Floristic homogenization of South Pacific islands commenced with human arrival

Biotic homogenization describes increasing similarity of species composition among regions, threatening biodiversity and ecosystem resilience. It is often linked to anthropogenic drivers such as habitat loss, introductions of non-native taxa, and extirpations/extinctions of native/endemic species. Islands, with high endemism and sensitivity to human-mediated changes, are key systems to study these dynamics. However, the timing of floristic homogenization and pre-human baseline patterns are poorly known because most studies rely on recent species lists. This study investigates when floristic homogenization began in Remote Oceania by analyzing fossil pollen records spanning 5,000 years from 15 sites on 13 South Pacific islands, quantifying temporal changes in floristic similarity and assessing the role of human settlement (two broad migrations ~3,300–2,700 and ~1,000–700 years ago) alongside natural drivers. The core research questions are: (1) Did floristic similarity among islands change over the last 5,000 years? (2) Is increased similarity associated with human arrival and activity? (3) Do site-specific factors such as elevation modulate homogenization trends?

Prior work links biotic homogenization to anthropogenic impacts, including urbanization, agriculture, and species introductions, often resulting in widespread generalists replacing narrowly distributed natives. Island ecosystems, rich in endemics, are especially vulnerable, yet assessments are incomplete (e.g., IUCN coverage ~3% for Pacific island plants). Traditional baselines derived from historical species lists cover only recent centuries and miss undocumented introductions/extirpations; palaeoecological records extend baselines and capture long-term dynamics. Studies in other regions using palaeo-data have documented homogenization in response to disturbances such as eutrophication and land-use change. For Remote Oceania, archaeological, linguistic, and genetic evidence outlines settlement history (Lapita, then Polynesian expansions). Palaeoecological studies on Pacific islands have documented increases in non-native taxa within the last millennium, human-induced extinctions/declines (e.g., palms, bird-mediated seed dispersers), and disturbance from burning and land-use, all potentially contributing to homogenization. These works motivate a long-term, standardized, multi-site analysis of floristic similarity trajectories in the South Pacific.

Study area comprised 15 pollen records from 13 islands in tropical, sub-tropical, and warm-temperate Southwest Pacific Remote Oceania, spanning a west–east gradient of ~8,300 km and an elevational range from 0 to 760 m a.s.l. Islands include continental-origin sites west of the Andesite Line (n=12) and strictly oceanic islands to the east (n=3). Site selection criteria: coverage of early (~3,000 cal years BP) and late (~700 cal years BP) human colonization phases, records ≥3,300 years to compare pre- and post-settlement assemblages, and elevation diversity to contrast coastal versus upland trends. Data acquisition: five datasets accessed via Neotoma (neotoma2 R package) and others from published sources; one record (Rano Aroi, Rapa Nui) digitized using ImageJ. Age-depth: several records already calibrated; remaining models recalibrated with SHCal20 using rbacon; one anomalous date in Volivoli excluded due to old-carbon shell. Taxonomic standardization: pollen/spore names matched to Plants of the World Online using taxize. Two harmonization approaches were applied: standardization-1 (preserves lowest identification level; synonyms updated; ambiguous names aggregated to one level; removes indeterminate/uncertain above family), yielding 383 taxa (81 families, 253 genera, 49 species); and standardization-2 (aggregates to higher level where appropriate), yielding 361 taxa (102 families, 218 genera, 41 species). Data processing: original pollen counts rarefied (where counts available) using vegan to enable temporal and cross-site comparisons; three percentage-only records could not be rarefied. Samples were binned into ten 500-year intervals over the last 5,000 years; within each bin, samples were averaged per site to create mean assemblages, producing 132 averaged assemblages. Percentages used for subsequent analysis. Pairwise similarity: All 15 sites were uniquely paired (105 pairs) excluding within-site comparisons, across each time bin, producing 810 pairwise comparisons. Bray–Curtis dissimilarity (0–1) computed and inverted to similarity (1=most similar). Temporal trends assessed using linear models estimating slope coefficients for each site-to-site comparison over time; negative slopes (<0) denote homogenization, positive (>0) denote differentiation; each slope from ≥4 data points. Human settlement effects evaluated by categorizing each pairwise observation as neither, one, or both islands settled in that time interval. Overall temporal trends visualized with nonparametric smoothing splines (npreg ss) for both standardization schemes. Sensitivity analyses: (1) comparisons within same taxonomic ranks, (2) exclusion of low-count samples (<300), and (3) exclusion of abundant taxa (Cyperaceae, Poaceae). Data and code are publicly available on GitHub and figshare.

• Across 105 site-to-site slope coefficients (standardization-1), a majority indicated homogenization for 13 of 15 sites. Sites showing most differentiation included Lake Tagimaucia (Taveuni), Waitetoke (Ahuahu), Tukou Marsh (Rapa), St. Louis Lac (Grande Terre), Bonatoa Bog (Viti Levu), and Lake Lanoto'o (Upolu). Sites showing strongest homogenization: Yacata (Yacata Island), Avai'o'vuna Swamp (Pangaimotu), Ngofe Marsh ('Uta Vava'u), and Finemui Swamp (Ha'afeva).
• Median slope coefficients across all pairings per site were negative (homogenizing) for all sites except Lake Tagimaucia, indicating most sites became more similar to others over time; Yacata had the lowest median slope (strongest homogenization).
• Temporal increase in similarity over 5,000 years: smoothing splines showed Bray–Curtis similarity rising from ~0.07 (standardization-1) and 0.08 (standardization-2) at 4,900–4,400 cal years BP to ~0.15 and 0.19, respectively, at 400 cal years BP–present.
• Human settlement association: median similarity for pairwise comparisons was 0.04 when neither island was settled, 0.04 when one was settled, and 0.08 when both were settled, indicating increased floristic homogeneity when both islands were occupied by people.
• Elevation patterns: higher-elevation sites (e.g., Lake Tagimaucia, 680 m; Lake Lanoto'o, 760 m) tended to be less homogenized, whereas several low-elevation coastal sites (Avai'o'vuna, 0 m; Ngofe, 4 m; Finemui, 7 m) showed strong homogenization; Waitetoke (1 m) was an exception, likely influenced by proximity to New Zealand’s North Island and ongoing propagule input.
• Abundant taxa dynamics: Before human colonization (prior to ~3,000 and ~700 cal years BP), Euphorbiaceae was most abundant (mean 7% of pollen sum). After colonization, Cyperaceae (17%) and Poaceae (10%) dominated, consistent with transitions from forests to more open vegetation, likely driven by anthropogenic burning. Small (<1%) increases in Arecaceae (possibly Cocos nucifera) and Casuarinaceae (possibly Casuarina equisetifolia) were noted post-settlement.
• Sensitivity checks indicated patterns were robust to excluding low-count samples and abundant taxa, and to alternative taxonomic standardization.

The analyses reveal a long-term, 5,000-year trajectory toward floristic homogenization across South Pacific islands, addressing the key question of timing: homogenization was already underway well before modern globalization, coinciding with human arrival. Pairwise comparisons show that similarity increases notably when both islands are simultaneously settled, implicating anthropogenic drivers—burning associated with agriculture, introductions of non-native taxa, and extirpations/extinctions of native/endemic plants and animal dispersers—in driving convergence of floras. Non-anthropogenic processes (sea-level changes, wetland succession, volcanism, cyclones/droughts, natural long-distance dispersal) likely influenced local composition but did not produce detectable large-scale homogenization in this dataset. Changes in dominant pollen taxa (from woody Euphorbiaceae-dominated to Cyperaceae/Poaceae-dominated assemblages) are consistent with human-induced opening of vegetation. Losses of birds and bats—key seed dispersers—along with rodent seed predation may have reduced recruitment of large-seeded trees on multiple islands, further aligning community composition across islands. Site characteristics modulate these outcomes: high-elevation and topographically complex sites experienced less homogenization, whereas low-elevation coastal sites were more exposed to human land-use. Proximity to mainland sources (e.g., Ahuahu near New Zealand) may maintain distinctiveness via continual propagule input. Overall, the results emphasize the primacy of human settlement and subsequent land-use in shaping long-term inter-island floristic similarity.

South Pacific island vegetation has become increasingly homogeneous over the last 5,000 years. Contrary to narratives attributing homogenization primarily to recent commerce and globalization, this study identifies initial human settlement as a major early driver. Future trajectories of floristic similarity will depend on the extent of ongoing ecosystem modification, rates of non-native introductions, and extinction/extirpation dynamics. The work underscores the value of long, standardized palaeoecological records integrated with contemporary ecological observations to inform conservation and management of island ecosystems.

• Taxonomic resolution limits of pollen identification necessitated two standardization approaches; although results were consistent, some uncertainty remains in taxon assignments and rank aggregations.
• Three records were available only as percentage data and could not be rarefied, potentially affecting comparability despite Bray–Curtis’s emphasis on abundances.
• Bray–Curtis similarity is sensitive to abundant taxa; although sensitivity analyses excluding Cyperaceae and Poaceae were performed, dominance patterns may still influence trends.
• Differences in sample sizes across time and sites could affect detection of rare taxa; however, analyses excluding low-count samples and using Bray–Curtis (which down-weights rare taxa) suggest large-scale patterns are robust.
• Spatial coverage is limited to 15 sites on 13 islands; while spanning an 8,300 km gradient, results may not capture all island contexts or microhabitats.
• Age-depth uncertainties and binning into 500-year intervals may smooth short-term dynamics around settlement events.