Seeing the Forest for the Tree(ring)s: Guarding Against False Discovery in Large-Scale Dendroclimatology

Abstract:

Measurements of tree-ring widths are the most widely-distributed and best replicated source of surrogate environmental information on the planet, and are one of the main archives used to estimate changes in regional and global climate during the past several centuries or millennia. Because the Northern Hemisphere ring-width network is now so large, it is more crucial than ever to ensure our understanding of tree-environment relations is not influenced by decisions to include or exclude certain records. It may be the case that a particular set of ring-width records are, for whatever reason, more tightly coupled to a particular climate factor than other records from the same region or species and, as a result, may be superior estimators of that factor’s past behavior. At the same time, it is known that selecting a small number of predictors from a large pool of potential candidates increases the likelihood of a Type I error. That effect may be particularly relevant to dendroclimatology because the total number of available ring-width records is often much larger than the number of records used to produce reconstructions of large-scale climate features. As an initial step, it would be helpful if paleoclimate reconstructions derived from tree rings described more explicitly the criteria used to select ring-width records as potential predictors and specified those records excluded by that screening. By comparing ring-width chronologies and their relations with climate against the standard set by thousands of records across the hemisphere, we should be better able to distinguish climate signals from proxy noise and produce more accurate reconstructions of climate during the late Holocene.