Spatial patterns and climatic drivers of leaf spring phenology of maple in eastern North America

Abstract

The resurgent frequency of extreme weather events and their strongly distinctive spatial patterns lead to a growing interest in phenology as an indicator of tree susceptibility. Using a long-term chronology of observations collected in situ, we predicted and investigated the spatial patterns and environmental drivers of spring leaf phenology across maple stand polygons dominated by Acer saccharum Marsh. and/or Acer rubra L. in eastern North America for 2000-2018. Model' calibration was based on Bayesian ordinal regressions relating the timing of the phenological events' observations to the MODIS vegetation indices EVI, NDVI and LAI. DAYMET data have been extracted to compute temperature and precipitation during spring phenology. Model accuracy increased as the season progressed, with prediction uncertainty spanning from 9 days for bud swelling to 4 days for leaf unfolding. NDVI and LAI were the best predictors for the onset and ending of spring phenology, respectively. Bud swelling occurred at the end of March in the early stands and at the onset of May in the late stands, while leaf unfolding was completed at the beginning of April for the early and in mid-June for the late stands. Early and late stands polarized towards a south-west-north-east gradient. In the south-western regions, which are also the driest, total precipitation and minimum temperature explained respectively 73 % and 25 % of the duration of spring phenology. In the north-eastern regions, precipitation and minimum temperature explained 62 % and 26 % of the duration of spring phenology. Our results suggest high vulnerability to extreme weather events in stands located in the south-west of the species distribution. The increasing incidence of drought in these locations might affect spring phenology, decreasing net primary production in these stands. Warmer nights might expose the buds to late frosts, events that are expected to become more frequent in the coming years.

Keywords: Bayesian inference; Bud phenology; Getis-Ord* statistics; Remote sensing.