You are on page 1of 12

ELSEVIER

On the Relation between NDVI, Fractional


Vegetation Cover, and Leaf Area Index
Tobu N. Carlson and David A. Riziley*
I

e use a simple

W
vegetation,

rudiative

soil, and atmospheric

hoW the normalized

diflerence

transfer

vegetation

leaf area index (LAZ), and fractional


dependent.

In particular,

tional vegetation

cover

model

components

with

index @DVl),

vegetation cover are

we suggest that LA1 and fracmay not be independent

titites, ut least when the fermer

quan-

is dejned without regard

of hare patches between plante, and that


of LAI with NDVI can be erplained as resulting from a variation in fractional vegetation cover. The following points are made: i) Fractional
to the presence
the customary

variation

vegetation

cover and LA1 are not entirely

yuantities,

depending

dependently

may partially

of the latter; ii) A scaled NDVI

tween the linzits of minimum


fractional

Gare must

vegetation cover in-

in a model because the forrner

take account

independent

on how LA1 is defined.

be taken in using LA1 and fractional

vegetation

cover

taken be-

(bare ,soil) and maximum

ia insenstive

to atmosphetic

correction for both clear und hazy conditions,

at least f3r

viewing angles less than aborct 20 degrees from


A ,simple relution
vegetation

cover,

between
previously

scaled

NDVI

described

nadir; iii)

and fractional

in the literature,

is ficrther c.on&-med by the .simulations; iv) The sensitice


dependence

of LA1 (~1 ND17 when the fonwr


is below
a value of about 24 may bc viewed as being dut to the

variation in the bar{: .soil component.


lx.,

Ol%evier

Sciencc

1997

BACKGROUND

The normalized
defined as

differente

vegetation

index

Departnrentof Meteorology,The Pennsylvania


Universitv

Park

(1)

to illustrate

(NDVI)

is

State University,

where amr and a.+ represent surface reflectances averaged


over ranges of wavelengths
in the visible (1-0.6 Pm,
red) and near infrared, IR (2-0.8 Pm) regions of the
spectrum, respectively. It is clear from its definition that
the NDVI (like most other remotely sensed vegetation
indices) is not an intrinsic physical quantity, although it
is indeed correlated with certain physical properties of
the vegetation canopy: leaf area index (LAI), fractional
vegetation cover, vegetation condition, and biomass. As
such, vegetation indices are highly useful measurements
despite their limitations.
The NDVI lias been criticized because of the following perceived defects:
Differences between the truc NDVI, as would
be measured at the surface, and that actually
determined
from space are sensitive to attenuation by the atmospheric and by aerosols.
The sensitivity of NDVI to LA1 becomes increasingly weak with increasing LA1 beyond a
threshold value, which is typically between 2
and 3.
Variations in soil brightness may produce large
variations in NDVI from one image to the next
(Liu and Huete, 1995).
Accordingly, various investigators
have addressed these
problems in light of indices that exhibit a better correlation with leaf area and less sensitivity to soil brightness
changcs or to atmospheric
attenuation
thall does NDVI
(Jasinski, 1996; L eprieur et al., 1996: Liu and Huete,
1995; Pinty and Verstraete, 1992).
That the relation between NDVI and L,AI undergoes
a marked decrease in sensitivity above a loosely defined
threshold is wel1 known from measurements.
Carlson et
al. (1990) stated that
NDVI incrcases

HEMOTE SENS. ENVIKON. 62241-252


(19973
OElswier Science Inc., 1997
655 Avw~~t~ of thr Atrwricas, New York, NY 10010

almost linearly

with increasing

LA1

00.3442~57/9/$17.00
PII Soo:~4-4570(K)57(97~~~~104-1

[\Jl
:lV[

l[J!,\\

[,\([N

JO LlO!~t?!.lt?,\ .ilt?lllO~StI.~

[<10[8 1 11?.>0[ l[+U

[,\(]N

Jl($ $CL[J &

,jO lN~!~t?!.lt?.l aL[J [

-t?~ax,A
Iv[

In!

[t?llo!~.>t?.lj
p?l[$

I! aA[oALI~

sLlo!~t!Amsqo

~~>[<[!ss!lll_rad d.lI? axj.ms


a[tIlrl!s

.>lIIOS

~U<mIcI

O$ S! _l.d?CI

S!l[J

JO

as0d.I11d

9L[L

-t?JaX>A
.,.Il?([ jo

JI[$

01
UlO.IJ

JI[J

II! SFqOL[ [[L

sappd

[,WJ

.,(I

[[IlO,-p-z
1? ,Z\O[dL[

-s1?\>.11?

S!

[V,J

,tO

,jo

[[0[<>11[$

[\?[O[;i

arp

[>IJ<[
d.l_)Lp\

~[[t?g_tt?I LI! [v[

[,dp?p;i,,.\

Ji\<[N

lIO~J?!.lt?A

JLlJ

$t?I[J

ipspxrdul~

Sll0~;iJ.l

[l[Ol;i
J[q!SIlR[d

dL[J

LI! S!
04

q$

~.x&~.l

SLlIrh>S

[two!$3t?y

-Sr>.4

SF [VT

-s11.7

mcpns-jtxq

iwt?

q[lLsa~
[>Llt!

q
] v[

xlp?1

[LU!
1lL111l[.

uo~~t?~dk~A

[!OS

[t?Llo~sLlarll~[,-~llo
jo

Ll~RLllO[)
jl?[

jI?[

JT> ![[1J)

[[I?

a!S

01

SdI?[

;frryy$

s[>LloIs,>.I

JO x11

al[J

.i([

04

J!

a[[t!.I[>ylo*

-[

smtl,J.l~~j~![

JO

[[KJ

~,idOUt?J

<,l[J

Sh?M[t?
arp

IVI

Jl[J

ll!

Kl3l?dS

[>UR
XS7

IFI

l?

[m:
.IaJ$t?[

[VI

a[q!s!\

S!

Skhl!LdO

JO

a_lt?c[

S~>ll[l?.\
llO!~

OJ

[tqo~

[V7

04

0 [

lrtx[$

ssq

,,[<[?!al

IIIIS put?

1!OS

l? 10 &lI?d

(iLll

.Irx40

JO
S! J[

IchZO.~ llO!J

s~IIt?[<I lWhq.X[

I?

1 9.,)

[ml1lst?alLl

p3~1pp

si:

.$.nxw~

dl[J

[A([N

JO

.l.).\O.)

lp!Lp\

.LOJ

OS[?

lll?.J

r)IIIos

.&&Id

S!LiJ

.I&w[

_lt?[os

.to

JO

mvj

mUt?~.qj

.qY!l(

JI?

