A multiscale assessment of general resilience capacity in China’s coastal social-ecological systems: a national baseline and nested diagnosis in Rongcheng

1 Introduction

Coastal zones are among the most intensively coupled interfaces among terrestrial, marine, and human systems. Riverine transport, tidal exchange, sediment movement, and biological migration interact in these areas, while port transportation, mariculture, coastal tourism, urban expansion, and ecological conservation compete within a limited shoreline space. Wetlands, shallow seas, estuaries, bays, and islands continuously provide primary production, shoreline protection, carbon sequestration, water purification, nursery habitat, and cultural services, while simultaneously experiencing sea-level rise, extreme weather, nutrient inputs, shoreline hardening, habitat fragmentation, and increasing resource-use intensity (Holling, 1973). Unlike a single hazardous event, these pressures combine abrupt and gradual change, interact across land- and sea-based sources, and amplify one another through ecological and social feedbacks. Their consequences can propagate across administrative boundaries, the land-sea interface, and sectoral jurisdictions (Walker et al., 2004). The central challenge for coastal governance is therefore not limited to forecasting a particular hazard. It also involves determining whether a social-ecological system can retain its essential structures, functions, and feedbacks when disturbance types are uncertain, scales of influence change, and governance objectives are adjusted, and whether learning, reorganization, and transformation can create new adaptive pathways when necessary (Folke et al., 2010; Carpenter et al., 2012).

Resilience provides a comprehensive perspective for understanding system persistence under compound pressures, but a valid assessment must define exactly what is being measured. Specified resilience concerns exposure, sensitivity, and recovery processes in relation to an identified disturbance. General resilience does not presuppose a single hazard; instead, it focuses on the conditions that preserve diversified responses, functional backup, learning, collaboration, and feedback-based governance (Holling, 1973; Walker et al., 2004; Folke et al., 2010; Carpenter et al., 2012; Biggs et al., 2012). Accordingly, this study evaluates general resilience capacity in coastal systems and does not interpret the composite score as observed recovery after a particular event. The conceptual distinction and measurement boundary between general resilience and specified resilience are summarized in Figure 1.

Previous studies have used composite indicators, network structures, scenario simulation, and spatial statistics to measure social-ecological resilience (Zhang et al., 2023; Wu et al., 2024). From the perspective of sustaining ecosystem services, Biggs et al. (2012) proposed seven principles: maintaining diversity and redundancy, managing connectivity, managing slow variables and feedbacks, fostering complex adaptive systems thinking, encouraging learning and experimentation, broadening participation, and promoting polycentric governance. Carpenter et al. (2012) summarized a broader set of conditions that support general resilience to extreme and unfamiliar events. Together, these studies demonstrate that general resilience can be operationalized through proxy indicators linked to system function, but indicators must remain grounded in process-based understanding. In large river ecosystems, Bouska et al. (2019) translated diversity and redundancy, connectivity, and controlling variables into ten indicators, showing that the same theoretical principle must be contextualized according to ecosystem structure, data availability, and management scale. Coastal assessments, however, still frequently conflate exposure, vulnerability, recovery rate, and general resilience, or compare land and marine indices directly after constructing different indicator systems. Systematic reviews have found substantial differences in the spatial scales, variable selection, and assessment purposes adopted in research on coastal social-ecological resilience, with participatory integration and dynamic updating remaining insufficient (Ferro-Azcona et al., 2019; Li et al., 2020). Indicator-based assessments are also sensitive to standardization, weighting, classification thresholds, and data gaps. Unless indicator sources, observation years, directions, weights, and spatial supports are reported, a visually clear map may still have an irreproducible measurement meaning.

Scale is another central constraint on coastal resilience assessment. Large-scale analyses can identify regional gradients using harmonized data and provide baselines for interregional resource allocation and monitoring priorities (Zhang et al., 2020; Wang et al., 2023). However, the requirement for continuous coverage and consistent definitions tends to generalize local habitats, aquaculture structures, nutrient variability, land-sea connectivity barriers, and governance feedbacks. Small-scale analyses preserve patch boundaries, continuous gradients, and localized discontinuities, but their locally adapted indicators, sampling support, and boundaries may no longer be directly commensurable with a macro-scale index. Scale mismatch in social-ecological systems prevents ecological processes, management responsibilities, and information provision from aligning at the same level (Cumming et al., 2006; Cash et al., 2006). Spatial aggregation also produces a modifiable areal unit problem, through which means, variances, and associations change with unit size or zoning (Jelinski and Wu, 1996). A multiscale ecosystem-service study in the Taihu Lake Basin identified different relationships and dominant factors at 12- and 83-km scales, indicating that local differences in human activities can be smoothed by coarser aggregation (Liu et al., 2017). A multiscale study should therefore not be framed as the simple magnification of one map. Instead, it should specify the question addressed at each scale, the shared theoretical anchors, and the measurement boundaries that prevent direct comparison.

A national assessment can establish a spatial baseline of general resilience capacity for the coastal zone of mainland China under consistent data definitions. Yet nationwide continuous datasets necessarily generalize habitat fragmentation, aquaculture occupation, nutrient fluctuations, and interrupted land-sea connectivity within bays, estuaries, and intensively used shorelines. National results are therefore suitable for identifying broad spatial distributions and priority regions but are insufficient for directly defining local monitoring units or spatial-use controls.

Rongcheng was selected as a high-resolution nested window. Its convoluted shoreline, numerous bays, and intensive aquaculture cause ecological processes, production activities, and governance feedbacks to overlap across short shoreline segments (Weihai Municipal Civil Affairs Bureau, 2016; Xu et al., 2023; Tu et al., 2022). This setting allows local information that cannot be represented consistently in national datasets to be added under a shared theoretical framework. The Rongcheng assessment is intended to reveal internal local differences; it neither validates the accuracy of the national index nor compares local scores against national scores.

This study develops a multiscale assessment framework connecting national and Rongcheng analyses and coastal land and marine subsystems through general resilience principles. Its objectives are to: (1) define the measurement boundary between general resilience and specified resilience; (2) report the directions, observation years, sources, spatial resolutions, and scale-specific weights of the national core indicators and the locally enhanced Rongcheng indicators; (3) characterize continuous distributions, five-class structures, and spatial clustering within four subsystems; and (4) use classification sensitivity and explicit evidentiary boundaries to align macro-scale screening, local diagnosis, and governance translation with the level of support provided by the data.

