Floristic change in Brazil's Southern Atlantic forest biodiversity hotspot : From the last glacial maximum to the late 21st century

Abstract

Brazil's Atlantic Forest biome is one of the world's biodiversity hotspots, whose heterogeneous ecosystems are threatened by habitat loss and climate change. Palaeoecological research can provide essential context for the impacts of anthropogenic climate change in the 21st Century and beyond, but existing studies have notable limitations in the insights they can provide: vegetation proxy data are spatially and temporally skewed with inconsistent taxonomic resolution; existing modelling studies typically overlook individualistic species-level responses, are limited in temporal coverage, and lack close integration with empirical palaeoecological data. Here, we investigate the impact of major climate changes upon the species-level floristic composition of southern Brazil's Atlantic Forest, from the Last Glacial Maximum (LGM) to the late 21st century, by modelling the distributions of 30 key species at seven time slices since the LGM and comparing the assemblages they form with an unprecedented dataset of palaeoecological proxy data. We find notable compositional changes through time across our study area, especially during the early Holocene, which was characterised by extensive no-analogue plant communities. Aspects of these modelled floristic changes are captured in proxy records but many occur in data-sparse regions, highlighting geographic foci for future palaeoecological investigation to test these model predictions. Our findings highlight the individualistic responses of Atlantic Forest plant species to climate change and help resolve long-standing palaeoecological questions – explaining the dominance of highland grasslands at the Last Glacial Maximum (likely due to low atmospheric CO₂ concentrations), clarifying the LGM extent of coastal tropical forest (probably in a grassland matrix on exposed continental shelf), and explaining the origins of Araucaria angustifolia's western populations (from climatic (micro-)refugia rather than human-mediated dispersal). Our results also set the 21st Century's impending climate and vegetation changes in a 21,000-year temporal context, revealing that, under a high emissions scenario, more than 100,000 km² of the southern Atlantic Forest will experience more climate-driven floristic change in the coming decades than it has in the last 21 millennia.