3L[J,

[,aLpt?

l[;illOL[~
pllTloM

aot!j.IIls

Jl?[

f< WIl[I?A
~St?&IJJ[>

Ul[

Iv1

C [lt?

(zfiA[

04

SpUaJ

&u?L[

($)efjr

) JJ!.ld

~Ch3XJ

p!sv

hhlt?L~
S!

$rlt?n

JllJ

J!

JL[$

[VI

JUC[

[xqj

[[>~$~!lLI[>V

(<:

x,l[J
J?[

<x4oc[t?)
(CJ#j

[)

(j#j[)

[l?

ll.[p~[

s~~lLl~2a.l

JI?

0)

N
\H[J

d;lUt?I~~
(<>[l;Ut?

S&~LIl:3

put?

I0.j

I)J

[ ) 11 Ja
.)f

ll_IO.>

smp

nyq
~I?Jl[A\

Sll!xx[

OS[?

h\

[v[

Uw.lu~)

S[>lldJ

]A([N

;i[>qn&

~Ll3~110.~

LI0

s[>LIda[,

[>U?

S!l[L

UO!~
[v[

J~O~C[lll.iSt?

.IO

.I.dt[ll
,iC[ [>d.ld

sll!~!~.tad

llO@~.l

(CKfj[ ) llKl.rll~ )

.il.,.\

s.>st?,,.l.,ll!

111:

llo!gt?~J

.IIl()

.~go~dLLlrist!

LK~!$?~&hA
&W[

sanrg.rus

&[

sLroql?nq&I~?
JO

S!

SJ!.IO~S

Uo!~t?q>t?.t
arp

ll!

[VT

~>.lOlll

Il![

lllO.l,j

LI!

~lla.lt?ddt!
[[O[SL[~

ll!

~AOqt?)

S.qS

SJt?O

~t?ar(Az

40

s~\uN[s

[>Ll
Uogt?&h

.ladcIll

.i[[1?1l;l

Sd:\W[

1? 01

s~lrytl

$IlO

.ip\o~s

~J!y~~dlll.ist:

[AC[

_lJl11.l0~

IVI
!llt?UlJN

S~Il[t~\

JAOC[t?)

(()$gj

lUIlL[;i.K~s

p-g

JI?L[3

U~~:A\-1-[
acp?~
Gacu$

ylq
s.lIwx>

I,laN

JI?l[J

.1~~.40.~

lK~!pIaxacI

JO

JLLl$h

~LlIloUu?
lt,!L[A\

,>.n?

stroa;it!~Ut!.4[>l!

rp11s

Xp+lUIoo
S!

uroq
.IaJJt?[

i~1,2!~!SllClS

sU~t?~p?

i[~llm.l

[>t!

Jl[J

I1i(lN

p:rp
<,U[A

pULU

l[dJAQJ([

[L

ll!
Sl?

S! IA N

[>Il,>sqo

lro~~l?~a;laA

Jyln

s&I~Ilba.l

MiL[

[[

%l!PJ([

anUt?$.IocIUl!

S! IiC[N
$SIl[

2LIp&Lln[>!

Yll!.l.lajLl!

.iC[

JL[J

.IJL[J_IIlj

W~I?pt

/$.4Q!SLLJS
jo

~SOUl[t?

;iU~Lpt?a.I
dL[J

dL[$

.+dn?~
arp

s.I113 0
JO

1:

111\2OL[S
+>ItfJ

sanL1a.Lr~ja.l
l[]!A\

1 [>lmh[
J[

S!

[>Ut!

lLlO.tj
(gqfj;)
(z

,~Aoqt?)
.q

~~.l~J1M Ivy
[l

Oj

dL[J

[ ()[

JYI?

.ld4&401~
.ll?~lI

SIIS.I&\

L[khloL[q~

mt?j.Ins

dL[J
l[J!ZZ

Jrll!x>.I

i([[>!C[t?J

S[JJX

[lt?U

;ilL!UUIl~
T

.!f

[[t]

(yz

sln?x[k~s

/iC[ [>!f

[NL?

.( $&fj

[>Ut:

S!

8()-S()

~soUl[t?

.1aq

;irly?,mr!
II!

S~~l(nt?O.lddt!

sarlp?2

.lt?~>lU[
)

01

[ ) [t $2
pag
J.wqt?)

sIloyt?~
J.lOL[S

.J~Jo~cIlLtXsr~
z()

_lO~~

j?_)[

[V]

Ut?

JO

L[@k\
(fjKfi[
5;
sno~.lt?z

(286
[>ll
(z

.I_)y IIJd
wsst?.l;i

s<Io.l.>

Lpls
L10$h.l

.u~?q

.IJtpt?_l
.roj

~~LLIOS S! .&lL(J

.i~~~J[&llon

.xy~$cllLl#ist!

l(.,ll[.\\

[1?lKqx?.t~

[1?p_toz~0L[

papy-auo
UoynLp?p

Jpelrl

S!L[J

.,U[t?A
Itqo~

t! ,i[[N~~LlO[,L~Il

sazt?a[

-lx+/

[>~1?11#
;i1q [,,[~,l11
asIlt? a<[
[!\C[N

JI?L[J

(CJj(iI)

pp3
[VT

Llgwn

~+.l~?.qc[.It?

c*$;

[,\([N
A\OI[S
;ilU[Wd~[>

9
.roj

[!OS
.t,,Jl?,\4

,,[L[\?!.ll?.\

LlO!~t!~J;i,,A

as1q>

-1:&h\

llo!~~!!.It:,\

- 10.1

[!OS

ssq

,&lld[)
q$

_r~[>~sUo~)

ll1?1[~

13AO.)

.,!~O~dLU St?

r,;\oc[t?

110

[,azqt?ap!

[l?O[

<isal[~

~Iloqt?

.I~LLI.IOj

[>[IlO\

[,mlst?alLl

s~lIr?[cI llaaA\J

01

LlJ&i&>([

rloy?n~~s

V-F

jo

LI!

WlllO2&[

.,11[1?.2

-->[

rpp\
[!JLlIl
J1?1[$

4-1 IlOSLI!~?!~)
<>Jl?,I

2lI~[m:~s.I~[>UIl

[>[llOM

t?

GILIJ O$ [I?~

i[[l?_IIl~

&~Lp
sMoJ

sl;U!LIado
SJUt?[d

ll!

.>I[J

10/

anLlasr>.IcI

paa.3xa
011

O$ .lo
st?Ul

L~3!L[M

pqI;:

IvrI
jo

dL[J

Spm[

rrado

[1<&.1.I0~

L[%tO.lL[~

->L[J

IVI

p[Il>M

LI! syt?aq

[hxnIst?.xU

11

04

Lf?!qM

J.ll?

lIRL[] ssq

J/\!~~SLl~~Sll!
fG()()[

N JO

sst!ulo~c[

llaI[

.,I[$

04

S!L[J
[>Ut?

