In the cerrado physiognomies (campo cerrado, cerrado sensu stricto, and cerradão), 360 species were found, belonging to 236 genera and 69 families (Tab. 1 and 2). The richest families were: Fabaceae, Asteraceae, Poaceae, Rubiaceae, Bignoniaceae, Myrtaceae, Malpighiaceae, Malvaceae, and Apocynaceae, together accounting for 57.8% of collected species (Fig. 2).
These families are also the most representative ones in other cerrado areas, such as Lagoa Santa (Warming 1892), Triângulo Mineiro (Goodland 1979), Brasília (Ratter 1980), Moji Guaçu (Mantovani & Martins 1993), and Pirassununga (Batalha et al. 1997).
Of the total collected species, 16 (4.4%) are weeds that does not occur spontaneously in cerrado sites (Mendonça et al. 1998). In this vegetation type, this is mainly a consequence of the fragmentation and edge effect (Pivello et al. 1998, 1999b). Mendonça et al. (1998) listed for the whole Cerrado Domain 289 weedy species, or 4.5% of its total flora, the same proportion observed in the Pé-de-Gigante flora. Plant invasions are a major problem in cerrado fragments, occuring pratically in every one of them, dominating in patches and outcompeting native species (Pivello et al. 1999b). This situation occurs also in the Pé-de-Gigante Reserve, where African grasses, especially Melinis minutiflora P. Beauv. and Brachiaria decumbens Stapf, are spreading fast (Pivello et al. 1999c).
A total of 272 species were found in the campo cerrado; 308, in the cerrado sensu stricto; and 148, in the cerradão (Tab. 2). Following the "forest-ecotone-grassland" concept (Coutinho 1978), the ecotonal physiognomies are expected to be richer, because they contain species from both the herbaceous and the woody components. Indeed, the existing ecotonal physiognomies in the Pé-de-Gigante Reserve, campo cerrado and cerrado sensu stricto, were richer than one of the extremes, cerradão. The woody to herbaceous species ratio for all three physiognomies together (cerrado sensu lato) was approximately 1:2 (Tab. 2). This ratio increased from cerradão, in which the number of woody species was higher than those of herbaceous species, to campo cerrado (Tab. 2). Once again, this is expected according to the "forest-ecotone-grassland" concept (Coutinho 1978), which predicts a gradual decrease in the number of woody species from cerradão to campo limpo. In Brasília (Ratter 1980), this ratio was also 1:2, while in Lagoa Santa (Warming 1892) and Moji Guaçu (Mantovani & Martins 1993) it was 1:3. In Santa Rita do Passa Quatro and Brasília, there is a predominance of closed cerrado physiognomies (cerrado sensu stricto and cerradão). In Lagoa Santa and Moji Guaçu, on the other hand, open cerrado physiognomies (campo sujo and campo cerrado) prevail, explaining the higher proportion of herbaceous species in these two areas.
In the three cerrado physiognomies sampled in the Pé-de-Gigante Reserve, the highest similarity was between campo cerrado and cerrado sensu stricto (0.803). Between cerrado sensu stricto and cerradão, the similarity value was 0.592 and between campo cerrado and cerradão, this value was 0.519. The same pattern was also observed in Itirapina (Mantovani 1990), where campo cerrado and cerrado sensu stricto were the most similar physiognomies and the campo cerrado and cerradão, the most dissimilar ones.
Comparing the Pé-de-Gigante Reserve with cerrado sites in Campininha Farm and Emas, 76 families, 287 genera and 516 species occur in Campininha Farm (Mantovani & Martins 1993) and 69 families, 229 genera and 358 species in Emas (Batalha et al. 1997). Despite its larger area (1269 ha), the Pé-de-Gigante Reserve presented a floristic richness similar to that found in Emas (16 ha) and lower than that found in Campininha Farm (342 ha). This could be a consequence of recent human impact on the reserve, such as timber explotation and cattle-raising, and the invasion of weeds (Pivello et al. 1999a, 1999b), but also of the prevalence of open physiognomies in Emas and Campininha Farm. Besides, differences on sampling effort could have contributed also to these results.
