Autumn Tree Phenology in Northern Wisconsin: Humans Versus Photographs

Abstract

Ecosystem primary productivity halts when plants go dormant, and so the timing of dormancy as it relates to autumn phenology has been a focus of much interdisciplinary research. While monitoring plant phenology has its roots in directly observing specimens, digital sensors along with modern methods have also become a mainstay in phenology. Results from different methods often vary, so there is still a need to better understand how digital cameras record autumn phenology, especially in comparison with ground-based observations (Keenan et al. 2014). This study compared autumn phenology derived from direct ground observations with upward-facing fisheye photography, in the context of a larger research project (C.H.E.E.S.E.H.E.A.D.19), to precisely determine autumn tree phenology across 53 field sites in a heterogeneous temperate deciduous forest with over 220 individual trees and 1,000 digital photos sampled. Less-studied trees such as aspen (Populus spp.), birch (Betula spp.), and basswood (Tilia americana) were included in the project, as well as sugar maple and red maple (Acer spp.). The results show that inflection points from sigmoid curves and change point detection are in close agreement for critical transition dates including the start of leaf coloration (bias of change points later at i= -0.47 days) and end (i= -0.6), but with slightly less agreement for the start of leaf fall (bias of change points earlier at i= 3.8) and the end of leaf fall (bias of change points later at i= -3.39). While camera-derived transition dates correlated poorly with corresponding human-derived transition dates, the best relationship detected was between green chromatic coordinate (GCC) inflection points and leaf fall (when foliage is mostly absent from tree canopies). This work is intended as a pilot study for novel methodologies in the field of ground-based phenology.