You are here

Harvard Forest >

Harvard Forest Research Project 2022

  • Title: Resolving the multi-scale drivers of tree mortality from field and remote sensing data on co-located ForestGEO-NEON sites
  • Principal investigator: Daniel Johnson (johnson.daniel@ufl.edu)
  • Institution: University of Florida
  • Primary contact: Lukas Magee (mageel@ufl.edu)
  • Team members: Ross Barreto; Eben Broadbent; Caitlyn Cherro; Liam Halloran; Daniel Johnson; Albert Kim; Lukas Magee; Alexander Seeley; Jessica Shue
  • Abstract:

    Overview
    The rate of tree mortality has been increasing, but why and where? Understanding the scope and scale of tree mortality is important because tree death can have a large impact on carbon dynamics and reduce biodiversity quickly. Mortality events most often occur at the scale of the individual or a small group of trees in a forest. Our ability to detect shifts in tree mortality and determine the causal factors responsible is defined by the intensity, scale, and frequency of our observations. Currently, most observations are too coarse in spatial scale and too infrequent in time to determine causal factors of mortality or how they vary across space and time and lack detail of why a tree died.
    Determining the uncertainty in tree mortality associated with the scale of observation is critical for making well informed estimates of carbon dynamics and predictions of forest community trajectories. The macroecological challenge lies gathering enough data on the ground to allow scaling up to broad regions with remote sensing products that can detect mortality events and changes in biomass at landscape and continental scales while minimizing uncertainty in our estimates. Our objective is to address this challenge and build scalable models of tree mortality with underlying factors associated with tree death.

    Intellectual merit
    We are proposing to fill key gaps in knowledge about where and why trees die across ecoregions. We will rigorously pair annual detailed ground-based mortality censuses, including identification of mortality agents at five ForestGEO large-area mapped forests that are co-located with NEON sites across five NEON Domains, with high resolution unpiloted aerial vehicle (UAV) flights at those forests. The UAV imagery can be scaled-up with the NEON Airborne Observation Platform (AOP) products to the flight view-shed. Using both lidar and hyperspectral imagery will allow us to pinpoint changes in canopy structure and reflectance indicative of tree mortality and associated factors leading to death that are ground validated in ForestGEO plots. By leveraging the methodologies of these two networks, we will be able to make well validated models of tree mortality and associated factors leading to tree death.
    Ultimately, we will use machine learning algorithms trained and tested on the field collected data in this proposal and NEON AOP data to scale up to space based platform products (GEDI and Landsat) to determine tree mortality and associated mortality biomass dynamics. Leading to better estimates of the scale of biomass change and forest health risks and the associated uncertainty in those estimates. We will develop flexible models trained on unprecedented data sets by systematically performing the same work across multiple NEON Domains.