The story begins in the 1970s, when the Italian wolf population had collapsed to fewer than 100 individuals, clinging to survival in the Central and Southern Apennines. Decades of persecution, deforestation and prey depletion had pushed the apex predator to the brink. Today, thanks to legal protection, land abandonment, reforestation and a surge in wild ungulate numbers, Italy hosts an estimated 3,307 wolves, including more than 2,000 in the Apennine chain alone. The Cilento park, covering more than 1,800 square kilometers of mountains, coastline, forests and rural villages in the Campania region, has become one of the strongholds of this recovery. Where the park once sheltered only a handful of wolves during the population bottleneck, the population now exceeds two hundred individuals, a figure corroborated by the new density estimates.

Led by Maria Buglione and Domenico Fulgione of the University of Naples Federico II, the research team assembled an unusually rich dataset to track both predator and prey. Between March 2023 and April 2024, they combined daytime and nocturnal transect surveys, a full year of camera trapping, roadkill records supplied by veterinary and forest police services, and validated citizen science reports gathered from trained volunteers and public platforms such as iNaturalist. After expert vetting to exclude unreliable or imprecise observations, the team compiled 790 georeferenced occurrence points: 280 for wolf, 329 for wild boar, 126 for roe deer and 55 for red deer. These were spatially thinned in R to remove duplicates and reduce autocorrelation before modeling.

The core of the analysis rested on Maximum Entropy distribution modeling, a presence-only technique widely regarded as one of the most reliable approaches in conservation biology because it performs well even with modest numbers of records. Twelve to fifteen environmental variables were tested for each species, including elevation, slope, aspect derived from a digital terrain model, and distances to agricultural meadows, forests, grasslands, scrublands, urban settlements and waterways derived from Corine Land Cover data. Crucially, for the wolf, the team also fed in the predicted distributions of its three principal wild prey, allowing them to quantify how much of the wolf’s habitat choice is driven by prey availability rather than landscape structure alone. Multicollinearity was controlled using variance inflation factors and Pearson’s correlation screening, and model complexity was optimized with the ENMeval package, varying regularization multipliers and feature combinations and selecting the best configuration by the Akaike Information Criterion.

The models performed strongly. The wolf model, built on 132 occurrences after thinning, achieved an area under the curve of 0.811 for training data and 0.796 for test data, while the roe deer model reached AUC values above 0.80, and the red deer model 0.82 for training. Jackknife tests showed that when used alone, the roe deer’s predicted distribution gave the greatest gain in explaining wolf occurrence. In the full model, the roe deer’s potential distribution contributed 85.9 percent of the explanatory power for the wolf, dwarfing the contributions of wild boar at 8.4 percent and elevation at 2.0 percent. Red deer and wild boar distributions contributed just 4.6 and 4.2 percent respectively.

To convert presence records into hard numbers, the team deployed the Random Encounter Model, a camera-trap density estimator that requires no individual recognition. Twelve quadrats of 10 by 10 kilometers were placed across the wolf core area, each containing four camera traps on a 2 by 2 kilometer grid. Using the trapping rate, estimated daily movement speed of around 16 kilometers per day drawn from published Apennine telemetry studies, and the cameras’ detection radius and angle, the researchers calculated a mean density of 1.2 wolves per square kilometer in the most suitable habitat, with a 95 percent confidence interval spanning 0.41 to 2.99. The area of maximum suitability, 78.95 square kilometers with habitat suitability at or above 0.9, could theoretically support up to 95 wolves, and the team estimates that more than a hundred wolves occupy the 1,810 square kilometer study area overall. Despite partial isolation, the population remains connected to the wider Apennine metapopulation through two corridors, the Vallo di Diano plain and the Piana del Sele, where the researchers documented numerous live and road-killed wolves crossing what would otherwise be an ecological barrier.

The statistical evidence for the roe deer link was striking. A Mantel test comparing Euclidean distance matrices of wolf and prey occurrences found a significant, non-random spatial correlation only between wolf and roe deer, with a correlation coefficient of 0.8397 and a p-value of 0.001, a value that exceeded all upper quantiles of the null distribution generated by 999 permutations. In contrast, the wolf showed no significant spatial association with red deer, whose correlation was actually slightly negative at minus 0.0803, hinting at possible spatial segregation, and no significant association with wild boar. Schoener’s D overlap statistics on the predicted distributions painted a consistent picture: wolf-roe deer overlap of 0.9041, wolf-red deer of 0.8117 and wolf-wild boar of 0.7805, with Pearson’s and Spearman’s correlations across raster layers confirming the same ranking and all associations statistically significant.

The reasons behind the hierarchy are ecological as much as numerical. Wild boar is by far the most abundant ungulate in the park, with local densities often exceeding 10 individuals per square kilometer, well above the 2 to 3 animals per square kilometer that Mediterranean agroecosystems can typically tolerate. Yet boar are so ubiquitous and behaviorally flexible, occupying every habitat type including human-modified landscapes, that their distribution offers the wolf little spatial signal. Red deer, reintroduced along with roe deer on Mount Cervati between 2003 and 2006 as part of park restoration actions, remain anchored to their release site and strongly associated with mixed woodland, a habitat that barely figures in wolf habitat selection. Moreover, killing an adult red deer typically requires a large, well-structured pack, and wolves in the Cilento are mostly organized in pairs or small groups of dispersing juveniles, with large packs documented only around Mount Cervati and the Alburni Mountains. Roe deer, by contrast, prefer high, quiet mountain habitats away from settlements, precisely the environments the wolf favors, and camera traps repeatedly recorded wolves carrying roe deer carcasses.

The broader context is a landscape transformed. Over the past half century, rural abandonment, expanding forest cover and the return of wild ungulates have restructured predator-prey dynamics that had been broken for generations. The wolf’s preference for mountainous, less disturbed terrain is consistent with patterns seen elsewhere in the Mediterranean, and interestingly, the distance from urban settlements emerged as an important variable for the deer species and for the wolf, reflecting both the animals’ avoidance of people and the fact that human settlements in this region extend far up the elevational gradient. The researchers emphasize that their models capture potential distributions rather than direct dietary proof, and that diet analysis using non-invasive genetic and morphological methods on fecal samples is the next step to confirm how these spatial associations translate into actual predation.

The implications for management are immediate. As Europe formally downgraded the wolf’s status from strictly protected to protected under the Bern Convention and the EU Habitats Directive, debates over culling, livestock compensation and coexistence have intensified across the continent. This study argues that effective management must be grounded in prey ecology: because wolf distribution closely follows wild prey, protecting and planning for roe deer habitat means protecting the corridors and core areas wolves need. Understanding where wolves are likely to settle can also inform grazing management and land-use planning, helping shepherds avoid the highest-risk areas and reducing conflict in peri-urban and agricultural zones. Dispersing young wolves, tempted by easy food sources such as garbage, may approach inhabited centers when wild prey is unavailable along their routes, another reason to maintain prey availability across the landscape.

For a species that was nearly exterminated from the Italian peninsula within living memory, the wolf’s return to the Cilento mountains is a conservation success story, but one whose next chapter depends on the delicate geometry between predator, prey and people. By showing that the map of the wolf is, in large part, the map of the roe deer, this research provides managers with a practical compass for reconciling large carnivore recovery with sustainable forest use in one of the Mediterranean’s most biodiverse and rapidly changing landscapes.