Title: Carbon Pools in Desert Grasslands from EOS: First Meeting Jornada Experimental Range, Las Cruces, NM, October 28, 2004
1Carbon Pools in Desert Grasslands from EOS First
MeetingJornada Experimental Range, Las Cruces,
NM, October 28, 2004
Agenda
830 a.m. Welcome, Statement of Objectives 845
a.m. Summary of Activities Milestones
Results 945 a.m. Perspectives on RS for
Ecological Modeling 1015 a.m. Discussion on
Goals and Approaches 1045 a.m. Coffee 1100
a.m. Discussion on Goals and Approaches
(continued) 1200 a.m. Lunch 200 p.m. Remote
Sensing Techniques MISR 400 p.m. ...or before
if possible End.
2Overall Objective
The overall objective stated in the proposal was
to validate a new approach to parameterizing the
CENTURY model for the arid and semiarid
grasslands of the southwestern US (and beyond) by
exploiting the unique information content of
multi-angular remote sensing data from the EOS
MISR and MODIS sensors.
3Overall Objective
Our new overall objective is to validate a new
approach to deriving parameters for driving
biogeochemical cycling models for the arid and
semiarid grasslands of the southwestern US (and
beyond) by exploiting the unique information
content of multi-angular remote sensing data from
the EOS MISR and MODIS sensors.
4Approach
- We want to combine MISR and MODIS data to
exploit 1. physical semi-empirical measures
from the RPV and Li-Ross models 2. structural
measures from CR models and ANIX and 3. spectral
measures of canopy vigor (SVIs). These are highly
complementary. - We are not currently able to include process
modeling as a result of the recent budget changes
-- but we will strengthen the validation aspects
of our work and continue to seek funding for the
modeling.
5Specific Objectives
- to assess the ability of kernel-driven and EMRPV
BRDF models to yield parameters useful in
precision community type mapping - to assess the ability of these models to yield
information useful in determining bare
soilgrassshrub proportions - to assess the ability of geometric-optical models
and multi-angle metrics to yield useful canopy
structure information.
6Specific Objectives
- And
- to evaluate the minimum acquisition period
required for the construction of adequate
multi-angle data sets for model inversions over
SW rangelands. This is predicated on the need to
augment the angular sampling of either sensor as
well as increase the number of observations.
7Specific Objectives
- N.B. these objectives are no longer immediate
- to create a map of above and belowground C pools
for arid shrub-grasslands of S. NM, S.E. AZ and
S.W. TX as predicted by CENTURY. - to assess the reduction in uncertainty in C pool
estimates using an areally-weighted
species-specific parameterization of CENTURY
based on well-defined vegetation community types
and the proportions of grasses, shrubs and bare
soil.
8Summary of Activities Milestones Preliminary
Results (Chopping/Su)Timeline of Main Events
Dec '03 Proposal accepted. Mar '04 Search for
post-doc initiated. Jun '04 Funds received. Jun
'04 Dr. Lihong Su selected from a field of
11. Jul '04 Lihong appointed as Research
Associate. Jul '04 Data acquisition/programming
Aug '04 ARS funding now deemed infeasible. Oct
'04 Still no funds, so no CENTURY post-doc.
We have to revert back to the original remote
sensing scope suggested by the panel (OKd by GG
on 10/13).
9Summary of Activities Milestones Preliminary
Results (Chopping/Su)EOS Data Processing
Technical Overview
Screened surface bi-directional reflectance
estimates accumulated over a 9-day period. Max.
observations possible for RED wavelength 27
(9 x MODIS/Terra 9 x MISR and eventually 9
from MODIS/Aqua). Plus other MISR channels at
nadir.
