Sixteen forest regions, drawn as sixteen strips of timber — one
ring for every year from 1500 to 1995, dark and heavy where that region's trees
grew badly. Press play to lay the grain down. Most bad years are local: the
American Southwest blackens in droughts nobody in Fennoscandia noticed, and the
strips mostly disagree. The 1810s are the exception — 1813, 1819, 1820, 1821 and
1824 darken forest after forest on both sides of the equator, in the decade after
Tambora erupted. Then 1887 does the same thing with no famous cause. A narrow ring
means growth was poor, not that the year was cold. Tap or hover any year.
Data & source
Data: NOAA
NCEI Paleoclimatology — International Tree-Ring Data Bank, standardized site
chronologies · open access, US Government public domain · —
site chronologies contributed by — lead investigators,
listed below · fetched —. ITRDB asks that the original
investigators be cited alongside the archive, which is what that list is for.
Scope and gaps, plainly. This is a slice of the bank, not the
bank: the roughly 1,800 chronologies that are long enough to reach back into the
window and deep enough to composite. Coverage is badly uneven — western North
America and Europe are thick, the tropics and the Southern Hemisphere are thin,
and there is one African region and none at all in East Asia, southern Africa or
the Amazon. So "the world's forests" is a claim this data cannot support, and
the page does not make it: what is drawn is sixteen regions where enough trees
have been measured.
A ring index is growth, not temperature. Each chronology has
already been standardized by its contributors into a dimensionless index where
1000 is that site's average ring, which removes the age trend but not the local
climate: at a dry site a narrow ring means drought, at a treeline site it means
a cold summer. A year that is dark in six regions means growth was poor in six
regions, and the reason is different in each. This pipeline makes no attribution
to any eruption — the 1810s are simply the decade in which the most regions were
simultaneously in their own worst tenth, and 1887, which has no famous cause,
ties 1821 for the single widest year.
Method. Each site is z-scored against its own mean and spread
inside 1500–1995 before averaging, so a variable site cannot dominate its region;
a region-year is drawn only where at least eight sites contribute, which is why
strips begin at different years and why the ragged left edges are real. Each
strip's shading is then measured against that region's own composite spread,
because regional variance differs by nearly a factor of two and a shared scale
would render half of them blank; a ring at the dark end of one strip is not the
same anomaly as a ring at the dark end of another. Sample depth thins going
backwards, so the earliest decade of any strip rests on the fewest trees — the
site count for the year you are touching is in the tooltip, and the per-region
depth is in the table below.
—