2 Theoretical and conceptual framework

2.1 Theoretical origins of general resilience and its distinction from specified resilience

General resilience does not originate from a fixed indicator checklist proposed directly by Holling. Its theoretical basis emerged progressively from research on ecological resilience, complex adaptive systems, and social-ecological governance. Holling (1973) distinguished ecological resilience from stability: stability emphasizes the rate at which a system returns after displacement from equilibrium, whereas ecological resilience concerns the magnitude of disturbance that a system can absorb before its controlling structures change and it crosses a state threshold. This shift replaced the assumption that systems fluctuate around a single equilibrium with an understanding that multiple stable states, nonlinear feedbacks, and abrupt transitions are central concerns for ecological management. Walker et al. (2004) further argued that the future trajectories of social-ecological systems depend jointly on resilience, adaptive capacity, and transformability. Resilience does not imply preserving every existing condition; rather, it involves sustaining socially valued functions, structures, identity, and feedbacks. Adaptive capacity describes the ability of actors to adjust the system, whereas transformability is the capacity to create a new development pathway when the existing system becomes untenable. Folke et al. (2010), together with Levin et al. (2013), situated these concepts within complex adaptive systems, emphasizing self-organization, learning, cross-scale interactions, and uncertainty. Levin et al. (2013) directly supports the complex-adaptive-system framing and its policy implications. General resilience is therefore theoretically defensible not because any single variable predicts every disturbance, but because multiple system attributes repeatedly identified in theory and empirical research describe the conditions that preserve response options, constrain the spread of disturbance, detect change, and organize action. Carpenter et al. (2012) explicitly differentiated resilience to specified risks from general resilience to unfamiliar, exceptional, and extreme shocks, identifying diversity, modularity, openness, reserves, feedbacks, nesting, monitoring, leadership, and trust as supporting conditions. Through a literature synthesis, Biggs et al. (2012) proposed seven principles for enhancing the resilience of ecosystem services and provided a systematic bridge from theoretical attributes to management practice. The three-tier, ten-attribute framework used here reorganizes this theoretical lineage for coupled coastal land-sea systems and data operability; it is not presented as a universally accepted set of ten principles.

Specified resilience requires a clear statement of which component of which system is resilient to which disturbance. Its measurement typically centers on hazard intensity, exposure, sensitivity, loss trajectories, and recovery time and is appropriate for storm surges, sea-level rise, harmful algal blooms, or pollution incidents. General resilience does not prespecify a single disturbance; it estimates the conditions that enable a system to cope with a broad range of changes. The two concepts share some resource and governance foundations but differ in measurement logic. For example, greater seawall height may substantially increase specified resilience to a storm surge but does not necessarily strengthen the capacity to address nutrient accumulation, industrial volatility, or ecological threshold shifts. Conversely, habitat diversity, functional redundancy, and learning networks may not lower the water level during a particular storm, but they can provide alternative pathways, information, and coordination capacity. This study therefore excludes direct measures of hazard exposure from the general resilience capacity index and does not interpret a composite score as observed post-disturbance recovery performance. Chronic pressures such as aquaculture intensity in Rongcheng are theoretically better overlaid as a separate pressure layer to identify diagnostic combinations such as “high pressure–low capacity.” The resulting three-tier, ten-attribute framework for coastal general resilience is presented in Figure 2, while the definitions and operational meanings of the ten attributes are summarized in Table 1.

Table 1. Three analytical tiers, ten attributes, and operational meanings of coastal general resilience.

2.2 Operationalizing coastal general resilience through a three-tier, ten-attribute framework

Translating theoretical principles into spatial indicators requires an explicit bridge between conceptual completeness and data observability. Drawing on diversity and redundancy, connectivity, slow variables and feedbacks, learning and experimentation, participation, and polycentric governance (Carpenter et al., 2012; Biggs et al., 2012; Bouska et al., 2019), and incorporating coastal ecosystem-service provision, land-sea connectivity, spatial zoning, and cross-sectoral management, we organized the assessment into three analytical tiers: system attributes, dynamic adaptation, and governance coordination. These tiers are not independent causal modules. They are an organizational device that prevents the indicator system from being dominated by a single type of data. System attributes describe present structure and long-term ecological constraints; dynamic adaptation describes the availability of exchange, substitution, and learning pathways; and governance coordination describes the institutional conditions through which disturbances are contained, information is fed back, and responsibilities are aligned across levels. The ten attributes are diversity, slow variables, ecosystem services, openness, redundancy, innovation capacity, social capital, modularity, overlapping governance, and tight feedbacks. This organization separates “diversity and redundancy” in Biggs et al. (2012) to distinguish the richness of response types from functional backup; translates “managing connectivity” into openness, interpreted jointly with modularity as a balance between connection and isolation; operationalizes learning and experimentation through innovation capacity and feedback; and translates participation and polycentric governance into social capital and overlapping governance. It thus preserves the theoretical origins of the framework while permitting national and local indicators to represent the same functions at scale-appropriate levels.

Within the system-attributes tier, diversity describes the extent to which species, functions, habitats, or land and sea uses provide differentiated responses. Where redundancy is insufficient, damage to a single functional group can rapidly translate into system-wide functional loss. Slow variables are conditions such as soil properties, sedimentary environment, nutrient status, and shoreline morphology that change gradually but can alter state thresholds and feedback structures. Ecosystem services concern whether the system continues to provide primary production, regulation, shoreline protection, water purification, and nursery functions. Together, these attributes determine the ecological capital available to absorb disturbance, but higher values are not automatically better. High primary production, for example, may reflect a healthy food web or eutrophication; indicator direction must therefore be determined against ecologically appropriate ranges. At the national scale, broad-coverage proxies include habitat diversity, natural cover, soil organic carbon, wetland retention, the normalized difference vegetation index (NDVI), net primary productivity (NPP), and marine environmental stability. In Rongcheng, 20-m habitat patches, biodiversity, nutrient variability, aquaculture duration, and benthic habitat quality provide enhanced representations of local processes.

The dynamic-adaptation tier includes openness, redundancy, innovation capacity, and social capital. Openness refers to the cross-boundary exchange of matter, energy, organisms, and information, but its theoretical meaning is appropriate connectivity rather than unlimited openness. Insufficient connectivity impedes replenishment, migration, and information flow; excessive connectivity can accelerate the propagation of pollution, disease, or other disturbances (Bouska et al., 2019). Indicators with a defensible ecological optimum are therefore standardized according to their distance from an optimal range instead of being assigned a mechanically monotonic positive or negative direction. Pressure proxies for which increasing development is consistently detrimental are treated as negative indicators. Redundancy represents the capacity of different structures or actors to substitute for the same critical function; innovation capacity reflects technological, institutional, and resource-allocation learning; and social capital reflects the support provided by cooperative networks, public participation, mutual assistance, and trust for collective action. National indicators emphasize functional redundancy in land cover, livelihood diversification, research and development or technical services, public services, and fisheries organizations. Local indicators in Rongcheng additionally represent integrated multi-trophic aquaculture, low-disturbance technologies, village participation, cooperative networks, and insurance and mutual-aid arrangements.

