SEARCH

SEARCH BY CITATION

Keywords:

  • biodiversity hotspots;
  • complementarity;
  • conservation assessments;
  • endemism;
  • Mata Atlântica;
  • Myrtaceae;
  • predictive modeling
  • complementariedad;
  • endemismo;
  • evaluaciones de conservación;
  • modelos predictivos;
  • Myrtaceae;
  • Mata Atlântica;
  • sitios de importancia para la biodiversidad

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations
  8. Conclusions
  9. Acknowledgments
  10. Literature Cited

Abstract: Plant-diversity hotspots on a global scale are well established, but smaller local hotspots within these must be identified for effective conservation of plants at the global and local scales. We used the distributions of endemic and endemic-threatened species of Myrtaceae to indicate areas of plant diversity and conservation importance within the Atlantic coastal forests (Mata Atlântica) of Brazil. We applied 3 simple, inexpensive geographic information system (GIS) techniques to a herbarium specimen database: predictive species-distribution modeling (Maxent); complementarity analysis (DIVA-GIS); and mapping of herbarium specimen collection locations. We also considered collecting intensity, which is an inherent limitation of use of natural history records for biodiversity studies. Two separate areas of endemism were evident: the Serra do Mar mountain range from Paraná to Rio de Janeiro and the coastal forests of northern Espírito Santo and southern Bahia. We identified 12 areas of approximately 35 km2 each as priority areas for conservation. These areas had the highest species richness and were highly threatened by urban and agricultural expansion. Observed species occurrences, species occurrences predicted from the model, and results of our complementarity analysis were congruent in identifying those areas with the most endemic species. These areas were then prioritized for conservation importance by comparing ecological data for each.

Resumen: Los sitios de importancia para la diversidad de plantas están identificados a escala global, pero se deben identificar sitios más pequeños, locales, para la conservación efectiva de plantas a escalas global y local. Utilizamos las distribuciones de especies endémicas y endémicas amenazadas de Myrtaceae para indicar áreas de importancia para la diversidad y conservación de plantas en los bosques de la costa del Atlántico (Mata Atlântica) de Brasil. Aplicamos tres técnicas, simples y baratas, de sistemas de información geográfica (SIG) a una base de datos de especímenes de herbario: modelado predictivo de la distribución de especies (Maxent); análisis de complementariedad (DIVA-GIS) y mapeo de localidades de colecta de especímenes de herbario. También consideramos la intensidad de colecta, que es una limitación inherente al uso de registros de historia natural para estudios de biodiversidad. Dos áreas de endemismo separadas fueron evidentes: la Serra do Mar de Paraná a Río de Janeiro y los bosques costeros del norte de Espírito Santo y el sur de Bahía. Identificamos 12 áreas, aproximadamente de 35 km2 cada una, como sitios prioritarios para la conservación. Estas áreas tenían la mayor riqueza de especies y estaban muy amenazadas por la expansión urbana y agrícola. Las ocurrencias observadas de especies, las ocurrencias de especies pronosticadas por el modelo y los resultados de nuestro análisis de complementariedad fueron congruentes en la identificación de las áreas con el mayor número de especies endémicas. La importancia para la conservación de estas áreas fue priorizada posteriormente mediante la comparación de datos ecológicos.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations
  8. Conclusions
  9. Acknowledgments
  10. Literature Cited

Local hotspots within 34 global biodiversity hotspots (Mittermeier et al. 2005) must meet at least 3 criteria. They must be small enough for intensive management, representative of the hotspot in which it is located, and in a highly threatened yet underprotected area. Conservation of an entire biome is impossible, and strategies must focus on small areas that represent a maximum of the biome's diversity and endemism. Identification of such local hotspots can assist conservation planning, particularly in areas where species diversity is rich, but taxonomic knowledge and resources are scarce.

Harris et al. (2005) used bird endemism hotspots to identify 8 areas of approximately 30 km2 of conservation importance in the Atlantic forests of Rio de Janeiro, Brazil. They combined endemic and threatened bird species range maps with a map of remaining forest cover, but did not account for sample density, which imparted uncertainty to their results. We used a similar approach to that of Harris et al. (2005) with the plant family Myrtaceae. We hypothesized that this taxon is an indicator of overall patterns of angiosperm diversity across the Atlantic forest biome.

Despite a rapid decline in the rainforests originally covering the Brazilian Atlantic coast (SOS Mata Atlântica 1998), the remaining 5% makes up the second-largest area of tropical moist-forest in South America (Oliveira-Filho & Fontes 2000). The Atlantic rainforest—Mata Atlântica—is a global species hotspot, containing approximately 20,000 plant species (almost 50% endemic) and similarly high levels of diversity for other biological groups. There are 3 centers of endemism in tropical lowland rainforest in this biome: Pernambuco and Alagoas (PE–AL); Bahia and northern Espírito Santo (BA–ES); and the Serra do Mar mountain range between São Paulo and Rio de Janeiro (SP–RJ) (Thomas et al. 1998).

A principally woody group, Myrtaceae is the fourth largest plant family in Brazil (Giulietti et al. 2005). In some areas of Mata Atlântica, it is the most speciose family of trees (Mori et al. 1983; Oliveira-Filho & Fontes 2000). Neotropical Myrtaceae comprises 7 as-yet-unranked monophyletic clades, 1 of which, Myrcia s.l. (hereafter Myrcia), includes the following traditionally accepted genera: Calyptranthes, Gomidesia, Marlierea, and Myrcia (Lucas et al. 2007). This clade is restricted to the Neotropics and has a center of diversity in the Mata Atlântica. Of approximately 2500 Neotropical Myrtaceae species, about 790 belong to Myrcia (Govaerts et al. 2006). The abundance and diversity of Myrcia species in the Mata Atlântica suggests it may be an indicator of wider plant diversity patterns within the biome. The availability of a comprehensive specimen database allowed us to test this, and we used the results of this test to identify areas of particular conservation priority. Critics of the use of indicator taxa suggest their overall effectiveness depends on their size in relation to the area under survey, the specificity of the indicator relationships to the area (Hess et al. 2006), and their determination on the basis of survey rather than range data (Hurlbert & White 2005).

