Ecological communities surrounding artificial reefs in tropical waters exhibit distinct, group-specific spatial transitions along continuous distance gradients rather than a single uniform boundary. In an investigation conducted at the Wuzhizhou Island National Marine Ranching Demonstration Zone in Sanya, Hainan Province, China, researchers applied generalized additive models to fishery survey records to quantify how biological community indices shifted as distance increased from an artificial reef complex. Across multiple ecological groupings, statistical response curves identified transition response distances spanning from 830 meters to 1560 meters, demonstrating that the spatial footprint of artificial structures varies substantially depending on the ecological habits, mobility, and substrate preferences of marine organisms. The findings provide quantitative reference metrics for evaluating habitat enhancement projects while establishing that community structure does not respond along a single universal spatial threshold.
Artificial reefs are deployed globally to restore degraded habitats, improve structural complexity, and support marine ranching initiatives, particularly across coastal ecosystems in the northern South China Sea where natural coral formations and engineered structures coexist. At Wuzhizhou Island, artificial reef deployment, stock enhancement, and coral restoration have established an ecologically heterogeneous environment across an initial marine ranch area of approximately 66.75 hectares. The demonstration zone encompasses eastern and western artificial reef zones of 33.375 hectares each, positioned in water depths ranging between 10 and 30 meters. Although previous regional surveys documented high fish diversity and marked spatial variation across benthic habitats, the precise geographic distance over which catch-based community composition transitions outward from the reef had remained largely unquantified, creating operational uncertainty for ecological assessment and marine spatial planning.
To capture biological patterns across diverse aquatic zones, the research team synthesized catch records from five distinct survey cruises conducted between 2019 and 2021 across 28 sampling stations. Field sampling incorporated five distinct collection gears: angling, gill nets, traps, diving sampling, and bottom trawling. In 2019, nine bottom-trawling stations were monitored across April, August, and December, with three different stations sampled during each campaign. In June 2020, four diving sampling stations were completed within the reef complex. In April 2021, surveys deployed four angling stations, four gill net stations, four trap stations, and three bottom-trawling stations. For spatial reference, the multi-phase deployments were treated as a single unified artificial reef complex, with the shortest horizontal distance calculated in geographic information software from each sampling station to the outer perimeter, designating the reef boundary as zero meters.
Biological sampling collected a total of 263 catch species over the three-year investigation, comprising 231 fish species, 13 echinoderm species, eight cephalopod species, seven crustacean species, and four gastropod species, with seven taxa identified to the genus level. Dominant species were determined within each capture method using the Index of Relative Importance, which integrates relative numerical abundance, relative biomass, and percentage occurrence frequency, classifying species exceeding a score of 1000 as dominant. Bottom trawling yielded 206 species but only a single dominant taxon, the cephalopod Sepiella maindroni. Angling yielded 30 species, dominated by Scolopsis ciliata, Thalassoma lunare, and Cephalopholis boenak. Traps captured nine species, led by Ostorhinchus fleurieu and Cephalopholis boenak, while gill nets yielded nine species dominated by Abudefduf sexfasciatus. Diver collections yielded 26 species of benthic invertebrates, dominated by Metapenaeopsis barbata, Holothuria atra, Oratosquilla species, and Stichopus chloronotus, with species overlap analysis showing no common taxa shared across all five gear types.
Catches were classified into five non-overlapping analytical categories based on adult habitat descriptions compiled from taxonomic databases and peer-reviewed literature. Reef-dependent organisms were defined as species whose adult stages inhabit coral reefs, rocky structures, or reef crevices without regular occupancy of soft-bottom environments. Reef-approaching species included taxa associated with reefs or margins that also utilize adjacent sand, mud, or seagrass beds. Sand-mud bottom-preference species comprised taxa associated primarily with unconsolidated sedimentary floors showing weak reef affinity. Highly mobile pelagic or nektonic taxa lacking benthic substrate dependency formed the strong-swimming group, while a total-catches pool incorporated all recorded organisms. At each station, researchers calculated the Shannon-Wiener diversity index, Margalef richness index, Pielou evenness index, and summed Index of Relative Importance values to characterize taxonomic richness, abundance distribution, and structural dominance for each ecological category.
Statistical relationships were evaluated by fitting fifteen Gaussian generalized additive models with identity links using cubic regression splines to model reef distance as a continuous explanatory variable without assuming rigid linear trajectories. Shannon-Wiener diversity and Margalef richness values were log10-transformed prior to modeling, whereas Pielou evenness and Relative Importance scores were analyzed on their original measurement scales. Prior screening with variance inflation factors confirmed that continuous candidate variables avoided problematic multicollinearity, maintaining factor scores below five. For models demonstrating statistically significant distance smooths, researchers identified the model-derived transition response distance by calculating the first zero-crossing where fitted smooth values changed sign along the outward distance gradient, interpolating linearly between adjacent prediction intervals to pinpoint the primary inflection point without subjective selection.