The governance-coordination tier comprises modularity, overlapping governance, and tight feedbacks. Modularity is not synonymous with spatial fragmentation. It is the capacity to restrict the spread of localized shocks through relatively independent functional units while retaining necessary exchange. Land habitat networks and marine aquaculture units can support both replenishment and risk isolation only when they achieve an appropriate balance between connectivity and separation. Overlapping governance describes complementary and backup responsibilities among departments, administrative levels, protected areas, functional zones, and restoration projects rather than the simple spatial superposition of management designations. Formal overlap without information sharing or clear responsibilities may increase coordination costs. Tight feedbacks require short links among monitoring, early warning, decision-making, implementation, and reassessment so that changes in slow variables and local anomalies can promptly inform governance. National proxies include monitoring density, early-warning coverage, wastewater treatment, and overlaps among protection and restoration areas. Local indicators in Rongcheng incorporate water-quality sampling frequency, environmental-incident response time, aquaculture planning, and marine-ranching programs.

2.3 Indicator translation, direction functions, and measurement boundaries

Indicators were selected according to five constraints: they had to represent general resilience capacity rather than exposure to a single hazard; their direction or optimal range required an ecological or governance rationale; they needed adequate coverage and reproducible processing at the target scale; land and marine subsystems could use different observed variables only when their functional meanings were consistent; and semantically redundant indicators had to be excluded to prevent double weighting. These constraints are consistent with established guidance on the evaluation of ecological indicators (U.S. Environmental Protection Agency, 2000). This location links the reference directly to indicator selection and documentation. For each indicator, we recorded the name, variable, unit, direction, observation year, original resolution, data source, preprocessing, corresponding attribute, and weight. After min-max normalization, higher values of positive indicators represent greater capacity, whereas negative indicators are inversely standardized. Indicators with an optimal range use a piecewise distance-to-optimum function based on planning standards, historically stable ranges, or monitoring references: observations within the optimum range receive high scores, and scores decline as deviations increase.

The composite score is defined as a general resilience capacity index that represents the relative coping potential provided by system structure, function, and governance at the observation time. It describes spatial differences but is not a predicted probability of future functional persistence, an observed recovery rate, or a direct estimate of the causal contribution of any single indicator. Predictive validity will require multiple periods of system-state observations, trajectories of function before and after disturbances, and independent measures of management performance.

2.4 Multiscale conceptual linkage and propositions for the nested diagnosis in Rongcheng

General resilience is inherently scale dependent. Species and habitat diversity may be most evident at patch and intra-bay scales, whereas slow variables, resource flows, and polycentric governance operate across broader spatial and temporal extents. Aggregating the same process into administrative units or coarse grids changes its variance, boundaries, and class proportions. Cash et al. (2006) framed scale problems in terms of cross-scale interactions, scale mismatches, and divergent perceptions of scale among actors. Cumming et al. (2006) showed that system function may be compromised when the scale of ecological processes does not match that of management institutions. We therefore link scales through “shared theoretical principles–scale-appropriate indicators–within-scale interpretation,” without assuming that the national and Rongcheng indices have identical measurement scales. Shared principles ensure that the same general construct is evaluated; indicator mapping specifies functional correspondence; and within-scale statistics prevent unlike spatial supports from being treated as paired samples.

Multiscale linkage depends on shared theoretical attributes and explicit indicator mapping, not on identical variables, weights, or spatial units. National results are interpreted for macro-scale shoreline units and nearshore grids, whereas Rongcheng results are interpreted for fine-resolution local land and marine grids. The nested relationship describes how a macro-scale signal can trigger local investigation and how new local information enriches attribute-level diagnosis. It does not create a basis for direct score comparison.

Three analytical propositions follow. First, national-scale assessment can identify the broad context of general resilience capacity but cannot replace diagnosis within bays and shoreline segments. Second, fine-scale assessment can reveal spatial differences smoothed by aggregation, but the heterogeneity of each subsystem depends on its actual ecological processes. Third, when indicators, weights, and spatial supports differ, conclusions must remain within their respective scales and be connected through governance processes rather than score comparison.

3 Materials and methods

3.1 Multiscale research design

The analytical workflow comprised theoretical operationalization, scale-specific indicator construction, data quality control, composite assessment, five-class classification, spatial statistics, and governance translation (Figure 3). National coastal land and marine subsystems emphasized consistent coverage and regional patterns, whereas the corresponding Rongcheng subsystems emphasized differences among bays, shoreline segments, and habitat patches. Standardization and statistical interpretation were conducted separately for the four subsystems, while the national land and marine domains shared one ten-attribute weight vector and the Rongcheng land and marine domains shared another. The nested synthesis traced the progression of research questions and information; it did not calculate score differences or rankings between the national and Rongcheng assessments.

The framework excludes the intensity or inundation extent of specific hazards and observed recovery time. Instead, it follows a shared theoretical logic of response diversity, slow-variable stability, functional redundancy, appropriate connectivity, learning and collaboration, and feedback-based governance. Both the national and Rongcheng analyses therefore assess general resilience capacity rather than hazard zonation or disaster-recovery performance (Cui et al., 2025).

The nested design is not a cropped or magnified version of the national map. It changes spatial support and information density while retaining the same theoretical anchors. National coastal land was represented by named administrative units, and national coastal waters by cells of approximately 8.09 km. Rongcheng land and marine subsystems used cells of approximately 180.84 and 263.92 m, respectively. Scale transitions can reflect true heterogeneity, spatial aggregation, and indicator substitution simultaneously. We therefore report continuous statistics, five-class composition, and spatial patterns only within each subsystem; cross-scale interpretation is limited to the manner in which information is connected.

The national and Rongcheng analyses were linked through theoretical and data mappings. The theoretical mapping required every locally enhanced indicator to correspond to one of the ten attributes. For example, permeability through longline-raft structures represented openness, nutrient variability represented slow variables, integrated multi-trophic aquaculture represented redundancy, and water-quality sampling frequency and incident-response time represented tight feedbacks. A variable was excluded from the general resilience capacity index if this functional correspondence could not be explained. The data mapping recorded the observed object, direction, unit, year, resolution, and source used for the same attribute at the national and Rongcheng scales. This design allowed different variables to be used at different scales without permitting unbounded indicator proliferation simply because more local data were available. Local information was considered an effective enhancement of the national baseline only when its theoretical meaning was clear, its direction function was interpretable, and its coverage could be verified.

The workflow also distinguished assessment, diagnosis, and governance translation as three levels of evidence. The assessment level produced auditable composite scores. The diagnostic level combined continuous distributions, five-class structure, and spatial clustering to identify units for future field verification. The governance-translation level linked attribute-level evidence with monitoring, restoration, connectivity, and coordination measures. Specific limiting mechanisms were inferred only where local pressure layers, attribute scores, and monitoring feedback were all available.