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations
  8. Conclusions
  9. Acknowledgments
  10. Literature Cited

Myrtaceae as Indicators

We extracted data from a database of Atlantic forest trees (TreeAtlan 1.0; Oliveira-Filho 2006) that consisted of tree species checklists from 439 sites across the biome. We ranked families by species richness and correlated species numbers of the 12 most speciose families across the sites against the total number of species for those sites. We split data into the following forest subcategories: rainforest, semideciduous forest, deciduous forest, Araucaria forest, and restinga. In our experience richer areas of Mata Atlântica contain more species of any family, implying that species diversity for a family is not independent of total species diversity at a given station. Therefore, for each correlation, we subtracted the number of species for that family from the total number of tree species. The correlation was thus between numbers of species for a particular family at each station and the total number of species minus species of that family. We used a Spearman correlation.

Myrtaceae are a difficult family to identify, and it is likely that TreeAtlan 1.0 underrepresents Myrtaceae diversity. Thus, we carried out the same analysis on data collected from 26 Mata Atlântica checklist stations in the Alto Rio Grande river basin (Minas Gerais: 21°00′–22°20′S and 43°50′–45°30′W). The sampling and collecting efforts at these stations were similar, and the Myrtaceae were accurately identified by a taxonomic specialist (M. Sobral). The total number of tree species at each station was correlated with the number of species in the 4 most speciose tree families in the region: Lauraceae, Leguminosae, Melastomataceae, and Myrtaceae (Oliveira-Filho & Fontes 2000). Again, for each correlation, we subtracted the number of species of the family concerned from the total species number.

Empirical Data

We extracted the primary set of species distributions from a database of >12,220 Myrcia specimens in the herbaria of the Royal Botanic Gardens, Kew; New York Botanical Garden; National Herbarium of French Guiana; Jardim Botânico do Rio de Janeiro, and Nationaal Herbarium Nederland, Utrecht. We obtained additional records for species endemic to Mata Atlântica from on-line herbarium resources (Field Museum, Chicago, Illinois; Smithsonian Institution, Washington, D.C.; TROPICOS, Missouri Botanical Gardens, St. Louis, Missouri) and from original literature sources. Each record in the database represents a herbarium sheet and the ecological information from its collection label.

Once compiled, the data set was restricted to 7200 Brazilian collections. If a record lacked locality coordinates, we obtained them from geographical gazetteers (http://www.fallingrain.com; http://middleware.alexandria.ucsb.edu). Georeferencing was to the nearest latitude–longitude minute and yielded a possible error margin of approximately 2 km2, which reduced the chance of a species record being assigned to the wrong grid square to an acceptable level. We standardized species names according to Govaerts et al. (2006).

Selection of Endemic Taxa

Because we used Myrcia diversity as a surrogate for angiosperm diversity and Myrtaceae are present in primary and secondary rainforest, seasonal deciduous, and semideciduous dry forest, we considered Mata Atlântica sensu lato (Oliveira-Filho & Fontes 2000), as defined by the IBGE biome map (2004). A list of species endemic to Mata Atlântica was generated by querying the same IBGE map and verified by a Myrtaceae specialist. Here, these are the endemic species unless otherwise specified.

Preliminary IUCN Assessments

Threatened species were those identified by the International Union for the Conservation of Nature (IUCN 2001) as critically endangered (CR), endangered (EN), or vulnerable (VU). To identify endemic Myrcia species under threat, we assigned a preliminary IUCN category to each species. We based these categories on IUCN-provided range parameters but did not perform full assessments because some species did not meet all IUCN subcriteria (Willis et al. 2003). We included near-threatened (NT) species because many of these were cited as rare or threatened in the literature.

Maximum Entropy Modeling of Endemism

To investigate the bias of different collecting intensities across the Mata Atlântica, we used a species distribution model to produce predicted continuous distributions of endemic species. The maximum entropy model (Maxent version 2.3.; Phillips et al. 2006; M. Dudík, S. J. Phillips, and R. E. Schapire, unpublished data) was used, a presence-only algorithm that requires known species occurrence points and environmental variables. This method performed best across a range of species and regions in comparison with other well-established methods (Elith et al. 2006). We used default settings for the regularization value, maximum number of iterations, and convergence threshold. The minimum number of unique records required for modeling techniques such as Maxent remains uncertain. Small numbers of occurrence records may not adequately define the environmental conditions of a species' niche. Alternatively, large numbers of spatially autocorrelated occurrence records will add nothing new to the model. Pearson et al. (2007) found high success rates and statistical significance with as few as 5 occurrence points under the Maxent model, and we maintained that threshold.

We used 19 bioclimatic variables from the WorldClim database (Hijmans et al. 2005a) and a 1-km resolution digital elevation model derived from the GTOPO30 database (http://edcdaac.usgs.gov/gtopo30/gtopo30.html). World Wildlife Fund ecoregion data (Olson et al. 2001) were also used. Each species was modeled separately to determine good predictor variables for the final run of the model. Because collection density for Myrcia was not uniform across the biome, all the data were used for the model-fitting stage. Maxent displays cumulative probabilities across the study region as a percentage in which cells with values toward 100 indicate high species suitability and cell values toward zero indicate low suitability. Output suitability maps for each modeled endemic species were stacked and numbers of species summed to identify areas with the highest predicted endemism. To produce a second suitability map, we separately stacked and summed threatened endemic species of this subset. Cells of potentially high species diversity were selected from within those patches of endemism in the combined highest classes (Jenks algorithm; Jenks & Caspall 1971) from each of the predicted number of endemic species and endemic threatened species. Jenks optimization (or goodness of variance fit [GVF]) minimizes the difference of values within a class while maximizing the difference between these classes.