2L[$

-tJ!U~!Sll!
JO

J!lLU[

ap?lll

~JllOO~
[.It?[[Od)

ft?

s~llmdlllon

(.IajsLI~!~~-a~ar[~[soLll~t?-tlo~~t?~
~JLlt:A+I

!ySLl!St?l
;iLU[J.xI

[,a$uasa.IcI.xl
s<q3Ja_I~s

01

Jlll
(cJjf$[
[lt!

(CfjfjT

SUqt?.rac[o

s[GJ[>oll1
IV7

iq

11 sso.I.x?
1111

'(ULfj[ ) l[~!dJllOW

[~>[[t?!-os
pul!

llOS<[llIOl[~

S_lJ[[2S)
iI?[,OJ

Il!

1
1!

m?+Uls

~U~qt?A!tlT,,,

S!L[

SNVHflSd
(cjGc[

[Ut?
_IOJ

$?IlJ.U~

,,l[l

[>Ja;,XJ

dt[J

sa3cds

dlUOS
,khIR~

O$ pp[

UI u/o()()[

L[ map
110 ~U~[>Ld~[,

[[OL[S-.U[~

;ill!khlt?l~~

$1

lLo!~l!~dh

~.&qcIsolUJt?
p[ll

.,p?ll~lLlJa~a[3LL!
~.x-Nlt?~s!s&l

S!L/J

XdOUt?~~ r>L[J I[*>!l(itZII!

r3*>lleJS!Sx>.l

['~[xl[['? Llo~$hLIIlss~?
~.[I?

>]:)V?<[
Ilt?lU.lO[~
~\~[cllLlt?xa

LVllS

I!f

_It?.lsv

(fLG[)

rl!ulllo[,

llmr\y,x[

J>l?j

,>.lI? ~aoU~?n

SA\O[[O,J

-d[

.kIt?([

$q!l~X>J

an!p?.ld

.i;>Ar)MO~

[>[1101\4

~socIdIls

41: qyj
r[YuoL[qt

OLI YLL!,\l?I[ kIoLrt!n

dL[$ OJ slr~t?$.Iac[ .rSAO3 rlog?~a2aA


~?d.ll? ~~~t~jllls-[>lltlo1Y

-Xr>
.Il[p?.l

I[+N

IAC[
J!

hl+LJ[>

pos

-.lIlS

,,L[$

-.>.I

aup2

Il_>_)AqJq

;iLl!tiLIt?L[,J

[>[ot[s<w[~

-lV
-1102
,>Z!S
,}W[
-.lO>
03

(J(j(i[
IrYJc,fi)
(LK(j[
cJJ,Vg
-llS-[?[
-a2a,\-ps)

11103

1L1~[.l![[
$2

-cIlllAs~
-Ir!

[l?<[O[;i

~1loL[Jyz

-x>.I

2l[J

%>[<[[yf\
$RI[J

[!OS

su~~?~Uo~~

sJ.>t?~_ltls
\>1110S
Iv7

[,\<[

S.l,,Jll.>

Il<>l($

[lil?

..tapdd+
.......................
Figuw 1. Schematic representation of radiative flux coinponmts
in the simple model. Solid lincs
with arrows represmt
the direct
solar flux component at clrvaticm
angle q,,, dashed lines represent
diffuse flux coinponents
over a
bare soil fraction (right side) and
a vegetated fraction (Fr; left sidIA).
F111xrs are representecl
by tlle
s,ymbol F with subscripts denoting
various flux coniponents
either
ahsorhed in the atmosphere, ahsorbecl at the ground or within
the vegetation cdnopy, scattercxd
upwdrd or downward, or reflectrd
upward. (Al1 ahsorbed fluxes arc>
~mderlined.) Thr fil~u to spacc
measiirrtl bv a satellite radionietrr is inclicdted by tlke dashed lirie
lalde<1 F ,,,,, (sw Appen<l).

can l)e largely


tional
that

NDVI

creasing
cover
that

explaincd

vegetation

decreases

ding

most

much

LA1 wllen

less rapidly
the

with

fractional

NDVI

between

of thc atmospheric

values

qetation

mixed

kween

at thv surfacr

NDVI

for hare

cover

factors

correction;

1
LlIIC

~W&IK~ to confirm the existente

fiirtlwr

plr relation

in-

vegetation

calculated

fol- 100% vegetation

scaled NDVI

and

The

He

(0.5-0.7

heain

of a sinifractional

is incident
dent

MODEL

rect

is a very simple
resentation

tion

consisting
The latter

canopy
Fr,

fraction,
Figure
dergo
through

of bare

a one-layer

1)~ a local

total

domain

the local

LAI.
and

rep

incident

over

Ween

vegetu-

occupying
global

LA1 is

flux streams

reflection

and the vegetation

that

as they
lqer.

unmove

Scatter-

reflectante
infrared

radiation

values

(0.X1.9

direct

component

canopy.

along

on the
The

at tlie ground,
ii con~l~onent

absorhed

p,,

the

flux

<lownward

unattenuated

anti the

diffke

by the

as diffuse

is scattrrecl
ground

angle

is ahsorbed

upward

with

regions

Some of that inci-

heam

comhinrd
whose

Pm)

at solwr elrvation

(F,,),), a part is scattered

are

is as follows.

at the top of the atmosphere.

another

that

hands in the visi-

scheme

solar radiation

flux, is incident

top

as
di-

of thr

and direct

flux

all~do

is a,, is divided

be-

at thr

gromid

in the hare

soil area (F,) and a component


reflected upward (F,,.).
An identical incident flux is absorkd
hy the vegetation canopy (F,,,,,) and by the ground underneath
the vegetation c~mopv (F,,.). The remaining
fll~x, tltat not al)sorhecl

the various

scattering,

the atmosphere

LAI)

whosc

and near

flux(F,~,),which,

vegetation

layer of veg-

by an unbroken

of the

absorption,

atmosphere

soil and a single

is represented

1 illustrates

in the Appendix,

and downward)

(represented

to Fr times

equal

(upwarcl

of fluxes through

;i surface
etation.

two-stram

described

from

in two wavelength

flux in the downward

cliffusc
model,

pin)

of direct

atmosphere

transfer

and at the top of the atmospherp.

of tlw spectn1111. The

out

cover.