3.2 National study area

The national study area comprised coastal land units of mainland China and the corresponding nearshore functional waters. The land domain extended approximately from 107.48° to 124.36°E and from 18.17° to 41.44°N, while the marine domain extended from 108.07° to 124.23°E and from 17.98° to 41.07°N (Figure 4). This coast spans tropical, subtropical, and temperate climates and contains deltas, estuaries, bays, rocky coasts, sandy coasts, muddy coasts, and islands. It also includes urban agglomerations, ports, agricultural and aquaculture areas, protected areas, and tourism zones. Substantial variation in environmental background, development intensity, and governance capacity provides the contrasts needed to examine spatial expression of general resilience principles under diverse social-ecological conditions. A national rather than single-province extent also makes it possible to establish a consistent baseline for monitoring and to identify the relative positions of shoreline regions and land and marine subsystems.

The assessment boundary was determined by land-sea interaction integrity, indicator-scale compatibility, and the operational relevance of management units. Coastal county-level administrative units formed the basic national land units. Counties were screened and integrated according to shoreline adjacency, estuarine and catchment connections, land-based material inputs, coastal land use, and socioeconomic influence. This screening yielded 290 county-level land assessment units. County units were preferred to a uniform-width buffer because population, industry, and governance variables are reported administratively and can be linked to ecological processes at that level. The coastal land study area covered approximately 248,300 km². The marine domain extended from the shoreline across intertidal areas, estuaries, bays, tidal flats, and connected shallow seas. Its offshore boundary integrated nearshore hydrodynamics, the dispersion of land-based materials, ecological connectivity, resource use, and functional management, and continuous spatial variation was represented using a uniform grid. The marine assessment area covered approximately 304,700 km². Land and marine indicators were standardized, aggregated, and classified separately. Spatial relationships were interpreted within a shared framework, but absolute scores for county units and marine grids were not directly compared.

At the national scale, land assessment units were the complete boundaries of selected county-level administrative units in the coastal region, rather than a fixed-width landward buffer. This choice matches the reporting scale of socioeconomic and governance data and the jurisdiction of coastal management. The marine assessment covered nearshore functional waters, including estuaries, bays, tidal flats, and adjacent shallow seas, and was represented by grid cells approximately 8.09 km wide. A 100-km cross-shore distance was used to describe, not delineate, marine coverage: it approximates the selected counties’ measured maximum landward depth (96.2 km), and 98.7% of valid marine-cell centers were within 100 km of the land domain. The extent is therefore not a strict 100-km buffer.

3.3 Rongcheng study area and rationale for case selection

Rongcheng is located at the eastern end of the Shandong Peninsula, approximately between 122°08′01″ and 122°42′18″E and between 36°44′48″ and 37°27′09″N, with the sea bordering its northern, eastern, and southern sides (Weihai Municipal Civil Affairs Bureau, 2016). Its temperate monsoon climate is strongly influenced by the ocean, and its convoluted rocky-bay shoreline, approximately 492 km in length, contains bays, headlands, islands, reefs, sandy shores, and tidal flats.

The shoreline was used as the land-sea boundary. The land domain extended 5 km inland and primarily encompassed coastal settlements, agricultural land, woodland, estuarine wetlands, and development areas directly influenced by marine processes. The marine domain extended seaward to the 10-m isobath and included intertidal areas, shallow bays, estuarine waters, nearshore aquaculture areas, and waters surrounding islands (Figure 5). This extent captures both the effects of land-based inputs, land use, and socioeconomic activities on nearshore ecosystems and the areas in which aquaculture, hydrodynamic exchange, and marine ecological processes are concentrated. The final assessment area was approximately 3,304.33 km², including 1,256.27 km² of land and 2,048.06 km² of marine waters.

Rongcheng provides a substantively representative nested observation window. Its extensive kelp and shellfish aquaculture and marine-ranching activities form a tightly coupled production-ecology system. Aquaculture can support nutrient uptake by macroalgae, production, and carbon sequestration, but intensive raft structures, spatial occupation, altered local exchange, and production risks can also impose chronic pressure (Xu et al., 2023; Tu et al., 2022). Research on kelp mortality associated with anomalous coastal currents and extreme precipitation further demonstrates that distant land-based processes can influence local aquaculture ecosystems through marine circulation (Li et al., 2025). These cross-boundary, cross-scale, and multi-pressure interactions are precisely the conditions relevant to general resilience. Rongcheng also has comparatively rich local monitoring, aquaculture records, and high-resolution imagery, allowing nutrient variability, habitat fragmentation, raft permeability, and governance feedbacks to be examined at levels not consistently supported by national datasets.

The national and Rongcheng assessments serve complementary rather than repetitive purposes. The national baseline identifies where broad differences in general resilience capacity occur, whereas the Rongcheng diagnosis identifies which local bays, shoreline segments, or patches warrant investigation. This arrangement retains the strategic reference offered by nationally continuous data while using high-resolution habitat, nutrient, aquaculture, and governance information to define locally actionable units. The two assessments use separate weights and are interpreted independently.

The nested position of Rongcheng also permits asymmetric land and marine processes to be observed within one coastal social-ecological system. Terrestrial habitats, roads, and developed patches typically have relatively discrete boundaries, whereas water-mass exchange, nutrient transport, and the effects of aquaculture infrastructure exhibit greater continuity and directionality. At the national level, the maps depict broad coastal patterns; in Rongcheng, the nested maps show localized land patches and marine gradients within bays. This difference in spatial expression supports the use of local assessment to refine monitoring and verification priorities identified through national screening. Because indicator selection and weights also change, we interpret the contrast as a joint outcome of the assessment designs, rather than as evidence that spatial resolution alone generated the observed differences.

This progressive design also provides an organizational basis for hierarchical monitoring and indicator updating, enabling national rules and local observations to remain connected in subsequent governance cycles.

3.4 Data sources and temporal alignment

The nominal assessment baseline was 2025. Products from 2025 were prioritized for remote-sensing, marine environmental, and socioeconomic indicators. For soil, shoreline, protected-area, and aquaculture-distribution datasets with lower update frequencies, the nearest reliable products available around the baseline year were used, and actual observation years are reported in Table 2. National data included Landsat 8/9 and Sentinel-2 land-cover and habitat information; MODIS NDVI and NPP; SoilGrids and Chinese soil information products retrieved from the GeoCloud Platform of the China Geological Survey; shoreline and protected-area data; MODIS, Visible Infrared Imaging Radiometer Suite (VIIRS), or fused marine environmental products; bathymetry and bay geometry; a nearshore aquaculture dataset retrieved from GeoCloud; coastal county statistics; monitoring stations; and planning and restoration programs. Enhanced Rongcheng data comprised high-resolution imagery, biological monitoring, nutrient sampling, aquaculture records, local projects, and incident-response information.

Table 2. Multisource data and their temporal and spatial supports.

The GeoCloud Platform is maintained by the China Geological Survey as a national geoscience data-sharing service (https://geocloudsso.cgs.gov.cn; accessed August 15, 2026).