The similarity index values between the Pé-de-Gigante Reserve and Campininha Farm and between the Pé-de-Gigante Reserve and Emas were 0.558 and 0.659, respectively. Between Campininha Farm and Emas, this value was 0.629 (Fig. 3).
The Campininha Farm distinguished itself from the Pé-de-Gigante Reserve and from Emas by its greater floristic richness. This variation existed mainly at species level, since the genera and families remained approximately constant, as stated previously by Mantovani & Martins (1993).
Herbaceous component - The 236 herbaceous species comprised 65.6% of the flora as a whole. The families represented only in this component accounted for 40.6% of the total number of families, a value lower than those obtained in Campininha Farm (44.7%) (Mantovani & Martins 1993) and Emas (49.7%) (Batalha et al. 1997). The richest families in this component were: Asteraceae, Fabaceae, Poaceae, Rubiaceae, Bignoniaceae, Malvaceae, Apocynaceae, Euphorbiaceae, and Malpighiaceae, representing 66.5% of the total number of herbaceous species. These families are also the most representative ones in this component in other cerrado sites, such as Campininha Farm (Mantovani & Martins 1993) and Emas (Batalha et al. 1997).
The similarity index values between the Pé-de-Gigante Reserve and the Campininha Farm and between the Pé-de-Gigante Reserve and Emas were 0.472 and 0.609, respectively. Between Campininha Farm and Emas, this value was 0.580 (Fig. 3). The much higher richness found in Campininha Farm (391) contributed to separate it from Pé-de-Gigante and Emas (236 and 250 species, respectively).
There are few studies including the herbaceous species of the cerrado flora (Castro et al. 1999). The seasonality of the epigeal portion of the herbaceous species, that lasts from a few months to two years, contributes to the poor knowledge of this component. As a consequence of the short epigean cycle of many species and depending on fire frequency and intensity, the herbaceous component composition changes a lot throughout the year (Mantovani & Martins 1993).
Woody component - The 124 woody species represented 34.4% of the whole flora. Of all sampled families, 29.0% had species only in the woody component, a value between those found in Campininha Farm (34.7%) (Mantovani & Martins 1993) and Emas (24.6%) (Batalha et al. 1997).
The richest families in this component were: Fabaceae, Myrtaceae, Melastomaceae, Vochysiaceae, Malpighiaceae, Rubiaceae, Annonaceae, Bignoniaceae, and Erythroxylaceae, comprising 57.3% of collected species. These families are also the best represented in Campininha Farm (Mantovani & Martins 1993) and Emas (Batalha et al. 1997).
In this component, 108 species were sampled in Emas and 125 in Campininha Farm. The similarity index values between the Pé-de-Gigante Reserve, Campininha Farm and Emas were all between 0.77 and 0.81 (Fig. 3). The similarity index values in this component were higher than those found in the herbaceous one, showing that the heterogeneity is higher in the herbaceous than in the woody flora. This higher heterogeneity of the herbaceous component was evident only at species level, since at genus and especially at family levels there was no pronounced difference between the two components.
Castro et al. (1999) made an extensive compilation of many floristic and phytosociological studies carried out in cerrado vegetation and elaborated a check-list of woody species. Of the species sampled here, only Myrcia guianensis (Aubl.) A. DC. (Myrtaceae) was not reported by these authors.
Andira anthelmia (Vell.) J. Macbr. (Fabaceae), Byrsonima crassa Naud. (Malpighiaceae), Anadenanthera peregrina (L.) Speg. (Mimosaceae), Eugenia langsdorffii O. Berg (Myrtaceae), Myrcia pubipetala Miq. (Myrtaceae), Luehea divaricata Mart. (Tiliaceae), Syagrus romanzoffiana (Cham.) Glass. (Arecaceae), as well as Myrcia guianensis, were also not reported as ocurring in São Paulo State cerrado areas by Leitão-Filho (1992) and should be added to his list.
We are grateful to Paulo César Fernandes, for field assistance, and to FAPESP, for the financial support conceded to the first author (process 95/4290-3).