radiative

is derived
separately

(FI,,), and
THE

ing and reflection


are either upward or downward.
Reflectances and their derivative product, NDVI, are deter-

in fiac-

100%;
the

soil and

by a variation

11111

cover;

glol~l
reachcs

Fd
/ III

tqd vegddim
IIIIIIIIIII

in thcOvegetation
(F,,). Upwarci

a1ic1at thr gronnd,


flux streams

is refiected

l-eflected from the atmosphere. the hare soil, and the vrgetation canopy (and
its lmtkrlving soil) combiw
to fornr thr total upwad flux
upward

flux is further attenuated


by ahsorption, which rrmoves im amount of flux (F,,,,,). Mo accomlt is taken of backwartl scattering of this upward radiation stream.
The flux reaching the satellite is determined
indirectly by subtracting
al1 absorbed atmosphere
and surface components
from the incident flux at the top of the
atmosphere. The apparent reflectante is calculated by dividing the upward flux at the top of the atmosphere (F,,,,,)
by the exoatmospheric
solar flux [Eq. (AlO)]. Surface reflectance is determined
as a ratio of the sum of reflected
fluxes at the surface divided by the incident surface flux
[ Eq. (AS)]. Satellite angle is considered only insofar as it
affects the path lengt11 of the reflected flux component
F, toward the satellite. NDVI is calculated, by Eq. (1).
Model initialization
requires specifying the reflectances for bare soil and leaves, the fractional vegetation
cover, time of day, satellite viewing angle, local LAI, latitude, longitude, horizontal visibility (from which aerosol
optica1 depth is calculated), and a few other variables of
much less importance,
such as surface pressure
and
ozone concentration
(see Table Al).
As presented in the Appendix, the radiative transfer
formulation
does not constitute a rigorous treatment
of
the radiation physics; nor is the treatment
particularly
rtew.
The purpose of the radiation model is to provide a
reasonable estimate of the major radiation components
in the atmosphere,
in a vegetation layer, and at a bare
soil surfwce. Our confidence
in the models efficacy is
based on its long-term satisfactory performance
as a solar
radiation component within a SVAT model (Gillies et al.,
1997). Direct (unpublished)
comparisons
of simulated
and measured solar fluxes have been made, and these
comparisons show agreement within about 5%. More important, however, the SVAT model lias been widely employed in conjunction
with remote and in situ measurrments to investigate land-surface
processes, particularly
the role of soil moisture in modifying the surface-ener&?
budget (Gillies et al.. 1997), the determinatiou
of trainspiration fluxes over plants, and the intake of carbon in
plaut canopies (Olioso et al., 1996). Further validation by
with anothcr
model is presented
in the
comparison
ncxt section.
(F,). The total upward

RESULTS
Initial Conditions

Radiative transfer simulations were made for the latitude


and longitude of State College, PA (approximately 40 N
ad
6 W), for a July day over a range of times from
IIOOII until 2:00 P.M. and satellite
zenith angles from 0 to
20 degrees from nadir. Because results were similar for
all satellite and sun angles investigated,
al1 illustrations
refer to one time and one viewing angle-I:OO
P.M. locd
time 20 degrees
from nadir.

7&le Z. Albedos of Bare Soil and Leaves (%)


Visible

Near IK

11

50

Fractional vegetation cover (Fr) was varied from 0.0


to 1.0, the local LA1 remaining fixed. Global LAI, equal
to Fr times the local LAI, is not directly used in the calculations, but this parameter is referred to in the illustrations. For example, a local LA1 of 3 and a value of Fr
equal to 50% corresponds
to a global LA1 of 1.5. At
Fr=lOO%, global and local LA1 are identical. Values of
the local LA1 for an Fr value of less than 1.0 were fixed
at a value of 3.0 in most simulations, but a few simulations were made by using a local LAI of 2.0 and of 4.0.
Additional simulations were made at Fr= 1, by varying
LA1 in increments
from the local LA1 for the partial
cover case to LAI=lO.
Al1 calculations
refer to clear
sky conditions.
Albedos pertaining
to the bare soil surface ((lg, including the albedo beneath the canopy) and to the leaves
(ar) were fixed for al1 simulations; they are identical for
diffuse and direct flux. Table 1 refers to these fixed reflectances, thereby furing NDVI for bare soil and at the
limiting values in the asymptotic reg.ime (infinite LAI).
The purpose in choosing this particular combination
of
albedos is that they yield the range of NDVI typically
measured by satellite over a mixture of bare soil and vegetation (Cillies and Carlson,
1995). (The values are not
meant to be representative
of any particular soil or vegctation types.)
To simulate normal and hazy conditions, two differing
visibility values were used: 15 and 5 km. They were COIIverted into aerosol optical depth, as indicated in the Alpendix. A summary of parameters used to perform the
simulations and selected values of reflectante
obtained
thereof are presented in the Appendix in Table Al.
Comparisons

with MODTRAN

MODTRAN (Kneizys et al., 1996), a more recent version


of LOWTRAN (Kneizys, 198X), is considered a standard
1node1 for making atmospheric correction to satellite radiance mrasurements.
As such, it is normally rmployetl
to correct al1 om satellite and aircraft images for atmospheric attenuation for both the solar and the thermal IR
part of the spectrum (Gillirs et al., 1997). MODTRAN
does not account for vegetation, but it does requircx a
temperature
and humidity sounding and an cstimate of
horizontal visibility.
A comparison
between
NDVI simulated with the
simple model described
in the Appendix and MODTRAN was made as follows. Tables of surf&, and apparent (at sensor) reflectances were generated by using both
the simple model ad
MODTRAN
ancl were subst~-

NBVI,

Fractional Vegetation

Cowr,

245

crnd LA1

0.8

0.6

..
0.1 -0

0.1

.I
0.2

..

I
I.
0.3

I.
I.
I
0.4
0.5
Apparent NDVI
.

I
I..
0.6

1.
0.7

.
0.8

a
0.8 ,

0.7 -.
+

0.6 -*
$0.5
Figure 2. NDVI converted into atsurface values for atmospheric attenuation by using MODTRAN for four

atmospheric temperature and moisture soundings made over Pennsylvania on differing days (see key) and
the corrected NDVI based on the radiative transfer model simulations
(Nsim) described in the text (solid
rectaneles)
I,

versus

the

02Jul-89
-iF

E0.4-.
ar
t
0 0.3 -.
0

12Jul-89
-s
02-Aug-90
*
Ol-!%p-90

0.2 -.
0.1 -.

uncorrected

(apparent) NDVI for va!ues used in


the simulations: (a) 15-km visibility;
(b) 5-km visibility.