Coordinate systems and metadata were checked before spatial data entered the calculations. National remote-sensing and marine data were aggregated using a common equal-area or otherwise appropriate projection. Continuous variables were resampled bilinearly, categorical variables by nearest neighbor, and vector-derived ratios by area or length overlay. Rongcheng data retained their original or otherwise supportable fine resolution, and the actual pixel sizes of final score rasters were reported. County-level social and governance variables participated in weighting only at their true statistical unit and were not replicated across pixels to manufacture pseudo-observations. If downscaling was required for cartographic purposes, the value was treated as constant within the county and documented accordingly.

Temporal consistency was controlled by reporting both the nominal baseline year and the actual observation year. Sea-surface temperature, chlorophyll a, NDVI, and NPP exhibit substantial intra-annual variability and were therefore summarized from annual or monthly sequences rather than a single-day image. Soil, shoreline, protected-area, and aquaculture-distribution products with slower update cycles were treated as background layers nearest to 2025 and were not linearly extrapolated. Clouds, shadows, and anomalous values in multisource imagery were removed according to product-quality flags, while original ranges and NoData codes were retained before resampling continuous rasters. Socioeconomic and governance data were retained at their publication units; data across years were combined only when indicator definitions remained stable and spatial boundaries could be matched. The nominal 2025 baseline should be understood as a common temporal reference rather than a synchronized observation date. Products closest to 2025 were prioritized, whereas earlier datasets were retained for slowly updated variables when they provided the most complete and spatially consistent coverage. Temporal mismatch may obscure changes occurring between observation years and consequently affect standardized scores and class assignments; therefore, the results represent a multisource temporal window centered on 2025 rather than an exact snapshot of conditions in a single year.

3.5 Multiscale indicator systems and weights

The ten operational attributes served as shared theoretical anchors, while scale-appropriate indicators were selected separately for national coastal land, national coastal waters, Rongcheng coastal land, and Rongcheng coastal waters. The national core system prioritized coverage and consistency. The enhanced Rongcheng system incorporated direct biological observations, nutrient variability, aquaculture duration, raft permeability, multi-trophic production, cooperative networks, and incident-response time. Because the two systems differed in observed objects, sample domains, and information dispersion, their weights were estimated separately. Table 3 presents the national core indicators and national attribute weights; Table 4 presents the locally enhanced Rongcheng indicators and weights. A shared attribute denotes consistent theoretical function, not interchangeability of variables or weights across scales.

Table 3. National core indicators and attribute-level weights.

Table 4. Enhanced Rongcheng indicators and attribute-level weights.

At each assessment scale, the ten attribute scores were oriented so that higher values indicated greater general resilience and standardized to the range [0, 1]. Entropy weights were then calculated from the standardized matrix:

pij=zij∑izij

ej=−(1ln n)∑ipijlnpij

dj=1−ej

wj=dj∑jdj

with 0 ln 0 defined as zero. This yielded one weight vector for the national assessment and one for Rongcheng. Within each scale, the same vector was applied to land and marine units, and unit-level scores were aggregated as Ri=∑jwjzij. The processing steps and weight results are provided in the Supplementary Materials. Attribute-level weights for the national and Rongcheng systems are reported in Tables 3 and 4, respectively.

3.6 Indicator standardization and composite calculation

Positive indicators were standardized as z_ij = (x_ij − min x_j)/(max x_j − min x_j), and negative indicators as z_ij = (max x_j − x_ij)/(max x_j − min x_j), such that a higher standardized value consistently represented greater capacity. A monotonic function was not assumed for variables such as openness, development intensity, or nutrient status when a nonlinear optimum was theoretically expected. If the optimal range was [L_j, U_j], then z_ij = 1 within that interval; below L_j, the score increased as (x_ij − min x_j)/(L_j − min x_j); and above U_j, it declined as (max x_j − x_ij)/(max x_j − U_j).

The general resilience capacity index for spatial unit i was calculated as R_i = Σw_jz_ij, where Σw_j = 1. Linear aggregation supports auditing and spatial representation but assumes partial compensability among indicators and relative independence among spatial units. A low score for one indicator can be partly offset by a high score for another, and neither land-sea connectivity nor spatial autocorrelation is represented explicitly in the aggregation function. The index is therefore interpreted as a relative composite measure of general resilience capacity, not as a probability, recovery time, or distance from a system threshold.

3.7 Descriptive statistics, classification, and sensitivity analysis

For each subsystem, we calculated the minimum, first quartile, median, mean, third quartile, maximum, population standard deviation, and coefficient of variation. National coastal land was weighted by geodesic area; the three raster subsystems assigned equal weight to each valid cell. Each subsystem was classified separately using the five-class Jenks natural breaks method, with classes termed low, relatively low, moderate, relatively high, and high. Internal break values were 0.2871, 0.3375, 0.3880, and 0.4439 for national coastal land; 0.3067, 0.4185, 0.4915, and 0.5623 for national coastal waters; 0.5064, 0.5742, 0.6420, and 0.7514 for Rongcheng coastal land; and 0.3356, 0.4731, 0.5996, and 0.7262 for Rongcheng coastal waters. These classes indicate only relative positions within the corresponding subsystem and are not common thresholds for cross-scale or land-marine comparison.

Classification sensitivity was evaluated for national coastal land and marine subsystems by comparing five-class Jenks natural breaks, five equal intervals over the observed range, and quintiles. Natural breaks preserve internal data clustering, quintiles emphasize relative ranking, and equal intervals represent positions within the observed range. The maps and area composition reported in the main text use natural breaks, accompanied by continuous statistics and alternative schemes. Global Moran’s I was used to test spatial structure. Row-standardized spatial weights were constructed using eight nearest neighbors among centroids of national land units and four-neighbor contiguity among valid marine raster cells. Two-sided pseudo-p values were calculated using 999 random permutations.

4 Results

4.1 National patterns of general resilience capacity

General resilience capacity in national coastal land ranged from 0.2188 to 0.5254. The geodesic-area-weighted mean was 0.3590, the median was 0.3631, the first and third quartiles were 0.3040 and 0.4018, respectively, and the area-weighted standard deviation was 0.0642. Under the five-class Jenks natural breaks classification, the low, relatively low, moderate, relatively high, and high classes accounted for 9.68%, 30.07%, 26.55%, 25.93%, and 7.78% of the area, respectively. The low and relatively low classes together accounted for 39.75%, the moderate class for 26.55%, and the relatively high and high classes for 33.71%. National coastal land therefore exhibited a pronounced lower tail together with continuous middle and upper class structures.

Figure 6 shows a broad north-south gradient with localized mosaics. Low and relatively low levels formed continuous or discontinuous bands along parts of the Bohai coast, the Liaodong coast, and intensively developed eastern shoreline segments. Moderate levels were widespread in transitional areas. Relatively high and high levels became more frequent along parts of the coast south of the Yangtze River, with high values concentrated in southern and central Hainan, the western Beibu Gulf, and selected southeastern coastal units. These geographically continuous high- and low-value areas cannot be explained by a single natural or social factor.

