Full text of DOI:10.2307/1478997

Applied Vegetation Science 3: 189-195, 2000
© IAVS; Opulus Press Uppsala. Printed in Sweden
- ESTIMATION OF PRIMARY PRODUCTION OF SUBHUMID RANGELANDS FROM REMOTE SENSING DATA -
189
Estimation of primary production of subhumid rangelands
from remote sensing data
Paruelo, José M.1*; Oesterheld, Martín1; Di Bella, Carlos M. 2; Arzadum, Martín3;
Lafontaine, Juan4; Cahuepé, Miguel5 & Rebella, César M.2
1IFEVA, Departamento de Ecologia, Facultad de Agronomia, Universidad de Buenos Aires, Av. San Martin 4453,
1417 Buenos Aires, Argentina; 2Instituto de Clima y Agua, CIRN - Instituto Nacional de Tecnología Agropecuaria
(INTA), Los Reseros y Las Cabañas s/n, Castelar (1712), Buenos Aires, Argentina; 3Campo Experimental Pass-
man, CC204, Coronel Suárez (7540) Buenos Aires, Argentina; 4AACREA, Sarmiento 1236, 1041Buenos Aires,
Argentina; 5Facultad de Ciencias Agrarias, Universidad Nacional de Mar del Plata, CC276, 7620 Balcarce,
*
Buenos Aires, Argentina; Corresponding author; Fax +541145148730; E-mail paruelo@ifeva.edu.ar
Abstract. Above-ground Net Primary Production (ANPP) is
the main determinant of forage availability and hence of
stocking density. A tool to track the seasonal and interannual
changes in ANPP at the paddock level will be very important
for livestock management. We studied the relationship be-
tween field ANPP data and the Normalized Difference Veg-
etation Index (NDVI) for rangelands of the Flooding Pampa of
Argentina using spectral data provided by sensors on board of
two satellites: NOAA/AVHRR and Landsat TM. The relation-
ship between NDVI and ANPP was linear both for data
derived from NOAA/AVHRR and Landsat TM. Changes in
ANPP accounted for a large proportion of the temporal and
spatial variation of NDVI: 71% of NOAA/AVHRR data and
74% of Landsat TM data. By inverting these models, ANPP
may be inferred from NDVI data at a seasonal and paddock
scale. NOAA/AVHRR data captured better the seasonal vari-
ation in ANPP and were less sensitive to local variations than
Landsat TM data. In contrast, Landsat TM data were more
sensitive to inter-site differences in primary production, ex-
cept for the winter months. Thus, combining information from
these two sources offers a good alternative for monitoring
rangeland production at high temporal and spatial resolution.
Introduction
Primary production, the rate of biomass production
per area and time, is an important attribute of the ecosys-
tem. It determines, for instance, the total amount of
energy available for upper trophic levels. Herbivore
biomass is tightly related to primary production both for
natural (McNaughton et al. 1989) and human-modified
ecosystems (Oesterheld et al. 1992). In rangelands,
Above-ground Net Primary Production (ANPP) is the
main determinant of stocking density both at regional
(Oesterheld et al. 1992; 1998) or local scales (Golluscio
et al. 1998). In order to make decisions on animal
movements among pastures, forage storage, stocking
densities, etc., range managers need accurate estimates
of ANPP at the paddock level and at relatively detailed
temporal resolution (months to season).
Usually, ANPP is estimated from temporal changes
in biomass (Lauenroth et al. 1986), which are generally
measured by harvesting small plots (Sala et al. 1988a;
Sims et al. 1978). This is a time consuming approach
that clearly restricts the collection of ANPP data in
grassland areas. Many methodological problems have
been associated to the harvest method to estimate ANPP
(Sala et al. 1988a; Oesterheld & McNaughton in press).
