ABSTRACT:
weight (R? = 0.77).
1. INTRODUCTION
Accurate estimates of canopy height, canopy base height,
canopy bulk density, and canopy fuel weight would improve the
data layer creation process for wildfire simulation models such
" as FARSITE (Finney, 1998) or future fire spread models.
Previously, these data layers were generated using the, output
from stand-level growth models such as the Forest Vegetation
Simulator (FVS), which depend upon a tree list to drive the
simulations (Beukema et al, 1997). Since the stand-level
estimates generated from these models are based upon a
relatively sparse sample of inventory attributes, they will be
subject to sampling error and will be unable to capture
variability in stand structure at finer spatial scales over the
landscape. If such variables could be accurately estimated using
remotely-sensed data, in a spatially explicit format, the
application of fire spread models to landscapes would be
significantly improved.
Synthetic aperture radar (SAR) is an active sensing technology
that emits and records the reflection of microwave radio energy.
Airborne SAR systems typically collect data from very high
flying heights, allowing them to collect data at a rate of
approximately 1000 km?/hour with a cost that ranges from $10 -
$80/km? for typical X-band digital surface models (Mercer,
2001). The information content of radar data in forested terrain
varies depending upon the wavelength of the transmitted pulses
— energy with short wavelengths (— 1 cm) is reflected from the
canopy surface while radar energy with longer wavelengths (~ 1
m) penetrates the foliage in the canopy and reflects from large
branches, tree stems and the terrain surface. Microwave remote
sensing, in contrast to optical remote sensing, can penetrate
| | cloud and smoke cover, allowing data collection in most
* Corresponding author.
KEY WORDS: Forest fire, mapping, radar, SAR, interferometer
ESTIMATING CANOPY FUEL PARAMETERS IN A PACIFIC NORTHWEST CONIFER
FOREST USING MULTIFREQUENCY POLARIMETRIC IFSAR
Hans-Erik Andersen * *, Robert McGaughey °, Stephen Reutebuch ^, Gerard Schreuder ^, James Agee “, Bryan Mercer ©
* University of Washington, College of Forest Resources, Seattle, WA, 98195 USA —
(hanserik, gsch, jagee)@u.washington.edu
® USDA Forest Service, Pacific Northwest Research Station, Seattle, WA, 98195 USA —
(bmcgaughey, sreutebuch)@fs.fed.us
¢ Intermap Technologies Corp., Calgary, Alberta, Canada T2P 1H4 — bmercer@intermaptechnologies.com
ISPRS Commission III, WG I1I/3
Fire researchers and managers need accurate, reliable, and efficiently-obtained data for the development and application of crown
fire behavior models. In particular, reliable estimates of critical canopy structure characteristics, including canopy bulk density,
canopy height, canopy base height, and canopy fuel weight are required to accurately map fuel loading and model fire behavior over
the landscape. The use of polarimetric interferometric synthetic aperture radar (IFSAR), a high-resolution active remote sensing
technology, provides for accurate and efficient estimation of crown fire behavior variables over extensive areas of forest. In this
study, estimates of crown fuel variables were developed from the polarimetric backscatter and interferometric information
(elevation, coherence and phase) for an IFSAR dataset acquired within a coniferous forest in western Washington State, USA.
Multiple regression analysis showed that plot-level IFSAR-based canopy fuel estimates were highly correlated with field-based fuel
measurements of canopy height (R? 2 0.89), canopy base height (R? 2 0.85), canopy bulk density (R* = 0.74), and canopy fuel
weather conditions. The resulting image represents the intensity
of the radar backscatter throughout the illuminated region.
While previous studies have shown that SAR backscatter
amplitude can be used to estimate forest biomass (Hussin et al.,
1991), it has been noted that the biomass saturation limits for
even long-wavelength SAR systems (~ 100-150 tons/ha) are too
low to reach levels present in temperate closed forests (~300
tons/ha) (Imhoff, 1995; Mette et al., 2003).
The availability of three-dimensional interferometric radar
(IFSAR) data in recent years has the potential to significantly
expand the applicability of radar analysis for forest structure
analysis. Radar interferometry uses the difference in phase, or
phase shift, between two radar images acquired from slightly
different locations to acquire information relating to the
elevation angle to an imaged point, which is used in
conjunction with the range information to determine the three-
dimensional location of this imaged point (Hagberg et al,
1995). Varying the wavelength of the emitted energy will allow
for collection of different three-dimensional structure data -
sensors emitting pulses with short wavelength measure the
canopy surface while sensors with longer wavelength will
collect information about sub-canopy and terrain features. The
difference between the canopy elevation (X-band) and
underlying terrain elevation (P-band) yields a canopy height
model that represents a spatially-explicit description of canopy
structure (i.e. volume, height, biomass, canopy fuel density)
over a given area of forest. It has been shown that
inteferometric observables, such as coherence and phase, arc
more sensitive than radar power (backscatter) to forest
structural parameters and biomass over a large range of forest
types (Treuhaft and Siqueira, 2004). The use of IFSAR
Intern
dite
theref
more
a wid
The
systel
reseat
classi
al, 2(
to es
(And
Sique
appro
inclu
heigh
IFSA
2d :
The «
Capit
fores
(Psei
heter
(Alm
area
Figui
(cour
This
silvic
stanc
surve
evalt
fores
and t
2.2
IFSA
arca