Across 4,658 valid national marine cells, scores ranged from 0.1264 to 0.7083, with a mean of 0.4698, a median of 0.4826, an interquartile range of 0.4278–0.5323, and a standard deviation of 0.0959. Under the five-class Jenks natural breaks classification, the low, relatively low, moderate, relatively high, and high classes accounted for 7.54%, 14.47%, 32.76%, 32.72%, and 12.52% of the marine area, respectively. As shown in Figure 7, the moderate and relatively high classes dominated, jointly accounting for 65.48%. High-value areas formed continuous or semi-continuous patches east of the Liaodong Peninsula, in some nearshore waters of northern Jiangsu, along portions of the southeastern coast, and in several functional zones of the northern South China Sea. Low and relatively low values were concentrated around the Yellow River estuary–Laizhou Bay, waters adjacent to the Yangtze River estuary, and several localized bay mouths.

Spatial autocorrelation confirmed that the national distributions were nonrandom. Global Moran’s I was 0.6123 for national coastal land using the eight-nearest-neighbor matrix and 0.9744 for coastal waters using four-neighbor contiguity; two-sided pseudo-p values from 999 permutations were 0.001 in both cases. Figures 8A–C show the area-weighted land distribution, the distribution of valid marine cells, and the five-class area composition, respectively. Land scores were concentrated primarily between 0.30 and 0.40, whereas the marine distribution was centered near 0.48. In both subsystems, the continuous scores and class structures conveyed complementary information.

Classification sensitivity was evident in the proportions assigned to the two lower classes. For national coastal land, their area share was approximately 39.7% under Jenks natural breaks, 0.3% under equal intervals, and 40.4% under area-weighted quintiles. The corresponding shares of valid national marine cells were 22.0%, 10.9%, and 40.0%. Thus, changing the break rule altered the prevalence of categorical classes but not the underlying continuous scores.

4.2 General resilience capacity within the nested Rongcheng assessment

Across 38,640 valid land cells in Rongcheng, scores ranged from 0.3664 to 0.9241, with a mean of 0.5826, a median of 0.5809, an interquartile range of 0.5309–0.6192, a standard deviation of 0.0820, and a coefficient of variation of 0.1408. Under the five-class Jenks natural breaks classification, the low, relatively low, moderate, relatively high, and high classes contained 6,238, 11,582, 14,810, 4,406, and 1,604 cells, respectively, accounting for 16.14%, 29.97%, 38.33%, 11.40%, and 4.15%. The moderate and relatively low classes jointly accounted for 68.30% and dominated the land domain. Figure 9 also shows central relatively high- and high-value patches juxtaposed against low-value belts near the coast and domain margins, demonstrating that the mean cannot substitute for patch-scale diagnosis.

Across 29,558 valid marine cells in Rongcheng, scores ranged from 0.1485 to 0.8500, with a mean of 0.5544, a median of 0.5491, an interquartile range of 0.4299–0.6906, a standard deviation of 0.1565, and a coefficient of variation of 0.2823. The low, relatively low, moderate, relatively high, and high classes contained 2,064, 8,441, 6,518, 8,209, and 4,326 cells, respectively, accounting for 6.98%, 28.56%, 22.05%, 27.77%, and 14.64%. The relatively low and relatively high classes jointly accounted for 56.33%, while the two tails jointly represented 21.62%. The coexistence of belt-like gradients and high- and low-value patches in Figure 10 further demonstrates pronounced internal heterogeneity in the marine subsystem.

Valid-cell coverage corresponded closely to the current Rongcheng assessment boundary, with edge discrepancies below 1%. The continuous statistics therefore represent the valid score area within the current domain. For planning overlays, however, edge cells should still be clipped precisely using the vector boundary.

4.3 Within-scale synthesis of the nested assessment

Figure 11 summarizes the continuous distributions, five-class composition, and dispersion of the Rongcheng land and marine subsystems independently, without introducing national score benchmarks. The moderate class accounted for the largest share of land, whereas relatively low and relatively high levels represented larger shares of marine waters. The interquartile range, standard deviation, and coefficient of variation were all greater for Rongcheng coastal waters than for coastal land, indicating that local governance should distinguish discrete terrestrial patches from continuous marine gradients.

5 Discussion

5.1 From general resilience theory to the assessment of coastal capacity

The primary contribution of this study is to translate the distinction between general resilience and specified resilience into a clearly bounded measurement object. Coastal studies often develop hazard-specific indices by combining exposure, sensitivity, and recovery resources for storm surges, harmful algal blooms, sea-level rise, or flooding. Such approaches are essential for specified risk management (Cui et al., 2025). Coastal systems, however, also experience nutrient accumulation, habitat degradation, aquaculture volatility, industrial transition, and fragmented governance, which cannot be reduced to one hazard. General resilience emphasizes response diversity, functional backup, appropriate connectivity, learning, and polycentric governance and thus describes the foundational capacity to address uncertain combinations of disturbances (Carpenter et al., 2012; Biggs et al., 2012). The index developed here represents capacity potential rather than observed recovery performance; this boundary avoids claiming faster recovery when neither event intensity nor recovery trajectories have been measured. Research on “resilience potential” likewise emphasizes that indicator-based assessment can support decisions only when its objectives, aggregation method, uncertainty, and intended management use are transparent (Gudka et al., 2024).

The value of the three-tier, ten-attribute framework lies not in proposing a new universal law but in organizing dispersed theoretical principles into a data structure appropriate for coupled coastal land-sea systems. Diversity, slow variables, and ecosystem services provide structural and functional foundations. Openness, redundancy, innovation capacity, and social capital provide exchange, substitution, and collective-action pathways. Modularity, overlapping governance, and tight feedbacks connect shock containment with cross-level coordination. This grouping allows nationally consistent proxies and fine-resolution local observations to represent the same functional meanings. It also makes explicit that connectivity, openness, and modularity are nonlinear and involve trade-offs. As Bouska et al. (2019) noted for river systems, excessive connectivity can spread invasive species and pollution, whereas insufficient connectivity restricts replenishment and recolonization. Bay exchange and aquaculture-raft configurations must likewise be interpreted against appropriate ranges rather than maximized monotonically.

Indicator-based assessment still compresses important dimensions of general resilience. Social capital, innovation capacity, and overlapping governance cannot be fully represented by a single spatial proxy, and a favorable ecological state does not automatically imply high future resilience. Quinlan et al. (2016) distinguished measurement, which seeks reproducible quantities, from assessment, which places greater emphasis on system understanding and plural knowledge. The composite index in this study is a measurement tool, whereas the theoretical framework and Rongcheng case add an assessment function. Their combination reduces the risk of interpreting resilience solely from map colors, but it cannot replace participatory interviews, historical trajectories, or scenario experiments. Governance attributes in particular require independent validation through organizational networks, decision lags, and records of policy implementation.

