IWMMM poster 20026 - PowerPoint PPT Presentation

Title: IWMMM poster 20026


1
Validation and Inversion of a Simple Geometric
Canopy BRDF Model for Chihuahuan Desert
Grass-Shrub Transition Zones
Mark J. Chopping1, Lihong Su2, Albert Rango1,
Connie Maxwell1, Christoph C. Borel3 1 USDA,
ARS Jornada Experimental Range, Las Cruces, NM
88003, USA 2 Research Center for Remote
Sensing GIS, Beijing Normal University, Beijing
100875, P.R. China 3 Los Alamos National
Labs., Los Alamos, NM 87545, USA
Introduction The objectives of this research
were to validate and test the invertibility of a
simple, geometric bi-directional reflectance
distribution function (BRDF) model (SGM 1) in a
Chihuahuan Desert grass-shrub transition zone.
To this end, it was necessary to evaluate the
importance of overstory (mesquite, acacia,
ephedra and yucca shrubs), understory (broom
snakeweed with other small shrubs and forbs and
some grama grass), soil, grama grass and litter
in the BRDF. This was achieved by simulations
using the POVRAY and RGM 3-D graphics/raytracing
and radiosity packages 2, driven by
measurements of plant locations and dimensions in
the Jornada Experimental Range near Las Cruces,
New Mexico, USA. Method The 3-D models were
driven by the plant maps and the SGM was driven
by mean plant density, radius and crown vertical
to horizontal radius ratio for each plot,
calculated using the field data. A
non-Lambertian soil/litter background BRDF was
provided for all models by the Walthall model
3, adjusted against ground BRDF samples. The
modeled values were assessed against samples of
the BRDF at 650 nm acquired from the air at up to
six view zenith angles and three solar zenith
angles by a well-calibrated DuncanTech MS2100
digital camera. The SGM was inverted for mean
shrub density and leaf area index using the Hooke
- Jeeves direct search algorithm, with shrub
shape and height retained as fixed parameters,
using a least-squares criterion.
Top Aerial photograph chips (25 m2) for plots
107 and 108. Bottom views of the modeled
plants in plots 107 and 108 modeled as
spheroids and their shadows, generated by the
Forester/POVRAY package. The soil colors given
here are arbitrary. Both Forester and RGM allow
more detailed descriptions of plant
physiognomies, including specification of leaves,
branches and stems.
Location
Multi-Angle Reflectance Images
Conclusions Chihuahuan Desert grass-shrub
transition landscapes present important
challenges for canopy reflectance and BRDF
modeling. Although soil is the most important
component governing brightness and anisotropy,
with a fractional cover gt 0.7, the presence of
black grama grass (Bouteloua eriopoda) and its
litter plays an important role in relation to
that of the overstory of larger shrubs and the
understory of small forbs and sub-shrubs.
Assigning the understory proportions to soil has
a small but noticeable effect, mainly because of
the very large numbers of plants (380-660/25m2).
The higher proportion of grama grass and its
litter in plot 108 seems to be the most plausible
explanation for the lower observed values in the
forward-scattering direction relative to the
model outputs. To be accurate, models must not
only include a non-Lambertian soil and account
for the understory but must also account for the
grass and litter components. The radiosity
simulations show that a simplified
geometric-optics model (SGM) closely follows the
bi-directional reflectance values predicted by
the RGM 3-D model, with maximum absolute root
mean square errors of 0.039 and 0.015 with
respect to the multiangular observations and RGM,
respectively (minimum r2 of 0.91 and 0.98,
respectively). The inversion results show that
it is possible to retrieve measures of canopy
plant density and leaf area index distributions
which are reasonable, when the parameters least
sensitive to BRDF (plant shape and height) are
fixed. References 1 Chopping M.J., Rango, A.,
Havstad, K.M., Schiebe, F.R., Ritchie, J.C.,
Schmugge, T.J., French, A., McKee, L.,
and Davis, R.M. (2002), Canopy attributes of
Chihuahuan Desert grassland and
transition communities derived from multi-angular
0.65µm airborne imagery, unpublished
manuscript. 2 Qin, W., and Gerstl, S. A. W.
(2000), 3-D scene modeling of Jornada semi-desert
vegetation cover and its radiation regime.
Remote Sens. Environ. 74145162. 3Walthall,
C.L., J.M. Norman, J.M. Welles, G. Campbell, and
B.L. Blad (1985), Simple equation to
approximate the bidirectional reflectance from
vegetative canopies and bare surfaces,
Applied Optics 24, No.3 383-387. 4 Wanner,
W., X. Li, and A. H. Strahler (1995), On the
derivation of kernels for kernel-driven
models of bidirectional reflectance, Jnl. of
Geophysical Research, 10021077-21090.
BRDF Modeling and Inversion Results
Calculation of Reflectance _at_ 650 nm (red) 1.
skylightdirectlight 112 skylight
directlight 1 (skylight diffuse
irradiance) 2. SOILref soil BRDF (Walthall
model calibrated with BRFs from Grassland
PROVE) GUSAref Snakeweed reflectance
PRGLref (Mesquite_leaf_reflectance x 10
Mesquite_stem_reflectance)/11.0 EPTOref
Ephedra reflectance YUELref
(Yucca_leaf_reflectance x 3 Yucca_stem_reflecta
nce)/4.0 Mesquite_leaf 0.095 Mesquite_stems
0.075 Yucca_leaf 0.095 Yucca_stems
0.140 Snakeweed 0.110 Ephedra
0.200 Mesquite leafstem 101 Yucca leafstem
31 3. reflectance sunlit__soil x
(1-skylight) x soilref sunlit__GUSA x
(1-skylight) x GUSAref sunlit__PRGL x
(1-skylight) x PRGLref sunlit__YUEL x
(1-skylight) x YUELref sunlit__EPTO x
(1-skylight) x EPTOref shde__soil x
skylight x soilref
shde__GUSA x skylight x GUSAref
shde__PRGL x skylight x
PRGLref shde__YUEL x skylight x YUELref
shde__EPTO x skylight x
EPTOref
Solar Zenith Angle38
Solar Zenith Angle49
Solar Zenith Angle59
Overlapping multi-angle spectral reflectance
images (650nm) over the JORNEX transition site,
aligned here so that each row (column) has the
same latitude (longitude). The value scale is
constant across all images. In reading order the
VZA/SZA for the target centres are -19/38,
-7/38, 2/38, 15/38, 29/38 -36/49, -20/49, -2/49,
14/49, 35/49, 40/49 -35/59, -28/59, -9/59, 2/59,
19/59 and 26/59 (negative values indicate
backscattering). The co-registered 2 m ground
resolution multiangular images were used to
simulate the response of a sensor with a 50 m
footprint through convolution with a
pseudo-Gaussian kernel. The resulting data were
subsequently mapped at a 25 m sampling interval.
Landsat TM5 NDVI image 09/25/95 of the USDA, ARS
Jornada Experimental Range, New Mexico, pastures
outlined in blue, roads in red (center 106.9W
32.5N).
Solar Zenith Angle38
Solar Zenith Angle49
Solar Zenith Angle59