Nsim
8

-.

n-l
-

::::::::::
0.1

quently used to calculate corresponding


values of surface
and apparent NDVI. In this manner, a set of values of
corrected
and uncorrected
or apparent
NDVI were
made for both models over a range of values of fractional
vegetation cover and LAI. Four different sets of initial
temperature
and moisture soundings were used for the
MODTRAN calculations.
Figure 2 shows the results of the comparison of the
corrected (at-surface) NDVI and the apparent (at-satellite) NDVI for normal and hazy conditions (visibility=I5
km and 5 km, respectively). The relation is nearly linear
for both models, and the results are very similar, although agreement
was not quite as good for the hazy
conditions. As in Pinty and Verstraete (1992), the corrected NDVI is approximately 0.15-0.2 greater than the
apparent (measured) NDVI.
NDVI

as a Function of LAI

Local LA1 was set at a fixed value of 3, and the fractional


vegetation cover was varied from 0 to 1. In the asymp-

;:;
0.2

0.3

0.5
Apparti

0.6

:
0.7

0.8

NDVI

totic regime (Fr= l), LA1 was increased incrementdlly


from 3 to 10. Figure 3a (15-km visibility) shows that the
apparent NDVI increases from 0.54 to 0.61 and the corrected NDVI from 0.72 to 0.75 as LA1 increases from
the 100% vegetation cover threshold to LAI=lO in the
asymptotic
regime. Figure 3b is similar but shows a
lower NDVI in the asymptotic regime for this hazy case
(visibility=5 km).
Additional simulations for the 15-km visibility case
were made for a local (threshold) LA1 of 2 and of 4 (Fig.
4). These simulations
show an increase
of apparent
NDVI from 0.47 to 0.60 and the corrected NDVI from
0.57 to 0.76 in the asymptotic regime (LA1 greater than
2). For a threshold LA1 of 4, the increase in NDVI in
the asymptotic regime (LA1 between 4 and 10) was from
0.57 to 0.60 for the apparent NDVI and from 0.74 to
0.76 for the corrected NDVI. Little change occurred in
any- of the simulations for a LA1 greater than about 6.
Because the most likely values of local LA1 for vegetated surface just reaching 100% cover are between 2

0.8 ,
Ncorr(Fr=l)
0.6

10

LAI

a
0.8 ,

0.6

Ezg~re 3. Uncorrected (apparent; Napp)


NIWI ad NDVI corrected to surftic<,valWS (Ncorr) versus glohl LAI; [global
LA1 is cqual to the fractiorlal wgrtatio~~
cover (Fr) times a fixed value of local LA1
of 3 up to 100% vegetation cover, dmv~~
whicll locd ad globd LA1 are ecpd]: (a)

6
LAI

10

fix 15-km visibility; (17)for R 5-km visibilih-. Horizontal dotted lines denote thr
vulue of NDVI at the threshold to thr asymptotic rrgirnca in which fractional vegctation cu\vr is jllst rexhing 100%~(Fr= 1).

NDVZ,, and NDVZ, correspond


to the valurs of
NDVI fi)r bare soil (LAI=O) and a surface with ;t fractional vegetation cover of lOO%, respectively.
Using entirely different approaches and data sources
from this paper, Clwudhury
et al. (1994) and Gillies and
(1995) independently
obtained
UJ~ identical
(hrkon
square root rrlation lwtween
N and Fr, which is stated:

;md 4, we can infer froin Figures 3 and 4 that the apparent NDVI at the asyrnptotic threshold is likely to be between 0.05 and 0.10 helow the values pertaining
to au
infinitely lnrgc LAI. It seems reasonable
to suppose,
therefore, that the NDVT for areas in which thc vegetatiou cover is just reaching 100% wil1 be adequately reprrscnt(~d by vahles slightly less than the maximum tin~~l
in the image over densr vegetation.

where

Staling the NDVI

Eigure 5 shows N* versus Fr for both the cwrected and


uncorrected
NDVI vahws for the I5-km visibility and
S-km visibility cases. Figure 6 is identical in farm with
Figure 5a but fol- the other two threshold values (2 and
4) of LAl, both corrected and uncorrected.
In al1 cases,
the curves conform closely to Eq. (3), with the maximum
deviation in Fr hing
generally less than 0.1. An imporcorrection ot
tant implication
Henk is that atmospheric
the scaled NDVI is unnecessary
for drttwnining
fiactional vcgetarian
cmer and LAI, b~causr Ai [anti thew

Thr advantage of a scaled NDVI in minimizing


uncertainty in the initial conditions in models that yield soil
water content and surfidce energy fluxes by inversion of
a WAT model bas been pointed out by Gillies t al.
(1997). An added benefit in staling NDVI can now be
sren. Scaled NDVI (N) is defined as
N _ NDVI-NDK
NDVI,- NDVI,,

(2)

Fi4P.

(3)

I,AI

Figurc, 4. Same as Figure 3a (lrj-km visibility) hut for differing LA1 thrrshlds
rwted and uncorrected radiances.

fiere Eq. (3)] IS


. a 1)p roximately the same for both corrected
and uncorrected
NDVI. That the atmospheric correction
effectively cancels in making the staling from NDVI to N
is a consequente
of the linearity between corrected and
uncorrected NDVI, as indicated in Figure 2.

CONCLUSIONS
We show with the aid of a simple radiative transfer model
that the charactcristic behavior of NDVI as a function of
LA1 can be simulated by changing only fractional vegetation cover when it is less than 100%. NDVI is sensitive
to changes in the fractional vegetation cover until a full
cover is reached, beyond which a further increase in LA1
results in an additional smal1 and asymptotic increase in
NDVI. The change in regimes from one that is affect&
primarily by changes in fractional vegetation cover to an
asymptotic one can be simulated by varying fractional
vegetation cover while keeping LA1 fxed at a value of
about 3 in thc vegetated fraction.
The importante
of this finding, which by itself is not
very startling, is that the identification
of the NDVI
threshold between a full and a partial vegetation cover
allows one to scale NDVI between hare soil and 100%
vegetation cover, for which there is a simple square root
relation with fructional vegetation cover [Eq. (3)]. Moreover, our calclllations
suggest that this relation holds
equally wel1 for NDVI corrected or uncorrected
for atmospheric attenuation.
The latter is of practica1 significance in view of the seeming importante
of fractional
vegetation covcr. which may be more caasily obtainable
from satellite
tntasurenwnts
than is LAI, ad the diffi-

247

of2 anti 4, cor-

culty in accurately correcting for atmospheric


attenuation. Staling nut only minimizes
the importante
of
choosing the initial conditions when using land surface
model to obtain the surface energy fluxes and soil water
content (Gillies et al., 1997), hut also mav eliminate the
need to accurately correct the satellite radiances for atmospheric
attenuation.
Furthermore,
we suggest that
fractional vegetation cover, not LAI, is the key variable
in determining
surface enerm fluxes over partial vegetation cover. Indeed, varying both fractional vegetation
cover and LA1 in a land surface model may be somewhat
redundant when the vegetation cover is less than 100%.
Identification
of this full-cover vah~e of NDVI (ref&rcld to here as NDVI,) may not be straightforward
unlcss the image contains a full range of vegetation cover.
In this case, our calculations suggest that the likely value
of NDVI wil1 be about 0.05 below the largest values of
NDVI. In any case, a reasonable estimate of NDVI, based
on a qualitative inspection of a histogram wil1 likely leave
an uncertainty of ahout +0.05 in choosing NDVI,. For a
range of NDVI from 0.0 to 0.6, an error of 0.05 would
correspond to an error of less than 0.1 in Fr. This stil1
leakes some uncertainty in choosing the bare soil valse of
NDVI, however.