Performance Assessment and Complementarity

We used holdout validation to assess the distribution models. Two separate runs were carried out for each model in which a random sample of 20 and 25% of occurrences, respectively, were withheld for testing. To reduce problems of nonindependence caused by spatially autocorrelated points in training and testing, we removed testing points within a 10-km buffer of the training points and resampled remaining points. We used testing data and randomly generated pseudoabsences to generate receiver operating characteristic (ROC) curves within the Maxent program. Area under curve (AUC) values for each model were calculated and averaged.

To identify complementary sets of grid cells that captured the maximum number of species in as few cells as possible, we ran a complementarity analysis with all known specimen records for the Mata Atlântica. The program DIVA-GIS (Hijmans et al. 2005b) uses an iterative algorithm (Rebelo & Sigfried 1992) to prioritize areas, in this case cells, by conservation value, selecting the cell containing the most species first and then the cell containing the highest number of additional species. When 2 or more cells contain the same number of additional species, 1 is selected at random. This simple algorithm is sufficient because low commonality of species between cells means the data would yield similar results under the parameters of any standard complementarity technique. We ran the analysis with the reserve-selection output variable and rarity setting. For easy comparison with the 12 areas of high diversity identified by modeling and richness counts, the analysis was limited to 12 iterations and provided 12 cells containing a maximum amount of species diversity.

Grid-Cell Counts of Richness

To speed data-processing time, we used a geographic projection and a mean grid-cell size of 35 km2 without reprojecting the data. Cell area varied by only 4% throughout the study area.

Specimen density of all Mata Atlântica species was calculated to illustrate collection intensity per grid cell for comparison with species-richness data. Richness of all species, all endemic species, and all endemic threatened species was also calculated. To make species richness and endemism values directly comparable, we used only primary specimen occurrence data and 2 criteria to identify endemic species hotspots from grid-cell counts. For the first criterion, we regarded a cell on the endemic or endemic threatened species density maps or both as a potential hotspot if it contained records from the highest 2 natural-break categories of density (Jenks algorithm).

For the second criterion, we used counts of unique species per grid cell and designated cells with 2 or more species endemic to that cell as hotspots, following the definition of an area of endemism as an area containing at least 2 endemic taxa (Cracraft 1995). We limited the potential scale dependence of this criterion by using identically sized cells. Although we set a very strict criterion for endemism, which assumes a consistent sampling approach, other measures of endemism (e.g., weighted endemism, Williams & Humphries 1994) would not have been directly comparable with observed species density, and the issue of small species ranges has been tackled with predictive-distribution modeling (MaxEnt).

Forest Cover, Protected Areas, and Areas of Endemism

We combined grid cell counts and modeled suitability maps with current forest-cover imagery from Satellite Pour L'Observation de la Terre (SPOT VGT) (Harris et al. 2005). We used a raster calculator to convert the SPOT VGT image to percent forest cover/ 35-km2 grid cell. These values were assigned by spatial analysis, and the grid was the same as the one used in the cell counts. The IUCN protected areas within or intersecting a hotspot cell were identified with the World Database on Protected Areas shapefile (WDPA 2004).

Relative species composition in areas of endemism identified by Thomas et al. (1998) was examined by querying the Myrcia database and comparing species lists for each area. Because no Myrcia species are endemic to the PE–AL area of endemism of Thomas et al., only the remaining 2 areas were used.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations
  8. Conclusions
  9. Acknowledgments
  10. Literature Cited

Myrtaceae as Indicators

On the basis of TreeAtlan 1.0 data, Myrtaceae were the second richest family of tree species in the Mata Atlântica, just behind Leguminosae, and more than twice as diverse as the next family, Rubiaceae. Analysis of the whole database placed Myrtaceae as the sixth-best indicator of total tree species diversity in the Mata Atlântica sensu lato. But after separating stations into forest subtypes, Myrtaceae was consistently better correlated across each subtype than any family after Rubiaceae. Myrtaceae was the second-best correlated family for Araucaria forest (after Melastomataceae) and the restinga (after Rubiaceae). On the basis of the 26 stations in the Alto Rio Grande river basin, Myrtaceae was the best correlated of any family after Lauraceae (Fig. 1) (correlation coefficients: Lauraceae, 0.84 [p < 0.001]; Myrtaceae, 0.77 [p < 0.001]; Melastomataceae, 0.66 [p < 0.001]; Leguminosae, 0.38 [p < 0.01]). Although Lauraceae appeared a better family for indicating patterns of diversity in this region, it had more taxonomic problems at species and higher taxonomic levels than Myrtaceae. In both analyses Myrtaceae was a reliable indicator of patterns of species richness across the biome. The taxon was better correlated than the most speciose family, Leguminosae; more speciose than the best indicator family, Rubiaceae; and much richer and better known than Lauraceae.

image

Figure 1. Total number of tree species recorded at 26 Mata Atlântica checklist stations versus the number of species represented by the 4 most speciose tree families in the biome.

