PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
April 2016
271
Estimating Forest and Woodland
Aboveground Biomass Using Active and
Passive Remote Sensing
Zhuoting Wu, Dennis Dye, John Vogel, and Barry Middleton
Abstract
Aboveground biomass was estimated from active and passive
remote sensing sources, including airborne lidar and Landsat-8
satellites, in an eastern Arizona (USA) study area comprised of
forest and woodland ecosystems. Compared to field measure-
ments, airborne lidar enabled direct estimation of individual
tree height with a slope of 0.98 (R
2
= 0.98). At the plot-level,
lidar-derived height and intensity metrics provided the most
robust estimate for aboveground biomass, producing dominant
species-based aboveground models with errors ranging from
4 to 14
Mg ha
–1
across all woodland and forest species. Land-
sat-8 imagery produced dominant species-based aboveground
biomass models with errors ranging from 10 to 28
Mg ha
–1
. Thus,
airborne lidar allowed for estimates for fine-scale aboveground
biomass mapping with low uncertainty, while Landsat-8 seems
best suited for broader spatial scale products such as a nation-
al biomass essential climate variable (
ECV
) based on land cover
types for the United States.
Introduction
High uncertainty currently exists in regional, national and
global biomass and carbon stock estimates (Brown
et al
.,
1995; Houghton, 2005). Creating an operational biomass
product at the national scale to provide vegetation biomass
with low uncertainty at various scales is important for future
biomass and carbon balance projections, particularly when
considering the ongoing effects of climate change
.
Aboveground biomass is defined as an essential climate
variable (
ECV
) by the Global Climate Observing System
(
GCOS
) (Bojinski
et al
., 2014). Development of an effective
satellite-based methodology for a US national scale biomass
data product that can benefit from the more than four-decade
record of Landsat observations is among the objectives of the
ECV
activity sponsored by the US Geological Survey’s Land
Remote Sensing Program (Stitt
et al
., 2011). This investigation
contributes to that goal for finding the pathway to a Biomass
ECV
product by examining the performance of a passive, mul-
tispectral remote sensing approach with Landsat-8 relative to
an active remote sensing approach with airborne lidar, in a
pilot study area in the Southwestern US
.
The southwestern US serves as an ideal pilot study area in
the pursuit of a national biomass product given the challeng-
ing landscape changes the region is now facing, changes that
can be better addressed by having a Biomass
ECV
product. For-
ests dominated by ponderosa pine and woodlands dominated
by pinyon, juniper, and evergreen oak are the most common
higher elevation land cover types in the semi-arid southwest-
ern US (Daubenmire, 1978; Hicke
et al
., 2007). These forests
and woodlands sequester and store carbon in their biomass
and soils (Rasmussen, 2006), and play a significant role in
the regional carbon budget. During the past century, south-
western US forest and woodland ecosystems have undergone
significant changes including woody encroachment (Neff
et
al
., 2009; Rau
et al
., 2011), regional vegetation die-off due to
climate change (Allen
et al
., 2010; Breshears
et al
., 2005), and
anthropogenic land use change (Covington and Moore, 1994;
Covington, 1997). Sagebrush and grassland ecosystems are in-
creasingly influenced by pinyon and juniper expansion (Rau
et al
., 2011) while fire suppression and other factors have
turned historically open and park-like forests into overstocked
dense forests occupied by younger and smaller trees with
heavier fuel loads (Covington and Moore, 1994). The south-
western US is expected to experience further warming and
drying as a result of changing climatic conditions (Seager
et
al
., 2007). Increasing frequency and intensity of wildfires have
consumed vegetation biomass, and prolonged droughts have
led to regional vegetation die-off (Allen
et al
., 2010; Breshears
et al
., 2005), both of which put forest biomass and carbon
storage at risk (Breshears and Allen, 2002). Meanwhile, land
use and management practices such as grazing diminished
and altered herbaceous vegetation, which could potentially
affect fire regimes (Wu
et al
., 2015). This could limit the long-
term capacity to retain stored carbon and contribute to the un-
certainty of future regional carbon stock estimates. To address
these issues effectively, remote sensing-based methods for
quantifying, mapping and monitoring biomass (Skowronski
et
al
., 2014) that account for the particular structural properties
and landscape conditions of the dryland ecosystems of the
Southwest, including tree growth form, tree density (canopy
closure) and presence of understory vegetation, are needed
.
Conventional field inventory measurements provide accu-
rate and consistent means of assessing biomass, but involve
destructive sampling within a limited geographic area at a
very high cost. Remote sensing methods can be integrated
with field measurements to assess vegetation biomass and
carbon stocks across spatial scales (Anaya
et al
., 2009; Black-
burn and Milton, 1997; Carreiras
et al
., 2006; Chidumayo,
1990; Cohen and Goward, 2004; Franklin and Hiernaux, 1991;
Houghton, 2005; Patenaude
et al
., 2004; Van Tuyl
et al
., 2005;
Zhao
et al
., 2009). Methods for remote sensing-based biomass
Zhuoting Wu is with the US Geological Survey, Western
Geographic Science Center, Flagstaff, AZ, and the Merriam-
Powell Center for Environmental Research, Northern Arizona
University, Flagstaff, AZ 86001 (
).
Dennis Dye is with the US Geological Survey, Western Geo-
graphic Science Center, Flagstaff, AZ, and School of Forestry,
Northern Arizona University, Flagstaff, AZ 86001.
John Vogel and Barry Middleton are with the US Geological
Survey, Western Geographic Science Center, Flagstaff, AZ.
Photogrammetric Engineering & Remote Sensing
Vol. 82, No. 4, April 2016, pp. 271–281.
0099-1112/16/271–281
© 2016 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.82.4.271