APPENDIX
The following mathematica1 development
tire 1, which depicts streams of radiant
partial vegetation cover. Al1 fluxes mov~
or downward with respect to a horizontal
fluxes are distributed
as follows. A pencil

refers to Figfluxes above a


either upward
surface. The
beam of sun-

1
0.9

0.8

0.7
0.6
r
2 0.5
0.4
0.3
0.2
0.1
0
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Fractional Vegetation Cover


1

0.9

0.8
0.7
0.6

T!
4 0.5
0.4
0.3
0.2
0.1
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Fractional Vegetation Cover

Figure 5. Scaled NDVI squared (N)


versus fractional vegetation cover based
on the apparent [uncorrected,
N (a);
solid circles] and corrected radiances [N
(c); open rectangles] for a threshold LA1
of 3. The dashed line indicates the 1:l
correspondence;
(a) 15-km visibility; (b)
5-km visibility.

light (solar constant S0 adjusted for solar distance) at so-

the transmittances

lar elevation

ther diffuse or direct and in the visible or near IR. By

angle

p0 passes

The direct flux reaching


soil and vegetation

through

the atmosphere.

the surface just above the bare

canopy (Fd) is

Fd= S,,sinqq,T,hT,.
The

symbol

T represents

defined in general

(Al.1)

a broad-band

transmittance,

terms as
T=,

(A1.2)

where r is a normalized optica1 depth for that band and


m is a path length. Taband T, refer to absorption by the
direct

beam

(water vapor,

carbon

aerosols) and forward scattering


spectively.
Unlike Beers

dioxide,

ozone,

(by air and aerosols),

Law, which it resembles

and
re-

[Eq. (A1.2)],

definition,

refer to a broad-band

the remainder

radiant flux, ei-

of the incident flux transmitting

through a medium (1-T) is split between an absorptance


and a reflectante,
R. The latter is subdivided into a forward-scattering
component,

component

and

a backward-scattering

which refer to the fraction

beam scattered upward or downward.


The diffuse flux reaching the surface

of the incident

(F,,f)
is

Fdf=s,sinyl[T,b(l-T,)(l-T,~~)l,

L42)

where Tbsrefers to the backscattering component of the


transmittance.
Some of the flw incident on the bare soil
surface is reflected back and forth between the atmosphere and soil. If absorption by these multiple reflections and other second-order effects are ignored, the to-

NDVI, Fractional Veptation Cooer, n~tdLA1

249

-_
M(local)=P

(Uncomcted)

LAl(lmal)=2 (comcied)

lAl(local)=4 (Uncorrected)

LAl(local)=4 (Corrected)

0.2

0.4

0.6

0.6

Fl

Figure 6. Same as Figure 5a (15-km visibility) but for simulations with threshold values of LA1 equal to 2
and 4, for both apparent (uncorrected)
N and corrected radiances. (The 1:l line is omitted.)

tal direct plus diffuse flux absorbed at the surface in the


bare soil portion (F,) is
F,=y_+(l-u,),

(A3.1)

where uR is the bare soil albedo (identical for direct and


diffuse flux) and X is the correction for the internal
re&ctions:
X=a&L

I - ~~Ld)~abdsinv)o*

(A3.2)

TbSa, TSpd,and Tak1 respectively refer to transmittance


components for backward scattering of diffuse flux by air
and aerosols, forward scattering of diffuse flux, and the
absorption of diffuse flux. (Mathematical definitions for
the transmittances and the method of calculating them
are presented later.) The value of X tends to be very
smak-about
1 or 2% of the incident flux at the surface.
The flux reflected at the surface over the bare soil
(F,,) is approximately
F,,=(F<,+F&,.

(A3.3)

Both direct (FJ and diffuse (Fdf) flux components are


also incident at the top of the vegetation canopy. In accord with Taconet et al. (1986), the direct flux penetrating the plant canopy and absorbed at the ground beneath
the vegetation (F,i) is
Fgcd=F<l[(I-~O(I-ag)/(I-~~~gal)l,

(A4)

where ug and af refer to the albedo of the ground surface


and the vegetation, respectively.
The light attenuation factor (canopy transittance) for

direct solar flux (of) is calculated by using the formula


(if=e -KI/SII$0~

(A5)

where IC is assigned a value for isotropic leaf orientation


of 0.4. For the diffuse flux component (Fgd), the diffuse
solar angle (po) direction is assigned a value of 54 degrees (path length 1.7). Calculation of the diffuse flux absorbed at the ground beneath the canopy (Fgc$ is identical with that of Fg,d in Eq. (A4) but with F,f replacing Fd
and with a canopy transmittance factor (of) for diffuse
light (sinv0=1/l.7).
The direct flux absorbed by the plant canopy (FJ is
F,,=F~(l-a~)a~l+u,(l-a~)/(l-a~ugu~).

(A6.1)

The diffuse flux absorbed by the canopy (F& is similarly

calculated but with incident diffuse flux (FdJ replacing


the incident direct flux (F,!) on top of the canopy and the
diffuse path length in erf.
The total flux absorbed at the ground under the canopy (F,) is therefore the sum of F,d and F,, and the
total flux absorbed in the canopy (FC,,,) is the sum of the
diffuse and direct flux components absorbed in the canopy (Fd+FJ.
The total flux absorbed in and below the
plant canopy (F,) is the sum of Fo,,, and F,,..
Weighted for the fraction of vegetation cover (Fr),
the absorbed flux at the surface over both vegetated and
bare soil fractions (F,,J is
F,,,,=FrF,+(l-Fr)F,.

(A6.2)

The total reflected flux from the canopy (F,,) is not


calculated directly. Instead, it is assumed to be the dif-

f;m~ic.(> I)rtwcwr
Ol>\

Lln(l

tho iiicident
ill)SO~l>ftl

tllilt

ill

flur at thct top 01tlrcs cxiiIK~lO\~~


tIlc
calrop:

ml

1;,.,=l;,,+F,,,-F,.

(i\.l)

Thca total rrfiected flilx over the canopy


weiglitetl for vrgetatioii covcar (F,) is

of that surf&

1,,=~

Ixwt~ soil

F, = FrFs, + ( 1 - Fr)F,,
Thr reflectivity

Irailsiriitt~ur~(, l;)r iltIIIOS~>II(i~~


a siiriilar
Iii;iiin<7.