Remote sensing can be used to infer temporal and spa-
tial differences in biomass, LAI, plant cover, or produc-
tivity (Tucker et al. 1985; Kennedy 1989; Running
1990; Lloyd 1990; Prince 1991a; Paruelo et al. 1997;
Purevdorj et al. 1999) . The use of spectral data to assess
vegetation parameters is largely based on the differen-
tial reflectance of photosynthetic tissue in the red and
infrared portions of the electromagnetic spectrum. Green
leaves reflect a small proportion of the incoming radia-
tion in the red band and a high proportion in the near
infrared band (Guyot 1990). Spectral indices, such as
the Normalized Difference Vegetation Index (NDVI)
Keywords: Above-ground production; Argentina; Calibra-
tion; Flooding Pampa; LANDSAT; NDVI; NOAA-AVHRR;
Primary production; Rangeland.
Abbreviations: ANPP = Above-ground Net Primary Produc-
tion; APAR = Absorbed Photosynthetic Active Radiation;
NDVI = Normalized Difference Vegetation Index.

190
PARUELO, J.M. ET AL.
that combines the red and infrared bands can be calcu-
lated in order to infer several vegetation parameters:
temperate grasslands of North America based on aver-
age annual ANPP data. A non-linear model accounted
for 95% of the variability in ANPP. The model proposed
to estimate ANPP from the NDVI integral (NDVI-I)
was (Paruelo et al. 1997):
NDVI = (IR–R)/(IR+R)
(1)
Both basic and applied studies showed that NDVI is
a good estimator of the amount of photosynthetically
active radiation absorbed by the canopy and, thus, of
carbon fluxes through the ecosystem (Gamon et al.
1995; Sellers et al. 1992). NDVI is linearly related to the
Absorbed Photosynthetic Active Radiation (APAR) and
other canopy biophysical rates (Gallo et al. 1985;
Choudhurry et al. 1987; Goward & Huemmrich 1992;
Law & Waring 1994). Net primary production (NPP) is
directly related to APAR by the model proposed by
Monteith (1981):
ANPP (g.m–2.yr–1) = 3803 . NDVI-I 1.9028
(2)
This study also provided an estimate of the coefficient
of energy conversion in biomass. Across the precipita-
tion gradient analysed (300 - 1000 mm), increased from
0.10 to 0.20 g C/MJ.
Most of the calibrations reported in the literature
were based on annual data and on the use of coarse
spatial resolution information (NOAA/AVHRR satel-
lites). From a theoretical viewpoint this is a conse-
quence of adopting Monteith’s model, where NPP is
related to the annual integral of the absorbed radiation
(estimated from NDVI in these cases). But, to some
extent, the temporal resolution of these analyses is also
constrained by the availability of ANPP data (i.e. Paruelo
et al. 1997). Often, peak biomass is used to estimate
ANPP (Lauenroth et al. 1986) and consequently only a
unique value is available for the whole year. The tempo-
ral resolution of the ANPP data is also constrained by
the type of satellite being used. An annual value of
ANPP can only be related to an index integrated over the
entire year. NOAA/AVHRR is probably the best alter-
native to integrate over the year because it has a nominal
temporal resolution of 1 day.
NPP = ε APAR
(2)
[
]
∫
where ε is the energy conversion efficiency (g/MJ).
Thus, there is a strong theoretical basis to relate NPP
with NDVI. Several equations have been proposed to
relate the fraction of Photosynthetically Active Radion
(PAR) intercepted and NDVI (Dye & Goward 1993;
Sellers et al. 1994).
The positive relationships between Above-ground
Net Primary Production (ANPP) and NDVI have also a
strong empirical support. Studies performed on a number
of ecosystems and geographic areas showed that the
correlation between ANPP and NDVI is often higher
than 0.8 (Goward et al. 1985; Tucker et al. 1985; Box et
al. 1989; Burke et al. 1991; Prince 1991b). Paruelo et al.
(1997) presented a calibration of this relationship for
These strong relationships between ANPP and coarse
resolution satellite data are promising, but they fall short
of resolving many managerial problems that need sea-
sonal information at the paddock level. In this article,
Fig. 1. Map of the Argentine
Pampas showing the study area.

