Multivariate climate change, the climate niche, and the Holocene history of eastern hemlock (Tsuga canadensis)

Abstract

Forests in the eastern North America have changed progressively over the 11,700 years of the Holocene Epoch. To understand the dynamics involved, we focus on eastern hemlock (Tsuga canadensis), which shifted its distribution through time and, notably, exhibited a rapid range-wide decline at 5280{+/-}180 YBP. We consider how climate could have shaped this history by comparing fossil pollen records from eight New England sites with quantitative temperature and effective precipitation reconstructions and evaluating the realization of Tsugas climate niche through time. The comparisons indicate that multivariate climate change significantly influenced Tsuga abundance, including its abrupt decline and recovery. The comparisons show that the realized climate niche of Tsuga expressed today includes two important features that persisted through time. First, Tsuga pollen percentages reach their maxima (>30%) where July temperatures equal 18-20{degrees}C, but do so at two modes where annual precipitation equals either [~]1100 or [~]775 mm. The bimodality reflects Tsugas two geographic modes in the Great Lakes and Appalachian regions today, and explains past dynamics, such as short-lived peaks in Tsuga abundance associated with effective precipitation of [~]775 mm at ca. 10,000 years before CE 1950 (YBP). Second, the two peaks in Tsuga abundance follow negative correlations between temperature and precipitation such that the two modes shift toward high precipitation if temperatures are low (e.g., [~]1400 and [~]1000 mm at <18{degrees}C). Consequently, rapid cooling at 5200{+/-}100 YBP facilitated widespread Tsuga declines because cooling did not coincide with increased precipitation. Abundance declined as local climates departed from optimal temperature and precipitation combinations. Recovery only followed as effective precipitation increased by >150 mm over the past 4000 years. A regionally calibrated model of the relationship of Tsuga pollen percentages to temperature and precipitation explains 70-75% of the variance in the percentages at eight study sites. Iteratively excluding each site from the model shows that accurately representing the major features of the climate niche enables the model to predict the mid-Holocene decline and other past changes at the excluded site (site-level RMSE = 2.1-5.6%). Similar multivariate climate dynamics closely modulated the species abundance throughout the Holocene with no evidence of additional large-scale disturbances.