The CRC are found in the region of Carajás, located in the southeast part the State of Pará (Viana et al. 2016; Zappi et al. 2019), one of the largest mineral provinces in the world (Ab’saber 1986). At the Serra dos Carajás, the CRC appears atop a series of outcrops that form discontinuous island-like habitats of open, shrubby or grassy vegetation within a dense matrix of rainforest in the southeastern Amazon basin (Mota et al. 2018).

Most of the ferruginous island complex in the southeastern Amazon is within areas protected at different levels. The Serra Norte (SN1, SN2, SN3, SN4, SN5, SN6, SN7, SN8), the Serra Sul (S11A, S11B, S11C S11D) are located in the Floresta Nacional de Carajás, which is an area of sustainable use and thus subject to anthropogenic pressures, and iron ore mining currently occurs in areas SN4, SN5 and S11D. The Serra da Bocaina and Serra do Tarzan are the only fully protected areas, and are both inserted within the Parque Nacional dos Campos Ferruginosos (PNCF). However, the Serra Arqueada and Serra de Campos of São Félix do Xingu have no legal protection.

The Serra de Campos (SFX) is a canga outcrop found in the municipality of São Félix do Xingu, southeastern Pará state, Brazilian Amazon. It represents the westernmost limit of the Serra dos Carajás, a complex of ferruginous highland outcrops that extends eastwards to the Municipality of Curionópolis, totalling 126 km2. The plateaus previously studied in the scope of the FCC project (Viana et al. 2016) are found in the Municipalities of Parauapebas (Serra Norte – SN1 to SN8), and Canaã dos Carajás (Serra Sul – S11, Serra do Tarzan – ST and Serra da Bocaina – SB). The SFX comprises two plateaus measuring c. 9 km2, distant about 1 km from each other, known as SFX1 and SFX2 (Fig. 1). The largest of the two plateaus, known as SFX2, extends for 8.5 km and covers an area of 7.6 km2, while SFX1 is 2.5 km long, measuring 1.4 km2. The plateaus are located at 6°23'41"S, 51°52'25"W, with altitudes ranging from 580 to 730 m. a.s.l. (Fig. 1). Distant about 80 km west from SA, the SFX can be accessed through the Municipality of São Felix do Xingu first by crossing the Rio Fresco then taking a road that goes through farmland, leading, after a steep climb, to the canga plateaus.

Figure 1. 

a Geographic location of the present study site at SFX and the other study areas from Carajás complex b aerial view of an island of canga vegetation surrounding by the rainforest (Photo: Leonardo Vianna) c Serra de Campos of São Félix do Xingu (SFX) phytophysiognomy with shrubby and grassy vegetation.

Botanical specimens from SFX deposited in herbaria prior to this study were located through an online search at the Herbarium of the Museu Paraense Emílio Goeldi (MG) and Herbário Ezechias Paulo Heringer (HEPH) (acronyms according to Thiers, continuously updated). Prior to our expeditions, specimens at MG were collected in the 1990´s by João Batista Fernandes da Silva and include the type of Mimosa dasilvae A.S.L. Silva & Secco and several gatherings of Orchidaceae, while HEPH currently holds collections made by Annajulia Elizabeth Heringer Salles and J.B.F. Silva in 2001. All materials available in these collections were analyzed and included in this study.

Four plant collecting expeditions were carried out between 2016 and 2019 (May 2016, April 2017, March 2018, October 2019), aiming to collect fertile material of all vascular species. Collecting method followed Filgueiras et al. (1994) with random walks covering the accessible parts of both plateaus, attempting to stop every 1 km to sample the vegetation and collect fertile specimens. We aimed to visit diverse vegetation types, including open canga slabs, nodular canga, canga scrub, palm swamps (buritizais) and temporary lagoons (Mota et al. 2015)

The samples collected were identified to species by comparing their macroscopic and microscopic morphological features with available bibliography, against herbarium collections (physically and on-line) and also consulting key family specialists. Voucher specimens were deposited at MG. Only one collection number per taxon is cited in the present floristic list. A full specimen list is provided in supplement S1. Species names follow Flora do Brasil online (Flora do Brasil under construction), family delimitation followed APG IV (Angiosperm Phylogeny Group 2016) and author abbreviations follow IPNI (2019).

