Full text of DOI:10.2307/1479004

Applied Vegetation Science 3: 253-260, 2000
© IAVS; Opulus Press Uppsala. Printed in Sweden
- ENVIRONMENTAL CONTROLS ON NDVI AND SHEEP PRODUCTION IN THE TIERRA DEL FUEGO STEPPE -
253
Environmental controls of NDVI and sheep production
in the Tierra del Fuego steppe of Argentina
Posse, Gabriela1* & Cingolani, Ana María2,3
1Instituto de Clima y Agua, INTA, Las Cabañas y Los Reseros s/n (1712) Castelar, Provincia de Buenos Aires,
Argentina; 2Centro de Ecofisiología Vegetal (CONICET), Serrano 669 6to piso. (1414) Buenos Aires, Argentina;
3Present address: Instituto Multidisciplinario de Biología Vegetal (CONICET - Universidad Nacional de Córdoba),
CC 495, 5000 Córdoba, Argentina; *Corresponding author; Fax +541146215663; E-mail gposse@cnia.inta.gov.ar
Abstract. We analysed vegetation dynamics in Tierra del
Fuego steppes using Normalized Difference Vegetation In-
dex (NDVI) data provided by advanced very high-resolution
radiometer (AVHRR) on board the National Oceanic and
Atmospheric Administration (NOAA) polar satellite. Our
objective, at a regional scale, was to analyse the spatial
variability of NDVI dynamics in relation to parent material
and geographic location, representing the fertility and cli-
mate gradients respectively; at a local scale, it was to analyse
the inter-annual variability associated with climate and its
relation with sheep production indices. The general pattern
of NDVI dynamics was analysed with Principal Component
Analysis. We found that the geographic location was more
important than landscape type in explaining NDVI dynamics
despite the fact that the variation in landscape type reflects a
fertility gradient strongly associated with floristic composi-
tion and secondary productivity.
Discriminant Analysis was performed to identify the vari-
ables that better distinguish geographic units. The Northern
region (with the lowest precipitation and the highest tempera-
tures) had lower NDVI values over the year. In the Central
region, NDVI reached the highest value of the season, surpass-
ing both other regions. The Southern region (the coldest and
moistest) had its growth pattern displaced towards the sum-
mer. For the Central region we analysed 10 years of monthly
NDVI data with PCA. We found that precipitation from Au-
gust to December and winter temperature are the most impor-
tant determinants of overall NDVI values. Lamb production
was correlated with spring and early summer NDVI values.
Sheep mortality is affected by low NDVI values in late sum-
mer and high annual amplitude. Satellite information allowed
us to characterize the vegetation dynamics of three ecological
areas across the Fuegian steppe.
Introduction
Above-ground net primary productivity (ANPP) is
an integrative indicator of ecosystem functioning (Mc-
Naughton et al. 1989). It reflects processes throughout
the trophic web and is also an important indicator of
nutrient and gas flows in the biosphere. The analysis of
ANPP dynamics and its variation in space and time over
large areas is important for understanding regional and
landscape controls of ecosystem functioning (Paruelo &
Lauenroth 1995) and to interpret and predict timing of
ecosystem processes. Many studies showed that, at least
in arid to sub-humid ecosystems, ANPP is mainly regu-
lated by precipitation, both spatially (e.g. Lauenroth
1979; Sala et al. 1988; Le Houérou et al. 1988) and
temporally (e.g. Smoliak 1986; Lauenroth & Sala 1992).
However, ANPP is also depending on other factors. At
any level of precipitation, such variables as temperature,
distribution of precipitation, soil fertility, amount of
disturbance, succession status (Lauenroth 1979) and
soil water capacity (Sala et al. 1988; Nicholson & Farrar
1994) can influence productivity.
In the Fuegian steppe of Argentina temperature de-
creases towards the south, while precipitation increases,
both factors causing an increase in soil moisture with
latitude (Vallerini 1975). There is also a fertility gradi-
ent associated with different soil parent materials (Ancho-
rena et al. 1991). At the regional scale, floristic compo-
sition seems to respond mainly to soil fertility and
secondarily to soil moisture (Collantes et al. 1999). This
is probably why long-term sheep production, the main
land use in the steppe, is also strongly associated with
soil fertility (Cingolani et al. 1998). Inter-annual varia-
tion of animal productivity was more related with tem-
perature than with precipitation (Cingolani 1992). These
results on secondary productivity, a variable highly
related with ANPP (Coe et al. 1976; McNaughton et al.
1991; Oesterheld et al. 1992), suggested that tempera-
ture and fertility should influence vegetation productiv-
ity in the Fuegian steppe more than precipitation.
Keywords: Climatic gradient; NDVI; NOAA; Precipitation;
Remote sensing; Secondary productivity; Sheep mortality;
Temperature; Vegetation dynamics.
