Landsat Time-series for the Masses: Predicting Wood Biomass Growth from Tree-rings Using Departures from Mean Phenology in Google Earth Engine
Abstract
The terrestrial carbon cycle is perturbed when disturbances remove leaf biomass from the forest canopy during the growing season. Changes in foliar biomass arise from defoliation caused by insects, disease, drought, frost or human management. As ephemeral disturbances, these often go undetected and their significance to models that predict forest growth from climatic drivers remains unknown. Here, we seek to distinguish the roles of weather vs. canopy disturbance on forest growth by using dense Landsat time-series to quantify departures in mean phenology that in turn predict changes in leaf biomass. We estimated a foliar biomass index (FBMI) from 1984-2016, and predict plot-level wood growth over 28 years on 156 tree-ring monitoring plots in Minnesota, USA. We accessed the entire Landsat archive (sensors 4, 5 & 7) to compute FBMI using Google Earth Engine's cloud computing platform (GEE). GEE allows this pixel-level approach to be applied at any location; a feature we demonstrate with published wood-growth data from flux tower sites. Our Bayesian models predicted biomass changes from tree-ring plots as a function of Landsat FBMI and annual climate data. We expected model parameters to vary by tree functional groups defined by differences in xylem anatomy and leaf longevity, two traits with linkages to phenology, as reported in a recent review. We found that Landsat FBMI was a surprisingly strong predictor of aggregate wood-growth, explaining up to 80% of annual growth variation for some deciduous plots. Growth responses to canopy disturbance varied among tree functional groups, and the importance of some seasonal climate metrics diminished or changed sign when FBMI was included (e.g. fall and spring climatic water deficit), while others remained unchanged (current and lagged summer deficit). Insights emerging from these models can clear up sources of persistent uncertainty and open a new frontier for models of forest productivity.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2017
- Bibcode:
- 2017AGUFM.B13I..06F
- Keywords:
-
- 0426 Biosphere/atmosphere interactions;
- BIOGEOSCIENCES;
- 0439 Ecosystems;
- structure and dynamics;
- BIOGEOSCIENCES;
- 0480 Remote sensing;
- BIOGEOSCIENCES;
- 1632 Land cover change;
- GLOBAL CHANGE