- ESTIMATION OF PRIMARY PRODUCTION OF SUBHUMID RANGELANDS FROM REMOTE SENSING DATA -
191
we studied the seasonal and paddock-level relationship
between NDVI and ANPP. Our approach was to com-
pare field ANPP estimates of small paddocks of the
Flooding Pampa (Argentina) differing in species com-
position and topographic position with a spectral index
derived from two satellites sensors: NOAA/AVHRR
(low spatial resolution, high temporal resolution) and
Landsat TM (high spatial resolution, low temporal reso-
lution). The relationships derived allowed us to track
intra-annual changes in ANPP and capture differences
among paddocks with an accuracy similar to that ob-
tained by direct biomass sampling.
× 30 m). LAC AVHRR/NOAA images were obtained
from the Global Land 1km ×1 km AVHRR Project of
the EROS Data Center (Sioux Falls, SD USA) (http://
edcwww. cr.usgs.gov/landdaac/1kmhomepage.html).
From NOAA-AVHRR 10-day composite images, we
extracted the mean NDVI value for a 2000m ×2000 m
window for each site for the periods centred at the
following dates (4 sites ×8 dates):
31.08.1992
23.03.1993
25.10.1992
06.03.1995
03.12.1992
03.06.1995
18.01.1993
04.02.1996.
Sites 3 and 5 were so close to each other that they could
not be clearly separated in a NOAA/AVHRR image.
NOAA/AVHRR NDVI was integrated over the period
between biomass harvests (between 35 and 96 days).
From cloud-free Landsat TM scenes, we extracted the
mean NDVI for a 500m ×500 m window for each site
in August 1992, August 93, October 1993, July 1994,
November 1994, January 1996 and August 1996 (5 sites
×7 dates). The dates of the images corresponded closely
(± 10 days) to the average date of the period between
harvests. NDVI was calculated from band 1 and 2
reflectances of the NOAA/AVHRR sensor and from
bands 3 and 4 of the Landsat TM sensor. For data
derived from NOAA/AVHRR images, NDVI was inte-
grated over the period between clippings. For the NDVI
derived from Landsat TM images we used the instanta-
neous values.
ANPP was estimated from sequential biomass sam-
pling by the harvest method. Eight plots of 1 m2 each
were clipped at 3 cm height simulating an intense utili-
zation by livestock. Wire cages were installed in those
areas and total biomass was harvested two months later
by clipping at 3 cm height and oven-drying the material
collected. At the same time, new plots nearby were
clipped to 3 cm height and covered with cages to esti-
mate the ANPP of the following period. Thus, our
ANPP values represent a management situation similar
to a rotational grazing with intervals between grazing
events varying between 35 and 95 days and intense
forage removal. ANPP was estimated as the amount of
biomass harvested divided by the number of days be-
tween harvests.
Material and Methods
The study area was located in the Laprida county,
close to the western border of the Flooding Pampa
(Fig. 1). The Flooding Pampa, a subdivision of the Rio
de la Plata grasslands (Soriano 1991), is an extensive
area of rangelands in the Buenos Aires province (Ar-
gentina). The vegetation and soils of the specific area
under study have been described by Batista et al.
(1988). The region is characterized by a flat topogra-
phy and periodic floods (Paruelo & Sala 1990). Agri-
culture is constrained by the halomorphic and hydro-
morphic characteristics of the soils. Native grasslands
are characterized by a co-dominance of C3 and C4 grasses.
The climate is temperate-subhumid. Mean annual rain-
fall is approximately 900 mm, and mean monthly tem-
peratures range from 6.8 °C in winter to 21.8 °C in
summer. There is not a clear dormant season for the
vegetation. Based on climate, the expected ANPP of the
area ranged between 4500 and 5000 kg.m–2.yr–1
(McNaughton et al. 1993; Sala et al. 1988b). The most
thorough determination of ANPP for a widespread plant
community of this region was 5320 kg.m–2.yr–1 (Sala et
al. 1981).
