Method

To map the past, present, and uncertain future of the olive tree, the story begins on the ground, tracing exact coordinates tied to geotagged images of living groves. Before any predictive model could sketch the future, this raw dataset demanded meticulous refinement. Erroneous coordinates were discarded, duplicate entries erased, and dense clusters of sampled points thinned out—ensuring that an over-surveyed valley would not skew the digital narrative or blind the algorithm to less-frequented terrains.

With clean locations established, the landscape was infused with atmospheric context. High-resolution climate records from the CHELSA-bioclim dataset provided the environmental tapestry: granular measures of temperature swings, rainfall patterns, and seasonal shifts. To keep the mathematical engine sharp, strongly correlated climate factors were carefully pruned. If two variables whispered the exact same environmental clue, one was set aside, leaving only a lean set of distinct, non-redundant indicators.

Next came defining the stage itself. Drawing a boundary around known olive occurrences and extending it outward created a realistic dispersal arena—a spatial backdrop representing every sanctuary the trees might have reached over generations. Against this canvas, the predictive software Maxent ran a tournament of trial models. Evaluators looked first to the Continuous Boyce Index (CBI) as the primary gauge of success, measuring how accurately the predictions mirrored real-world presence distributions across unmapped terrains. To refine the field, secondary evaluations were brought to bear: the AICc statistic acted as a counterweight against overfitting by penalizing excessive complexity, while the AUC (Area Under the Curve) verified the model’s overall capacity to distinguish suitable habitats from unsuitable ground.

The winning configuration was ultimately unleashed across the entire dataset. Through Jackknife evaluations and response curves, researchers stripped away individual climate variables one by one, confirming that the model's mathematical logic aligned with the real biological thresholds of the olive tree.

The resulting map transformed abstract climate projections into vivid geography. Continuous gradients of habitat suitability were translated into clear zones of opportunity and threat. By comparing present-day conditions against future climate projections, the map illuminated retreating boundaries, emerging horizons, and crucial refuges—enduring sanctuaries where olive trees can stand resilient against a changing world.