The research team, led by Binod Kunwar and colleagues, combined two powerful modeling approaches: species distribution modeling using the Maximum Entropy algorithm, known as MaxEnt, and landscape connectivity analysis using least-cost pathways. The study drew on three years of intensive fieldwork between 2023 and 2025, during which 120 camera traps were deployed across Jeju Island in a systematic 1-by-1-kilometer grid. Each trap, positioned at least 30 meters from its neighbors, recorded bursts of three 20-megapixel photographs followed by 30 seconds of video whenever its passive infrared sensors detected movement. Combined with opportunistic sightings and field signs such as rubbing trees, feces, wallowing, and soil rooting, the surveys yielded 94 georeferenced wild boar occurrences, which were spatially filtered so that only one record was retained per 30-by-30-meter grid cell.

The environmental backbone of the model consisted of ten carefully selected variables, all resampled to a crisp 30-meter resolution. These included Euclidean distances to built-up areas, agriculture, forest, barren land, and water, all derived from the Ministry of Environment’s high-resolution land cover data, alongside elevation, slope, and a terrain ruggedness index calculated from the Shuttle Radar Topography Mission digital elevation model. Satellite-derived indices added another layer of ecological detail: the normalized difference vegetation index captured vegetation greenness, the modified normalized difference water index detected open water, and the normalized difference built-up index quantified development. Spearman’s rank correlation analysis screened out redundant variables, ensuring that multicollinearity would not distort the model’s predictions. To correct for uneven sampling intensity, the team generated a Gaussian kernel density from the occurrence records and used it to probabilistically draw 10,000 bias-weighted background points.

Rather than accepting MaxEnt’s default settings, the researchers tuned the model exhaustively, testing combinations of linear, quadratic, hinge, and product feature classes against regularization multipliers ranging from 0.5 to 5. The configuration that minimized the Akaike Information Criterion, a measure that balances goodness of fit against model complexity, emerged as the winner: linear, quadratic, and hinge features with a beta multiplier of 2.5. The final model, run in ten replicates with 70 percent of data for training and 30 percent for testing, delivered strong performance metrics, with an area under the curve of 0.76, a True Skill Statistic of 0.46, and a Continuous Boyce Index of 0.82, indicating reliable discrimination between suitable and unsuitable habitat.

The results were striking. Wild boar potential habitat covers 661.82 square kilometers, roughly 36 percent of Jeju Island’s entire surface. Of this, 265 square kilometers qualify as highly suitable and 396 square kilometers as moderately suitable. More than half of Hallasan National Park, the island’s ecological crown jewel with its unique vertical ecosystems, was projected as potential wild boar habitat, a finding that raises serious conservation concerns for the park’s native flora and fauna. The habitat overlapped overwhelmingly with forest, which accounted for nearly 62 percent of predicted suitable areas, followed by grassland at 28 percent. Notably, built-up areas also contributed to suitable habitat, underscoring the species’ notorious ecological flexibility and its capacity to thrive in human-modified environments.

Which environmental factors matter most? Distance to agriculture dominated the model, contributing a permutation importance of 0.64, far ahead of distance to built-up areas at 0.22, elevation at 0.14, and distance to water at 0.13. The response curves tell a vivid story: wild boar suitability rises sharply within 500 meters of agricultural land, reflecting the abundance and easy accessibility of forage in crop fields, and remains elevated near human settlements, a behavioral adaptation that inevitably increases the risk of human-wildlife conflict. The animals showed a strong preference for low elevations between sea level and 500 meters, moderately rugged terrain, proximity to forests that provide forage, shelter, and thermal refuge, and nearby water sources, with suitability declining exponentially as distance to inland water increases. Dense vegetation, indicated by NDVI values around 0.6, further enhanced habitat quality, likely offering both food and concealment.

Beyond mapping where the boars are, the study tackled a question that is arguably more important for management: how connected are their populations? The team converted the habitat suitability surface into a landscape resistance model using a nonlinear transformation calibrated for ecologically plastic species, then identified six core habitat patches larger than 3 square kilometers, a threshold based on known wild boar home range sizes. Patches separated by less than a kilometer, within the daily movement capacity of the animals, were merged. The result was a network of six patches totaling 181.51 square kilometers, dominated by the vast Hallasan patch at 123.88 square kilometers, with smaller strongholds at Saekdal-dong, Manjanggul, Geumakseo-gil, Gotjwal, and Aewol.