The generalized additive models revealed marked differences in spatial transitions across ecological groupings. For the total-catches assembly, all three diversity metrics showed significant relationships with reef distance: Margalef richness exhibited a transition distance of 830 meters, Pielou evenness shifted at 840 meters, and Shannon-Wiener diversity transitioned at 1120 meters, yielding an overall response band spanning 830 to 1120 meters. Across this gradient, species richness and diversity increased with distance, while evenness decreased. For reef-dependent organisms, diversity and richness indices showed nonsignificant relationships, but the Index of Relative Importance declined significantly with increasing distance, establishing a solitary statistical transition response distance of 890 meters. This pattern indicated that dominant reef-dependent species concentrated tightly near physical structures rather than undergoing broader community-level structural shifts across the wider survey domain.
Broader spatial distributions were recorded for organisms capable of utilizing multiple benthic habitats. For the reef-approaching group, Margalef richness and the Index of Relative Importance yielded significant distance responses with transition distances estimated at 1560 meters and 1420 meters, respectively, spanning a response zone between 1420 and 1560 meters. Organisms exhibiting sand-mud bottom preferences demonstrated significant positive increases across three metrics, with transition distances calculated at 1100 meters for Margalef richness, 1280 meters for the Index of Relative Importance, and 1450 meters for Shannon-Wiener diversity, establishing an overall group span of 1100 to 1450 meters. Conversely, the strong-swimming group showed nonsignificant distance smooths across all evaluated indices, with diversity, evenness, and richness yielding probability values of 0.707, 0.271, and 0.064, precluding the estimation of an operational response distance for highly mobile pelagic species.
Post-hoc consistency assessments dividing stations into spatial cohorts situated inside and beyond the derived transition distances supported nine of fifteen comparisons with statistically significant differences and large Cliff’s delta effect sizes. For total catches, Margalef richness was significantly higher beyond 830 meters, and Shannon-Wiener diversity was greater beyond 1120 meters, whereas Pielou evenness was higher within 840 meters. For reef-dependent catches, Relative Importance was significantly elevated within 890 meters with an effect size of 0.989. For reef-approaching species, Margalef richness was higher beyond 1560 meters, while Relative Importance was greater within 1420 meters. Sand-mud preference taxa showed significant increases beyond their respective splits of 1100, 1280, and 1450 meters. These non-parametric group comparisons corroborated the directional shifts captured by the non-linear regression curves while confirming that within-group contrasts align with the model estimates.
The study authors emphasize several critical methodological constraints that dictate how these statistical thresholds must be interpreted by marine managers and researchers. Gear selectivity and uneven spatial allocation between gears represented a major limitation; when supplementary models included fishing method as a categorical covariate, statistical support for distance effects declined, indicating sensitivity to sampling heterogeneity. Furthermore, reef stations were assigned zero-meter reference distances, leaving an unsampled spatial gap between 0 and 1419 meters that required smooth curves to interpolate across intermediate intervals. In addition, the strong-swimming group was collected at only 13 stations, including 12 distant trawl stations and a single zero-meter gill-net station. Consequently, the authors conclude that these transition response distances represent site-specific descriptive statistical patterns within a pooled multi-year dataset, rather than sampling-independent ecological radii or rigid boundaries for marine ranch zoning.
Subject of Research: Spatial response distances of reef-associated catch communities around tropical artificial reefs
Article Title: Estimating spatial response distances of reef-associated communities in the Wuzhizhou Island marine ranch using generalized additive models
Article References: Yang, C., Zhang, X., Li, J., Zhang, Z., Guo, Y., Xu, Q., Wang, H., & Yu, H. (2026). Estimating spatial response distances of reef-associated communities in the Wuzhizhou Island marine ranch using generalized additive models. Frontiers in Ecology and Evolution, 14, 1983890. https://doi.org/10.3389/fevo.2026.1983890
Image Credits: AI Generated
DOI: 10.3389/fevo.2026.1983890
Keywords: artificial reefs, marine ranching, Wuzhizhou Island, generalized additive models, spatial response distance, community structure, Index of Relative Importance, Shannon-Wiener diversity, Margalef richness, Pielou evenness, reef-dependent species, reef-approaching species, sand-mud bottom preference, strong-swimming species, South China Sea, fishery resources, gear selectivity
News Source: Drew Townsend. (November 12, 2026). Statistical Models Map Marine Community Transitions Around Tropical Artificial Reefs. Scienmag.