We selected five sites in the study areas: 1 (37.53 S;
61.20 W), 2 (37.5 S; 60.79 W), 3 (37.42 S; 61.29 W), 4
(37.61 S; 60.70 W), and 5 (37.45 S; 61.35 W). They
were commercial paddocks (modal size 70 ha) that
encompassed a large portion of the heterogeneity of the
forage resources of the region (alfalfa pastures, site 1, C3
grasses- legumes pastures, sites 2, 3, 4 and 5) occupying
different topographic positions (convex, well drained
areas, sites 1, 2 and 3; concave, poorly drained areas,
sites 4 and 5), and soils (moderately acid mollisols, sites
1, 2 and 3; alkaline mollisols and alfisols, sites 4 and 5).
NDVI data were correlated with simultaneous field
estimations of ANPP based on biomass harvests at each
site. Satellite data were obtained from two different
sources: NOAA-AVHRR Local Area Coverage (LAC,
1km ×1 km spatial resolution) and Landsat TM (30 m
We studied the relationships between ANPP and
NDVI by means of regression analysis. We assumed
that ANPP was the independent variable and that NDVI
thedependentvariablebecauseremotedatareflectground
processes. We then calculated the inverted models to
make them useful as a calibration tool that allows one to
infer ANPP from NDVI.

192
PARUELO, J.M. ET AL.
Results
Changes in ANPP accounted for a large proportion
of the temporal and spatial variation of NDVI (Fig. 2).
ANPP accounted for 71% of variation of NOAA/
AVHRR data and for 74% of variation of Landsat TM
data. The models fitted to both data sets showed similar
slopes (0.0052 for NOAA/AVHRR and 0.0053 for
Landsat TM) but quite different Y-intercepts (0.424 for
NOAA/AVHRR and 0.126 for Landsat TM). The in-
verted models that can be used to infer ANPP from
NDVI were:
ANPP = 192.3 NDVI – 81.5 for NOAA/AVHRR data;
ANPP = 188.7 NDVI – 23.8 for Landsat TM data.
The use of Landsat TM images from different dates did
not introduce much additional variability in the relation-
ship with ANPP.
The variation in NDVI and ANPP has two major
sources: seasonality and site-related differences. Both
types of satellite data captured the seasonal variation in
ANPP, and Landsat TM data additionally captured site
differences. The relationship between NOAA/AVHRR
NDVI and ANPP for each site across dates was highly
significant and similar for all sites: Y-intercepts and
slopes were statistically similar, and differed by no
more than 11% and 33% respectively (Table 1). For
Landsat data the relationship between NDVI and ANPP
across dates was also significant for each individual site
(Table 1). However, Y-intercepts and slopes differed
among some of the sites (Table 1).
For NOAA/AVHRR data the relationship between
NDVI and ANPP across sites was significant only for
one date (Table 2). Landsat TM NDVI was significantly
related to the ANPP across sites, except for the winter
months (July and August) when biomass and ANPP
showed the lowest seasonal values.
Fig. 2. Relationship between NDVI and ANPP estimated from
AVHRR/NOAA (NDVI AVHRR = 0.424 + 0.0052. ANPP (n
= 33; r2 = 0.71; F = 73; p < 0.001) and Landsat TM satellites
(NDVI TM = 0.126 + 0.0053 .ANPP (n = 35, r2 = 0.74; F =
92; p < 0.001). Data points for each site correspond to
different dates.
The use of different type of images (Landsat TM and
AVHRR/NOAA) was associated with a change in the
spatial resolution of the observation (window’s size).
For a Landsat TM image the minimum size of the
windows is 30 m ×30 m and for an NOAA/AVHRR,
1000 m ×1000 m, the pixel size of each of the sensors.
We explored the effect of changing the size of the
window on the spatial variability of the NDVI estimate
among study sites. To do that, we obtained the NDVI
for each site from the seven Landsat TM images, using
windows of different size, from 200 m ×200 m (4 ha)
up to 3200 m × 3200 m (1024 ha). The NDVI values
for the windows of different sizes corresponded to the
average value of the pixels included. The standard de-
viation of NDVI among sites decreased as the spatial
resolution of the observation (window’s size) became
coarser (Fig. 3). The modal size of a paddock in the
Fig. 3. Relationship between the NDVI standard deviation of
the five sites considered for the seven Landsat TM images
used, and the size of the window used to integrate data (reso-
lution of the observation).

