suggested that species composition affects the NDVI-LAI relationship through
leaf-level properties (leaf optics, leaf structure and orientation) and canopy-level
structural properties that influence the vertical and horizontal distribution of leaf
area within a canopy. Relationships between RSR and LAI in closed canopy
regimes suggest that the inclusion of the short-wave band decreases the effect of
understory reflectance on the retrieval of LAI below a certain threshold value of
crown-closure (Nemani et al. 1993; Rautiainen 2005). It is evident that NDVI/
RSR-LAI empirical relationships do vary across different species and are sensitive
to canopy structure and fractional ground cover. These empirical relationships can
also vary both seasonally and inter-annually with respect to phenological development of the vegetation. Thus, a relationship established between LAI and NDVI
in a particular year may not be applicable in other years (Wang 2004). Consequently, the empirical relationships will be site-, time-, and species-specific, and,
therefore, poorly suited for large-scale operational use (Houborg et al. 2007).
An alternate approach is to use physically based models that describe the
interaction of radiation inside a canopy based on physical principles and provide
an explicit connection between biophysical variables and canopy reflectance
(Combal et al. 2002). The physical models of radiation transfer and interaction in
vegetation canopies are usually categorized into four broad types: (1) radiative
transfer models (Knyazikhin et al. 1998; Myneni et al. 1989), (2) geometrical
optical models (Li and Strahler 1992), (3) hybrid models that incorporate both
radiative transfer as well geometric optics (Welles and Norman 1991), and (4)
Monte-Carlo simulation models (Lewis 1999; Ross and Marshak 1988). In
Sects. 2.1 and 2.2, we describe in brief two state-of-the art physical algorithms in
retrieving LAI and FPAR that have evolved over time.
2.3 Modis LAI/FPAR Algorithm: Scaling to AVHRR
and Landsat
The MODIS LAI/FPAR algorithm retrieves LAI and FPAR values given sun and
view directions, Bidirectional Reflectance Factor (BRF) for each MODIS spectral
band, uncertainties in input BRFs, and land cover classes based on a 8-biome
classification map (Myneni et al. 2002; Yang et al. 2006). The retrieval technique
compares observed and modeled BRFs stored in a Look_Up_Table (LUT) for a
suite of canbiome-opy structures and soil patterns that represent an expected range
of typical conditions for a given biome type. The modeled BRFs are simulated
using a canopy 3D stochastic radiative transfer model. All canopy/soil patterns for
which modeled and observed BRFs differ within a specified uncertainty level are
considered acceptable solutions. The mean values of LAI averaged over all
acceptable solutions and the dispersion are reported as the output of the algorithm
(Knyazikhin et al. 1998). The algorithm currently requires: (a) atmospherically
corrected surface reflectances at Red and NIR bands, and (b) an 8-biome Land
46
S. Ganguly et al.
leaf-level properties (leaf optics, leaf structure and orientation) and canopy-level
structural properties that influence the vertical and horizontal distribution of leaf
area within a canopy. Relationships between RSR and LAI in closed canopy
regimes suggest that the inclusion of the short-wave band decreases the effect of
understory reflectance on the retrieval of LAI below a certain threshold value of
crown-closure (Nemani et al. 1993; Rautiainen 2005). It is evident that NDVI/
RSR-LAI empirical relationships do vary across different species and are sensitive
to canopy structure and fractional ground cover. These empirical relationships can
also vary both seasonally and inter-annually with respect to phenological development of the vegetation. Thus, a relationship established between LAI and NDVI
in a particular year may not be applicable in other years (Wang 2004). Consequently, the empirical relationships will be site-, time-, and species-specific, and,
therefore, poorly suited for large-scale operational use (Houborg et al. 2007).
An alternate approach is to use physically based models that describe the
interaction of radiation inside a canopy based on physical principles and provide
an explicit connection between biophysical variables and canopy reflectance
(Combal et al. 2002). The physical models of radiation transfer and interaction in
vegetation canopies are usually categorized into four broad types: (1) radiative
transfer models (Knyazikhin et al. 1998; Myneni et al. 1989), (2) geometrical
optical models (Li and Strahler 1992), (3) hybrid models that incorporate both
radiative transfer as well geometric optics (Welles and Norman 1991), and (4)
Monte-Carlo simulation models (Lewis 1999; Ross and Marshak 1988). In
Sects. 2.1 and 2.2, we describe in brief two state-of-the art physical algorithms in
retrieving LAI and FPAR that have evolved over time.
2.3 Modis LAI/FPAR Algorithm: Scaling to AVHRR
and Landsat
The MODIS LAI/FPAR algorithm retrieves LAI and FPAR values given sun and
view directions, Bidirectional Reflectance Factor (BRF) for each MODIS spectral
band, uncertainties in input BRFs, and land cover classes based on a 8-biome
classification map (Myneni et al. 2002; Yang et al. 2006). The retrieval technique
compares observed and modeled BRFs stored in a Look_Up_Table (LUT) for a
suite of canbiome-opy structures and soil patterns that represent an expected range
of typical conditions for a given biome type. The modeled BRFs are simulated
using a canopy 3D stochastic radiative transfer model. All canopy/soil patterns for
which modeled and observed BRFs differ within a specified uncertainty level are
considered acceptable solutions. The mean values of LAI averaged over all
acceptable solutions and the dispersion are reported as the output of the algorithm
(Knyazikhin et al. 1998). The algorithm currently requires: (a) atmospherically
corrected surface reflectances at Red and NIR bands, and (b) an 8-biome Land
46
S. Ganguly et al.