You are here

Mapping aboveground biomass annually across the northwestern USA from LiDAR and Landsat image time series

Status: 
Complete
Dates: 
January, 2015 to October, 2019

Northwestern USA study area of Washington, Oregon, and Idaho, with lidar coverages and field plot locations overlaid.
Fig. 1 - Northwestern USA study area of Washington, Oregon, and Idaho, with lidar coverages and field plot locations overlaid.

The cumulative area of LiDAR collections across multiple ownerships in the northwestern USA has reached the point that land managers of the US Forest Service (USFS) and other stakeholders would greatly benefit from a strategy for how to utilize LiDAR for regional aboveground biomass inventory. The need for Carbon Monitoring Systems (CMS) can be more robustly addressed by using not only available NASA satellite data products, but also commercial airborne LiDAR data collections.

  • We developed a “living” LiDAR and field plot database of reference observations that can continue to be updated as new project-level forest inventory data are collected. This strategy actively engaged users by utilizing existing data collected by and maintained by land managers of the USFS and other public and private stakeholders.
  • The reference database of field and LiDAR observations of initial conditions is in a format ready for ingestion into the latest version of the Forest Vegetation Simulator with climate change projection capabilities.
  • Our CMS provides an objective, accurate, repeatable, and transparent system of carbon monitoring, reporting, and verification (MRV) for USFS and other forest planners.

Approach

Map of predicted aboveground biomass (AGB) across the study region
Fig. 2 - Map of predicted aboveground biomass (AGB) across the study region

  • We used multiple airborne lidar datasets previously acquired at the project level in conjunction with field plot datasets to predict aboveground biomass across the diverse vegetation types of the northwestern USA.
  • We published two papers demonstrating that aboveground biomass (AGB) estimates in LiDAR project landscapes are transferable across time (Fekety et al. 2015) and space (Fekety et al. 2018) to other project landscapes with similar forest types.
  • Project-level biomass maps were post-stratified and pixels randomly selected to train a regional model predicting regional biomass carbon annually from Landsat time series imagery processed through LandTrendr.
  • Regional biomass maps were validated against U.S. Forest Service Forest Inventory and Analysis (FIA) plot data and a simple linear model fit to correct for bias and therefore calibrate the 2000-2016 AGB maps.
  • A 2009 forest/non-forest classification map derived from globally available PALSAR radar data was used to mask non-forest areas from the 2000-2016 AGB maps.
  • AGB maps for project landscapes and annual (2000-2012) regional AGB maps have been submitted to the Oak Ridge National Laboratory Data Active Archive Center (ORNL-DAAC).

Key Findings

Map of predicted aboveground biomass (AGB) across the study region
Fig. 3 - Map of predicted aboveground biomass (AGB) across the study region (e.g., 2010); similar maps were produced annually from 2000 through 2016.

  • We have assembled and consistently processed field plot (N=3,805) and LiDAR datasets (greater than 13M hectares) from project landscapes distributed along a broad climate gradient across the northwestern USA from temperate rainforest to cold desert (Fig. 1) and contributed by 29 public and private stakeholders.
  • 1984-2016 Landsat image time series were processed through LandTrendr across the entire study region. Landsat image time series have been found to explain more structural variation in forest canopies and stand attributes than a single Landsat image.
  • By also including topographic and especially climate metrics in the predictive models, we more accurately predicted biomass than by using only Landsat-derived predictors, which unlike LiDAR, saturates and loses sensitivity in high biomass forests. 
  • We used the Random Forests machine learning algorithm as our predictive modeling approach. The models explained 78-80 percent of variance in AGB at both the landscape and regional levels without asymptote (saturation) at either scale (Fig. 2), meaning the field estimates of biomass are scalable to project landscapes and the entire study region.
  • Having calibrated our AGB maps with unbiased FIA data (Tinkham et al. 2018), the AGB maps can also be considered unbiased and spatially and temporally consistent (Fig. 3), which is important for forest planning and monitoring, reporting, and verification (MRV) of carbon sequestration for mitigation of greenhouse gas emissions.

Publications

Tinkham, Wade T. ; Mahoney, Patrick R. ; Hudak, Andrew T. ; Domke, Grant M. ; Falkowski, Mike J. ; Woodall, Chris W. ; Smith, Alistair M. S. , 2018
Fekety, Patrick A. ; Falkowski, Michael J. ; Hudak, Andrew T. ; Jain, Terrie B. ; Evans, Jeffrey S. , 2018
Fekety, Patrick A. ; Falkowski, Michael J. ; Hudak, Andrew T. , 2015


Project Contact: 

Principal Investigators:
Co-Investigators:
Michael J Falkowski - Colorado State Univeristy
Robert E Kennedy - Oregon State University
Alistair M.S. Smith - University of Idaho

Collaborators:
Nancy Glenn - Boise State University
Grant Domke - Northern Research Station
Van Kane - University of Washington

Research Staff:
Patrick Fekety - Colorado State Univeristy

Funding Contributors:
NASA CArbon Monitoring Systems Program