Abbreviations: ANPP = Above-ground Net Primary Produc-
tivity; NDVI = Normalized Difference Vegetation Index;
NOAA = National Oceanic and Atmospheric Administra-
tion.

254
POSSE, G. & CINGOLANI, A.M.
As in other areas with scarce field data, remote
but a strong soil fertility gradient (Collantes et al. 1999).
The strong relation found between ANPP and NDVI in
other regions encouraged us to use the seasonal dynam-
ics of the NDVI as an estimator of vegetation photosyn-
thetic activity in our study area, at local and regional
scales. From the knowledge of environmental constraints
of ANPP we could learn about ecosystem functioning
and thus design sustainable strategies of sheep manage-
ment. Overgrazing and bad management decisions have
led to a decrease in productivity of the system.
Our objective, at a regional scale, was to analyse the
spatial variability of NDVI dynamics in relation to
landscape type and geographical location, which reflect
the regional fertility and climatic gradients respectively
(Collantes et al. 1999). At a local scale, the objective was
to analyse the inter-annual variability associated with
climate and its relation to sheep-production indices. Our
predictions are: (1) differences in NDVI seasonal curves
between landscapes will be greater than differences be-
tween geographic regions; (2) inter-annual variation of
integral NDVI and secondary productivity will be closer
related to temperature than to precipitation.
sensing data seem appropriate for the rapid evaluation
of vegetation-functional aspects (Tucker 1980; Taylor
et al. 1985; Paruelo & Golluscio 1994; Reed et al. 1994).
Despite the loss in precision and detail, satellite informa-
tion has the advantage of providing regional information
and allowing for the evaluation of large areas as a whole
without accessibility restrictions. The Normalized Differ-
ence Vegetation Index (NDVI), a parameter derived from
red and near-infrared reflectance, has proven to be a good
indicator of the intercepted photosynthetic active radia-
tion, both in crops (Kumar & Monteith 1981) and in
natural vegetation (Gamon et al. 1995). In fact, the annual
integration of NDVI is highly correlated with primary
productivity (which is a function of iPAR and radiation-
use-efficiency)forawidevarietyofvegetationtypes(Taylor
et al. 1985; Goward et al. 1985; Burke et al. 1991; Paruelo
et al. 1997). This suggests that a relatively constant
radiation-use-efficiency may be a reasonable assumption
for vegetation where canopy structure and light absorp-
tionchangeinsynchronywithcanopylevelfluxes(Gamon
et al. 1995), as in grasslands (Paruelo et al. 1997).
Our aim was to study the relationship between preci-
pitation, fertility and temperature with ANPP and the
seasonal dynamics of vegetation in the Fuegian steppe, a
region with a short precipitation gradient (300-400 mm)
Methods
Study area
The Fuegian steppe occupies the northern extreme
of Tierra del Fuego Island (Argentina and Chile) (52°
45' to 54° S), but our study was restricted to Argentina.
The climate of this region is semi-arid to subhumid, with
oceanic characteristics (Walter & Box 1983; Koremblit
& Forte Lay 1991). Mean annual precipitation is 300 -
400 mm (Fig. 1). The climate is cold, with a mean
annual temperature of 5.4 °C in Rio Grande City, and
no frost-free period. A temperature decrease from north
to south accompanies a precipitation increase (Vallerini
1975) (Fig. 1). Snowfall is common in winter, and
strong winds are present almost all the year. There is a
water deficit in the summer (Koremblit & Forte Lay 1996).
Soil parent material, the main determinant of vege-
tation (Collantes et al. 1989, 1999) is indicated by
different landscape types: Quaternary meltwater plains
and valleys; Quaternary moraines; Tertiary hills and
plains and mixed landscapes (Tertiary geoforms with
glacial cover) (Fig. 1). The chemical and textural char-
acteristics of the different substrates in each landscape
type produce a soil fertility gradient to which vegetation
responds. There is a gradual shift from dwarf shrub
heaths (with Empetrum rubrum as the dominant spe-
cies) in meltwater plains with very infertile soils, to
grasslands in Tertiary landscapes with fertile soils, pass-
ing through a range of shrubby steppes with different
Fig. 1. Map of the surveyed area with regions dominated by
different landscape types. MEL = Quaternary meltwater plains
and valleys; M = moraines; TER = Tertiary hills and plains;
MIX = Mixed landscapes (Tertiary landforms with glacial drift
cover). Mean July isothermes (___) and isohyetes (- - ) are
shown. RedrawnwithmodificationsfromCollantesetal. (1999).

- ENVIRONMENTAL CONTROLS ON NDVI AND SHEEP PRODUCTION IN THE TIERRA DEL FUEGO STEPPE -
255
densities of mid-size shrubs (Chiliotrichum diffusum) in
moraines and mixed landscapes (Collantes et al. 1999).