(A7.2)

ii(~rosoIs (TI,) is cd~ikkl~tl

iii

J ?S,,,[f,~il)Jf/~,
s,,A3
ii 1
1

dierc, t,, is thr riormalizcd opticd tlepth i)]. WJ-osols


(dwt). In accord with Paltridgr and llwtt, we writca tlrcl
Augstron~ turbidity law:

(H,,,) is
s=/O?.
(As)

Th

upwding

dar

flux at the

top

of the

atmo-

(F,,,,,) is calculated indirectly. It is assumed to be


the differente
betweeii the incoining solar flux ancl al1
thc flux components
absorbed in the atmosphere ad at
thc surfk.
sphere

F,,,,,=Sl,sinyl,,-F,,~,-F,,~,,-F,,,~,.

(AY)

Here, F,,,, = [S,,sin(o,,(1 -T,,J] is the al)sorbed flux hom th


tlownwelling direct beam, F,,1,, is the absorbed flux fronl
the upvelling
diffuse radiation streani reflected by tlie
caiiopy ad
bare soil surfaces (F,), and F,<,,,is the total
absorbrd flux at the surface.
Tbr reflectivity at the top of thc atlnosphcm~ (H,,,,,) is
thrn

H,,,,,=-

!L-

(A 10)

S,,sinq,\

wllich is to lw compared with that at thr ground, as elpressed by Eq. (A8). Equation (AlO) is the at-sensor (alpurent) reflectivity measured by satellite, whereas Eq.
(AX) is cousidered to be corrected reflectitity
that is
independent
of the intrrvening
atmosphere.
Trwnsrnittances
are calculated separately for scattering ad absorption.
For Hayleigh scattering by air mokcules. the norinalized optica1 depth rII is expressrd ils
(All.l)

7,{=O.OOXX?.,

where s(i), a function of wavelength A (in micrometers),


is qua1 to (-4.15+O.eA)P,/1013,
where P, is the surf&
pressurt in millibars. Path kmgth, 1)~ [sec Eq. (A1.2)] is
coriiputecl from the ecpition
llr~~=sin~,,+O.lS(y7,,+3.8H)-.
Then

the transmittance

u\l2.1>

for Hayleigh

scattering

(All.2)
(Tl<) is
(A11.3)

where po is taken in degrees, AA. is the radiation band


width in micrometers pertaining to the limits of the integal, and S,,, is the extraterrestrial
solar flux as a function
of wavelength.
S,, represents
the integrated flux in the
band. Th<> integral thereforc represents a weighted mean
transrnittance
over the wavelength band, 1,=O.5-0.7 pul11
(visible fxd) ad A=O.-0.9 pn (near IR band).

(412.2)

where ,8 is the Angstrom turbidity coefficient and Ibr exponent a depends on the type of aerosol: a valer of I .O
was chosen for CL,which is typicd fi)r natura1 continental
aerosols. Coefficient p is related to the arosol 0ptic;il
dept11 at 05 Pm through Eq. (12.2):
~=t,,(X=o.FjlUI/1)/(~.~5~ ),
where 7,1(Eb=0.5 pin)
visibility (1: km) iis

is deterinined

(A12.3)

from the horzoi~tal


( A 12.4)

t,,(i,=o.s~,,,)=:3.~lAi.

Thus aerosol transmittance


is varied as ;I function of the
aerosol optica1 depth at O.Ejpirl as indica~ed by tlic horizontal visibility. We divide 5,) into two componeiits,
oiit~
fol- scattering (t,) and ene fol- absoi-ption (7,). wlicrc
t,=O.ii-ir,, and r,=0.25sl,,
wliich yield thr tr~iiisinittailct,s
for scaterring antl absorpfion
by- (lust, rrspr~ctivrl\~ TI,>
the scattering transmittancr
(7:) is wlwa11<1 Tl>,,. Then
lated as the product of scattcring tntnslrlitt;nlces.
I:= ,{,>\.

~\AlXI)

Absorption of radiation by wwter vapor, carbon dioxide,


and oxygen is neglecM,
bccausc these effects are assiiirid to he small-betwfwi
O.Fj and 0.9 pin, altholq$
a weak water vapor band cbxists nar 0.74 pm a11t1 hm,
is a very weak oxygen band at 0.68 Pm. A smal1 correction is made for ozoii (norrn;ilized clepth 0.:3 CJ~I) in thr
Chappuis (vis&)
band by using a method outlined 1)).
Paltridge ad Platt (1976): ser Lacis ad H~SPJI ( 1Y4).
Ths

correction

wil1 not be tliscllssed

cxcept

to indicak

that T,,, is computrd only for the visible band and its n11~iierical ~duc~ is vcry nearly ~yi~il to I.O.
The only significant absorber therefi)re i u cither thr
visible or tht, ncar IK band is acrosol. Thus,
1:,,>=K,,,,.

(AIX2)

111calculating thc diffuse transmittance


functions in Eq.
(A3.2), we calculate Z,,.,,and ?,J,<~
in a marmer identical
with tliat for : anti Yy,,,,,vxcept tliat ;i dar anglc of Fi4
degrees is used for thtx quivalent
solar angle of diffuse
isotropic solar radiation.
For backscattered
flux (Ft,,), which does not reach
the surhce,
thr transmittance
(Y,,,) is calculated by assluning that half of thrh K&igh scattering by air rnolrcults is directed backward (toward the top of the ;itnlosphclr) and half downwarcl. For clllst, \v( ;LssIllII( tIlat

the

fraction

scattered
hulk

backscattered

flux directed

hackscattering

is only 0.2, leaving

toward

the ground.

transmittance

0.8 of the

Therefore,

for dust

the

und air is

(A14.1)
where

the eosine

tropie

nature

amounts

of scatteting

the solar elevation


of the

function

of aerosol

scattered

zon (FjO% direct&


Values

occur

angle

from which

tained

are given

for the highly


in which

in the

decreases

flux directed

upward
from

upward)

aniso-

increasing
direction

solar noon
to near

the

as
(20%
hori-

downward).

of parameters

tions

accounts
scattering

used

the illustrations
in Tahle

to perform

the

in this article

simulawere

oh-

2.

REFERENCES
Asrar, C.. Fucbs. M., Kanymasu, E. T., and Hatfield, J. L.
(1984), Estimating absorbed photosynthetic
radiation and
lcaf area indtls from spectral reflectancr in &xit. Agrm. J.
76:30(1-:306.
Best. E. G.. and Harlan, J. C. (1985), Spectra1 estimation of
17:
grern leaf anx index of oats. Kemo~~ Sms. Ettzjirm.
27-36.
<:arlson, T. N., Perry, E. M., and Schmugge, T. J. (1990), Kemate estimation of soil moisture availability and fractional
qetation
cover for agricultural fields. Agric. For. A-l&ord. 52:415-69.
Cboudbm~, B. J., Ahmed, N. IJ., ldso, S. B., Reginato, K. J.,
and I~aughtry, C. S. T. (1994), Relations brtween evaporation coeffcitxlts and vegetation indices studied by model
sirnulations. &~noio Sens. Enuiron. SO:l-17.