- ESTIMATION OF PRIMARY PRODUCTION OF SUBHUMID RANGELANDS FROM REMOTE SENSING DATA -
193
Table 1. Regression analysis of the ANPP and NDVI relation-
ship for individual sites and for the different sites pooled
together for NOAA/AVHRR and Landsat TM data. Eight
dates were available for the NOAA/AVHRR dataset and seven
for the Landsat TM. ** indicates p < 0.001. Different letters
indicate significant differences between slopes or Y-intercepts
(p < 0.05) (t-test).
Table 2. Regression analysis of ANPP and NDVI relationship
for individual dates (Landsat TM dataset) and periods (NOAA/
AVHRR dataset) for NOAA/AVHRR and Landsat TM data.
Four sites were available for the NOAA/AVHRR data set and
five for the Landsat TM. ** indicates p < 0.001.
Date
r2
Y-intercept
Slope
Sign.
NOAA/AVHRR data
Site
r2
Y-intercept
Slope
Sign.
AUG92
OCT92
MAR92
JAN93
MAR93
MAR95
JUN95
FEB96
0.66
0.32
0.92
0.85
0.61
0.99
0.01
0.45
0.37
0.81
0.71
0.46
0.41
0.43
0.51
0.47
0.003
n.s.
n.s.
n.s.
n.s.
n.s.
**
-0.004
-0.002
0.004
0.003
0.005
0
NOAA/AVHRR data
1
2
3
4
0.51
0.69
0.78
0.93
0.82
0.37 a
0.38 a
0.38 a
0.34 a
0.36
0.005 a
0.006 a
0.005 a
0.007 a
0.006
**
**
**
**
**
n.s.
n.s.
0.004
All
Landsat TM data
Landsat TM data
AUG92
AUG93
OCT93
JUL94
NOV94
JAN96
AUG96
0.09
0.35
0.88
0.08
0.95
0.85
0.16
0.20
0.15
0.28
0.17
0.27
0.05
0.11
-0.003
n.s.
n.s.
**
n.s.
**
1
2
3
4
5
All
0.92
0.67
0.57
0.88
0.83
0.89
0.16 a
0.09 ab
0.21 a
0 b
0.05 ab
0.36
0.005 a
0.006 bc
0.004 ac
0.009 b
0.009 b
0.006
**
**
**
**
**
**
0.005
0.003
0.005
0.004
0.009
-0.002
**
n.s.
study area is around 70 ha. If the size of the window
corresponds to the AVHRR/NOAA pixel (1000 × 1000
m = 100 ha), the observation will probably integrate
information of the reflectance of more than one pad-
dock. This results in a greater variability within the
window and a lower variability among windows. The
variability among sites or single paddocks will also
change among seasons. The standard deviation of the
NDVI among sites decreases with an increase not only
in the plot size (Fig. 3) but also with an increase in the
average NDVI value of the window. This indicates that
the spatial variation is larger in spring and summer than
in winter.
for a better representation of the spatial heterogeneity of
forage resources. Landsat TM images, because of their
fine spatial resolution and given the grain of the hetero-
geneity of forage resources in the study area, are best
suited to capture the inter-site differences. Such differ-
ences could be related, for example, to inter-site differ-
ences in the energy conversion efficiency (ε, g/MJ) or
allocation associated with the specific composition of
the pasture or to the canopy architecture. For the study
area, site and site × date differences among forage
resources revealed by Landsat TM data were relatively
stable among years. Thus, we envisage a monitoring
system of ANPP based on real time, readily available
NOAA/AVHRR data, weighed by a site × date specific
coefficient derived only once for each site from Landsat
TM images.
