Investigating Pre-Columbian floristic legacy effects using machinelearning in the southern Atlantic Forest, Brazil

Barney Harris^1*Hermann Behling²Jonas Gregorio De Souza³Antonia Lena Reinhardt²Patrick Roberts⁴Philip Riris¹

• ¹Bournemouth University, Poole, United Kingdom
• ²Department of Palynology and Climate Dynamics, Faculty of Biology and Psychology, University of Göttingen, Göttingen, Lower Saxony, Germany
• ³Department of Humanities, Pompeu Fabra University, Barcelona, Balearic Islands, Spain
• ⁴Max Planck Institute for Geoanthropology, Jena, Thuringia, Germany

The scope and scale of past human impacts on both historic and current vegetation is of widespread interest in the historical sciences. In the Atlantic Forest of southern Brazil (Portuguese: Mata Atlântica), previous work has identified Amerindian settlement and land-use as a probable driver of the extent and composition of forest cover, with time-extended legacies that remain detectable in modern floristic inventories. Previously published investigations into the ecological history of the southern Atlantic Forest have either eschewed the role of humans or, where anthropogenic drivers are explicitly examined, utilized spatially restricted environmental datasets, necessarily limiting the generalizability their conclusions. This study aims to redress this gap, and to quantify the impact of past Amerindian Pre-Columbian settlement and associated land use on the modern-day distribution of several key plant species across the entire southern Atlantic Forest. We fit Maxent species distribution models (SDMs) using Indigenous archaeological site locations (Tupi-Guarani and southern Jê) and modern plant species occurrence data (35 unique species) in a comparative analytical framework to investigate Indigenous influence on the likelihood of occurrence of culturally significant or medicinal plant species. Our results indicate that i) the inclusion of archaeological settlement location data and SDM predictions as covariates can improve the performance of contemporary floristic species distribution modelling and should be incorporated into ecological models of plant species in landscapes with long-standing human presence, especially when they are used to inform policy that explicitly aims to preserve 'natural' biomes and; ii) a synanthropic relationship can be demonstrated between the southern Jê and Araucaria angustifolia, a finding that complements previously published phylogeographic and palaeoenvironmental studies exploring the same link.