The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
■ SPEC
OBNC
■ CRDR
■ NBOI
1 angle 1 angles 3 angles 4 angles 5 angles
Figure 4: Model coefficient of determination (R 2 ) of water
content regressed on the data sets SPEC, BNC, CRDR and
NBDI, respectively. R 2 represents the mean of all models with
the same number of viewing zenith angles involved. For
instance R 2 of one angle stands for the mean of five
monodirectional models (± 36°, ±55° and nadir).
3.3 Multi-temporal LAI Estimation
The third case study aimed at exploiting synergistically the
spectral, spatial and temporal information dimensions contained
in multi-temporal CHRIS/PROBA observations. Such Earth
observations represent a rich source of information for
monitoring the dynamic vegetation status. For the assessment of*
vegetation phenology the leaf area index (LAI) is essential as it
is a key variable for the understanding and modeling of several
eco-physiological processes within a vegetation canopy (eg.
(Myneni, 1997). In this study, a radiative transfer model (RTM)
is coupled to a canopy structure dynamics model (CSDM). The
coupled models are used to exploit the complementary content
of the spectral and temporal information dimensions for LAI
estimation over a maize canopy. The resulting estimation of the
temporal and spatial variation of LAI is improved by
integrating multi-temporal CHRIS/PROBA data and ground
meteorological observations. Further, the presented method
provides the continuous LAI time course over the season.
CHRIS/PROBA multi-angular data sets were acquired in Mode
5 over the earlier described study site Vordemwald (VOR) on
eight different dates between 26 May 2005 and 22 September
2005. Out of these data sets, four dates that represent major steps
in phenology of the selected agricultural fields were selected for
further processing and data exploitation. The selected dates are
26 May 2005 (day of year (DOY) 171), 20 June 2005 (DOY
196), 17 August 2005 (DOY 229) and 22 September 2005 (DOY
265). The data sets were geometrically and atmospherically
corrected.
A coupling scheme to combine two models, the joint radiative
transfer models (RTM) PROSPECT/SAIL and the canopy
structure dynamics model (CSDM), was implemented to
estimate LAI based on the multi-temporal remote sensing
observations (Koetz et al., 2005). The joint RTM provide an
explicit connection between the canopy biophysical variables,
the view and illumination geometry and the resulting canopy
reflectance by exploiting our knowledge of the involved
physical processes (Baret et al., 2000). The RTM have to be
inverted to retrieve the biophysical variables from the measured
canopy reflectance (Bacour et al., 2002; Kimes et al., 2000;
Weiss et al., 2000).
The use of a canopy structure dynamics model (CSDM) allows
us also to derive a continuous estimation of LAI which is
required in some applications, particularly those based on the
forcing of agricultural growth or land surface models (Delecolle
et al., 1992; Moulin et al., 1998). The used CSDM is a simple
semi-meachanistic model describing the LAI dynamics (Baret,
1986). Concerning the radiative transfer models in this study,
the turbid medium radiative transfer model SAIL (Scattering
from Arbitrarily Inclined Leaves (Verhoef, 1984; Verhoef,
1985)) was used to describe the canopy structure. The
PROSPECT model (Jacquemoud and Baret, 1990) was used to
describe leaf optical properties. The coupling of the RTM and
CSDM models was based on the hypothesis that the remotely
sensed observations of LAI had to be consistent with the time
profile of LAI generated by the CSDM. Consequently, the
remotely sensed LAI was recalibrated, where necessary,
relative to the phenologically sound LAI provided by the
CSDM. The integration of the CSDM to the retrieval algorithm
allowed a continuous description of the LAI time course over
the growing season. For the evaluation of the LAI retrieval
performance estimated LAI was compared to field measured
LAI.
This case study showed the successful coupling of a joint RTM
to a CSDM to exploit the complementary content in the spectral
and temporal information dimensions for the LAI estimation
over a maize canopy. The knowledge of the canopy structure
dynamic provided by the CSDM is used as ancillary
information to achieve an improved robustness of the RTM
inversion. Further, the presented coupled models integrate
spacebome remote sensing data with ground meteorological
observations providing a continuous LAI time course over the
season along with start, end and length of the growing period.
Crop growth as well as surface process models require such a
continuous description of the vegetation evolution. The
proposed methodology prepares for the assimilation of remote
sensing observations into land surface process models.
4. CONCLUSIONS AND OUTLOOK
The above discussed case studies in Switzerland all present
improved use of spectro-directional (and in one case multi
temporal) measurements exceeding the classical use of the
directional dimension for canopy structure retrieval and the
spectral dimension for biochemistry. In particular improved
measures for canopy heterogeneity (Minnaert’s k),
unprecedented estimates of C w and C N , and temporal evolution
of vegetation structure for integration in dynamic vegetation
models have been demonstrated. The six years of
CHRIS/PROBA operation in space have fostered closer
collaboration of various Earth System sciences and allowed
working at various scales that could be validated in the field.
However, the realization of further spectro-directional imagers
in space remains a challenge. Proposed missions had generally
not found a majority for support (e.g. SPECTRA (Rast, 2004)).
Currently, combined instrumented approaches (e.g. FLORA,
FLEX) or approaches with more flexible acquisition pointing
have been suggested (e.g. EnMAP), but true spectro-directional
concepts to widen the fields of ecological monitoring and
modeling remain sparse also in the future.
ACKNOWLEDEGMENTS
The work presented has been supported by the Swiss National
Science Foundation (SNF, project 200020-101517), the
Netherlands Organisation for Scientific Research (NWO SRON
GO, grant EO-080) and the European Commission (contracts
EVK2-CT-2002-00136 and GOCE-CT-2003-505376).
CHRIS/PROBA data were acquired in the frame of the ESA
AO proposal No. 2819. The continuing effort and support of
SSTL (formerly SIRA) for CHRIS/PROBA is gratefully
16