Download figure to PowerPoint

Selection of Endemic Taxa, IUCN Assessments

The Myrcia databases contained 2858 Mata Atlântica specimens. Spatial analysis indicated 235 species in the biome of which 121 are endemic. Fifty-one endemic species were threatened according to the provisional conservation assessment algorithm (1 species EX; 9 CR; 20 EN; 10 VU; 11 NT). Twenty-three species were considered data-deficient and thus levels of threat were not assigned.

Maximum Entropy Modeling of Endemism

Species known from fewer than 5 unique localities were excluded from entropy modeling (64 endemics including 31 threatened species). Nevertheless, grid-cell counts of all endemic species were well correlated with grid-cell counts of only those endemics used in the modeling (rs= 0.97; p < 0.001). Average AUC values from the validation procedure were 0.97 and 0.96 when 25 and 20% of the data were, respectively, withheld. This represents good model discrimination. Six cells were predicted to contain the highest presence of endemic or endemic threatened species in the Mata Atlântica in the Maxent model (Fig. 2 & Table 1).

image

Figure 2. Predicted distributions of Myrcia s.l. species as portrayed in summed suitability maps of (a) 57 modeled endemic species and (b) 31 modeled threatened endemic species (each pixel within the grid has a probability value on an arbitrary suitability scale from blue [very unsuitable conditions] to red [highly suitable conditions]). (c) Combined suitability maps of endemic (b) and endemic threatened (c) species (red, regions of the highest natural-break category [goodness of variance fit, Jenks algorithm]; cells A–F contain the highest number of predicted species for each area of endemism taken from the map of endemic threatened species [b]). (d) Twelve highest priority cells from the complementarity analysis (cells G–L correspond to those described inFig. 3).

Download figure to PowerPoint

Table 1.  Ecological data for 6 hotspot cells of the Mata Atlântica (on the basis of modeled data).
CellStateaCoordinatesMax. no. of endemic speciesNo. of endemic threatened taxaVegetation classProtected areasProtected-area IUCN statusbProtected area (%) (approx.)Forest cover (%)
  1. aAbbreviations: BA, Bahia; RJ, Rio de Janeiro; PR, Paraná; ES, Espírito Santo.

  2. bProtection categories (IUCN 1994): Ia, strict nature reserve, managed mainly for science; II, national park, managed mainly for ecosystem protection and recreation; III, managed mainly for conservation of specific natural features; V, protected landscape/seascape, managed mainly for landscape/seascape conservation and recreation; unset, status unknown.

ABA13°58′–14°19′S183Atlantic dense rainforestMorro do Cururupê Ecological ReserveIa0.00332
38°53′–39°13′W       
BRJ19°55′–20°16′S276Atlantic secondary rainforestMassambala Environmental Protection AreaV100.5
40°17′–37′W   Massambala Ecological ReserveIa  
Serra da Sapiatiba Environmental AreaV  
Guapimirim Environmental Protection AreaV  
CSP24°7′–24′S309Atlantic dense rainforest;Itariri State Forest Reserveunset2669
 Atlantic secondary rainforest    
46°59′–47°16′W   Itatins State Forest Reserveunset  
    Chauás State Ecological StationIa  
    Juréia Ecological StationIa  
DPR25°10′–31′S266Atlantic secondary rainforestPico do Marumbi State ParkII5772
49°1′–11′W   Graciosa State ParkII  
    Marumbi (area of interest)III  
ESC27°58′–28°19′S328Atlantic secondary rainforestSerra do Tabuleiro State ParkII7022
49°1′–48°40′W       
FSC/RS29°1′–22′S317Atlantic savannah with shrubsSerra Geral National ParkII18 7
49°43′–50°4′W   Aparados da Serra National ParkII  

Stacked and summed predicted distributions of all endemic and all threatened endemic species (Figs. 2a & 2b) indicated that areas of high endemism occur along the coast and are concentrated in Paraná, Rio de Janeiro state, and Bahia. The combined highest natural-break categories for the combined endemic and threatened endemic species maps indicated 6 main patches of contiguous coastal forest in the states of Rio Grande do Sul, Santa Catarina, Paraná, São Paulo, Rio de Janeiro, and Bahia (Fig. 2c). Cells A–F each represented 1 of the 6 patches of highest predicted endemism (Fig. 2b) and were selected as having the highest endemism of all cells within each of these patches.

Complementarity and Grid-Cell Counts of Richness

After 12 iterations, the complementarity analysis accounted for 140 species out of 235. The 12 resulting cells (Fig. 2d) matched those determined by other approaches in this study as having high species or endemic species richness.

The best sampled area, just outside Rio de Janeiro, contained 220 specimens of Myrcia species (Fig. 3a). Areas of high collection density correlated with those with high species richness (Fig. 3b), up to a maximum observed number of species of 49, again just outside Rio de Janeiro. We manually compared the attributes (Tables 2 & 3) of the 6 most diverse 35-km2 cells in terms of endemic species (Fig. 3c) with endemic threatened species richness (Fig. 3d). On the basis of the first criterion described in the methodology, cells G–I contained the most endemic species, whereas cells J–L were identified on the basis of criterion 2. Cell H met both criteria. Cells G–I contained 13, 26, and 15 endemic species, respectively (Table 2), of which 7, 9, and 2 species were threatened. Of the cells that met hotspot criterion 2, H and J each contained 3 endemics. Cell K contained the highest number of species endemic to a single cell of the Mata Atlântica. Cell L (which bordered cell H) contained 2 endemic species. Cells J–L contained numbers of endemic and endemic threatened species comparable to cells G–I, with 11, 13, and 13 endemic species, respectively, of which 5, 3, and 6 species, respectively, were threatened.

image

Figure 3. Grid cell densities for (a) 2858 Mata Atlântica specimen collections, (b) all 235 Myrcia s.l. species in the Mata Atlântica, (c) 121 Myrcia s.l. species endemic to the biome, and (d) 51 endemic and threatened Myrcia s.l. species (RN, Rio Grande do Norte; PB, Paraíba; PE, Pernambuco; AL, Alagoas; SE, Sergipe; BA, Bahia; ES, Espírito Santo; RJ, Rio de Janeiro; SP, São Paulo; PR, Paraná; SR, Santa Catarina, RS, Rio Grande do Sul). Cells G–L in (c) and (d) contain the most endemic and the most endemic threatened species in the biome.