• author: Wilson, Oliver J. ; Mayle, Francis E. ; Walters, Richard J. ; Lingner, Débora V. and Vibrans, Alexander C.
• organization: Centre for Environmental and Climate Science (CEC); BECC: Biodiversity and Ecosystem services in a Changing Climate; Biodiversity
• publishing date: 2021-07-15
• type: Contribution to journal
• publication status: published
• subject: Climate Research; Ecology
• keywords: Anthropocene, Atlantic forest, Biodiversity hotspot, Climate change, Conservation, Late quaternary, Palaeoecology, Present, South America, Vegetation dynamics
• in: Quaternary Science Reviews
• volume: 264
• article number: 107005
• pages: 22 pages
• publisher: Elsevier
• external identifiers: scopus:85107887091
• ISSN: 0277-3791
• DOI: 10.1016/j.quascirev.2021.107005
• language: English
• LU publication?: yes
• additional info: Funding Information: At 12,000 BP, much of our study area was somewhat drier and more seasonal than the present; the north-east was 1?2 ?C cooler than present and the south up to 1 ?C warmer, though most areas were within 1 ?C of modern average temperatures. Proxy evidence from Botuver? and Santana supports the early Holocene (12,000?9000 BP) being the driest in the study period, with the lowest contributions from the summer monsoon. Subsequent gradual and relatively minor increases in precipitation in the proxy records are reflected in our downscaled data. Temperature proxies generally record 12,000 BP as being similar to, or warmer than, the preceding time slices; several also show slight increases in temperatures through the Holocene which are less clear in our downscaled data.Most of the predicted changes between the LGM and Late Glacial occur in areas not covered by palaeo-data sites, making it difficult to discern their accuracy. Most sites in our study area do show modest increases in forest pollen between 18,000 and 15,000 BP, in line with predictions of more range expansions for forest species than herbs, though fossil pollen from the growing cluster of south-eastern plateau sites remains ?90% grassland taxa. The largest increases in forest pollen are found in GeoB2107 and Curucutu, reaching 33% of the total in each. GeoB2107's arboreal pollen also becomes more diverse, with notable increases of Moraceae/Urticaceae, Melastomataceae and Celtis, while previously established taxa such as Alchornea, Araucaria and Podocarpus are maintained. This might support the predicted development of assemblage 7 at the expense of assemblage 9 on the north-central continental shelf: the former is more dominated by warm-adapted tropical trees, with cold-adapted Campos, Cloud Forest and AMF species less prevalent than in assemblage 9. A notable proxy site which first covers the 15,000 BP time slice is Buriti (Bertoldo et al., 2014): at this time period, 91% of Buriti's pollen comes from tree taxa ? a proportion far higher than any other site at this period ? with Araucaria (ca. 20%), Ilex and Podocarpus (15?20%) the main contributors. This would suggest that at least some western areas in our study area ? around assemblages 29, 26 and 34 ? had AMF-like forest cover at 15,000 BP.The cold-adapted assemblages of cluster c also experience significant changes between the late glacial and early Holocene. Here again, however, few palaeoecological sites can test these predictions. Phytoliths and soil carbon isotopes at Aroeiras (Silva, 2018), Guarapuava (Calegari, 2008) and another nearby site (Silva et al., 2016; no precise location) suggest the presence of woody AMF-like vegetation in the early Holocene, which subsequently declined. Further soil profiles at CPCN Pro Mata (D?mig et al., 2008; Silva and Anand, 2011) show that smaller-than-modern forest blocks were also found in the highlands' south-east by 9000 BP, but any changes in their composition are invisible in the region's grassland-dominated pollen records. Further north, there is little change in the grassland-dominated pollen record from Serra do Aracatuba (Behling, 2006), but a short buried peat sample near the Cerro do Touro Watershed (de Oliveira et al., 2008) shows a 30% decline in Campos pollen within a few centuries of 12,000 BP as Weinmannia, Myrtaceae, Myrsine and Dicksonia sellowiana pollen increase. Combined with changes in AMF taxa from GeoB2107 (Araucaria, Schinus/Lithraea, Podocarpus and Myrsine all respond differently), this may provide general support for predicted compositional changes in highland vegetation through the early Holocene, though forest coverage apparently remained patchy.Most palaeo-proxy sites in the region's west are dominated by grassland/C4 signals at the mid-Holocene, apart from Trincheira Reserva (Alcantara dos Santos, 2013) to the north-west, where phytoliths and ?1?C records show dominant woody C3 vegetation, and Buriti, where forest pollen increases to make up 70% of the total. As before, there is compositional change in Buriti's arboreal pollen with Araucaria's contribution decreasing, being overtaken by Alchornea. There are also increases in Dicksonia sellowiana spores and pollen from Drimys, Podocarpus and Luehea, with Sloanea occurring at 6000 BP for the first time in several millennia. This combination of AMF and more warm-adapted tropical taxa may suggest a no-analogue forest assemblage at Buriti in the mid-Holocene; although our modelled species do not cover the location well at this period, this could support our prediction of the site occurring between floristic assemblages resembling tropical forest and AMF 6000 years ago.This work was supported by a University of Reading Graduate Teaching Assistant PhD studentship (OJW), the Funda??o de Amparo ? Pesquisa e Inova??o do Estado de Santa Catarina (IFFSC ? DVL and ACV), and the Conselho Nacional de Desenvolvimento Cient?fico e Tecnol?gico research grant 312075/2013?8 (ACV).Speleothem data were kindly provided by Francisco W. Cruz. This work made use of the Reading Academic Computing Cloud. OJW thanks M. Jane Bunting (University of Hull) for logistical support while undertaking this research. We thank Bronwen Whitney and an anonymous reviewer for their comments which helped to improve this manuscript. Funding Information: This work was supported by a University of Reading Graduate Teaching Assistant PhD studentship (OJW), the Fundação de Amparo à Pesquisa e Inovação do Estado de Santa Catarina (IFFSC – DVL and ACV), and the Conselho Nacional de Desenvolvimento Científico e Tecnológico research grant 312075/2013–8 (ACV). Publisher Copyright: © 2021 Elsevier Ltd Copyright: Copyright 2021 Elsevier B.V., All rights reserved.
• id: 77d301f0-8266-4b4b-87b2-3af876a6aa27
• date added to LUP: 2021-06-28 15:06:46
• date last changed: 2022-04-27 02:36:34