P R O C E SS I N G
HDF-EOS MODIS Observations (MOD09-- 250 m ISIN)
HDF-EOS MODIS Angles (MODPTQKM -- 1,000 m ISIN)
HDF-EOS MODIS QC (MODGST -- 1,000 m ISIN)
EMRPV, Li-Ross, ANIX, NDAX
GORT/SGM/other non-linear model
HDF-EOS MISR Observations (MI1B2T, includes QC --
275 m )
Physical Structure (FVC, radius/height, gap,
fiPAR, LAI)
Empirical Surface Metrics (iso, geo, vol k, f,
q ANIX)
HDF-EOS MISR Angles (GEOM -- 17,600 m)
HDF-EOS MISR Cloud Mask (RCCM -- 1,100 m)
1st level classification (Community Types on
soils)
2nd level classification (condition)
HDF-EOS MISR Aerosols (17,600 m)
Comm.Type subdivisions
C Pools AGC BGC
10Summary of Activities Milestones Preliminary
Results (Chopping/Su)EOS Data Processing
Technical Overview
HDF gt TOC reflectance for working region (ISIN
gt UTM w/MODIS Reprojection Tool)
HDF-EOS MODIS Observations (MOD09-- 250 m ISIN)
Collate observations, angles, and screen by QC
on an orbit (IDL -- WIP)
HDF-EOS MODIS Angles (MODPTQKM -- 1,000 m ISIN)
HDF-EOS MODIS QC (MODGST -- 1,000 m ISIN)
Accumulate observations from multiple orbits (9
days) (IDL -- WIP)
HDF-EOS MISR Observations (MI1B2T, includes QC --
275 m )
HDF gt TOA radiance, mask for cloud (IDL SOM
gt UTM)
HDF-EOS MISR Cloud Mask (RCCM -- 1,100 m)
Estimate surface reflectance from TOA radiance
(IDL/C/6S)
Combine MISR and MODIS data for each 9-day period
HDF-EOS MISR Aerosols (17,600 m)
HDF-EOS MISR Angles (GEOM -- 17,600 m)
Merge the observations from the 9 cameras for one
orbit (IDL)
11Summary of Activities Milestones Preliminary
Results (Chopping/Su)EOS Data Processing
Technical Overview
ISIN gt UTM for working region (MODIS
Reprojection Tool)
HDF-EOS MODIS Observations (MOD09-- 250 m ISIN)
Collate observations, angles, and screen them by
QC on an orbit (IDL -- WIP)
HDF-EOS MODIS Angles (MODPTQKM -- 1,000 m ISIN)
HDF-EOS MODIS QC (MODGST -- 1,000 m ISIN)
Accumulate observations on multiple orbits (9
days) (IDL -- WIP)
Obtain TOA radiance after cloud mask navigate
(IDL SOM gt UTM)
HDF-EOS MISR Observations (MI1B2T, includes QC --
275 m )
HDF-EOS MISR Cloud Mask (RCCM -- 1,100 m)
Estimate surface reflectance from TOA radiance
(IDL, C, Fortran 6S)
Combine MISR MODIS during a period( C -WIP)
HDF-EOS MISR Aerosols (17,600 m)
Merge 9 cameras observations on one orbit (IDL)
HDF-EOS MISR Angles (GEOM -- 17,600 m)
Collect multiple orbits during a period (C)
12Summary of Activities Milestones Preliminary
Results (Chopping/Su)EOS Data Processing
Technical Overview
Lihong -- anything to add here? Note last 2
slides are the same material organized / colored
differently!
13Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
We have been investigating the potential for
using a model based on geometric-optics (GO) to
retrieve information on shrub cover, density,
size and shape, initially using 4 x 631nm (red)
multi-angle reflectance images _at_25m from CHRIS on
Proba. Principles
14Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
How does a GO model respond to very heterogeneous
canopies? -- are GO principles violated in
this case? -- do GO models operate on mean
parameter values?
15Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
Can GO models work for very heterogeneous
canopies which have a highly variable and bright
understorey?
16Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
Can GO models work for very heterogeneous
canopies which have a highly variable and bright
understorey, on different soils?
17Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
GO models have been demonstrated as useful tools
for forested environments but are more
challenging in arid environments here, the
magnitude and anisotropy of the remotely-sensed
signal is dominated by the "background" comprised
of varying proportions of exposed soil, grasses,
litter and forbs. How to obtain the background
BRDF in order to isolate the effects of the
larger canopy elements?
18Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
The first approach taken was by setting the large
plant parameters estimated from high resolution
imagery and inverting the SGM for the background
(represented by the 4-parameter Walthall model).
This was done for a wide range of conditions and
a linear relation between near-nadir 631 nm
reflectance and the four Walthall model
coefficients was obtained. This relation was
used on inverting the SGM for large canopy
parameters over the entire imaged area (September
28, 2003 scene).
19Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
A look at some CHRIS imagery from 08/22/03
(SZA26
25 m CHRIS FCC _at_ 24 Back (748 nm, 631 nm, 530 nm)
25 m CHRIS 631 nm reflectance (56Fwd, 40Back,
24Back)
20Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
The 25 m CHRIS 631 nm reflectance _at_4 viewing
angles Note low SZA for 08/22/03.
21Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
IKONOS pan image 05/23/01 over the transition
zone showing the 7 points selected for
contrasting over- and understorey
configurations
22Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
IKONOS chips showing locations selected to obtain
Walthall1
23Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
IKONOS chips showing locations selected to obtain
Walthall2
24Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
IKONOS chips showing locations selected to obtain
Walthall3
25Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
IKONOS chips showing locations selected to obtain
Walthall4
26Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
IKONOS chips showing locations selected to obtain
Walthall5
27Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
IKONOS chips showing locations selected to obtain
Walthall6
28Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
IKONOS chips showing locations selected to obtain
Walthall7
29Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
Another look at the IKONOS pan image 05/23/01
over the transition zone
30Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
25 m CHRIS 631 nm reflectance _at_ nadir viewing
Walthall model scaled with near-nadir brightness
(SZA26 VZA24 RAA33) for the seven test
plots (solid lines). Ground-measured sand BRDF at
the transition site (dotted line). Not sure if
the scaling is accurate
31Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
Inversion results for the points selected to
obtain Walthall good.
a measured by counting 1 m pixels in the May 23,
2001 IKONOS panchromatic image. b kG, kC, kZt
are the fractions of viewed and sunlit background
and crown and shaded components in the
IFOV, respectively.
32Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
- Results (vs. gt 100 random point samples from
QuickBird and IKONOS) .poor! Potential
Reasons - The scale of the observation 25 m may be too
small -- it may violate the principles of GO
models which assume a Poisson distribution. - Walthall model terms are not orthogonal -- this
makes it difficult to obtain consistent relations
with nadir brightness by setting large canopy
parameters and inverting for Walthall (also too
many coefficients?). - The SGM may be inadequate (Goel 35 error in k?)
- The CHRIS data or processing may be flawed.
33Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
Soon we will be able to test the last one (CHRIS
data flawed or inadequate) by using MISR and
MODIS data. The and range of the angular
sampling are better.
34Summary of Activities Milestones Preliminary
Results (Chopping/Su)BRDF / CR Model Inversion
(ongoing work)
Field photograph _at_ nadir (January 2004)
35Perspectives on RS for Ecological Modeling
(Peters)
Perspectives on RS for Ecological Modeling
(Peters) Perspectives on RS for Ecological
Modeling (Peters) Perspectives on RS for
Ecological Modeling (Peters) Perspectives on RS
for Ecological Modeling (Peters) Perspectives on
RS for Ecological Modeling (Peters) Perspectives
on RS for Ecological Modeling (Peters) Perspectiv
es on RS for Ecological Modeling (Peters)
Perspectives on RS for Ecological Modeling
(Peters) Perspectives on RS for Ecological
Modeling (Peters) Perspectives on RS for
Ecological Modeling (Peters) Perspectives on RS
for Ecological Modeling (Peters) Perspectives on
RS for Ecological Modeling (Peters)
36Discussion on Goals and Approaches
- temporal element (when to sample?
transitions/stages of maturity?) - what to measure, with respect to our goal?
- hard / soft, or hard soft classification?
- spatial scales and BRDF / CR modeling
- solutions to CR modeling background problems
- validation plan
37Carbon Pools in Desert Grasslands from EOS--
First Meeting --Jornada Experimental Range, Las
Cruces, NMOctober 28, 2004
38Summary of Activities Milestones Preliminary
Results (Chopping/Su)Canopy Reflectance Modeling
(with CHRIS/Proba)
HDF-EOS MODIS Observations (MOD09-- 250 m ISIN)
HDF-EOS MODIS Angles (MODPTQKM -- 1000 m ISIN)
HDF-EOS MODIS QC (MODxxx -- 1000 m ISIN)
HDF-EOS MISR Observations (MI1B2T, includes QC --
275 m )
HDF-EOS MISR Angles (GEOM -- 17600 m)
HDF-EOS MISR Cloud Mask (RCCM -- 17600 m)
HDF-EOS MISR Aerosols (xxx -- 17600 m)
39Welcome, Statement of Objectives
- SAVED We are taking a mechanistic approach to
estimating spatial distributions of C pools in
the arid and semi-arid SW using multi-angle data
from MISR MODIS. - Our approach is based on the premise that there
is unique information in the BRDF as sampled by
these EOS instruments which complements spectral
measures to aid mapping objectives. - We want to combine MISR and MODIS data to
exploit 1. semi-empirical measures from RPV and
Li-Ross models 2. structural measures from CR
models and ANIX and 3. spectral measures of
canopy vigor (SVIs). These are highly
complementary. - We are not currently able to include process
modeling as a result of the recent budget changes
-- we will strengthen the validation aspects of
our work and continue to seek funding for the
modeling.