Download figure to PowerPoint

Table 2.  Collection density, species richness, and number of Mata Atlântica (MA) endemics and threatened endemics as a proportion of total species for each of 6 priority hotspot cells of the MA.
CellNo. of collectionsNo. of speciesNo. of MA endemic taxaNo. of modeled taxaNo. of MA endemic threatened taxaNo. of taxa endemic to cellMA endemics as proportion of total species (%)MA threatened endemics as proportion of total species (%)
G50191310706836
H222492619935318
I822715132055 7
J532311 6534721
K291913 8346815
L37241310625425
Table 3.  Ecological data for 6 hotspot cells in the Mata Atlântica (on the basis of observed data).
CellStateaCoordinatesVegetation classProtected areasProtected-area IUCN statusbProtected area (%) (approx.)Forest cover (%)
  1. aAbbreviations: BA, Bahia; RJ, Rio de Janeiro; PR, Paraná; ES, Espírito Santo.

  2. bProtection categories (IUCN 1994): Ia, strict nature reserve, managed mainly for science; II, national park, managed mainly for ecosystem protection and recreation; V, protected landscape/seascape, managed mainly for landscape/seascape conservation and recreation; unset, status unassigned.

GBA15°1′–22′SAtlantic secondary rainforestUna Biological ReserveIa723
38°53′–39°13′W Mico-leao Biological Reserven/a  
HRJ22°2′–22°22′SAtlantic secondary rainforestTijuca National ParkII60.3
43°1′–23′W Chacrinha State ParkII  
  Municipal Ecological ParkII  
  Cidade de Niterói Ecological StationIa  
  Reconcavo Envionmental AreaV  
IPR25°10′–31′SAtlantic secondary rainforestCuritiba State Forestunset0.75
49°19′–22′W Metropolitana State Forestunset  
JBA14°40′–15°1′SAtlantic secondary rainforestIguape Ecological ReserveIa0.0623
38°53′–39°13′WAtlantic savannah    
KES19°55′–20°16′SAtlantic secondary rainforestSanta Lúcia Biological Stationunset341
40°17′–37′W Mestre Álvaro Biological ReserveIa  
LRJ22°22′–43′SAtlantic dense rainforest;Tinguá Biological ReserveIa1448
Atlantic secondary rainforest    
42°43′–43°4′W Serra dos Orgãos National ParkII  

Areas of Endemism

The SP–RJ area includes cells B, C, D, E, F, H, I, and L, and the BA–ES area included cells A, G, J, and K. Myrcia species composition differed between these two areas, supporting the view that they are significantly distinct, with 38 (31%) and 58 (48%) species occurring in the SP–RJ and BA–ES, respectively, and only 25 (21%) occurring in both. Therefore, not 1 of the 12 identified cells represented the endemism of the entire Mata Atlântica biome; rather, the cells represented endemism of either the SP–RJ or BA–ES area.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations
  8. Conclusions
  9. Acknowledgments
  10. Literature Cited

Modeled versus Observed Patterns of Diversity

Four main areas of diversity for Myrcia species were identified (Figs. 2a–c): coastal Bahia (cell A); the coast outside Rio de Janeiro city (B); coastal São Paulo (C); coastal Paraná (D) and Santa Catarina (E and F). These cells were well correlated with remaining forest cover, and the high species diversity was likely due to the habitat diversity in these topographically diverse areas of Mata Atlântica that are unsuitable for agriculture. Only cell D was indicated by complementarity analysis as potentially rich in endemic species (Fig. 2c). Narrow-range endemic species are more likely to be discovered in species-rich areas because these areas are well surveyed (Fig. 3a), which suggests a low overall dependence on collecting effort.

Observed data also indicated total (Fig. 3b) and endemic (Fig. 3c) species to cluster along the Atlantic coast. Areas of observed species richness of southern Bahia (cells G and J), central Espírito Santo (K), outside Rio de Janeiro (H and L), and in Paraná (I) are all close to large towns and, importantly, active local herbaria. Results of the predictive modeling supported the hypothesis that these areas are genuinely species rich by highlighting potential distribution areas that satisfy a species ecology, but from where no specimens have been collected. Complementarity analysis recovered all 6 cells of high species richness and endemism (Fig. 2d), which indicates that most endemic species occurred in cells also rich in species overall (Figs. 3b & 3c), so most Mata Atlântica Myrcia diversity occurred in a small number of cells.

Pimm (2005) used counts of endemic and endemic threatened species within individual grid cells to identify areas of high orchid endemism. With a uniform grid, it is difficult to select a suitable cell size. The cell size of 125 km2 that was used for orchids (Pimm 2005) is too large to suggest areas for effective conservation. Cells of 35 km2, which we used here, allowed for finer resolution of distribution patterns in a manageable number of units and identified areas large enough to have a real impact on regional conservation efforts (the average protected area in the southeastern part of this biome is 57 km2). To map species richness, Harris et al. (2005) stacked and summed species range maps and did not use an explicit grid-cell size or density of bird sightings. We had the advantage of being able to predict species distributions on the basis of more complete species ranges because we knew collection intensity within the region and had access to historical collection localities.