Seed plant species distribution data were assembled from the FCC project (Mota et al. 2018), with the compilation of a database comprising 3228 occurrences of 823 species (Zappi et al. 2019). The updates included 23 recent new occurrences for SN1, SN4, SN5, SN7, S11D, and the Serra da Bocaina based on recently collected herbarium material; 149 species for SA (Fonseca-da-Silva et al. 2020); and the newly prepared dataset of SFX. The assembled database comprises 909 seed plant species recorded in CRC at the Carajás complex, including 16 sites (SN1, SN2, SN3, SN4, SN5, SN6, SN7, SN8, S11A, S11B, S11C, S11D, ST, SB, SA and SFX). For the purpose of our analyses, exotic, invasive and weedy species were removed from the dataset as identified in (Giulietti et al. 2018), resulting in 893 species analysed. The code assigned for each site is found in Table 2.

To perform the biogeographical analysis of the CRC of the Carajás complex, the species database was used to investigate the floristic similarity and shared endemicity between different mountaintops across canga sites. Invasive exotic species recorded in each site were excluded from this analysis, as well as specimens with imprecise identification, Lycophytes, and Monilophytes. Floristic similarity between sites was calculated using a presence-absence Matrix (S2, Suppl. material 1) to perform multivariate analysis using ordination and group multivariate methods using the Vegan package in R (Oksanen et al. 2010). We constructed a matrix showing the presence of each species in each site and subjected it to ordination and grouping analyses using a Non-metric Multidimensional Scaling (NMDS) and Unweighted Pair Group Method with Arithmetic mean (UPGMA), respectively. Both analyses used Sorensen (Bray-Curtis) index (Legendre and Legendre 2012) to reflect beta diversity between sites.

To investigate the floristic richness of sites in relation to the size of each outcrop we used the species count for each canga outcrop and, employing GIS, we calculated the area of each outcrop in square kilometres. A linear model of the recorded richness versus area of each outcrop using the ‘glm’ function with Gaussian model was prepared in R. Because the outcrops were subjected to a large collecting effort during the ‘Flora of Carajás’ Project, we assumed that they were adequately sampled. We also evaluated whether the total number of species and of endemic species shared between sites were significantly related with the geographical distance between them. We computed the centroid of each outcrop using GIS and calculated the geographical distance between the centroids of all outcrop pairs. We tested the normality of the residuals of the models with the Shapiro-Wilk test to see whether the residuals significantly departed from normality. If these did not significantly differ from normality, we accepted the p value of the model. If the residuals differed from normality, we analysed the data using non parametric Spearman’s correlation to evaluate if the correlation was significant.

This study recorded a total of 254 species, of which 248 are seed plants, five ferns and one lycophyte in the SFX (Table 1). The richest families recorded are Fabaceae (22 species), Poaceae (21 spp.), Cyperaceae (15 spp.), Orchidaceae (12 spp.) and Rubiaceae (12 spp.). The five richest genera are Mimosa (Fabaceae), with 5 species, Cyperus and Rhynchospora (Cyperaceae), with 4 species each, and Borreria (Rubiaceae) and Aechmea (Bromeliaceae), with 3 species each. Thirty-seven species are new records for the CRC of the Carajás complex. From these new records, seven belong to the family Orchidaceae, five are new records of Fabaceae, three Annonaceae, and three Sapindaceae. A yet undescribed species of Lauraceae was found in SFX, belonging to the genus Dicypellium (Dicypellium aff. caryophyllaceum (Mart.) Nees – PLV 6100, Table 1; Fig. 2).

Figure 2. 

