Patterns of tree species composition and richness across the principal biomes of lowland tropical South America and their underlying environmental drivers
Silva De Miranda, Pedro Luiz
MetadataShow full item record
Lowland tropical South America encompasses some of the most species-rich and threatened ecosystems in the world, spanning across countries such as Brazil, Bolivia, Colombia, Ecuador, Peru and Venezuela, which are known for their biodiversity. However, due to its incredible environmental and ecological complexity and that most of its area has yet to be scientifically studied in any depth, controversy surrounds its biomes’ identities, the limits of their geographic and environmental distributions and estimates of their tree species richness. The main objective of this thesis is to study the phytogeography of lowland Tropical South America by delimiting its biomes through a floristic approach, by investigating these biomes’ environmental controls and dynamics and by assessing their tree species richness and endemism. In order to fulfil this objective, we have employed a dataset of thoroughly checked tree species checklists, the NeoTropTree (NTT) dataset, which covers more than 8000 locations across South, Central and southern North America and encompasses occurrence records for more than 12000 tree species. Firstly, I defined and mapped the main biomes in lowland tropical South America (LTSA) through the means of a hierarchical clustering analysis based on tree species composition associated with an a priori classification of 4103 NTT sites into vegetation types. I then proceeded to map these biomes geographically and to assess their environmental overlaps (both climatic and edaphic) through a classification tree approach (random forest analysis). I was able to delimit five main biomes in LTSA: Amazon Forest, Atlantic Forest, Chaco, Savanna and Seasonally Dry Tropical Forest (SDTF). I also show that there is an important environmental overlap amongst biomes. Error rates for site classification into biome using solely environmental data ranged from 19-21% when only climate was considered and 16-18% when I also took edaphic variables into account. I conclude that it is viable and advisable to use tree species composition to determine biome identity, at least within individual continents. In the case of LTSA, there is high biome heterogeneity at small spatial scales, which explains why it is so challenging to use climatic and/or interpolation based edaphic data, or remotely-sensed imagery, to map tropical biomes. Because of this, I then conclude that biome delimitation using floristic information may enable more efficient biome conservation and management efforts. Secondly, I investigated the environmental controls distinguishing biome limits for two regions of LTSA with high biome heterogeneity – eastern Brazil and Bolivia. To this end, I selected 182 NTT sites in these two regions, collected detailed soil data from the field and extracted climate and fire data from publicly available GIS data layers. I assigned these sites to one of three states based on their tree species composition: moist forest (including both Atlantic and Amazon Forests), SDTF or savanna. Selected environmental variables were organized into three distinct categories describing functional environmental regime: water availability, soil fertility and fire, and their significance as predictors of biome identity was assessed within a structural equation modelling framework. I found that environmental controls behind biome distribution differ between the two studied areas and according to the biomes involved. I concluded that water availability, soil fertility and fire are all important determinants of biome limits. Amongst the three categories, water availability was the most important one in determining biome identity at our study sites, with soil fertility differentiating eastern Brazil SDTFs from the other biomes, and fire representing an important determinant of savanna’s environmental limits. Thirdly, I estimated and compared tree species richness and endemism levels of LTSA’s main biomes using NTT’s tree species checklists and incidence (i.e., occurrence) data. To do so, I extracted tree species information for 4540 sites registered in NeoTropTree distributed across four biomes: Amazon Forest, Atlantic Forest, Savanna and SDTF. I first compared how tree species accumulated with number of sites sampled for biomes and then estimated biomes’ total tree species richness using non-parametric approaches (species extrapolation curves). I also estimated the number of endemic tree species to these areas with two approaches: indicator species analyses and absolute unique/shared species counts. I was able to show that the Amazon Forest is the most tree species-rich environment in LTSA, followed by the Atlantic Forest, Savannas and then SDTFs. In relation to endemism levels, the Amazon and Atlantic Forests’ tree flora are mainly composed of endemic tree species whereas that is not the case for the savanna and SDTF. The estimation of total tree species richness through extrapolation curves revealed that around 94% of the tree flora of the Amazon forest, the Atlantic forest and the SDTF have already been recorded. According to the same analysis, only around 70% of the savannah tree flora has been recorded. However, this pattern might be related to the high number of biome intrusions into this biome. The differences in richness and endemism between the moist (Amazon and Antlantic forests) and drier biomes (savanna and SDTF) suggest that drought-sensitivity and biogeographic history are drivers of tree species distribution in LTSA. Finally, by integrating biome delimitation based on floristic composition with knowledge on these environments’ environmental correlates and tree species richness, I was able to describe LTSA’s main phytogeographic features in a way that has never been done before, drawing attention to its complexities and performing novel cross-biome comparisons. My study shows that LTSA’s biomes are interspersed across geographic space, especially in the Dry Diagonal located between the Amazon and Atlantic Forests, and that environmental controls driving these ecosystems’ distributions can vary according to the biomes being considered and the geographic location. I also show that LTSA’s most tree species-rich biomes are the ones with the highest quantity of endemic tree species and that taxonomic expeditions to the Amazon Forest can potentially lead to more species being described in these environments. To summarize, I was able to highlight LTSA’s main floristic patterns and link them to environmental drivers and tree species richness, thereby substantially transforming how these biomes are perceived by biodiversity scientists and conservationists.