Incorporating fossil data into ecological niche models enhances predictions of the past distribution of Macaronesian‐Mediterranean strawberry trees

Hindcasting is a widely used method to predict the past distribution of plant species. However, its reliability is often constrained by the scarcity of fossil records, which are fundamental for validating and refining model projections. Without sufficient palaeobotanical evidence, hindcasts may lack accuracy and ecological realism. Here, we incorporate palaeobotanical occurrences in the modeled distribution for the genus Arbutus L. (Ericaceae family) in the Macaronesian-Mediterranean region to calibrate the ecological models for seven bioclimatic periods from the Last Glacial Maximum to the Late Holocene and to compare them with the traditional hindcast method. Species distribution models were developed using the biomod2 R package that applies an ensemble forecasting approach to produce species-environment relations and to obtain spatiotemporal predictions. We used two calibration approaches: (i) a model trained with present-day bioclimatic conditions, hindcasted for each past period with known presence-only data plus random pseudo-absences generated by biomod2, and (ii) models trained for each past period with fossil data in different formats (presence only, presence-absence, and presence-absence complemented with pseudo-absences). We evaluated the accuracy of hindcasts for each time period and assessed the spatial coherence among different methods over time by applying the Jaccard similarity index. Different modeling approaches projected partially overlapping suitable areas for Arbutus. The balanced accuracy indicator revealed large discrepancies between the hindcasted predictions and the projections calibrated with fossil data for each past period. The spatial coherence between different training methods shows an overall time-increasing reliability. Our study suggests the following future directions: (i) incorporating modern and past datasets in modeling design, (ii) comparing different modeling approaches to predict past environmental suitability, and (iii) interpreting and validating past spatial distribution patterns using independent lines of evidence (e.g., palaeoclimatic and palaecological proxy data). This framework fosters collaboration and integration of multiple disciplines (e.g., palaeoecology, climatology, geology, ecological modeling, and biogeography).