "nSeries", "nA", "nB", "pctFlagged", "interseriesCor")]
file.show(file.path("reports/correlation-stats/txt", x$subdir, paste0(x$file, ".txt")))

A segment is flagged B when another position correlates better with the master than the dated one, and A when the dated position is the best tested but is under the critical value. xdate.report() departs from COFECHA in two places, on purpose. On a record too short for 50-year segments it shortens the segment length, and that report's notes say so (506 files). And it does not score partial segments at the ends of a series.

Current build

Built 1 October 2026 from the ITRDB as it stood then. Figures from code/04_readRWLs.R.

• 10,168 .rwl files from 6,862 ITRDB studies, of which 10,166 read
• 6,807 total ring width files, covering 6,781 studies
• In those ring-width files: 275,873 series and 53.4 million ring measurements
• 358 species
• Years spanned: 6000 BCE to 2025

Two files are not in the data. WY081 NOAA has temporarily withdrawn at the contributor's request. KYRG014 will not read: one of its series carries two different precision flags, so there is no single reading of it to hand back.

This build is not compatible with builds before October 2026. Those held one file per study, keyed by upper-case study code (rwls[["CA671"]]). The file was chosen by the shortest name, and for 304 studies that was not ring width; 253 were maximum density. If you used an earlier build, check which file you were reading.

How it is built

Scripts in code/, run in order:

1. 01_indexITRDB.R asks NOAA's paleo-search API for every tree-ring study, keeps the ITRDB ones, and lists their .rwl files with the variable each holds. This is the route NOAA recommends.
2. 02_syncFiles.R brings data_files/ up to date with that list: it fetches new and changed files and deletes any NOAA no longer lists. Rdatafiles/sync_manifest.csv records what the server reported for each file.
3. 03_metaITRDB.R adds family, order and group to each species, and fills the few missing elevations from a terrain model.
4. 04_readRWLs.R reads every file into an rwl object.
5. 05_xdateReports.R runs dplR's xdate.report(), a COFECHA-style crossdating report, on every ring-width file. The reports and an overview are in reports/correlation-stats/.

Reading is done by dplR::read.tucson() (dplR 1.8.0 or later, on CRAN), which records what it found in each file and returns that alongside the data. Each run writes three reports to reports/:

• index-flags.csv: problems in NOAA's metadata, such as a study with no species recorded.
• sync-log.csv: which files the last run fetched, refreshed or deleted.
• rwl-read-report.csv: one row per file, with series and measurement counts, year span, precision, interior gaps, and any warning the reader raised.

We send what the first and last of these find to NOAA.

What is left out, and why

• Studies with no .rwl file. Chronology-only studies are not included, nor are contributed datasets published only as NOAA template .txt files. The latter is deliberate on NOAA's part: that is how they archive tree-ring contributions that do not meet ITRDB standards, such as subfossil material with no calendar dating, or collections mixing several species.
• Files that will not read. Currently one, KYRG014. Rdatafiles/rwls_bad.rds holds every file that did not read, with the reason.

Things to know about the metadata:

• Species, coordinates and elevation are NOAA's. Where NOAA records no species (23 studies in this build), scientificName is NA. Where it records no elevation (12 studies), the value comes from a terrain model, and AltitudeSource says so.
• Some files hold years that are not calendar years. Seven files are floating chronologies that NOAA dates in calibrated years BP; timeUnit marks them.
• Some coordinates are wrong in the archive itself. This repo copies what the ITRDB holds rather than silently correcting it.

This is a snapshot, not a live mirror. The ITRDB is the authority. When the two disagree, the ITRDB is right and this repo is stale.

Citing the data

The measurements are not ours. Cite the original investigators and the ITRDB, following NOAA's guidance. For each file, rwls_meta gives the study's doi and landingPage, which carries the citation.

If this repo saved you time, a mention is welcome, but the credit belongs to the people who collected and contributed the cores.

Related

• dplR — the R package for tree-ring analysis these objects are built for
• dplPy — the Python counterpart
• openDendro — the wider project

License

MIT, for the code in this repository. The measurement data comes from the ITRDB and carries its own terms of use.