357
In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B
surface temperature, surface albedo, burnt areas, soil moisture,
products, provided at various spatial resolution depending on
the sensor data used as input (Table 3).
Most of the products will be delivered in NRT in the sense of
few days (less than one week), as requested by the final users
who need to know the surface conditions within a few days of
delay in order to react appropriately in case of anomaly, and to
anticipate and manage the potential resulting problems.
Detecting anomalies by comparing the current observation with
a reference requires having consistent long time series. BioPar
will provide such time series for some Essential Climate
Variables (LAI, FAPAR, soil moisture) taking advantage of
existing EO data archive, and developing sensor-independent
methodologies.
4.2 LAI, FCover, FAPAR, NDVT derived from
SPOT/VEGETATION sensor data
The leaf area index (LAI) is defined as half the total foliage area
per unit of ground surface. The FCover is the fraction of ground
unit covered by green vegetation. The FAPAR is defined as the
fraction of photosynthetically active radiation absorbed by
green vegetation for photosynthesis activity. The instantaneous
FAPAR value at 10:00 solar time is used as a very good
approximation to the daily integrated value under clear sky
conditions. The Normalized Difference Vegetation Index
(NDVI) corresponding to the SPOT-5/VEGETATION-2 sensor
characteristics for its Red (B2) and NIR (B3) bands, is also
provided.
The algorithm is based on already existing LAI, FAPAR, and
FCover products to capitalize on the efforts accomplished and
get a larger consensus from the user community. Following the
published literature on products validation (Weiss, et al. 2007;
Garrigues, et al. 2008), the best performing products were
selected and combined to take advantage of their specific
performances while limiting the situations where products show
deficiencies. The selected products are re-projected onto the
VEGETATION plate-carree 1/112° grid, smoothed through
time and interpolated at the 10 days frequency. Then the
products are combined, and eventually scaled, to provide the
fused product that is expected to give globally the ‘best’
performances. The fused products are generated for few years
over the BELMANIP2 set of sites that is supposed to represent
the possible range of surface types and conditions over the
Earth (Baret, et al. 2006). Neural networks are then calibrated
over this set of sites to relate the fused products to the
corresponding atmospherically-corrected and directionally-
normalized top of canopy SPOT/VEGETATION reflectances
(Baret, et al. 2007).
Figure 3: LAI derived from SPOT/VEGETATION, July 2002
Such methodology has been defined by INRA, and the
processing line has been adapted by CNES, based upon an
existing chain developed previously by Medias-France in the
framework of the FP5/CYCLOPES project. CNES has also
generated two years of global, 10-daily products (example on
Figure 3). Before the end of the year, the processing line will
run in NRT at VITO. These vegetation products are being
validated by EOLAB according to the protocol defined by the
Land Product Validation (LPV) group of CEOS (Morissette et
al., 2006).
The methodology will be adapted to the historical AVHRR
surface reflectances made available by the LTDR project. The
archive from 1981 will be processed to get a long time series
(about 30 years) of vegetation variables fully consistent with the
SPOT/VEGETATION products.
4.3 The set of biophysical variables derived from the FR
MERIS sensor data.
The MERIS Full Resolution (FR) biophysical products, and the
High Resolution (HR) biophysical products, contain a set of
variables including estimates of the green, brown & soil cover
fractions, the LAI, the FAPAR, the chlorophyll content, a
canopy shadow factor, and the water & snow cover fractions.
Figure 4. FAPAR over the Adour-Garonne river basin for June
2008 (left), and the Guadalquivir river basin for June 2006
(right) derived from MERIS FR data. The values vary from 0
(white) to 1 (red).
The baseline vegetation model developed for processing the
MERIS data uses the SAIL/PROSPECT model as core
component (Verhoef, 1984; Jacquemoud et Baret, 1990). This
model was upgraded and completed in order to include the
contribution of brown vegetation, the modelling of “rough”
canopies, the computation of vegetation cover fractions in
reference directional conditions, and the computation of
FAPAR from the SAIL model. Then, to restore the
heterogeneous nature of the MERIS pixels, a further modelling
step is applied consisting in having a composite canopy model
made of two components: 1) a main canopy component made of
predominantly green vegetation, that may have all range of
conditions from crops to forest/shrub canopies; 2) a second
canopy component made of low brown vegetation, primarily
designed to model either senescent crops/grasslands or bare soil
conditions. Another important component of the developed
model is the soil modelling, the soil reflectance being an input
of the SAIL model. In this approach, the variation of soil
brightness in relation with soil roughness and humidity is
handled by performing an initial learning of the soil mean
spectral signature in the MERIS bands at a regional level. Then,
this signature is exploited through a physical model of soil
variability, accounting for the soil shading and surface humidity
effects. This scene model is then coupled with a model of the
atmospheric transmission from Earth surface level to the sensor
(Verhoef and Bach, 2003). The detection of water and snow
covers is done through additional modelling of the