Note Y-axis scales!
RGM and SGM model outputs compared with the soil
BRDF function, multiangle observations from the
airborne tilted Duncan MS2100 camera, and RGM
with no snakeweed. Modeling was for sparse (top
set) and dense (bottom set) plots.
Polar plot of a typical angular sampling at the
center of the target area provided by flights at
three times of day, providing observations at
three solar zenith angles. Radial axis is zenith
angle polar axis is azimuth angle. The
observations marked with a square and circle were
acquired at a hot spot and a specular
configuration, respectively.
RETRIEVALS
REFERENCE
Transition Zone aerial photograph (0.33m
resolution) showing location of the 3 x 3 grid of
25 m2 plots. Note the brighter areas along the
roads.
Transition Zone, aerial photograph chip (0.33m
resolution) of the 3 x 3 grid of 25 m2 plots
(numbered). The large dark objects are honey
mesquite shrubs. Mottled areas have more
snakeweed.
Leaf Area Index (scaled 0.0 - 1.0) plot 107
6.9 plot 108 7.7
Plant Density (scaled 0.5 - 1.36 / m2) plot 107
0.777 plot 108 0.952
Airphoto (Leica mapping camera)
MSAVI2 (IKONOS) Greener more green leaves
Acknowledgments the staff of the USDA, ARS
Jornada Experimental Range (Eddie Garcia, Jim
Lenz, Dave Thatcher and Rob Dunlap) Wenhan Qin
at Science Systems and Applications, Inc.,
Lanham, MD Betty Walter-Shea at the School of
Natural Resource Sciences at the University of
Nebraska-Lincoln and Jerry C. Ritchie, Charlie
Walthall and Rob Parry at the USDA, ARS Hydrology
and Remote Sensing Laboratory, Beltsville, MD.
Courtesy Dr. Jerry C. Ritchie.
Data from the JORNEX campaign of September 2000