São Paulo–Rio de Janeiro Area of Endemism

Cell C in São Paulo state was predicted to have more endemics than any cell in Rio de Janeiro state (Table 1). The second-highest predicted diversity of endemic species was in cell B, which despite its location in the coastal forests of Rio de Janeiro, did not correspond precisely to either of those identified by cell counts.

The 3 cells with the greatest number of observed endemic and endemic threatened species (Table 2) all occurred within the SP–RJ area of endemism. Of these, cell H consistently scored highest for total number of species and for the proportion of these endemic or endemic and threatened and was therefore also the single highest priority cell from the complementarity analysis. The second-highest-scoring cell, I, differed significantly in that—despite containing 27 species—only 2 of these were endemic and threatened. Of these, 1 was endemic to the cell, falling entirely within already protected areas, whereas the other extended into reserves within Santa Catarina state. Cells predicted by modeling to have the most endemic species should have been independent of collecting bias. Cell L, however, was directly above cell H, but included only half as many species and endemic species. This difference may be due to collecting intensity because cell H contained more than 3 times as many collections as cell I and more than 5 times as many as cell F.

Northern Espírito Santo–Southern Bahia Area of Endemism

Species endemic to the ES–BA area typically had smaller range sizes than did SP–RJ endemics and were therefore less likely to be represented by sufficient records for modeling (26% ES–BA endemics modeled vs. 36% SP–RJ; 56% Mata Atlântica endemics modeled). Our model was therefore biased against this region, and for cell A to register here suggests a particularly high diversity of endemics are underrepresented in herbaria worldwide. The concentration of 3 hotspot cells within an area of 6000 km2, 2 of which contained the highest proportions of single-cell endemics, strengthens this hypothesis.

In the ES–BA area of endemism, the cell with most observed endemic and endemic threatened species was cell G. Cells G and K both had the highest proportion of threatened endemics of any cell in either area of endemism. A broad area of diversity in coastal Bahia was indicated by cell G and the adjacent cell J. These were identified under criterion 2 (see Methods) and by cell A, the only cell from the ES–BA area highlighted by the modeling analysis, which was 1 cell removed from J. The high number of endemic and endemic threatened species in cell K was remarkable because this cell was the least well collected of cells G–L.

Other Areas of Endemism

Cells E and F contained the highest predicted number of endemic species of all cells identified in the model. The high scores for these cells may be because of complete Myrtaceae accounts for the southern states of Rio Grande do Sul (Sobral 2003) and Santa Catarina (Legrand & Klein 1967–1977). In these states most endemics are known, whereas in states that have not been surveyed as well more species go unnoticed. Modeling amplifies this factor, underestimating the suitability of poorly sampled areas that then appear less diverse relative to well-sampled areas. Of the 5 cells identified as species rich only by complementarity analysis (Fig. 2d), each contributed only 10–18 additional species to the total complement set. An additional complementarity analysis conducted on the modeled surface data (not shown) indicated a similar pattern, identifying cells H, J, and D, respectively, as containing the most unique endemic species.

Limitations

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations
  8. Conclusions
  9. Acknowledgments
  10. Literature Cited

Although herbarium specimens provide an abundant source of collections, the number of locations over a large area is often small. The principal limitation of using natural-history or herbarium records for studies of diversity and endemism is nonuniform sampling, with areas of high endemism often coinciding with areas of high collecting intensity. Cell H, which had the most collections, contained the highest number of total species and number of endemic and endemic threatened species. In fact, areas of high diversity were all close to local centers of botanical research. This may be why the AL–PE area of endemism (Thomas et al. 1998), a poorly surveyed area, was not identified as one of the most species-rich areas here. Modeling potential species distributions mitigates against this, but is more appropriate when the taxonomy of the group under investigation is relatively well understood and sampling in a given area is uniform. Modeling is advantageous because it provides a continuous prediction of plant diversity, allowing conclusions to be drawn about areas from which little data are available. Our predictive modeling approach may have been better suited to the SP–RJ area of endemism than the ES–BA area, where there was a greater number of narrow-range species. Thus, results from the ES–BA area should be interpreted cautiously.

The Myrtaceae collections we used emphasize the bias of modeling techniques against poorly collected areas. These collections contained additional species from Espírito Santo, Bahia, and Pernambuco, but they were collected from too few locations to be used in the model. Many of these collections are of species new to science that were overlooked in grid cell counts and the model. It is these poorly known, unrecorded species that are most likely to have restricted ranges and be in need of conservation. Our reliance on presence–absence data may have somewhat favored selection of species-rich ecotones rather than core habitats as areas of high diversity (Araújo & Williams 2001) because the grid squares used were relatively large and likely included degraded areas as well as primary Mata Atlântica forest. The use of these data did, however, allow identification of regions of the highest species richness and under the greatest threat, areas that might otherwise be missed. The value of ecotonic areas and areas of core habitat to conservation has been described across habitat biomes (Araújo & Williams 2001), but at the scale of our study, we believe problems related to ecotones were minimal.

Priority Areas and Decision Making

Our results supported the recognition of 2 centers of endemism (Thomas et al. 1998), but showed that 1 (ES–BA) is undercollected relative to the other (SP–RJ). This complements the results of Harris et al. (2005) for birds and Pimm (2005) for orchids in identifying parts of the Serra do Mar, particularly coastal forests in São Paulo and Rio de Janeiro states, as hotspots of species diversity. These are, however, also areas where most collections have been made, something not addressed in either previous study. Knowledge of locations that have and have not been intensively surveyed is vital for assessing conservation priorities, and predictive species modeling can be a useful tool for planning further survey work. Areas predicted to contain the most endemic species are in São Paulo state, spilling south into Paraná. Cell B in Rio de Janeiro contained the second-highest predicted density of endemics and threatened endemics of any cell and is the area identified as the richest in Mata Atlântica endemic bird species and of the highest conservation priority by Harris et al. (2005). The highest priority cells (H, I, J, and L) from the complementarity analysis were well distributed throughout the Mata Atlântica, which suggests that few species are shared between them.

