摘要：We described vegetation heterogeneity and its relationship with environmental variables over an area of the basaltic region of central and north-western Uruguay dominated by shallow soils. Natural grasslands under continuous grazing by cattle and sheep are the predominant vegetation type (90%) of this geomorphologic region that comprises approximately 1.5 millions hectares. Vegetation data were obtained using the relevé method. Stands were sampled preferentially to represent all the perceptible physiognomic vegetation types. In quadrats of 10 x 10 m, located in the centre of each stand, floristic composition and structural characteristics of vegetation were recorded. Each species was visually assigned to a cover – abundance class following the scale proposed by Braun-Blanquet. Additionally, for each stand, we measured environmental variables (soil depth, top 10 cm of soil texture, slope, percent cover of rock and pebble, macro and micro topographical characteristics and livestock faecal abundance). Normalized Differential Vegetation Index (NDVI), a surrogate for net primary production, was derived from Landsat TM images. The species by relevés matrix with presence/absence data was subjected to multivariate analysis to reveal vegetation patterns. Agglomerative classification, based on Ward´s method and Jaccard’s distance, was used to identify main groups in species relevés and the hierarchical relationships among them. The optimal levels of clustering and the indicator species for each cluster were determined by indicator species analysis. Correspondence analysis was carried out to describe floristic gradients. Relationships between ordination axes and environmental variables were explored by correlation analysis. Two hundred seventy four vascular plant species, representing 43 families, were encountered in 46 relevés. The families with the highest number of species were Poaceae (78) and Asteraceae (44). The best represented genera were: Stipa (8), Paspalum (7), and Aristida (7). A structure of six communities grouped in three main vegetation units emerged from this procedure: Meso–xerophytic grasslands, Litophytic steppes and Meso–hidrophytic grasslands. The main vegetation units showed different physiognomies and occurred on sites with different combinations of topography and soil properties. Meso–xerophytic grasslands showed predominantly a two layer structure with high-to-mid plant cover. The upper layer was dominated by grasses and sub-shrubs and the lower layer by forbs and grasses. This unit occurred mainly on shallow soils on steep and gentle slopes and convex interfluves of hills. Litophytic steppes showed one open layer dominated by Selaginella sellowii Hieron, a small pteridophyte. This unit was associated to flat erosion surfaces at high and middle topographical positions. The Meso–hidrophytic grasslands unit was a closed vegetation type, with two layers dominated by grasses and graminoids. This unit occurred predominantly on deep soils on gentle low slopes, valleys and plains. The first ordination axis explained a 62.7% of the variance and detected a water availability gradient going from Meso–hidrophytic grasslands to Litophytic steppes and Meso–xerophytic grasslands. The environmental controls of this water gradient were soil depth, texture, slope and microtopography. The position of the relevés along the main floristic gradient was strongly correlated to NDVI (p < 0.05).The second axis explained a 20.3% of the variation, and separated Litophytic steppes from Meso–xerophytic and Meso-hidrophytic grasslands. The present study provides information that could be useful for the design of grassland management practices and selection of conservation areas