Tertiary landscapes have the highest animal productiv-
ity and Quaternary meltwater plains the lowest (Cingolani
et al. 1998). Associated with soil moisture, a second
vegetation gradient occurs; it is mainly related to the
regional climatic gradient (Fig. 1) and secondarily to a
topographic gradient (Collantes et al. 1999).
the values of its six pixels. From these monthly values, four
‘synthetic variables’ (variables which summarize annual
information) were calculated for each sample: the annual
integratedNDVI(astime-weightedaverage);themaximum
and minimum annual values, and the amplitude (difference
between the last two values). In this way, we obtained a
matrix of 57 samples × 11 variables. Seven of the 11
variables were monthly data, and four were synthetic.
We used Principal Component Analysis (PCA, Afifi
1984) to obtain the main directions of variation of NDVI
seasonal curves among sample points. PCA was per-
formed on the variance-covariance matrix calculated
from the data set containing only the seven monthly data
as variables. We described the main variables that dif-
ferentiate geographic regions with Discriminant Analy-
sis (DA, Afifi 1984). Variables were correlated with the
two first axes of DA for better interpretation of results.
Spatial data analysis
As an estimation of the seasonal course of primary
production, we used monthly data of NDVI (Paruelo &
Lauenroth 1995) provided by the AVHRR/NOAA 14
satellite. The timing of the growth season was achieved
by calculating the maximum, minimum and the ampli-
tude of annual NDVI. A multivariate approach was
used, which has been proven to be useful for these types
of data (Paruelo et al. 1991; Paruelo & Lauenroth 1995).
The spatial variability in the NDVI dynamics was ana-
lysed over the entire Argentinian portion of the Fuegian
steppe (ca. 5500 km2). Four landscape types and three
geographical regions were considered. The geographi-
cal regions were distinguished according to the precipi-
tation gradient: Northern (300 mm mean annual pre-
cipitation, 52°45'to53°15'S);Central(350mm, 53°16'to
53° 45' S), and Southern (400 mm, south from 53° 46' S).
Combining landscape types and geographical regions
(Fig. 1), nine sub-regions were obtained: Northern, Cen-
tral and Southern moraines; Northern, Central and South-
ern meltwater plains; Central and Southern mixed land-
scapes and Central Tertiary landscapes.
Temporal data analysis
Inter-annual variability of NDVI dynamics was ana-
lysed for the central region (Fig. 1) where climatic data
were available. Sheep productivity data were supplied
by the administrator of a representative ranch of the
central region (María Behety ranch, 62 000 ha). For this
analysis, NDVI was computed from Global Area Cov-
erage (GAC) images, provided by the Global Inventory
Monitoring and Modeling Systems (Gimms) group at
NASA’s Goddard Space Flight Center. Despite the loss
of resolution (49 km2), the possibility of obtaining
monthly data (composition of maximum NDVI) for
many years (1980-1990) made these images very useful.
In the central area, a sample of 12 pixels was taken per
date and the average was the NDVI value per sample. The
synthetic variables (integrated NDVI, maximum, mini-
mum and amplitude) that characterized each growing
season (September - April) were also calculated. The
basic matrix was finally constituted by 10 samples (grow-
ing seasons) and 12 variables, eight of them were monthly
NDVI values and four were synthetic variables.
Data were analysed with PCA based on a variance-
covariance matrix obtained from the whole data set.
Pearson correlation coefficients were calculated between
each axis and climatic and sheep production data. Cli-
matic variables used were: annual, seasonal and monthly
precipitation; maximum, minimum and mean tempera-
tures for each month; and mean annual temperature.
Sheep production data were: number of lambs produced
per ha; wool production (kg/ha); stocking rate (number
of sheep/ha); and sheep winter mortality (%). Lamb data
are obtained in December (at lamb marking) and wool
data are obtained in January (date of shearing). Stocking
rate and mortality correspond to the winter following
the growth season under study.
We used NOAA-14 LAC (Local Area Coverage)
images (1 km2 of resolution at nadir). From channel 2
(near-infrared, 725-1100 nm) and channel 1 (red, 580 -680
nm) we computed the NDVI as:
(Channel 2 - Channel 1) / (Channel 2 + Channel 1).
Monthly images were composed using the maximum
value composite technique applied to the daily NDVI
values to obtain cloud-free images. This technique mini-
mizes atmospheric and other degrading effects (Holben
1986). We analysed data from September 1996 to March
1997. Winter months were discarded because the asso-
ciated effects of latitude, solar angle, and cloudiness
produce too much noise. Image processing was per-
formed using ERDAS 8.2 software (ERDAS Incorpo-
rated, Atlanta, GA). The processing included geometri-
cal and panoramic corrections. Samples of six pixels (1
km × 1 km each) were taken in each one of the nine sub-
regions. The number of samples taken for each area was
directly related to its relative surface. Samples were
taken at least two pixels from the limit between sub-
regions to minimize spatial error. For each sample, a
unique NDVI value per month was obtained, averaging

256
POSSE, G. & CINGOLANI, A.M.
Results
among samples could be mainly attributed to geographi-
cal location. From both analyses, we learned about the
seasonal dynamics of the three geographic regions, which
is summarized in Fig. 4. The NDVI values and the
change rates were similar for the three regions between
September and October, but thereafter strong differ-
ences emerged. The Northern region did not have a clear
NDVI peak but remained in a low plateau till December
and then decreased. The Central region NDVI reached
its highest value in November, surpassing those of the
other two regions, and after December, it decreased
abruptly. The Southern region had a lower maximum
NDVI value, and it was reached about a month later than
in the Central region, but in January and February NDVI
values remained relatively high in relation to the other
regions. Regions had significant differences in their
integrated NDVI values (one way AOV, F = 12.72, P <
0.0001). This value was significantly lower in the north-
ern region (mean ± SE = 0.323 ± 0.0062) than in the
other two (mean ± SE = 0.369 ± 0.0049 for the central
region and 0.365 ± 0.0046 for the southern region),
which did not show significant differences between
them (Scheffé pairwise comparisons). A DA with sam-
ples grouped by landscape type didn’t add new insights.
Spatial variability
The first and second PCA axes accounted for 52 %
and 20 % of the total variance, respectively. Samples
were well grouped by geographical location and not by
landscape type (Fig. 2). The first axis separated northern
samples, with the lowest monthly NDVI values, from all
others. The second axis separated Central from South-
ern samples. Central samples had higher NDVI values
from September to December and Southern samples had
higher values from January to March.
Samples were well discriminated according to geo-
graphic region by DA, which showed a canonical corre-
lation of 0.82 (P = 0.000) and 0.77 (P = 0.000) for the
first and second axes (i.e. canonical discriminate func-
tions), respectively. These high values and the position
of the samples on the DA axes indicated that geographic
regions have significantly different NDVI seasonal
curves (Fig. 3). The correlation of the variables with the
axes showed that Northern samples had the lowest
monthly values in late spring and summer and the low-
est maximum, integral, and annual NDVI amplitude.
Central and Southern samples were differentiated in
spring (especially November) and late summer (Febru-
ary) NDVI values, Central samples with the highest
values in spring, and Southern samples with the highest
values in late summer. The patterns obtained from both
analyses (PCA and DA, Figs. 2 and 3) were very similar,
indicating that the variation in seasonal NDVI curves
Inter-annual variability
Axes I and II of the 10 growing-season ordination
accounted for 55 % and 25 % of the variance, respec-
tively (Fig. 5). Axis I describes most of the monthly and
yearly variation of NDVI, grouping early months of the
Fig. 2. Samples-variables biplot for first and second PCA
axes. Variables (NDVI monthly data) are identified with a
symbol (x) and labeled with the name of the month. Other
symbols represent samples according to their geographic re-
gion and landscape type: Northern moraines; ꢀꢀ Central
moraines; Southern moraines; Northern meltwater plains;
Fig. 3. Diagram of DA ordination (Axes I and II) of NDVI
monthly data grouped according to geographic region.
ꢀ Northern samples, Central samples, Southern samples.
ꢁ
Arrows represent proportional values (× 10) of correlation
coefficients for variables that showed significant (P < 0.05)
correlation with the axes.
ꢀꢀ Central meltwater plains;
Southern meltwater plains;
ꢀCentral mixed; ꢀ Southern mixed; ꢀꢀCentral Tertiary.

- ENVIRONMENTAL CONTROLS ON NDVI AND SHEEP PRODUCTION IN THE TIERRA DEL FUEGO STEPPE -
257
Fig. 4. Seasonal curves of mean monthly NDVI (with standard
errors) for the three geographic regions: ---ꢀ--- Northern
region; ---ꢀ---Central region; ---ꢁ---Southern region
season as well as integrated, maximum and minimum
values at the negative extreme. Axis II places late months
of the season at the negative extreme and NDVI annual
amplitude at the positive end. Climatic and animal pro-
duction variables that were significantly correlated with
the two first PCA axes are shown in Table 1. The most
relevant variables were plotted with arrows in Fig. 5a.
Fall, winter, spring, and annual precipitation were nega-
tively correlated with Axis I and positively correlated
with almost all variables related to NDVI. Precipitation
of late winter and spring (from August to December)
Fig. 5. Diagram of PCA ordination (Axes I and II) of the 10
growing seasons (from 1981 to 1990) relative to NDVI monthly
data and synthetic variables for the Central region. a. Ordina-
tion variables (eight were monthly NDVI values and four
synthetic variables); arrows represent the values of the corre-
lation coefficients of the most relevant climatic and animal
production variables (see Table 1). b. Ordination of the 10
Table 1. Pearson correlation coefficients of climatic and sheep
production variables with PCA Axes I and II. Only significant
relationships are shown.
growing-season samples.
NDVI annual peak in Novem-
Axis I
Axis II
ber, NDVI annual peak in December, ꢀꢀNDVI annual peak
in January.
Late winter-spring precipitation (Aug-Dec)
Fall-winter-spring precipitation (Mar-Dec)
Winter-spring precipitation (Jun-Dec)
Spring precipitation (Sep-Dec)
–0.90 ** +0.10 ns
–0.87 ** +0.08 ns
–0.84 ** +0.00 ns
–0.76 ** +0.04 ns
Annual precipitation (Mar-Feb)
Fall precipitation (Mar-Jun)
Winter-spring-summer precipitation (Jun-Mar) –0.71 *
Winter precipitation (Jun-Sep)
Fall-winter precipitation (Mar-Aug)
June precipitation
–0.75 *
–0.75 *
–0.06 ns
+0.15 ns
–0.38 ns
–0.04 ns
+0.11 ns
–0.12 ns
was the most strongly correlated variable, indicating the
importance of precipitation falling just before and dur-
ing the first months of the growth season. This is con-
firmed by the strong correlation found between August-
December precipitation and integrated NDVI (R = 0.91,
P < 0.01). A similar correlation with Axis I was found
with winter mean temperatures. Axis II was positively
and strongly correlated with spring temperature (Table
1), and negatively with summer precipitation.
Lamb production was negatively correlated with axis
I (Table 1), indicating that seasons with higher NDVI
values in spring and early summer were the most produc-
tive. The individual factors most strongly associated with
lamb production were September and minimum NDVI
(R = 0.76, P = 0.010 and R = 0.74, P = 0.014, respec-
–0.68 *
–0.64 *
–0.63 *
August mean temperature
Winter mean temperature (Jun-Set)
July mean temperature
–0.69 *
–0.66 *
–0.64 *
+0.34 ns
+0.27 ns
+0.24 ns
Summer precipitation (Dec-Mar)
–0.24 ns
–0.63 *
Spring mean temperature (Sep-Dec)
December maximum temperature
November mean temperature
November minimum temperature
Spring maximum temperature (Sep-Dec)
+0.07 ns +0.76 **
–0.22 ns +0.67 *
+0.04 ns +0.67 *
–0.04 ns +0.73 *
–0.33 ns +0.64 *
Lambs/ha
Mortality
–0.63 *
+0.38 ns +0.65 *
+0.04 ns
* P < 0.05, ** P < 0.01.

258
POSSE, G. & CINGOLANI, A.M.
tively). Other factors with significant correlation were
the integrated NDVI (R = 0.67, P = 0.032), November
NDVI (R = 0.65, P = 0.040), and October NDVI (R =
0.64, P = 0.047). Mortality in the following winter was
correlated positively with Axis II (Fig. 5a, Table 1). The
individual factors which showed the stronger correlation
with sheep mortality were NDVI amplitude (R = 0.73,
P = 0.016), April NDVI (R = – 0.73, P = 0.016) and
minimum NDVI (R = –0.66, P = 0.038).
of integrated NDVI northward, as temperature decreases,
in the United States. Yang et al. (1998), for the northern
and central Great Plains in North America, found that
temperature is a limiting factor in those grasslands. In
Tierra del Fuego, the oceanic characteristic of the cli-
mate produces a cold summer without a frost-free pe-
riod (Koremblit & Forte Lay 1991) and this accentuates
temperature limitations.
Lack of the expected relationship between land-
scape units and NDVI dynamics could be due because
secondary productivity (which is related with landscape
units) and primary productivity (which is not related
with landscape units) are actually not linked, as we had
assumed. Landscapes have strong differences in species
composition (Collantes et al. 1999), which probably
cause differences in forage quality, but at our scale these
differences were not reflected in NDVI dynamics. Pre-
vious works in the region suggest that secondary pro-
ductivity is more related with the forage quality than
with the primary productivity. Posse et al. (2000) pointed
out that lawn patches supported more sheep than sur-
rounding tussock grassland of Festuca gracillima. This
was because lawn patches had lower productivity but
higher quality (Posse 1997). At our landscape scale, the
short grassland community (similar to lawn patches as
to floristics and productivity ) supports higher stocking
rates than the tussock grassland (Anchorena & Cingolani
unpubl.) The lack of evident differences in NDVI dy-
namics between landscapes, despite their strong struc-
tural differences, could also emerge because of high
intra-landscape variability in habitat types.
The temporal analysis showed that the variability in
NDVI integral in the Central region is mainly explained
by precipitation. The strong correlation of integrated
NDVI with August to December precipitation shows the
importance of rain during the growth season (see Table 1).
Yang et al. (1998) and Briggs & Knapp (1995) have
found the same in North American central grasslands.
The high correlation with fall and winter precipitation
showed the importance of winter recharge of the soil.
When summer precipitation was added to late-winter
and spring precipitation, the relation with axis I was
not significant, indicating that productivity at the end
of the growth season is not important over total pro-
ductivity, as was demonstrated by February, March
and April low contribution to axis I (Fig. 5). Winter
temperature was also related with the first ordination axis
(Fig. 5, Table 1) suggesting that it influences primary
productivity. However both variables were correlated in
our data set (e.g. August-December precipitation with
August mean temperature: R = 0.73 P = 0.015), so we
could not separate both effects.
Discussion
Contrarily to our expectation, we found more varia-
tion in NDVI dynamics among geographic regions than
among landscapes (Fig. 2). Geographic regions, with
similar patterns of vegetation structure and relatively
small differences in climate, had large differences in
NDVI monthly values (Fig. 3). The timing of the growth
season reflects the constraint between regional tempera-
ture and precipitation gradients. The balance between
high temperatures and low precipitation in the Northern
region implies restrictions over vegetation photosyn-
thetic activity (Fig. 4). The Southern region has its
growth pattern displaced towards the summer. This
region has the lowest water restriction and the lowest
temperatures. The Central region, with a balance be-
tween temperature and precipitation constraints, has the
highest amplitude and maximum NDVI values. The sud-
den decrease of photosynthetic activity that occurs in sum-
mer (over all three regions) may be attributed to soil water
depletion during the growth season (Paruelo et al. 1991;
Cihlar et al. 1991; Yang et al. 1998). Thus, even though the
climatic gradients among the regions are shallow (100 mm
in precipitation and 2 ºC in mean temperature), they are
sufficient to generate important differences in the func-
tional characteristics of these regions.
At our scale of analysis, annual ANPP (in NDVI
integrate units) did not increase with increasing soil
moisture, as it does in other regions of the world (e.g.
Sala et al. 1988; Nicholson & Farrar 1994; Paruelo &
Lauenroth 1995; Yang et al. 1998). The Southern region
(with the highest precipitation and the lower tempera-
tures, and hence, with a better water balance than the
other two regions) did not differ in NDVI integrate from
the Central region, with lower soil moisture. Appar-
ently, the higher NDVI values in summer of the South-
ern region compensate for the lower values in previous
months, being annual integrated NDVI similar to that of
the central region. Precipitation and temperature seem to
be, respectively, the limiting factors of annual ANPP in
Northern and Southern regions. Temperature limitations
to vegetation growth with increasing latitude were re-
ported by Godward et al. (1985), who found a decrease
Lamb production was related positively with spring
and early summer NDVI values. The strong relationship

- ENVIRONMENTAL CONTROLS ON NDVI AND SHEEP PRODUCTION IN THE TIERRA DEL FUEGO STEPPE -
259
Agropecuaria supported this research. A. Cingolani enjoys a
fellowship of CONICET. We thank Carlos Di Bella for assist-
ance with satellite data acquisition. The manuscript was greatly
improved by the comments of Juan Anchorena, Susana
Stoffella, Karen Braun and Carlos Di Bella and two anony-
mous reviewers. We thank Daniel Renison and an anonymous
reviewer for the critical reading of the English version.
between lamb production and September NDVI suggests
the importance of availability of green forage in this
month, when lambing ewes have high nutritional re-
quirements (Russel 1971). In contrast to our initial
hypothesis, lamb production was not related to precipi-
tation or temperature variables. A more detailed study
of the factors influencing spring NDVI would be neces-
sary for understanding the relations between climate
and lamb production.
References
Winter mortality is affected by low NDVI values in
late summer and high annual amplitude. Higher ampli-
tudes are logically correlated with summer droughts,
which diminish the summer NDVI. High spring tem-
peratures promote a higher and earlier peak, in Decem-
ber (positive years on Axis II), followed by lower values
in late summer probably owing to accelerated soil water
depletion. Therefore, if there is low availability of green
forage in summer, ewes could not reach in that season
the necessary weight to withstand the winter period. Ewe
live weight at the onset of the winter was negatively
correlated with winter mortality in the Central region
(Anchorena pers. comm). Both lamb production and
mortality are more correlated with minimum NDVI rather
than with maximum NDVI. Probably this is a conse-
quence of land use: the traditional management of con-
tinuous grazing requires stocking the land according to
winter capacity, but later this small number of animals
cannot take advantage of the high peaks of production in
spring.
At the temporal scale, in the Central region, effects
of precipitation and temperature could be discriminated,
the first contributed more to total annual ANPP (i.e. to
axis I) and the second to the timing of the growth season
(axis II). At the spatial scale, however, discrimination is
not possible because precipitation and temperature are
linked, and the conjunction of both factors explained
both aspects of NDVI dynamics. It is possible that a
temporal analysis in the Northern region, with higher
water limitations, would highlight even more the impor-
tance of precipitation for NDVI dynamics, while a tem-
poral analysis in the Southern region, would show an
increased importance of temperatures. We suggested
that animal and vegetal productivity could be not linked
because of strong differences in forage quality. But if
the spatial variability was eliminated, taking into ac-
count the temporal variation over a relative constant
vegetation structure and forage quality, relations be-
tween ANPP and sheep production became more evi-
dent, although the timing of the growth season (i.e. what
happens in early spring and in late summer) seems to be
more important than annual ANPP. Then, both spatial
and temporal analysis suggested that factors other than
total annual ANPP would govern animal productivity in
northern Tierra del Fuego grasslands.
Afifi, A.A. 1984. Computer-aided multivariate analysis. Life-
time Learning Publications, Belmont, CA.
Anchorena, J., Bianciotto, O. & Collantes, M.B. 1991. Los
suelos del norte de Tierra del Fuego. In: Actas del XII
Congreso Argentino de la Ciencia del Suelo, pp. 193-195.
San Carlos de Bariloche.
Briggs, J.M. & Knapp, A.K. 1995. Interannual variability in
primary production in tallgrass prairie: climate, soil mois-
ture, topographic position, and fire as determinants of
aboveground biomass. Am. J. Bot. 82: 1024-1030.
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.
Cihlar, J., St. Laurent, L. & Dyer, J.A. 1991. Relation between
the Normalized Difference Vegetation Index and Ecologi-
cal Variables. Remote Sens. Environ. 35: 279-298.
Cingolani, A.M. 1992. La producción ovina en Tierra del
Fuego. Incidencia de los factores ecológicos. Seminario
de Licenciatura, Universidad de Buenos Aires.
Cingolani, A.M., Anchorena, J. & Collantes, M.B. 1998.
Landscape heterogeneity and long-term animal productiv-
ity. A case study of a sheep ranch in Tierra del Fuego. J.
Range Manage 51: 79-87.
Coe, M.J., Cumming, D.H. & Phillipson, J. 1976. Biomass
and production of large African herbivores in relation to
rainfall and primary production. Oecologia (Berl.) 22:
341-354.
Collantes, M.B., Anchorena, J. & Koremblit, G. 1989. A soil
nutrient gradient in Magellanic Empetrum heathlands.
Vegetatio 80: 183-193.
Collantes, M.B., Anchorena, J. & Cingolani, A.M. 1999. The
steppes of Tierra del Fuego: floristic and growthform
patterns controlled by soil fertility and moisture. Plant
Ecol. 140: 61-75.
Gamon, J.A., Field, C.B., Goulden, M.L., Griffin, K.L., Hartley,
A.E., Joel, G., Peñuelas, J. & Valentini, R. 1995. Rela-
tionships between NDVI, canopy structure and photosyn-
thesis in three Californian vegetation types. Ecol. Appl.
5: 28-41.
Goward, S.N., Tucker, C.J. & Dye, D.G. 1985. North Ameri-
can vegetation patterns observed with NOAA-7 advanced
very high resolution radiometer. Vegetatio 64: 3-14.
Holben, B. 1986. Characteristics of maximum value compos-
ite images from temporal AVHRR data. Int. J. Remote
Sens. 7: 1417-1434.
Koremblit, G. & Forte Lay, J.A. 1991. Contribución al estudio
agroclimático del norte de Tierra del Fuego (Argentina).
An. Inst. Patagonia Ser. Cienc. Nat. 20: 125-134.
Acknowledgements. The Instituto Nacional de Tecnología

260
POSSE, G. & CINGOLANI, A.M.
Koremblit, G. & Forte Lay, J.A. 1996. Características agro-
hidrológicas del norte de Tierra del Fuego. Libro de
resúmenes del VII Congreso Argentino de Meteorología,
Buenos Aires.
Kumar, M. & Monteith, J.L. 1981. Remote sensing of crop
growth. In: Smith, H. (ed.) Plant and the daylight spec-
trum. Academic Press, New York, NY.
Lauenroth, W.K. 1979. Grassland primary production: North
American grasslands in perspective. In: French N.R. (ed.)
Perspectives in grassland ecology. Ecological studies,
Vol. 32, pp. 3-22. Springer-Verlag, New York, NY.
Lauenroth, W.K. & Sala, O.E. 1992. Long-term forage pro-
duction of North American shortgrass steppe. Ecol. Appl.
2: 397-403.
Le Houérou, H.N., Bingham, R.L. & Skerbek, W. 1988.
Relationships between the variability of primary produc-
tion and the variability of annual precipitation in world
arid lands. J. Arid Environ. 15: 1-18.
McNaughton, S.J., Oesterheld, M., Frank, D.A. & Williams,
K.J. 1989. Ecosystem-level patterns of primary productiv-
ity and herbivory in terrestrial habitats. Nature 341: 142-
144.
McNaughton, S.J., Oesterheld, M., Frank, D.A. & Williams,
K.J. 1991. Primary and secondary production in terrestrial
ecosystems. In: Cole, J., Lovett, G. & Findlay, S. (eds.)
Comparative analyses of ecosystems. Patterns, mecha-
nisms and theories, pp. 120-139. Springer-Verlag, New
York, NY.
by sheep grazing. J. Veg. Sci. 11: 43-50.
Reed, B.C., Brown, J.F., VanderZee, D., Loveland, T.R.,
Merchant, J.W. & Ohlen, D.O. 1994. Measuring pheno-
logical variability from satellite imagery. J. Veg. Sci. 5:
703-714.
Russel, A.J.F. 1971. Relationship between energy intake and
productivity in hill sheep. Proc. Nutr. Soc. 30: 197-204.
Sala, O.E., Parton, W.J., Joyce, L.A. & Lauenroth, W.K. 1988.
Primary production of the central grassland region of
United States. Ecology 69: 40-45.
Smoliak, S. 1986. Influence of climatic conditions on produc-
tion of Stipa-Bouteloua over a 50 years old period. J.
Range Manage. 39: 100-103.
Taylor, B.F., Dini, P.W. & Kidson, J.W. 1985. Determination
of seasonal and interannual variation in New Zealand
pasture growth form noaa-7 data. Remote Sens. Environ.
18: 177-192.
Tucker, C.J. 1980. A critical review of remote sensing and
others methods for non-destructive estimation of standing
crop biomass. Grass Forage Sci. 32: 177-182.
Vallerini, J. 1975. Relevamiento expeditivo de recursos natu-
rales de la zona cordillerana de Patagonia. INTA. Est.
Exp. Bariloche, Río Negro (mscr.).
Walter, H. & Box, E.O. 1983. Climate of Patagonia. In: West,
N.E. (ed). Temperate deserts and semi-deserts, pp. 432-
435. Elsevier Scientific Publishing Company, Amster-
dam.
Yang, L., Wylie, B.K., Tieszen, L.L. & Reed, B.C. 1998. An
analysis of relationships among climate forcing and time-
integrated NDVI of grasslands over the U.S. Northern and
Central Great Plains. Remote Sens. Environ. 65: 25-37.
Nicholson, S.E. & Farrar, T.J. 1994. The influence of soil type
on the relationships between NDVI, rainfall and soil mois-
ture in semiarid Bostwana. I. NDVI response to rainfall.
Remote Sens. Environ. 50: 107-120.
Oesterheld, M., Sala, O.E. & McNaughton, S.J. 1992. Effect
of animal husbrandy on herbivore-carrying capacity at a
regional scale. Nature 356: 234-236.
Paruelo, J.M. & Golluscio, R.A. 1994. Range assessment
using remote sensing in northwest Patagonia (Argentina).
J. Range Manage. 47: 498-502.
Received 24 August 2000;
Revision received 6 November 2000;
Accepted 12 November 2000.
Coordinating Editor: J. Bakker.
Paruelo, J.M. & Lauenroth, W.K. 1995. Regional patterns of
Normalized Difference Vegetation Index in North Ameri-
can shrublands and grasslands. Ecology 76: 1888-1898.
Paruelo, J.M., Aguiar, M.R., Leon, R.J.C. & Batista, W.B.
1991. The use of satellite imagery in quantitative phyto-
geography: a case study of Patagonia (Argentina). In:
Nimis, P.L. & Crovello, T.J. (eds.) Quantitative approaches
to phytogeography, pp. 183-204. Kluwer Academic, The
Hague.
Paruelo, J.M., Aguiar, M.R., Golluscio, R.A., León, R.C. &
Pujol, G. 1993. Environmental controls of NDVI dynam-
ics in Patagonia based on NOAA-AVHRR satellite data.
J. Veg. Sci. 4: 425-428.
Paruelo, J.M., Epstein, H. E., Lauenroth, W.K & Burke, I.C.
1997. ANPP estimates from ndvi for the central grassland
region of the United States. Ecology 78: 953-958.
Posse, G.1997.Interacciónaniveldecomunidadentrelahetero-
geneidad de la vegetación y el pastoreo ovino en la estepa
magallánica. Ph.D. Thesis, University of Buenos Aires.
Posse, G., Anchorena, J. &. Collantes, M.B. 2000. Spatial
micro-patterns in the steppe of Tierra del Fuego induced