'ed
the
del
im
nts
tal
'ea
m
he
he
cal
he
he
ht)
4).
he
he
he
1al
For the specific application to the POLDER data, parameters of
the PROSPECT model have been adjusted to account for the
chlorophyll concentration, Cab, the senescent pigments
concentration, Cs, the dry matter content, Cdm, the water
equivalent thickness, Cw, and the effective number of layers
inside a leaf, N. More, only director factors, al and a2, of the 2
first functions of Price [9], which have been optimized, are
considered; others are set to 0. The learning base of the neural
network has been produced by sampling LAI [0 - 6.5], Cab [15 -
0pg/em?], N [1 - 4.5], Cs [0 - 2], al [0.1- 0.8], whereas other
parameters are fixed (Cw = 0.01g/cm?, Cdm = 0.015g/cm?, 1* =
0.1 and a2 = 1).
The network inputs are a single orbit of 11 directional
reflectances in 3 spectral bands (565nm, 670nm, and 865nm)
and their angular configurations. The output is the LAI
estimated for each POLDER track. Then, a simple merging
algorithm, with a Gaussian temporal weighting, averages these
retrieved LAI over 30 days to get a monthly value mainly
characteristic of the central 10-day period. Then, the Fraction of
Vegetation Cover (FVC), defined as the fraction of ground
covered by vegetation, is derived following the relationship:
FVC = exp (- 0.5 * LAI) (1)
3. VALIDATION OF THE BIOPHYSICAL
PI
PARAMETERS
s» X ki
E : : ;
a c « pd à
2 5 = z = E 3
BAG y À or 1v xf : 3
S MM > sé at & 5
= 8,2 * à w 5 5 15
TF td) 3; i I
TEL LA UE m Sane LE if i
Lx. Ro e i
i de oM i
Jet E d ; ver
z up i :
S ud frappe St ae A den eee rm de ue ve
Z us dee
‚Ar 4. N fr fetes scene
À Î
up 4 i '
A.. IL. N A a Sn i i
aal Ti: L Mo eo celeste Joop taies deci Linee -J
m + bibe eben meme eren tttm ven j
— F —
aem tnc tte pede td d — 1 i Stns tte
BE fof ep es boe] i be tegere
B debupeevbeenbesee disce) v aar art Sv ; ,
to nn na + + + mm
v | H
>. i haod...
9o | " y t
7 t £s
(M. nil-i-- due HA { i
= 4 4 :
ec I sj E11 eer re em ak
Le
is: | e As eae anin
Le ;. -— ala À 4 ee d À. il E
i 4 a rie
f ?
= Ene i ae mde fm i. "a i
- H " r?
e ! + a
se |
rast lal l4 "00 ee ls eer ecrire ene
= wa Present iit
zZ. ASS mem
t vena Puy
i elei i i
E 43 - -5 0 5 10 15 20 24 30 3%
Lauch:
Figure 2 : Spatial variations of visible and NIR DHR, and
NDVI along 25°Est over Africa for 3 months
(April: blue, July: red, October: green)
The Level 3 biophysical parameters represent the properties of
the continental ecosystems. They are dedicated to various
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
515
environmental studies, whose results are partly controlled by the
relevance of the input biophysical parameters. Then, they must
be validated, first to estimate their accuracy for the user
community, and also to provide feedback so that retrieval
algorithm can be improved.
The first step of the validation procedure consists in analyzing
the spatial variability of the parameters, i.e. the representation of
the gradients at the continental scale, and their temporal
evolution over the7 months of acquisition. Figure 2 displays the
spatial variations of visible and NIR DHR, and NDVI over
Africa. Maxima of DHR appear over the Sahara with DHR
670nm equal to DHR 865nm. We can note the variety of surface
on this area, sand with large DHR and rocks with twice lesser
DHR. On the both sides of the equator, the season inversion is
well marked, especially on the NDVI profile. The minima and
maxima values are similar on “Sahelian woodland” on North
and “Equatorial Wooded Grassland” on South. These profiles
display local ground characteristics such as the Kalahari arid
area around 21° South, where the NDVI is rather small in April
comparing to the surrounding regions. Over equatorial forest,
values are realistic with visible DHR lesser than 0.05 and PIR
DHR around 0.25, and NDVI close to 0.8. The NDVI enhances
the seasonal variations, especially over intermediate
vegetations, woodland, bushland, and grassland.
Arctic Tundra Praire Weiland
- na —
66 de cé &
ca £ E 4
i vg x
na t y
» a
co OH eut d
<a Aa TID A primate Up p
rr se TEETER OT WE TE NE N I'WURUWUIAUGCISUWU
Tergorate Faliow Gressland Telkyess Parle
c6 — oar ry ge
cs gas c uH : ;
p rt ns ES E 4
gal, ov 4 + f oF ;
Me de : * i + ]
52 pit # 4 ca, y 4
a Nimm i F LE
^ ox a =
gi t d i
te n A DAL oua rur uc pe 100 N rx nd
TTY EIT UTYWEWOCDEC RE NU pP WOCEUNSWS
Hedllermnena Craps Tam perate Crops Subp: Crops
Barr OR aa na REP TOP
m i i
GE 2e La f os! 1
ee [T1 i * 1
ua y ; f+ mem os 1
k^ fy A USE tem | |
eel? SR oen P axi e 62 i
e Seo uei 1 Ga { 1
sù i ni voi x i
JTMAU Sa ED AO JFRANW JI JAIN DT JFRAVIJABSGNE
Deckluous Forest Beckluous Forest Decklaous Forest
16 . í MÀ a Kp emen ee
Geb g^ i
cet £l M 1
2 i
&4 Fa i.
E 1
ex 1
i i
CO ln 33 io ot
TINA TTEFE ES ITF RAVI TREE ITWEYWIJARSGKRI
Laces Pores Senikieciduous Rafr Fonst Prünare Rein Forest
ni e : d Np = :
ca H
tak I m tél 1
s 3€ vesc ^s. i x 1
Py wet. "Pe À cel i
c T ag os; Nha y= T 7 e-- i
= us i uat 3
eer 1 oz i os i
soit ont : ool E DUE
TT MANGIA SD N 2 J FTRMANM JIA SDN D IrMRANIJR SOND
Evergreen Broadleaf Forest Open Sevaona Woody Saver
GR USE Dr c Bigot o]
t wh i
£ i { na} & ea
e T * + get
a? 5 ! 4” Ri }
Cah # ti Ds * 02 «
bare A Í p de a T
* att cit
$42 où 5 ü pe RR i
i i i
4 > i Oil. 1 Se BU oru rne th
a an nr rw TrTs es
Figure 3 : Temporal variations of NDVI from POLDER-1
(green) and POLDER-2 (magenta)
A second step consists in comparing the POLDER-1 and
POLDER-2 parameters. This exercise allow to check the
consistency of the products over stable ecosystems and also to
display the changes in the vegetation distribution due to climatic
variations or anthropogenic actions. Figure 3 shows seasonal
variations of NDVI over 18 sites characterizing the main