Discussion
The coefficients of determination of the temporal
and spatial relationships between ANPP and NDVI of
our study were similar to those reported at an annual and
regional scale in the Sahel and in the Central Grasslands
of North America (Paruelo et al. 1997a; Prince 1991b).
The models fitted to both data sets showed similar
slopes (0.0052 and 0.0053 for NOAA/AVHRR and
Landsat TM respectively) but quite different Y-inter-
cepts (0.424 and 0.126 for NOAA/AVHRR and Landsat
TM respectively). NDVI values derived from different
satellites are not directly comparable because they were
calculated from values of red and infrared reflectance
covering different portions of the electromagnetic spec-
trum. Galvao et al. (1999) showed that NDVI increases
Our results indicate that NDVI and the calibration
equations presented may provide reliable estimates of
the ANPP of rangelands of the Flooding Pampa. Al-
though Landsat TM images suffice in capturing both
seasonal and spatial variation in ANPP of single small
paddocks, they are more expensive and have a lower
temporal resolution than NOAA/AVHRR images. Thus,
a combination of both sources may be more efficient.
NOAA/AVHRR data were able to capture seasonal
changes in APAR because of their high temporal resolu-
tion. Within a NOAA/AVHRR pixel it is likely that
more than one type of forage resource is included. A
correction based on Landsat TM information will allow

194
PARUELO, J.M. ET AL.
Acknowledgements: This project is part of the AACREA
SUDOESTE – FA UBA contract. It was funded by grants from
CONICET, FONCYT, UBA and Fundación Antorchas. Martín
Garbulsky helped in gathering the NOAA-AVHRR images.
with the spectral distance of the red and infrared bands
of the sensors. The difference between the centre of the
red and infrared bands for NOAA/AVHRR is almost
twice than that for Landsat (300 nm vs. 170 nm) (Galvao
et al. 1999). The effect of the difference between the
center of the bands on NDVI increases as the green
component of the vegetation decreases. Geometric cor-
rections, re-sampling and compositing of the images
also vary between satellite sources. Differences in post-
processing and in the spectral information used to calcu-
late the NDVI may account for the bias observed be-
tween data obtained from the two satellite sources.
Several studies have shown a saturation response of
NDVI as ANPP or plant cover increases (see, e.g., Box
et al. 1989; Paruelo et al. 1998; Purevdorj et al. 1999). In
this study we did not detect a significant non-linear
signal, even though we included in the analysis high
values of ANPP. A saturation response of the NDVI
may occur above 60 kg.ha–1d–1 in the study area as
suggested by the Landsat TM data set. However, ANPP
values of 60 kg.ha–1d–1 sustained over several weeks are
exceptionally high for grasslands or sown pastures in
temperate areas (Ratcliffe & Baar 1987). In this analy-
sis, values over this threshold were recorded with a low
frequency (less than 10%). The NOAA/AVHRR data
did not cover any period including values of ANPP
higher than 60 kg.ha–1d–1.
Landsat imagery has been extensively used in range
assessment (Boyd 1986; Paruelo & Golluscio 1994).
However, its use as a predictor of ANPP is less frequent.
Our results showed that these images offer a good
alternative for monitoring rangeland production at high
spatial resolution. Primary production is the main deter-
minant of forage availability and hence of stocking
density. A tool able to track intra-annual variability of
ANPP at the paddock level may help to improve live-
stock management. Ranchers and extensionits will be
able to track the spatial and temporal changes in forage
availability. This will provide them with better informa-
tion to plan for the use of forage resources. The use of
Landsat images to estimate ANPP will also provide
range scientists with a crucial set of data to develop
predictive models of forage availability based on mete-
orologicalandmanagementvariables. Moreover, thecom-
bination of both data sources will provide more precise
descriptions of forage resources. NOAA/AVHRR data
may provide a regional ANPP context with a high tempo-
ral resolution, meanwhile Landsat TM data will allow
one to downscale ANPP estimates at the paddock level.