5.2 National high- and low-value patterns and their potential determinants

National high- and low-value areas displayed pronounced geographic continuity. Low-value land belts extended across multiple county units, whereas marine low values were concentrated around estuaries and semi-enclosed bays. Significant positive Moran’s I values confirm spatial clustering under the respective neighborhood definitions, but the different mapped forms direct attention to landscape continuity on land and water exchange in coastal waters as scale-appropriate hypotheses. The national maps thus identify regional configurations for follow-up rather than establishing the causes of individual low-value units.

High and low values in national coastal land can be interpreted through three complementary pathways: system attributes, dynamic adaptation, and governance coordination. Continuous low-value belts in the north may reflect the combined influence of coastal wetland loss, salinization or relatively low soil-organic-carbon backgrounds, limited natural-habitat diversity, and landscape fragmentation caused by urban and industrial shorelines. These constraints correspond, respectively, to slow variables, diversity, ecosystem services, and modularity. Localized low values in intensively developed eastern shoreline segments may also be associated with built-up expansion, interruption of river-sea corridors, and insufficient public ecological space. Conversely, relatively high values in Hainan, the western Beibu Gulf, and some southeastern shoreline units are consistent with higher vegetation productivity, more intact natural shoreline-wetland configurations, and greater habitat diversity, which provide buffers and alternative response pathways. These interpretations are derived from indicator functions and mapped patterns rather than indicator-level contribution or causal models and should therefore be treated as priority hypotheses for verification, not as confirmed drivers.

Interpretation of national marine low values should prioritize slow variables, openness, ecosystem services, and use redundancy. The Yellow River estuary–Laizhou Bay and waters adjacent to the Yangtze River estuary experience substantial land-based inputs, shifting sediment dynamics, and compound port, aquaculture, and navigation use. Nutrient variability, substrate stability, water exchange, and retention of natural shoreline may jointly constrain the corresponding attributes. High-value patches east of the Liaodong Peninsula, in parts of northern Jiangsu, along the southeastern coast, and in the northern South China Sea may reflect favorable water exchange, diverse habitat configurations, primary production within an appropriate range, and functional backup associated with protection, monitoring, and multiple compatible uses. Because marine exchange can connect neighboring cells, these clusters motivate joint examination of land-derived inputs and water transport rather than isolated cell-level explanations.

National governance should distinguish continuous low-value belts, class-transition zones, and high-value maintenance areas. In northern land areas, wetland restoration, habitat connectivity, and management of slow variables should be integrated with industrial and urban renewal. In low-value estuarine and semi-enclosed marine areas, cross-basin monitoring of nutrients and sediments should be combined with assessment of bay exchange. In Hainan, the Beibu Gulf, and southeastern high-value areas, governance should evaluate whether favorable natural and institutional conditions are being maintained and prevent new development from eroding existing buffers. The principal purpose of the national maps is to set priorities for interregional monitoring and verification; specific interventions still require local attribute decomposition and field evidence.

5.3 High- and low-value patterns in Rongcheng and their potential determinants

Within Rongcheng, the land assessment shifts the diagnostic unit from national county-level belts to habitat patches and shoreline segments. This finer spatial support, together with locally enhanced indicators and different weights, identifies more specific priorities for future field verification but does not isolate a resolution effect. Within the recorded Rongcheng weight structure, slow variables (0.1916) and diversity (0.1875) had the largest weights, followed by ecosystem services (0.1066) and redundancy (0.1086). Habitat-patch diversity, perennial vegetation status, soil salinity, and shoreline erosion are therefore priority attributes for interpreting high and low land values. Central relatively high- and high-value patches may correspond to more intact woodland-wetland-farmland-water configurations and stable vegetation productivity. Low-value coastal and edge belts warrant investigation of development-related fragmentation, interrupted river-sea corridors, salinization, and insufficient ecosystem-service provision.

In Rongcheng, marine scores show greater within-domain dispersion than land scores and form gradients within and among bays. This pattern points to connected waters, rather than isolated cells, as the relevant support for testing exchange- and aquaculture-related explanations. Given the Rongcheng indicators and weights, priority factors for verification include slow variables such as DIN/DIP variability, duration of aquaculture occupation, and sediment stability; phytoplankton, zooplankton, algal habitat, and substrate diversity; bay exchange and raft permeability; and the openness and functional redundancy provided by integrated multi-trophic aquaculture. Low-value areas may represent combinations of restricted exchange, dense raft structures, nutrient fluctuations, or deteriorated benthic habitat quality. High-value areas may possess better water exchange, habitat quality, production backup, and monitoring feedback.

The analysis of potential determinants in Rongcheng remains a diagnostic hypothesis based jointly on the weight structure, indicator meanings, and spatial morphology; it is not the result of a driver model. Entropy weights determine how the ten attributes enter the composite index, but their magnitudes alone cannot identify which attributes account for local high or low scores. Identifying these spatial contributions requires validated attribute scores and maps of their weighted contributions for the land and marine subsystems separately. Rongcheng’s mapped low-value areas are therefore treated as provisional priorities for field verification, focusing on slow-variable anomalies, habitat and ecosystem-service conditions, and conflicts between water exchange and coastal use.

5.4 Linking the national baseline and nested diagnosis in Rongcheng

The Rongcheng diagnosis follows the national baseline to convert macro-scale screening into locally verifiable units, not to place two indices in a common ranking. The strength of the national dataset is continuous coverage under consistent definitions, which helps identify regions and shoreline segments requiring further attention. The strengths of the Rongcheng dataset are its high-resolution habitat information, nutrient data, aquaculture structure, and governance feedbacks, which help determine where local monitoring should be deployed and which attributes should be investigated. The nested assessment therefore changes the diagnostic question from where regionally coherent signals occur to which local habitat, nutrient, aquaculture, and governance conditions warrant verification.

Scale linkage depends on shared theoretical attributes rather than identical variables or weights. After national and Rongcheng weights are estimated separately, each index has a relative meaning only within its own standardization domain. A locally enhanced indicator qualifies as a refinement of the general resilience framework only when it maps onto the three-tier, ten-attribute structure. The nested design can therefore examine how national screening triggers a Rongcheng diagnosis and how local monitoring may revise a national proxy, but it cannot use mean differences, differences in class proportions, or cell-level subtraction to infer which system is more resilient.

A defensible cross-scale evidence chain is “national spatial signal–local attribute verification–monitoring feedback–model updating.” If a local low-value area can be explained by attributes already represented nationally, the corresponding monitoring should be increased in resolution. If it arises from local aquaculture configuration, nutrient variability, or response delays not covered by national indicators, the applicability boundary of the national proxy should be recorded and the feasibility of broader data coverage evaluated. The nested evidence shows scale dependence in what the assessment can observe and test: discrete land patches and connected marine gradients call for different local verification designs. Because spatial support, indicators, and weights change together, these qualitative contrasts neither isolate a numerical scale effect nor validate the national index.