Least-cost pathway analysis revealed eight linkages threading through the island’s landscape. The mean least-cost path between patches stretched 11.41 kilometers, with individual corridors ranging from a short 1.92-kilometer hop between Geumakseo-gil and Gotjwal to a demanding 12.23-kilometer route between Saekdal-dong and Gotjwal. Connectivity was weakest between the Manjanggul area and Hallasan, where high landscape resistance impedes movement, while the Saekdal-dong area emerged as the most permeable landscape. Centrality analysis, which measures each patch’s contribution to overall network connectivity, crowned the Hallasan and Geumakseo-gil areas as the linchpins of the entire system. The corridor connecting Hallasan to Saekdal-dong, traversing mountainous forest, proved the most robust pathway, consistent with the species’ documented affinity for forested cover that offers food, shelter, and reduced human disturbance.

The practical implications are considerable. By pinpointing the corridors and core patches that sustain the invasion, the study gives South Korean managers a strategic map for intervention: targeted population monitoring, focused removal efforts, strategic fencing, and habitat modification concentrated where they will do the most good. Similar approaches, integrating fencing with population control along key movement routes, have been deployed in European countries grappling with expanding boar populations and the looming threat of African swine fever, though their success depends heavily on fence design, maintenance, and landscape configuration. On an island free of large predators and harboring four invasive mammal species, including sika deer and Siberian chipmunks, such spatially informed strategies may be the best hope for protecting Jeju’s vulnerable native ecosystems.

The authors are candid about their model’s limitations. Climatic variables were excluded because downscaling them to the island’s fine spatial scale would have introduced additional uncertainty, and independent validation of the connectivity analysis was not performed. Small, isolated habitat patches, which may serve as stopover sites, were left out of the network. Least-cost modeling identifies optimal pathways but cannot fully capture the stochastic, dynamic nature of real animal dispersal. The researchers recommend that future work integrate telemetry-based movement data, long-term population monitoring, and genetic assessments of gene flow to validate and refine the picture. Even so, the study stands as a template for proactive invasive species management: by predicting where an invader will thrive and how it will spread before it does, conservationists gain the precious commodity of time, and on an island as ecologically isolated and biodiverse as Jeju, time is everything.

Subject of Research: Habitat suitability and landscape connectivity modeling of invasive wild boar on Jeju Island, South Korea

Article Title: Modeling the Invasion: Habitat Suitability and Landscape Connectivity of Wild Boar (Sus scrofa) on Jeju Island, South Korea

Article References: Kunwar, B., Baral, S., Jeong, Y.-H., & Oh, H.-S. (2026). Modeling the Invasion: Habitat Suitability and Landscape Connectivity of Wild Boar ( Sus scrofa ) on Jeju Island, South Korea. Ecology and Evolution, 16(10), Article e74398. https://doi.org/10.1002/ece3.74398

Image Credits: AI Generated

DOI: 10.1002/ece3.74398

Keywords: wild boar, invasive species, Jeju Island, species distribution modeling, MaxEnt, landscape connectivity, least-cost pathways, habitat suitability, Hallasan National Park, island ecology, human-wildlife conflict, South Korea

Cycads are often described as living fossils, a lineage of gymnosperms that has persisted for roughly 280 million years, and they now hold the grim distinction of being the most threatened plant group on Earth. Cycas micronesica is the only cycad in Micronesia and the only gymnosperm on Rota, an 85-square-kilometer island whose tropical climate receives nearly 3,550 millimeters of rain each year. Until 2007, the species flourished across the island’s coastal forests and uplifted karst benches. That year, the armored scale A. yasumatsui, native to mainland Southeast Asia where it coexists harmlessly with its local Cycas hosts, was first recorded in Rota’s northeastern littoral forests. Within three years, three separate invasion fronts had spread across the island, and by 2011 every cycad habitat on Rota was infested.

The researchers were unusually well prepared for the catastrophe. In December 2007, before the scale had reached most of the island, the team collected leaflet tissue from 116 cycads across seven localities, sampling along transects that captured the full environmental and geographic range of the species on Rota. Using ten microsatellite loci developed specifically for this species, they characterized the pre-invasion genetic structure of the population. The genetic work revealed considerable variation, with 75 percent of alleles occurring at low frequency, and identified four distinct genetic clusters across the island. Pairwise differentiation among sites was moderate to strong, and a Mantel test found no significant isolation-by-distance pattern, suggesting that historical founder events, limited seed dispersal, and localized disturbance rather than simple geography shaped the island’s genetic architecture.

That genetic baseline proved critical for interpreting what came next. In January 2008, the team established 28 permanent monitoring plots across seven sites, each site containing square plots and belt transects covering roughly 100 square meters. At the time of the first census, the population stood at 11,254 stems per hectare, of which 9.2 percent were adult trees. Mean stem height was a mere 30 centimeters, because the population was dominated by seedlings and juveniles, a signature of vigorous regeneration. The seedling-to-juvenile-to-adult ratio of 70:21:9 described a population with a prosperous future. Within a single year of the scale’s arrival at each site, that future had collapsed.