There was little difference between the average percent forest cover of the 6 hotspot cells in the SP–RJ area (32%) and the 4 in the ES–BA area (30%). In contrast, the percentage of protected areas estimated by IUCN varied considerably between the 2 areas, averaging 30 and 2% per cell, respectively, in the SP–RJ and ES–BA areas of endemism. Cells must therefore be considered individually but could be prioritized for conservation on the basis of the following criteria: number of species and endemic species (which should be high); type of vegetation (primary rather than secondary forest); IUCN status of the areas already protected (poorly protected reserves over those already well protected); remaining forest cover per cell (high); and how much forest cover is already within protected areas (low).

Under these criteria, cells A, C, and K most urgently require further surveying efforts and new conservation actions. All 3 have large amounts of remaining forest, but in cells K and A, little of this is protected, whereas only 26% of forest in C is well protected. In cell A the little forest protected is well protected, but in cells K and C, the status of many protected areas is undetermined. Of these, only cell C occurred in the RJ–SP area of endemism, and it did not contain a particularly high number of endemics. But it warrants further investigation because it includes primary forest, is 200 km from the botanical centers in São Paulo, and is not an obviously overcollected cell. Cell K contained several species known only from that cell, and a similar situation was predicted for cell A (Table 1), which suggests a high proportion of narrow-range endemics in southern Bahia and northern Espírito Santo. This is of particular concern because the ES–BA area of endemism has less remaining forest than the SP–RJ region. Of the 12 cells highlighted by this analysis, G and J were also in the ES–BA area of endemism; J (2 cells to the south of A) contained 3 species found outside the only ecological reserve within it (Table 3), whereas G (just south of J) contained the highest proportion of threatened Mata Atlântica endemics (Table 2).

Our results suggest that cells G, J, K, and A and surrounding areas of the ES–BA area of endemism contain a high number of threatened endemic species in only small amounts of remaining forest that have little official protection. The Una Biological Reserve within cell G and the Biological Station of Santa Lúcia in cell K (not officially protected) are subject to high collection and study, but surrounding areas are less well known. Protection of all areas highlighted by this analysis is important; priority conservation of as many flagged forest fragments as possible in southern Bahia and northern Espírito Santo would protect the highest number of endemic and threatened endemic species.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations
  8. Conclusions
  9. Acknowledgments
  10. Literature Cited

Critics of the hotspot concept contend that a lack of objective targets for conservation areas once they are defined may result in political inertia. We suggest hotspots are important basic units of extreme diversity that allow analysis of spatial patterns of biodiversity and permit prioritization of conservation activities and financing.

Our results form the basis for a wider project that will apply these protocols to the other top-performing indicator families in this biome (Fig. 1) and identify areas of outstanding diversity to the Brazilian Institute of Environment and Renewable Natural Resources (IBAMA) as candidate areas for legal protection. IBAMA protects the remaining Mata Atlantica diversity, and economic (e.g., expanding agriculture, aggressive tourism) factors and limited financial resources restrict IBAMA's ability to meet its goals. In the face of socioeconomic, financial, and political limitations, underexploited resources, such as museum collections, provide an alternative and relatively cheaply available source of environmental data that accurately indicates candidate areas for priority protection.

Our results highlight the need for continuing research, collection, and identification of plant specimens in the Mata Atlântica, particularly within southern Bahia and northern Espírito Santo. As loss of natural habitats and species extinctions becomes evermore apparent and the time in which and resources with which entire floras can be studied decreases, identification of indicator taxa becomes increasingly necessary to identify areas of species diversity. Our techniques can be applied to other global hotspots and species groups, and their accuracy and relative ease of implementation make them attractive for rapid priority setting. Further analysis in the light of practical and political issues such as cost is required, but these conservation recommendations are a first step toward protection of these areas.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations
  8. Conclusions
  9. Acknowledgments
  10. Literature Cited

We thank G. Harris for the use of SPOT/VG data on forest cover, M. Sobral for verifying the list of Mata Atlântica endemics, and the database curators at herbaria who provided data. We are grateful for field support and discussions with colleagues at the University of São Paulo herbarium in Piracicaba (ESA) and for the constructive criticisms of the reviewers. We thank British Airways for a flight to Brazil under their Community and Conservation program.