Representative species of canga in new dataset, SFX a Axonopus longispicus (Döll) Kuhlm b Dicypellium aff. caryophyllaceum (Mart.) Nees c Inga heterophylla Willd d Ipomoea decora Meisn e Matelea microphylla Morillo f Mimosa dasilvae A.S.L. Silva & Secco g Nepsera aquatica (Aubl.) Naudin h Ouratea cearensis (Tiegh.) Sastre & Offroy i Pachyptera incarnata (Aubl.) Francisco & L.G. Lohmann j Passifora picturata Ker Gawl. k Phyllanthus minutulus Mull.Arg. l Rodriguezia lanceolata Ruiz & Pav.

Among the 38 edaphic endemic species of canga, defined according to Giulietti et al. (2019), 17 (c. 50%) were recorded in SFX. Two of these, Erythroxylum nelson-rosae Plowman (Erythroxylaceae) and Matelea microphylla Morillo (Apocynaceae) were not previously recorded for SFX in the list of endemic edaphic species of the canga of Carajás (Giulietti et al. 2019). One species, Mimosa dasilvae (Fabaceae), is only known to occur in SFX.

Around 25% (60) of the 248 angiosperms registered for SFX are restricted to the Amazonian Rainforest biome, but the majority of the flora is widely distributed in open habitats throughout South America.

Regarding the phytophysiognomies listed by Mota et al. (2015) for the region, the canga vegetation of the SFX has a predominance of vast spreads of scrub composed of closely disposed treelets and shrubs. Amongst them, treelets and shrubs such as Byrsonima chrysophylla Kunth, Cordiera myrciifolia (K.Schum.) C.H.Perss. & Delprete, Anemopaegma carajasense A.H. Gentry ex Firetti-Leggieri & L.G. Lohmann*, Cuphea annulata Koehne, Lippia grata Schauer, Erythroxylum nelson-rosae Plowman*, Syagrus cocoides Mart., as well as several species of Myrcia and Eugenia, the palm Syagrus cocoides Mart. and scramblers and climbers such as Norantea guianensis Aubl., Cissus erosa Rich., Mandevilla scabra (Hoffmanns. ex Roem. & Schult.) K. Schum. and Smilax irrorata Mart. ex Griseb. On more exposed, bare canga slabs, the plants grow mostly in rock crevices with presence of monocots such as Vellozia glauca Pohl, Sobralia liliastrum Salzm. ex Lindl., Dyckia duckei L.B. Sm. and the tuberous, low growing Mandevilla tenuifolia (J.C. Mikan) Woodson, as well as the herbaceous Borreria semiamplexicaulis E.L.Cabral, Perama carajensis J.H.Kirk.*, Begonia humilis Dryand and Brasilianthus carajensis Almeda & Michelangeli*. The nodular canga has more or less continuous covering of grass and sedge, with occasional specimens of Riencourtia pedunculosa (Rich.) Prusky. During the expeditions we did not come across low forest groves, and our impression was that between the canga edge and the surrounding rainforest there was not much transition but a sharp substitution of the open vegetation by the associated forest types. Regarding the hydromorphic vegetation found in SFX, temporary shallow ponds with Utricularia species, Burmannia flava Mart., Cabomba furcata Schult. & Schult. f., Syngonanthus caulescens (Poir.) Ruhland and Xyris brachysepala Kral.* were visited. However, perennial, larger ponds of the magnitude found in the Serra Sul were lacking and temporary streams were not observed. There were also Palm swamps (buritizais), with margins occupied by Mauritia flexuosa Mart. and Mauritiella armata (Mart.) Burret, harbouring aquatic Oryza glumaepatula Steud., Helanthium tenellum (Mart. ex Schult. & Schult.f.) Britton and Eleocharis spp. (edaphic endemic species marked with *).