References
Batista, W.B., Leon, R.J.C. & Perelman, S.B. 1988. Las
comunidades vegetales de un pastizal natural de la Region
de Laprida, Prov. de Buenos Aires, Argentina. Phyto-
coenologia 16: 465-480.
Box, E.O., Holben, B.N. & Kalb, V. 1989. Accuracy of the
AVHRR Vegetation Index as a predictor of biomass, pri-
mary productivity and net CO2 flux. Vegetatio 80: 71-89.
Boyd, W.E., 1986. Correlation of rangelands brush canopy
cover with Landsat MSS data. J. Range Manage. 39: 268-
271.
Burke, I.C., Kittel, T.G.F., Lauenroth,W.K., Snook,P., Yonker,
C.M. & Parton,W.J., 1991. Regional analysis of the Cen-
tral Great Plains. Bioscience 41: 685-692.
Choudhury, B.J. 1987. Relationships between vegetation indi-
ces, radiation absorption, and net photosynthesis evalu-
ated by a sensitivity analysis. Remote Sens. Environ. 22:
209-233.
Dye, D.G. & Goward, S.N. 1993. Photosynthetically active
radiation absorbed by global land vegetation in August
1984. Int. J. Remote Sens. 14: 3361-3364.
Gallo, K.P., Daughtry, C.S.T. & Bauer, M.E. 1985. Spectral
estimation of absorbed photosynthecally active radiation
in corn canopies. Remote Sens. Environ 17: 221-232.
Galvao, L.S., Vitorello, I. & Filho, R.A. 1999. Effects of band
positioning and bandwidth on NDVI measurements of
tropical savannas. Remote Sens. Environ 67: 181-193.
Gamon, J.A. Field, C.B., Goulden, M.L., Griffin, K.L., Hartley,
A.E., Joel,G., Penuelas, J. & Valentini, R. 1995. Relation-
ships between NDVI, canopy structure, and photosynthesis
in three Californian vegetation types. Ecol. Appl. 5: 28-41.
Golluscio, R., Deregibus, V. & Paruelo, J. 1998. Sustainability
and range management in the Patagonian steppe. Ecol.
Austr. 8: 265-284.
Goward, S.N. & Huemmrich, K.F. 1992. Vegetation canopy
PAR absorptance and the Normalize Difference Vegeta-
tion Index: An assessment using the SAIL model. Remote
Sens. Environ 39: 119-140.
Goward, S.N., Tucker, S.J. & Dye, D.G. 1985. North Ameri-
can vegetation patterns observed with the NOAA-7
advanced very high resolution radiometer. Vegetatio
64: 3-14.
Guyot, G. 1990. Optical properties of vegetation canopies,
Applications of remote sensing in agriculture. Butter-
worths, London.
Kennedy, P. 1989. Monitoring the vegetation of Tunisian
grazing lands using the normalized difference vegetation
index. Ambio18: 119-23.
Lauenroth, W.K., Hunt, H.W., Swift, D.M. & Singh, J.S.
1986. Estimating aboveground net primary productivity
in grasslands: a simulation approach. Ecol. Model. 33:
297-314.

- ESTIMATION OF PRIMARY PRODUCTION OF SUBHUMID RANGELANDS FROM REMOTE SENSING DATA -
195
Law, B.E. & Waring, R.H. 1994. Remote sensing of leaf area
index and radiation intercepted by understory vegetation.
Ecol. Appl. 4: 272-279.
Lloyd, D. 1990. A phenological classification of terrestrial veg-
etation cover using shortwave vegetation index imagery. Int.
J. Remote Sens. 11: 2269-2279.
ships between percent vegetation cover and vegetation
indices. Int. J. Remote Sens. 19: 3519-3535.
Running, S.W. 1990. Estimating terrestrial primary produc-
tivity by combining remote sensing and ecosystem simu-
lation, Remote sensing of biosphere functioning, pp. 65-
86. Springer-Verlag, New York,
McNaughton, S.J., Oesterheld, M., Frank, D.A. & Williams,
K.J. 1989. Ecosystem-level patterns of primary productivity
and herbivory in terrestrial habitats. Nature 341: 142-144.