(:urran, P. J. (1983), Multispectral rernotc sensing for the estimation of green leaf area index. Pidos. Trms. R. Land.
Sr. A 309:257-270.
I)irkinson,
R. E., Henderson-Sellers,
A., and Kennedy, P. J.
(1993), Biosphere-atinosphere
transf& scheme (RATS)
vcrsion le as coupled to the NAAR community climate
model. NCAR Tc& Notc~ NCAWN-3N7+ S7R, NCAR,
Boulder, CO.
Gdlo. T. F., Daughtry, C. S. T., and Hauer, M. E. (1985), Spectral cstimation of absorbrd photosynthicallv active radiation
in corn carropies. Retruttc SLIII..Encinttr.
1?:221-232.
Gillies, R. R., Cni. J., Carlson, T. N., Kustas. \\. P.. ;md Humes,
K. S. (1997). Vcrification of a mctbotl for obtaining suriace
wil water content and enrrgy floses froin rfmote nifwurf~nwnts of NW1 wd surfxe radiant temperature. I&. I. Retttotc Scttc.(in pres),
Gillies, K. K., anti (:arlson, T. N. (199.5). Thcrl~lal rrmotr sensing of sln-f:ice soil water content with partial vegctation
cover for incorporatioll into clinrate models. J. Al$. ilile~
orol. 34:745-756.
Holben, B. N.. Tucker, C. J. aml Fan, C;. J. (19YO), Spectra1
assrssment
of soybean leaf area :md 1~1 biolnass. Phoiogrcrttm
Ettg. 45:651-656.
Jasinski, M. F. ( 1996), Estimation of sllbpUc~1 \egetation density
of Iiatliral regions lising satellite nniltisprctral
iniagcry.
IEEE Tt-crrts.Geosci. Rett~otc, Srm. :33:X04-81:3.
Kneizys, F. X.. Ahreu, 1,. kV., Anderson. G. P., Chetwynd.
1. Il., Shuttle, E. P.. Berk, A., Bernstein, 1,. S., Robertson,
11. (:., Achqa.
P., Rothman. I ,. S.. S&Y, .[. E. A., Gallexy,
Ll.. 0.. ml C:lough,S. A. (1996). +Uw ,1IODTRAR2/3 Kr:jm-t mrl JLWTRAN
7 Motkl (L. 11. ~\tmqr and 6. P.
4ndcwon. Eds.), Ixeparrd hy Ontar Corp., North Andoxxbr, MA, for Phillips Laboraton;. (~eoplrvsical Directoratr,
1Iansco1n AFB, MA.. Contract No. FlY&&Hl-C:-0132.
Knei+x F. X., Shrttle, E. P., Abrcn. L. \\.., (:hr+ylid,
J. H.,
Anderson, G. P.. Gallery. \V. 0.. Sclb!,. J. E., an(l Clough,
S. A. ( 198S), C!YCJYY
Grieks io Lowfrcttt-7.
Air Force Geophysics Laboratorv Research Paper No. 40 1010. Project
No. x70.
Lacis, A. A., x~d Ha~~sen, J. E. (lY4). A panl~rl<~trnn~tion for
tlir absorption of solar radiation in tlic, f,arths atmospherra.
/. Al-tttos. Sh. 31:11K-13i3.
Lqrieur,
C:., Kerr. Y. II., and Pithon. 1. M. (IYYGI, Crtical
:usessment ofvegetaton indices froln AVHRR in a semiarid enviromnent.
Ittt. J, Rc~no~~~
Srw. 17:2Cj94-2563.
Liu, Q.. and Inetr, A. (199s). A feedback basrd modification of
thc ND\1 to minimize wnop bxkgrolmd and atmospheric
noisr. IEEE Trms. Ow~:i. Rrtttntc Sctc.~.33:457465
Montrith, J. L. (lYi3), Prittciph
of Ett1-ircttttttc,tttcII
Phyrics.
Arnold, 1). 197.
Nemani, R.. a11d Kunning, S. \\. (1989). Tcsting 21theoretical
clillrate-soil-leaf
area hydrological cc~~~ilibri~ur~of forests
llsing satfsllitc data and ecosvstrilr simlilation. AgriC. For.
~V<~tcwrol. 44:245-260.
Olioso. A.. Carlson, T. N., and Brisson, N. (1995), Simulation
Of diurilal transpiratiorr
anti photosynthesis
of a water
stresx~d soybeun crop. Agric,. For. Mc~trctrol. 81:4 1-59.
Ialt ridgc, CC. & Berber, J. (19811). Monitoring grassland dryIWSS and fire potential in Australia with NOAA/AVHRR
data. Ikwrof~~Sms. Ettcirott.
28::384-:3$X3.

CG.\V., and Platt, C. M. H. (IYX), Rndiatiw Proccs.ws


arul Clittu~tologj. Elsevier, Amstrrdain.
Pinty, H., and \erstraete, H. M. (1!%2), GEMI.: a non-hnrwr
index to monitor glol>al vegetatioii from satllites. 17qyta-

I&ridgc.

in ikteorolog~j

tior 101:15-20.

Pollard, D., and Thompson,


S. (1995), Use of land-surface
transfer schemer (12.X) in a global climatr model: tlrr response to doubling stomata1 resistance.
~Zolx~l klauwt.
Chnr2g~ 10:X30-161.
Price, J. C. (1992), Estimating vegetation amount from visible
and near infrared reflectances. &wwclte Sens. Enoirorr. 41:
29-34.
Sellers, P. J. (1985), Canopy reflectante,
photosynthesis
and
transpiration. Int. ]. of Remde
Sens. 10:855-867.
Sellers, P. J., and Dorman, J. L. (1987), Testing the simple hio-

Taconet, o., Carlson T. N., Kernard, K.. ikh~-hfd&LJ~.


1).
(1986), Evaluation of a sllrface/vegetatiori
rnodei
msing
satellite infared surfke teqrratwcs.
J. Clirr~. AppI. .Vf,fcord. 25: 1752-1767.
Wetzel, P. J,, and Boone, A. (1995). A parameterization tOr landatmospherc-clond
exchange (PLACE): documentation
and
testing of a detailed process model of the partl~. cloudy
houndary

18Ub1837.

layer

over

hetcrogeneons

land.

J.

olim.

8: