Ecosystems (2001) 4: 683–698 DOI: 10.1007/s10021-001-0037-9 ECOSYSTEMS © 2001 Springer-Verlag Current Distribution of Ecosystem Functional Types in Temperate South America Jose´ M. Paruelo,1* Esteban G. Jobba´gy,2 and Osvaldo E. Sala1 1IFEVA. Departmento de Recursos Naturales y Ambiente, Facultad de Agronomı´a, Universidad de Buenos Aires, Av. San Martı´n 4453, Buenos Aires 1417, Argentina; and 2Departmento de Produccio´ n Animal, Facultad de Agronomı´a, Universidad de Buenos Aires, Av. San Martı´n 4453, Buenos Aires 1417, Argentina ABSTRACT We described, classified, and mapped the functional species). More than 25% of the area showed an heterogeneity of temperate South America using NDVI peak in November. Around 40% of the area the seasonal dynamics of the Normalized Difference presented the maximum NDVI during summer. The Vegetation Index (NDVI) from NOAA/AVHRR sat- pampas showed areas with sharp differences in the ellites for a 10-year period. From the seasonal timing of the NDVI peak associated with different curves of NDVI, we calculated (a) the annual inte- agricultural systems. In the southern pampas, NDVI gral (NDVI-1), used as an estimate of the fraction of peaked early (October–November); whereas in the photosynthetic active radiation absorbed by the northeastern pampas, NDVI peaked in late summer canopy and hence of primary production, (b) the (February). We classified temperate South America relative annual range of NDVI (RREL), and (c) the into 19 ecosystem functional types (EFT). The date of maximum NDVI (MMAX), both of which methodology used to define EFTs has advantages were used to capture the seasonality of primary over traditional approaches for land classification production. NDVI-1 decreased gradually from the that are based on structural features. First, the NDVI northeastern part of the study region (southern traits used have a clear biological meaning. Second, Brazil and Uruguay) toward the southwest (Patago- remote-sensing data are available worldwide. Third, nia). High precipitation areas dominated by range- the continuous record of satellite data allows for a lands had higher NDVI-1 and lower RREL values dynamic characterization of ecosystems and land- than neighboring areas dominated by crops. The cover changes. relative annual range of NDVI was maximum for the northern portion of the Argentine pampas (high Key words: remote sensing; land cover; ecosystem cover of summer crops) and the subantarctic forests functioning; South America; Normalized Difference in southern Chile (high cover of deciduous tree Vegetation Index; NDVI. INTRODUCTION and focused on potential rather than current vege- tation. The attributes most frequently used to clas- Global-scale environmental problems are challeng- sify vegetation units were the abundance of plant ing the traditional approaches used to describe eco- functional types or physiognomy (Mueller-Dom- systems at large scales. Traditionally, the character- bois and Ellenberg 1974). Potential vegetation units ization of heterogeneity at regional or continental were often defined on the basis of climate, but the scales relied on the structural features of ecosystems correspondence between vegetation and climate was seldom tested empirically (Holdridge 1947; Box Received 6 October 1999; accepted 2 April 2001. 1981; Prentice and others 1992; but see Stephenson *Corresponding author; e-mail: [email protected] 1990). Mapping potential rather than current veg- 683 684 J. M. Paruelo and others etation was the aim; consequently, anthropogenic land and others 1991; Loveland and Beldward ecosystems, such as crops, cultivated pastures, tree 1997; De Fries and others 1998; Loveland and oth- plantations, or modified rangelands, were system- ers 2000). The IGBP-DIS global land-cover data set atically excluded. is one of the best documented of such products The structural attributes of ecosystems, such as (Loveland and others 2000). The seasonal dynamics vegetation physiognomy or the composition of of spectral indexes and individual bands provided plant functional types change slowly in response to by the AVHRR sensor on board the NOAA satellites human disturbances. These features tend to register were used to define the land-cover classes. Aside the effects of climate change or pollution much from the usefulness for global studies, the IGBP-DIS later than alterations in ecosystem functioning. In- land-cover database and most of the other land- ertia often characterizes the response of vegetation cover products available on the Internet (http:// structure to environmental changes, as reported in edcdaac.usgs.gov/glcc/glcc.html) fail to represent studies of several ecosystems at different time scales the current distribution of both agricultural areas (Pennington 1986; Malanson and others 1992; Mil- and vegetation types in South America. For ex- chunas and Lauenroth 1995). Thus, a characteriza- ample, the main agricultural areas of southern tion of ecosystems that is based exclusively on South America were classified as grasslands and structural attributes may not be sensitive enough to assess the impact of current environmental changes the areas dominated by natural grasslands as if the response of vegetation structure has a large croplands. The IGBP-DIS scheme also fails to de- time lag. Ecosystem functioning, the exchange of scribe the main differences among the semiarid matter and energy between the biota and the envi- and arid land-cover types of central and southern ronment, has in some cases a shorter response time South America. These inaccuracies may be due to than structure (Myneni and others 1997). Today the fact that this global classification, like many most terrestrial ecosystems are far from their orig- others, used training sites located in the Northern inal nonaltered or potential state. Therefore, if we Hemisphere to create classification rules that want to make a valid assessment of the effects of were applied globally. environmental change, our understanding of the In this paper, we characterized the functional current condition of these systems needs to be heterogeneity of temperate ecosystems of South based on sensitive and ecologically meaningful at- America. We described, classified, and mapped cur- tributes. rent vegetation using traits derived from the sea- Remote sensing is a valuable method that can sonal dynamics of the NDVI, which captures the be used to describe the spatial heterogeneity of amount and seasonality of ANPP. In addition to the ecosystems functioning at regional and global regional classification, we evaluated the association scales. Information derived from remotely sensed between the functional ecosystem traits and cli- data can accurately represent functional at- mate. tributes of the ecosystem such as aboveground Based on our functional classification of ecosys- net primary production (ANPP) (Tucker and oth- tems, we developed the concept of ecosystem func- ers 1985a; Prince 1991; Paruelo and others 1997). tional types (EFTs). These units are defined inde- Lloyd (1990) proposed the use of phenology, de- pendently of vegetation structure and focus on the rived from the seasonal course of the Normalized exchange of the energy and matter of ecosystems. Difference Vegetation Index (NDVI) obtained EFTs are conceptually related to plant functional from NOAA/AVHRR satellites, to describe ecosys- types (PFTs); however, EFTs are defined at a differ- tem functioning. Soriano and Paruelo (1992) pro- posed the use of biozones, units defined from ent level of organization than PFTs. EFTs group functional ecosystem traits derived from satellite similarly functioning ecosystems independent of imagery. Nemani and Running (1997) devised a structure, whereas PFTs group similarly functioning scheme to classify land cover into six classes species independent of phylogeny (Chapin 1993). based on the NDVI and temperature data derived As PFTs may be defined according to different func- from the NOAA/AVHRR satellites. Functional tional dimensions (for example, relative growth analyses based on remotely sensed data allow for rates, nitrogen fixation, tolerance or resistance to a top-down characterization of ecosystem heter- herbivory, and so on), EFTs can be defined on the ogeneity (Wessman 1992). basis of different aspects of matter/energy flows. Most of the attempts to describe land-cover pat- Here we focus on the dynamics of primary produc- terns at the global scale in recent years have been tion, one of the essential and most integrative func- based on data obtained via remote sensing (Love- tional attributes of ecosystems. South American Ecosystem Functional Types 685 Figure 1. Phytogeographic units of temperate South America. Redrawn from Ca- brera and Wilkins (1973), Paruelo and others (1991), and Soriano (1991). MATERIALS AND METHODS 1976; Paruelo and others 1991; Soriano 1956, 1991; Leo´ n and others 1998) (Figure 1). Our study focused on the temperate portion of We based our analysis on the NDVI, as deter- South America. We set the northern boundary of mined from AVHRR/NOAA satellites. NDVI com- the temperate zones at 30°S of latitude, which ex- bines the spectral data of channel 1 (red, 580–680 cludes most of the subtropical areas of South Amer- nm) and channel 2 (near infrared, 725–1100 nm): ica (Cabrera 1976) (Figure 1). From a structural NDVI ϭ (channel 2– channel 1)/(channel 2 viewpoint, our work examined forests (subantarc- tic); woodlands (Chaco, Espinal); grasslands (la ϩ channel 1) Pampa, Campos, Patagonia); scrublands (Chilean Matorral); shrub, grass