The biogeographical database from the CRC of the Carajás complex was updated by our study (see supplementary data) and includes now a total of 893 angiosperms distributed in 121 families and 441 genera. For the Carajás flora (FCC), Poaceae was the most species-rich family (75 species in the FCC), followed by Fabaceae (66 spp.), Cyperaceae (57 spp.), Rubiaceae (49 spp.), and Melastomataceae (40 spp.). The richest genera were Rhynchospora (24 spp.), Miconia (18 spp.), Paspalum and Solanum (17 spp. each), Myrcia and Ipomoea (13 spp. each), while 64% (284 genera) were represented by only a single species. The inclusion of SFX in our database increased the number of known taxa by 18 genera and 37 species not previously recorded for the canga of Carajás.

The mean angiosperm species richness for each outcrop of the Serra dos Carajás was 218 species. The NMDS and UPGMA analyses included 3451 records of 893 species across 16 sites (Fig. 3a, b). The UPGMA analyses produced statistically significant clusters (Fig. 3b) with the same major groups found by Fonseca-da-Silva et al. (2020), one comprising four of the eight areas of the Serra Norte (SN2, SN6, SN7, and SN8), while the remaining four (SN1, SN3, SN4, and N5) appear closer to the areas of Serra Sul (S11A, S11B. S11C, S11D), along with SB and ST. SA also emerged as the least similar to the Carajás complex, and SFX was found to be more similar to the group comprising SB, ST, Serra Sul and the four most species rich sites in Serra Norte (SN1, SN3, SN4, and SN5). A similar result was obtained by the NMDS analysis (Fig. 3a), also showing SA as the most dissimilar from other areas.

Figure 3. 

UPGMA (a) and NMDS (b) multivariate analysis clustering areas from FCC and SFX (see Table 2 for area codes). UPGMA cophenetic coefficient: 0.902. b. NMDS stress: 0.1859.

Species richness was significantly correlated with site area (r = 0.806094, P = 0.001548). The larger the area of each individual mountaintop (site), the larger the number of species recorded. The total number of shared species between mountaintop outcrops did not differ significantly with geographical distance across sites (r = -0.16; P = 0.08). There was a tendency of distant sites to share less species, but this trend was not significant. When the residuals of this model were evaluated they significantly departed from normality. Spearman’s correlation was not significant either (p-value = 0.2972). However, when focusing on the number of shared endemic edaphic species versus the geographical distance between sites, we found a significant correlation, where closer sites shared more edaphic endemic species than with more distant sites (r = -0.45872; P = 1.37e-07) (Fig. 4).

Figure 4. 

a Species richness plotted against area of Carajás. Pearson correlation coefficients: r = 0.806094, P = 0.001548 b the number of species shared between site pairs does not change significantly with geographical distance between regions. r = -0.16; P = 0.08 c the number of shared endemic species between site pairs declines with geographical distance between regions. r= -0.45872; P = 1.37e-07.

Regarding the total of species of the canga, the Carajás iron islands share an average of 40% of their flora with each other. SFX has, on average, 30% of shared species with each other area. The percentage of similarity between sites was a minimum of 30% and a maximum of 55%.

The most species-rich families and genera found in the SFX coincide with those found in the Flora das cangas de Carajás (Mota et al. 2018) and SA (Fonseca-da-Silva et al. 2020), where Cyperaceae, Fabaceae, Poaceae, and Rubiaceae are among the richest plant families. Interestingly, SFX has a much higher number of Orchidaceae species than other surveys of canga in the Amazon (Koch et al. 2018; Mota et al. 2018; Fonseca-da-Silva et al. 2020). The participation of botanical specialists during collecting expeditions enhances floristic studies in the Amazon (Medeiros et al. 2014) and elsewhere, and the high number of Orchidaceae in SFX possibly reflects the specific search for this group by J.B. Silva in the region from the 1990’s onwards, which may have resulted in a greater sampling effort for this group when compared to other areas.

There is a large turnover of species between outcrops (Zappi et al. 2019; Fonseca-da-Silva et al. 2020) and very few species are widely distributed across these disjunct, isolated habitats. Similar to what was found by (Costa et al. 2019) in Amazonian White Sand Campinas, the isolation of the patchy canga outcrops limits dispersal and increases floristic differentiation, and the adverse conditions, such as high temperature, extreme exposure to sunlight and winds, and a relatively well defined dry season represent ecological filters for the species that occupy the canga, partly explaining the high number of endemic species in the CRC of Carajás.