5.5 From maps to scale-matched land-sea governance

At the national level, the first priority is an updateable data infrastructure. Baseline years, product versions, coordinate systems, valid coverage, and missing-data codes should be harmonized, and original observations, standardized matrices, national weights, and processing scripts should be retained. National coastal land monitoring should stratify the continuous low-value belts in the north, localized low-value patches along the eastern coast, and high-value clusters in the south. National marine monitoring should be intensified in the Yellow River estuary–Laizhou Bay, the Yangtze River estuary, and other low-value clusters and class-transition zones. A single classification threshold should not be used as a rigid performance target. Policy priorities should instead combine continuous scores, area composition, spatial clustering, and temporal trends.

In Rongcheng, high-resolution results can support a zoning sequence of maintenance, monitoring, verification, and restoration. Because moderate and relatively low levels account for most of the land domain, low-value belts and class-transition zones should be overlaid with land cover, river-sea corridors, soil salinity, and field observations, while the connectivity and ecosystem services of central relatively high- and high-value patches are maintained. The juxtaposition of relatively low and relatively high classes in local marine waters requires focused monitoring of nutrients, substrate, exchange, and use intensity, together with tests of whether multi-trophic production, habitat quality, and feedback capacity persist in high-value areas. Before specific engineering interventions are proposed, attribute scores and low-value area should be summarized within formally recognized bays and aquaculture management units.

Aquaculture pressure should be separated from the general resilience capacity index before the two are coupled. The capacity index describes the structures, functions, and governance conditions a system possesses; aquaculture density, nutrient loading, and exposure to extremes describe the pressures imposed on it. Overlaying these dimensions can yield four diagnostic types: high pressure–low capacity, high pressure–high capacity, low pressure–low capacity, and low pressure–high capacity. These correspond, respectively, to pressure reduction and restoration, capacity-based early warning, reinforcement of foundational capacity, and functional maintenance. Large-scale kelp cultivation in Rongcheng may remove nutrients but may itself be vulnerable to anomalous exchange and harmful algal blooms (Xu et al., 2023; Li et al., 2025). Pressure and function therefore cannot be represented by assigning a single indicator a simple positive or negative direction.

Cross-scale feedback is essential for keeping the governance framework operational. Local monitoring should update slow variables and anomalous events seasonally or annually, while a national platform periodically compiles changes in indicator versions, weights, and spatial distributions. When local evidence identifies low-value patches that cannot be explained by national indicators, national proxies should be adjusted or their applicability boundaries documented. Overlapping governance should not mean adding institutions; it should define data responsibilities, trigger thresholds, and response deadlines among natural-resource, ecological-environment, agriculture and rural affairs, maritime, and local-government authorities. Through a cycle of “national rules–local diagnosis–implementation feedback–model updating,” general resilience principles can move from a static index toward adaptive governance capacity.

5.6 Uncertainty, limitations, and future research

The main uncertainties arise from classification, weighting, indicator proxies, and temporal support. Natural breaks, equal intervals, and quintiles alter class-area composition; class labels are therefore cartographic language valid only within a subsystem. National and Rongcheng indicators differ and require separate weights. The values reported in Tables 3 and 4 are the final entropy weights used for the national and Rongcheng assessments, respectively. A single temporal window cannot test proximity to thresholds or functional reorganization after disturbance.

A reproducible sensitivity assessment requires the original indicator observations, their standardized values, scale-specific weights, and versioned metadata for both the national and Rongcheng systems. With these records, future studies could compare equal, entropy, and expert weighting; test ±10% weight perturbations and the omission of individual indicators; and map the spatial contribution of each attribute. Dynamic validation requires at least three to five assessment periods and independent pressure and performance layers outside the general resilience capacity index. These data would allow tests of whether areas with higher capacity experience smaller functional losses, faster organizational adjustment, or more stable ecosystem services under compound disturbances. Strict numerical comparison across scales is defensible only under identical indicators, weights, and spatial supports. Otherwise, within-scale interpretation and nested governance linkage should remain the governing principles.

Future research should examine whether similar resilience levels arise through different regional pathways. Following Xu et al. (2026), geographically informed nonlinear models could assess how habitat configuration, land–sea connectivity, hydrodynamic exchange, use intensity, and governance responses relate to resilience trajectories in different coastal settings. Coastal units could then be grouped according to similar combinations of locally influential factors, and these groups overlaid with resilience states. For example, low-resilience areas associated primarily with habitat fragmentation would prioritize habitat and corridor restoration, whereas those associated with restricted water exchange and intensive coastal use would prioritize coordinated control of land-based inputs and marine-use intensity. Where governance feedback is more influential, monitoring and interdepartmental coordination would receive greater emphasis. Linking resilience states with spatially differentiated driving pathways would support zoning strategies tailored to regional mechanisms rather than assigning the same intervention to every area within a resilience class.

6 Conclusion

This study developed a three-tier, ten-attribute framework for assessing general resilience capacity in China’s coastal social-ecological systems, linking a national baseline with a nested diagnosis in Rongcheng. The composite index represents capacity to cope with diverse disturbances rather than observed post-disturbance recovery.

At the national scale, the mean index values were 0.3590 for coastal land and 0.4698 for coastal waters. Low and relatively low classes covered 39.75% of the assessed land area, whereas moderate to high classes covered 77.99% of coastal waters. Both distributions showed significant positive spatial clustering (Moran’s I = 0.6123 and 0.9744, respectively). In Rongcheng, the mean values were 0.5826 for land and 0.5544 for waters, and the local maps depicted land-patch and within-bay variation. Because the national and Rongcheng assessments differ simultaneously in indicators, weights, and spatial support, the additional local detail cannot be attributed to spatial resolution alone. Their scores are interpreted within their respective assessment domains rather than compared directly across scales. Classification-dependent changes in class proportions also indicate that categorical maps should be considered alongside continuous scores and spatial clustering.

The national results can inform regional monitoring priorities, while the Rongcheng results identify priority units for future field verification and local monitoring. Aquaculture pressure should be assessed separately from general resilience capacity. Future research should repeat the assessment over time, evaluate weight and threshold uncertainty, and test the index against independent observations of disturbance and system performance. Such evidence would strengthen the use of national screening and local diagnosis in integrated land–sea governance.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. During the preparation of this manuscript, the authors used generative artificial intelligence tools solely for two limited purposes: English-language polishing and translation. These tools were not used for research design, data generation, data analysis, interpretation of results, or the formulation of conclusions. All AI-assisted content was carefully reviewed, verified, and revised by the authors, who take full responsibility for the accuracy and integrity of the manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fevo.2026.1969265/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Chao Fan, Zhejiang Ocean University, China

Reviewed by: He Xu, Shandong Normal University, China

Yifei Xu, Yunnan University, China

Biao Liu, North University of China, China