McNaughton, S.J., Sala, O.E. & Oesterheld, M. 1993. Compa-
rative ecology of African and South American arid to
subhumid ecosystems. In: Goldblatt, P. (ed.) Biological
relationships between Africa and South America, pp. 548-
567. Yale University Press, New Haven, NY.
Monteith, J.L., 1981. Climatic variation and the growth of
crops. Q. J. R. Meteorol. Soc. 107: 749-774.
Oesterheld, M. & McNaughton, S.J. 2000. Herbivory in ter-
restrial ecosystems. In: Sala, O.E., Jackson, R.B., Mooney,
H.A. & Howarth, R. (eds.) Methods in ecosystem science,
pp. 151- 157. Springer Verlag, New York, NY.
Sala, O., Deregibus, V., Schlichter, T. & Alippe, H. 1981.
Productivity dynamics of a native temperate grassland in
Argentina. J. Range Manage. 34: 48-51.
Sala, O.E., Biondini, M.E. & Lauenroth, W.K. 1988a. Bias in
estimates of primary production: An analytical solution.
Ecol. Model. 44: 43-55.
Sala, O.E., Parton, W.J., Joyce, L.A. & Lauenroth, W.K.
1988b. Primary production of the central grassland region
of the United States. Ecology 69: 40-5.
Sellers, P.J., Berry, J.A., Collatz, G.J., Field, C.B. & Hall, F.G.
1992. Canopy reflectance, photosynthesis, and transpira-
tion. III. A reanalysis using improved leaf models and a
new canopy integration scheme. Remote Sens. Environ.
42: 187-216.
Oesterheld, M., DiBella, C.M. & Kerdiles, H. 1998. Relation
between NOAA-AVHRR satellite data and stocking rate
in grasslands. Ecol. Appl. 8: 206-212.
Oesterheld, M., Sala, O.E. & McNaughton, S.J. 1992. Effect
of animal husbandry on herbivore-carrying capacity at a
regional scale. Nature 356: 234-236.
Paruelo, J.M. & Sala, O.E. 1990. Caracterización de las inun-
daciones en la depresión del salado (Buenos Aires, Argen-
tina): dinámica de la capa freática. Turrialba 40: 5-11.
Paruelo, J.M., Epstein, H.E., Lauenroth, W.K. & Burke, I.C.
1997. ANPP estimates from NDVI for the central grass-
land region of the US. Ecology 78: 953-958
Prince, S.D. 1991a. Satellite remote sensing of primary pro-
duction: comparison of results for Sahelian grasslands
1981-1988. Int. J. Remote Sens. 12: 1301-1311.
Sellers, P.J., Tucker, C.J., Collatz, G.J., Las, S.O., Justice
C.O., Dazlich D.A. & Randall, D.A. 1994. A global 1o by
1NDVI dataset for climate studies. Part 2: The generation
of global fields of terrestrial biophysical parameters from
NDVI. Int. J. Remote Sens. 15: 3519-3545.
Sims, P., Singh, J.S. & Lauenroth, W.K. 1978. The structure
and function of ten western North American grasslands. J.
Ecol. 66: 251-285.
Soriano, A. 1991. Rio de la Plata grasslands. In: Coupland,
R.T. (ed.) Natural grasslands. Introduction and Western
Hemisphere, pp. 367-407. Elsevier, Amsterdam.
Tucker, C.J., Vanpraet, C.L., Sharman, M.J. & Van Ittersum,
G. 1985. Satellite remote sensing of total herbaceous
biomass production in the Senegalese Sahel: 1980-1984.
Remote Sens. Environ. 17: 233-249.
Prince, S.D. 1991b. A model of regional primary production
for use with coarse resolution satellite data. Int. J. Remote
Sens. 12: 1313-1330.
Received 15 October 1999;
Revision received 4 July 2000;
Accepted 31 August 2000.
Purevdorj, T., Tateishi, R.T.I. & Honda, Y. 1998. Relation-
Coordinating Editor: T.A. Watt.