As an example, only three species were recorded in all surveyed areas: the widely distributed Riencourtia pedunculosa, an Asteraceae common in open areas in the Amazon (Flora do Brasil under construction, Bringel 2014), and two species associated with Amazonian canga outcrops: Brasilianthus carajensis and Perama carajensis. Perama carajensis is a confirmed canga edaphic endemic species, and Brasilianthus carajensis has been collected also on granite, being locally endemic to Carajás, but not a canga edaphic endemic (Giulietti et al. 2019; Silva et al. 2020). Other four species also present wide occurrence across campos rupestres on canga of Carajás: Bulbostylis conifera (Kunth) C.B. Clarke, Rhynchospora barbata (Vahl) Kunth, Rhynchospora seccoi C.S.Nunes et al., and Syngonanthus discretifolius (Moldenke) M.T.C. Watanabe were recorded for SFX and many other FCC areas, except for one of them missing in SN3, SN7, SN7 and SA, respectively. Their absence in these four sites may be related to the more modest canga surface found in these areas.

Some widely distributed species from the canga of Carajás, found at more than 10 of the 16 sites surveyed, were not recorded at SFX. The absence of the common treelets Callisthene microphylla Warm. and Mimosa acutistipula var. ferrea Barneby (Mota et al. 2015) at SFX may be partially explained by differences in the micro-habitats between SFX and the other canga outcrops considered here. For Brasilianthus carajensis, distinct adaptive genetic clusters have been found in the SFX (see Silva et al. 2020), increasing the argument for the protection of the site.

The canga is typically a mosaic of different vegetation types (Mota et al. 2015, Viana et al. 2016). Some of these vegetation types are infrequent in SFX, as for example low forest groves (Mota et al. 2015), and in consequence some of the species found in these groves elsewhere are absent at SFX: Callisthene microphylla, Mimosa acutistipula var. ferrea, and Cereus hexagonus (L.) Mill. Although forest groves are closely associated with the lower scrub vegetation, the latter is more abundant in the canga plateau of SFX than the former. In plateau SFX2 of SFX the shrubby vegetation is dominant, and there are large stands of Syagrus cocoides Mart., a palm emerging from the impenetrable shrubbery. In the context of CRC of Carajás, this palm forms large populations only in SA and SFX.

Despite having the lowest number of species registered in the FCC, the hydromorphic vegetation found atop the plateaus is the habitat with the highest proportion of exclusive species (Pereira et al. 2016; Mota et al. 2018). Seasonal lakes and palm lakes in the SFX ensure the presence of annual aquatic species such as Eriocaulon carajense Moldenke, Oryza glumaepatula Steud., Syngonanthus caulescens (Poir.) Ruhland, and Xyris brachysepala Kral.

As a relatively large canga site isolated from the active iron mines further to the east, the SFX has been found to harbour a rich and unique vegetation, representing a suitable area for the implementation of conservation strategies. On the other hand, this canga outcrop is currently threatened by surrounding deforestation, land transformation and frequent fires, and is not included within any type of protected area.

The mosaic of landscapes typical of CRC of Carajás may also explain the low floristic similarity between the sites. The number of shared species represents less than half the local richness from each site separately. This brings attention to the high beta diversity among sites (Zappi et al. 2019), with a large species turnover across these disjunct outcrops. Habitat diversity associated with the size of the island-like habitats is also related to the beta diversity in French Guiana´s inselbergs (Henneron et al. 2019), similarly to what is found in Andean alpine flora (Sklenář et al. 2014) and South American tepuis (Riina et al. 2019). This confirms the association between area and habitat diversity found here for the canga vegetation as an important factor for determining plant biodiversity.