The demographic data collected through 2023 reveal a merciless selective culling ordered by plant size. Seedling mortality was immediate and complete: every seedling in the plots died within roughly a year of local scale arrival, and no seedlings were observed from 2014 through the end of the study. Juveniles followed more slowly but no less surely, with more than 70 percent dead by the 2011 census and 100 percent mortality reached by 2023. Adult trees, buffered by their larger stores of parenchymatous tissue and non-structural carbohydrates, held out longer, but from 2013 to 2023 the island lost an average of 80 adult trees per hectare each year, culminating in 86 percent adult mortality. Across all age classes, total population mortality reached 98.6 percent within fifteen years.

The plant health metrics tell an equally dramatic story. In 2008, adult trees carried an average of 83 leaves, and some individuals bore more than 150. The scale’s herbivory killed standing leaves and then each new flush of foliage as it emerged, driving the mean down to just 8 leaves per tree by 2013. A slow recovery began around 2017, and by 2023 surviving adults averaged 20 leaves, a 2.5-fold improvement over 2013 but still only 24 percent of the pre-invasion benchmark. Mean stem height, meanwhile, rose from 30 centimeters to 245 centimeters, an eightfold increase that reflects not growth but the selective death of every small plant, a severe left-truncation of the population’s size distribution that the authors describe as an irreversible reduction in future survival potential.

Infestation intensity followed a characteristic invasion curve. Scores rose sharply to a peak around 2011, when the mean infestation score reached 8.4 on a 10-point scale and 100 percent of the island’s cycad habitats were colonized. Scores then declined and stabilized at roughly 3 to 4 from 2016 onward. The authors caution against reading this apparent easing as recovery, noting that the cause of the reduced scale damage remains unidentified and that the healthier appearance of surviving adults in 2023 conceals the absence of any regeneration whatsoever. In 2008, Rota’s habitats held more than 10,000 cycads under 100 centimeters in height per hectare; in 2023, there were none.

Comparisons with neighboring Guam, where the scale arrived in 2003, show strikingly parallel outcomes despite different ecological contexts. Guam’s population suffered 96 percent total mortality by 2020, with adult mortality of 87 percent, closely matching Rota’s figures. The main difference was demographic: Rota’s higher density of seedlings and juveniles produced faster initial population-level losses. The timing of a second invader, the cycad blue butterfly Luthrodes pandava, which was already established on Rota before the scale arrived but reached Guam two years after the scale, appears to have made little difference, indicating that the timing of the A. yasumatsui invasion alone governs plant mortality. The consistency across islands demonstrates that the scale is the dominant driver of decline, overwhelming any buffering effect of genetic diversity, since even sites with high expected heterozygosity suffered extreme mortality.

The genetic findings carry profound implications for conservation. Because most variation is partitioned within populations and several sites harbor unique private alleles, each extirpated habitat represents an irreversible loss of evolutionary potential. Site 7, an isolated western peninsula fragment that was the last to be infested and the only site to reach 100 percent extirpation before 2023, exemplifies the stakes. The authors argue that ex situ seed collections, such as the established conservation program on Tinian, are now essential for any reintroduction, and that provenance integrity must be maintained to preserve Rota’s fine-scale genetic structure. They further identify effective biological control of A. yasumatsui, something repeatedly urged by the IUCN Cycad Specialist Group since 2003 but never formally funded and achieved, as the single most critical intervention. As a foundation species central to island ecosystems and CHamoru cultural traditions, the selective removal of C. micronesica threatens cascading consequences for dependent insects, fruit bats, and human communities alike, making this study both a eulogy and a roadmap for what remains.

Subject of Research: Population genetics and demographic collapse of the cycad Cycas micronesica following invasion by the armored scale Aulacaspis yasumatsui on Rota Island

Article Title: Genetics, survival, and demographic decline of Cycas micronesica due to invasive insect species on an oceanic island

Article References: Marler, T. E., Calonje, M., Rosiles-Loeza, P. Y., Leong, R. N. F., & Cibrián-Jaramillo, A. (2026). Genetics, survival, and demographic decline of Cycas micronesica due to invasive insect species on an oceanic island. Discover Conservation, 3(1), Article 19. https://doi.org/10.1007/s44353-026-00089-9

Image Credits: AI Generated

DOI: 10.1007/s44353-026-00089-9

Keywords: Cycas micronesica, Aulacaspis yasumatsui, cycad conservation, invasive species, population genetics, Rota Island, demography, seedling mortality, microsatellites, biological control, island ecology, extinction risk