A 2024 study measuring signal-to-noise ratios across a hemispheric network of tree-ring records found that maximum latewood density contains nearly twice the climate signal of ring width. The difference is not subtle. Width-based chronologies showed SNR values around 0.3 in boreal forests; density-based chronologies from the same sites exceeded 0.55. The denser the wood laid down in late summer, the more faithfully it tracks temperature — and the less it is contaminated by the biological noise that plagues width measurements.
The reason is physiological. Ring width integrates the entire growing season — spring moisture, summer temperature, autumn length, insect defoliation, competition from neighbors — into a single number. It is a sum of too many variables. Maximum latewood density, by contrast, responds primarily to late-summer temperature during the brief window when cells are thickening their walls. The measurement is narrower in what it captures, and that narrowness is its strength. A proxy that listens to fewer things hears each one more clearly.
But the study also revealed a troubling asymmetry across timescales. On short timescales — years to decades — density outperforms width decisively. On longer timescales — centuries to millennia — both proxies suffer from non-climatic trends that reduce their similarity to neighboring records. Biological memory effects, where a severe drought year suppresses growth for two or three subsequent years, dampen high-frequency signals. Age-related growth curves distort low-frequency ones. The tree remembers its own history, and that memory interferes with its recording of the climate's history.
The implication reaches beyond dendrochronology into any field that relies on natural archives. Every proxy is a filter, and every filter has a bandwidth. The question is never whether a record is accurate — it is accurate at which frequencies, and over which durations, and at what cost to the signals it cannot carry.