Evaluating the performance of past projection in Species Distribution Modelling through the combined use of phytogeographic and paleobotanical data: the case of Arbutus in the Old World

Species Distribution Models (SDMs) are often used to predict present and past geographic ranges, although they are rarely validated with paleo data. By combining current occurrence data, fossil records, and SDMs, we investigated the postglacial dynamics of Arbutus and the bioclimatic constraints that shaped its demographic history as an important representative of the Mediterranean broadleaved evergreen element. We applied an ensemble-modelling approach (BIOMOD2) to obtain present and past predicted distributions for the whole genus Arbutus in the Mediterranean regions, along the Atlantic coast of Europe, and in Macaronesia, The SDMs were based on double-checked occurrence records from literature and bioclimatic variables that were retrieved from CHELSA and PaleoClim databases. Spearman correlation allowed to select three subsets of varibales with moderate/low correlation (|r|<0.7) for model development. Past distribution maps for the last 30 ka at 1000-year intervals were obtained from >1000 chronologically controlled pollen and macrofossil records. First, we trained the current model with modern occurrences to provide hindcasts for each Paleoclim period (LGM, HS, BA, YD, EH MH, LH) which were further validated through metrics retrieved by a 2x2 confusion matrix (modelled presences/absence vs observed presences/absence). Secondly, we trained models for each PaleoClim period using past occurrences and then validated them through the 2x2 confusion matrix. Results show excellent performance for the BIOMOD2 models (TSS>0.9) and the Balanced Accuracy of the confusion matrix, which does not show a linear correlation between the hindcast performances and the distance in time. This work indicates that the combination of fossil records and modelling techniques is an effective tool to depict the Quaternary demographic history of tree taxa, to explain the modern biogeography of the Mediterranean evergreen ecosystems, to anticipate potential range shifts in relation to the ongoing climate change, and to raise hypotheses to inform phylogeographic research.