The greater similarity between SFX, SB and ST, along with Serra Sul (S11A, S11B, S11C, and S11D) and SN1, SN3, SN4 and SN5 reflected in the UPGMA clustering patterns (Fig. 3b) suggests there is more similarity of species richness between the largest sites rather than among geographically closest areas, as observed by Fonseca-da-Silva et al. (2020) for SA. In fact, the correlation between the shared species of each canga site and their geographical distance was significant. Considering the size of each of these areas and their positive correlation with floristic richness (Fig. 4), we interpret the canga’s overall surface as being more important for floristic composition than the distance between sites in the Serra dos Carajás. Thus, the larger a canga outcrop is, the greater the number of micro-habitats it can harbour, reflecting an increased species richness and unique floristic composition of each canga site. On the other hand, that relationship (distance between areas vs shared flora) holds true when analysing shared endemic species, where shared endemic species decrease with distance at different rates (Fig. 4C).

The low number of species restricted to the Amazon (25%) and the high number of species widely distributed in South America (75%) recorded at SFX, may explain the discrepancy in the correlation between shared species and distance being negative when all species are considered, whereas it is positive for endemic species only. On a macro-scale, the majority of the species recorded in SFX have a broad distribution, occurring beyond the Amazon Rainforest, and the distance factor between different outcrops may not matter so much. On the other hand, when observing only the species endemic to Carajás, and especially edaphic endemic species, the trend is the opposite, possibly due to the local scale of observation, as elsewhere the distance between areas tends to affect the floristic similarity between island vegetations (Sklenář et al. 2014; Schrader et al. 2020). A genomic study revealed that gene flow in two endemic species of Carajás is mainly influenced by geographic distance between mountain pairs, as the rainforest surrounding different mountaintops constitutes an important barrier (Carvalho et al. 2019). Therefore, gene flow also decreases with the increase of the barrier represented by the rainforest (Carvalho et al. 2019).

Another factor that may have an impact on the contrasting effects of floristic similarity vs. distance from canga islands is the different environmental requirements of herbs, shrubs and trees, that shape their biogeographical patterns and affect species-area and richness-environment relationships (Schrader et al. 2020). Herbs, shrubs and trees have contrasting strategies in different environmental conditions with potential implications for community assemblage on islands. For example, herbs can form larger populations on small islands due to their smaller size, and as a result face less risk of extinction and greater dispersal capacity (Moles 2005; Thomson et al. 2010), while shrubs are associated with more stable environmental conditions, and therefore have more success on larger islands (Chiarucci et al. 2017).

Recent analyses of open vegetation in the Amazon reinforce the insular character of Amazonian canga and their low similarity to other vegetation types in the Amazonian biome (Devecchi et al. 2020). While there is some evidence that canga in Southeastern Brazil may be influenced by the surrounding Atlantic Rainforest and Cerrado (Zappi et al. 2017) these biomes are known to have a more varied life-form balance (respectively 1: 4 and 1: 7 proportion of trees over other life forms) than the Amazon Rainforest, where the life form balance is less extreme (1: 2) (Brazil Flora Group [BFG] 2015), thus it may have less floristic influence over the open vegetation found in the CRC of Carajás (Zappi et al. 2019). Therefore, in order to colonize the Amazonian CRC, shrubby or herbaceous plant species may have to come from further afield through long distance dispersal, and, if established, they may remain genetically isolated from their original populations, leading over a period of time to the patterns of endemism observed today.

Different evolutionary processes of the species occurring in CRC may also have led to different floristic composition in the outcrops. Although evolutionary studies involving species of canga in the Brazilian Amazon are just beginning (Zappi et al. 2017), the phylogeography of a species of Gesneriaceae distributed in humid rock formations in the Cerrado reveals its recent expansion into CRC vegetation during the Pleistocene (Fiorini et al. 2020). Recent and rapid radiations have been observed in mountaintops ecosystems (Salerno et al. 2012; Pirie et al. 2016; Vasconcelos et al. 2020) but more phylogenetic and phylogeographic studies are necessary to establish dating for plants species groups found in the CRC in order to understand their diversification and colonization processes.