Perspectives on Plant Phenology in Deciduous Forest Ecosystems at Multiple Spatial Scales

Abstract

Deciduous forest phenology is a sensitive indicator of the ecological effects of global climate change, and in particular of changes in temperature. Phenology may also affect climate change through the changing length of the photosynthetic growing season, and hence the annual amount of atmospheric carbon dioxide sequestered by plants. To help understand the interplay between phenology and climate change, this dissertation links the organismic scale at which the physiological mechanisms of plant phenology must be understood, to the ecosystem scale at which phenology is measured and modeled for global applications. The second chapter uses records of repeat digital photography of deciduous forest phenology from 13 phenocam sites in Eastern North America to establish the connection between phenology metrics calculated from digital image analysis and events identified by human observers. We then link these phenology metrics from near-surface imagery to the larger scale of satellite remote sensing. We find that the difference between remote sensing and near-surface phenology depends on how representative the near-surface observations are of the surrounding landscape: less representativeness leads to a larger difference. The third chapter uses aerial drone photography to obtain near-surface phenology observations at the resolution of tree crowns (10 m), but with ecosystem-wide coverage; the study area encompasses a 250 m resolution remote sensing pixel. Analysis of spatial variance reveals that species and land cover variability explains most of the variance in phenology. By synthesizing drone imagery with high (30 m) and medium (250 m) resolution remote sensing, we find a logarithmic relationship between spatial variance in phenology dates and spatial resolution of observation, and that most of the phenological variability that can be gained through higher spatial resolution is at pixel sizes finer than the scale of medium resolution remote sensing. Chapter four examines the link between phenological metrics derived from aerial drone photography and the leaf life cycle events of trees. We find that plant area index of the forest canopy correlates with a canopy greenness index during spring green-up, and a canopy redness index during autumn senescence. Greenness and redness metrics are also significantly correlated with the timing of budburst and leaf expansion on individual trees in spring. However we note that the specific color index for individual trees must be carefully chosen if new foliage in spring appears red, rather than green. In autumn, both decreasing greenness and increasing redness correlate with leaf senescence. Maximum redness indicates the beginning of leaf fall, while the progression of leaf fall correlates with decreasing redness. We also find that cooler air temperature microclimates near a forest edge bordering a wetland advance the onset of senescence. The fifth chapter leverages a diverse set of phenological data sources, including phenocams, direct in situ observation, and aerial drone photography, to study the link between the timing of the start of spring, and the “velocity of green-up”, or duration between the beginning and end of leaf expansion. We find a significant association between later start of spring and faster green-up in all modes of observation. Using the in situ observational data, we fit degree-day models for the start and end of canopy development in spring time. We find that the default phenology parameters of an ecosystem model of carbon and water cycles make biased predictions of leaf initiation (39 days early) and maturity (13 days late) for red oak, while the fit parameters have biases of 1 day or less. Springtime productivity predictions using fit parameters are closer to results driven by observational data (within 1%) than those of the default parameterization (17% difference).