Literature Cited

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Limitations
  8. Conclusions
  9. Acknowledgments
  10. Literature Cited
  • Araújo, M. B., and P. H. Williams. 2001. The bias of complementarity hotspots toward marginal populations. Conservation Biology 15:17101720.
  • Cracraft, J. 1995. The urgency of building global capacity for biodiversity science. Biodiversity and Conservation 4:463475.
  • Elith, J., et al. 2006. Novel methods improve prediction of species' distributions from occurrence data. Ecography 29:129151.
  • Giulietti, A. M., R. M. Harley, L. P. De Queiroz, M. G. L. Wanderley, and C. Van Den Berg. 2005. Biodiversity and conservation of plants in Brazil. Conservation Biology 19:632639.
  • Govaerts, R., et al. 2006. World checklist of Myrtaceae. The Board of Trustees of the Royal Botanic Gardens, Kew. Available from http://www.kew.org/wcsp/myrtaceae (accessed May 2006).
  • Harris, G. M., C. N. Jenkins, and S. L. Pimm. 2005. Redefining biodiversity conservation priorities. Conservation Biology 19:19571967.
  • Hess, G. R., R. A. Bartel, A. K. Leidner, K. M. Rosenfeld, M. J. Rubino, S. B. Snider, and T. H. Ricketts. 2006. Effectiveness of biodiversity indicators varies with extent, grain, and region. Biological Conservation 132:448457.
  • Hijmans, R. J., S. E. Cameron, J. L. Parra, P. Jones, and A. Jarvis. 2005a. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25:19651978.
  • Hijmans, R. J., L. Guarino, A. Jarvis, R. O'Brien, P. Mathur, C. Bussink, M. Cruz, I. Barrantes, and E. Rojas. 2005b. DIVA-GIS version 5.2. Biodiversity International, Rome .
  • Hurlbert, A. H., and E. P. White. 2005. Disparity between range map- and survey-based analyses of species richness: patterns, processes and implications. Ecology Letters 8:319327.
  • IBGE (Brazilian Institute of Geography and Statistics). 2004. Mapa de bioma do Brasil. IBGE, Rio de Janeiro. Available from http://www.ibge.gov.br/mapserver (accessed February 2006).
  • IUCN (World Conservation Union). 1994. Guidelines for protected areas management categories. IUCN, Gland, Switzerland, and Cambridge, United Kingdom. Available from http://www.unep-wcmc.org/protected_areas/categories/index.html (accessed June 2006).
  • IUCN (World Conservation Union). 2001. IUCN Red List categories and criteria. Version 3.1. IUCN, Gland, Switzerland and Cambridge, United Kingdom. Available from http://www.iucnredlist.org/info/categories_criteria2001 (accessed May 2006).
  • Jenks, G. F., and F. C. Caspall. 1971. Error in choroplethic maps: definition, measurement, reduction. Annals of the Association of American Geographers 61:217244.
  • Klein, R. M. 1990. Espécies raras ou ameçadas de extinção: Estado de Santa Catarina. Volume 1. Mirtáceas e Bromeliáceas. Brazilian Institute of Geography and Statistics, Rio de Janeiro .
  • Legrand, C. D., and R. M. Klein. 19671977. Mirtáceas. Pages 1730 in R.Reitz, editor. Flora ilustrada Catarinense. Itajai, Santa Catarina , Brazil.
  • Lucas, E. J., S. A. Harris, F. F. Mazine, S. R. Belsham, E. M. Nic Lughadha, A. Telford, P. E. Gasson, and M. W. Chase. 2007. Suprageneric phylogenetics of Myrteae, the generically richest tribe in Myrtaceae (Myrtales). Taxon 56:11051128.
  • Mittermeier, R. A., P. R. Gil, M. Hoffman, J. Pilgrim, T. Brooks, C. G. Mittermeier, J. Lamoreux, and G. A. B. Da Fonseca. 2005. Hotspots revisited: earth's biologically richest and most endangered terrestrial ecoregions. University of Chicago Press, Chicago , Illinois .
  • Mori, S. A., B. M. Boom, A. M. De Carvalino, and T. S. Dos Santos. 1983. Ecological importance of Myrtaceae in an eastern Brazilian Wet Forest. Biotropica 15:6870.
  • Oliveira-Filho, A. T. 2006. Catálogo das arvores nativas de Minas Gerais—mapeamento e inventário da flora nativa e dos reflorestamentos de Minas Gerais. Universidade Federal de Lavras, Minas Gerais , Brazil .
  • Oliveira-Filho, A. T., and M. A. L. Fontes. 2000. Patterns of floristic differentiation among Atlantic forests in southeastern Brazil, and the influence of climate. Biotropica 32:793810.
  • Olson, D. M., et al. 2001. Terrestrial ecoregions of the world: a new map of life on earth. BioScience 51:933938.
  • Pearson, R. G., C. J. Raxworthy, M. Nakamura, and A. T. Peterson. 2007. Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar. Journal of Biogeography 34:102117.
  • Phillips, S. J., R. P. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modeling 190:231256.
  • Pimm, S. L. 2005. It's a new century: do you know where your orchids are? Selbyana 26:513.
  • Rebelo, A. G., and W. R. Sigfried. 1992. Where should nature reserves be located in the Cape Floristic Region, South Africa? Models for the spatial configuration of a reserve network aimed at maximizing the protection of diversity. Conservation Biology 6:243252.
  • Sobral, M. 2003. A família Myrtaceae no Rio Grande do Sul. 1st edition. UNISINOS, São Leopoldo , Brazil .
  • SOS Mata Atlântica. 1998. Evolução dos remanescentes florestais e ecossistemas associados do domínio Mata Atlântica no período 1990–1995. Fundação SOS Mata Atlântica/INPE, São Paulo .
  • Thomas, W. W., A. M. V. De Carvalho, A. M. A. Amorim, J. Garrison, and A. L. Arbeláez. 1998. Plant endemism in two forests in southern Bahia, Brazil. Biodiversity and Conservation 7:311322.
  • WDPA Consortium 2004. World database on protected areas. Version 2. World Conservation Union and UNEP-World Conservation Monitoring Centre, Cambridge , United Kingdom .
  • Williams, P. H., and C. J. Humphries. 1994. Biodiversity, taxonomic relatedness, and endemism in conservation. Pages 269287 in P. L.Forey, C. J.Humphries, and R. I.Vane-Wright, editors. Systematics and conservation evaluation. Oxford University Press, Oxford , United Kingdom .
  • Willis, F., J. Moat, and A. Paton. 2003. Defining a role for herbarium data in red list assessments: a case study of Plectranthus from eastern and southern tropical Africa. Biodiversity and Conservation 12:15371552.