Sampling bias overestimates climate change impacts on forest growth in the southwestern United States

Stefan Klesse,R Justin Derose,Ann M Lynch,Christopher H Guiterman,Christopher D O’Connor,Margaret E K Evans,John D Shaw

doi:10.1038/s41467-018-07800-y

Abstract

Climate−tree growth relationships recorded in annual growth rings have recently been the basis for projecting climate change impacts on forests. However, most trees and sample sites represented in the International Tree-Ring Data Bank (ITRDB) were chosen to maximize climate signal and are characterized by marginal growing conditions not representative of the larger forest ecosystem. We evaluate the magnitude of this potential bias using a spatially unbiased tree-ring network collected by the USFS Forest Inventory and Analysis (FIA) program. We show that U.S. Southwest ITRDB samples overestimate regional forest climate sensitivity by 41–59%, because ITRDB trees were sampled at warmer and drier locations, both at the macro- and micro-site scale, and are systematically older compared to the FIA collection. Although there are uncertainties associated with our statistical approach, projection based on representative FIA samples suggests 29% less of a climate change-induced growth decrease compared to projection based on climate-sensitive ITRDB samples.

Highlights

Climate−tree growth relationships recorded in annual growth rings have recently been the basis for projecting climate change impacts on forests
Large-scale inference based upon tree-ring time series sampled from old, targeted trees overestimates the impact of climate change, especially on southwestern U.S forest growth
That our tree-by-tree analysis revealed significant differences between International Tree-Ring Data Bank (ITRDB) and Forest Inventory and Analysis (FIA) samples in the variance explained by climatic parameters verifies that the practice of targeting open-grown trees occurring under ecologically marginal conditions, in order to maximize climate signal with a few samples, achieved that goal

Summary

Introduction

Climate−tree growth relationships recorded in annual growth rings have recently been the basis for projecting climate change impacts on forests. We show that U.S Southwest ITRDB samples overestimate regional forest climate sensitivity by 41–59%, because ITRDB trees were sampled at warmer and drier locations, both at the macro- and micro-site scale, and are systematically older compared to the FIA collection. The ecological and geographic distributions of individual tree species encompass a range of climatic and edaphic conditions (e.g., deeper soils, cooler aspects) across which climate may be a more or less important factor limiting tree growth[14,15,16] Representative sampling across these gradients is necessary to characterize the response of a forest ecosystem to climate variability. We augment the FIA sample with two densely sampled, landscape-scale collections including 828 trees in Arizona and New Mexico[19,20] (Fig. 2, cf. Methods section “Tree-ring data”) and refer to this combined dataset as the FIA or “inventory” collection

Methods

Results

Conclusion