Assessing Urban Flood Risk and Identifying Critical Zones in Xiamen Island Based on Supply–Demand Matching

Abstract

The supply–demand relationship of flood regulation services (FRS) plays a vital role in mitigating urban flooding. Yet, existing studies still fall short in the comprehensiveness of FRS indicators, the accuracy of assessment scope, and the fine-scale analysis needed to delineate spatial supply–demand features and precisely identify critical areas. Using Xiamen Island as a case study, we first quantify ecosystem-based FRS supply with the InVEST model and assess socioeconomic FRS demand under the H-E-V framework; second, we perform parcel-level supply–demand matching to identify spatial patterns and typologies; then, we diagnose FRS status via the coupling–coordination degree model (CCDM); and finally, we delineate flood-risk hotspots through priority-intervention grading. The results indicate that (1) higher FRS supply clusters in the south, southwest, and northeast, whereas demand is markedly higher in the central–northern sector, yielding an overall pattern of “pronounced mismatch in the central and north, with relatively sufficient supply along the periphery.” (2) Low supply–high demand zones exhibit the highest flood risk and contain higher proportions of industrial, transportation, and residential land. (3) These low supply–high demand zones are further subdivided into five priority-intervention levels, for which we propose tiered, differentiated risk-management strategies. Collectively, the findings clarify supply–demand mechanisms and mismatch characteristics, providing decision support for urban flood safety and sustainable development.

1. Introduction

As climate change and urbanization continue to intensify, urban flooding has become a widespread and serious problem in many countries and regions [1]. Over the past 50 years, global flood events have caused 58,700 deaths and approximately USD 651 billion in socioeconomic losses [2]. From 2016 to 2035, the direct economic losses from floods are projected to reach USD 597 billion [3]. The rapid expansion of impervious surfaces, disruption of natural hydrological processes, and high concentrations of population, facilities, and industry make cities more vulnerable to flooding than other areas [4]. Therefore, accurately identifying the spatial distribution of flood risk in high-density urban areas is essential for formulating risk-prevention measures and reducing flood losses.

In flood risk research, numerous theoretical frameworks have been developed [5,6]. Among them, the ecosystem’s FRS are widely regarded as pivotal in mitigating flood risk [7]. From the perspective of human well-being, FRS capture the role of ecosystems in reducing flood damage and providing flood prevention and detention; this regulatory function, dynamically transmitted between ecosystems and human society, can be evaluated through the relationship between supply and demand. FRS supply refers to an ecosystem’s capacity to retain stormwater runoff, attenuate peaks, and alleviate the hazardous effects of intense rainfall [8]; FRS demand denotes the level of prevention and detention capacity required by society to reduce flood risk and associated losses [9]. The degree of supply–demand matching determines the realized regulatory effectiveness of FRS [10,11], whereas supply–demand imbalance is often a key driver of the emergence and amplification of flood risk [12]. Accordingly, systematically assessing the FRS supply–demand relationship helps gauge a city’s overall capacity to cope with pluvial hazards and the risks it faces, thereby providing a scientific basis for flood mitigation and risk-zoning management.

A substantial body of work has examined FRS supply–demand balance and sustainability at national and river-basin scales [13,14,15], yet at the urban scale, the comprehensiveness and coherence of integrated assessment frameworks remain limited. On the supply side, urban ecosystems regulate flooding through two service types—flood prevention and flood mitigation [16]. However, prior studies have emphasized mitigation, commonly using distributed hydrological models such as the soil conservation service curve number (SCS-CN) method [17], the stormwater management model (SWMM) [18], and the soil and water assessment tool (SWAT) [19], to simulate and compute the rainfall loss index [20,21]. In practice, prevention services are equally important to FRS supply [22]: soil erosion degrades soil structure, reduces surface detention, and clogs drainage systems, thereby weakening water storage and drainage capacity [23]; elevated pollutant loads degrade water quality and indirectly reduce the total volume available for detention [24]. Neglecting any class of regulatory service—or its associated ecological processes—impedes accurate characterization of urban FRS supply [25]. Accordingly, there is an urgent need for methods that integrate multiple flood-relevant ecosystem functions; the integrated valuation of ecosystem services and tradeoffs (InVEST) model offers a more holistic approach for assessing FRS supply capacity [26]. On the demand side, existing studies often represent demand via the disaster sensitivity of human society [27]. For example, Zhang et al. [28] used population, GDP, points of interest, and road networks to evaluate demand for flood resilience in Nanjing; Xu et al. [29] assessed coastal cities’ demand for typhoon-protection services using typhoon frequency, impervious-surface ratio, and population density. Such approaches typically infer demand from the spatial clustering of population, assets, and infrastructure; however, in reality, these clusters do not necessarily coincide with inundation extents during floods [30]. Therefore, demand should be reflected by explicitly evaluating losses to social elements within graded ranges of flood depth. The hazard–exposure–vulnerability (H-E-V) risk-assessment framework [31,32], which couples flood physical characteristics with the exposure and vulnerability of recipients such as population, assets, and infrastructure [33], enables integrated measurement of actual reliance on FRS under specific flood scenarios. This yields a more accurate depiction of spatial variability in hazard intensity and is particularly well suited to identifying flood risk in dense cities with high heterogeneity and complex spatial patterns.

Moreover, interactions of varying intensity between FRS supply and demand within cities give rise to complex and diverse risk signatures across areas [34]. Analyzing the internal mechanisms of the supply–demand subsystems and their interaction effects helps build a more comprehensive understanding of risk-generation processes. The CCDM characterizes the strength of associations and the state of coordinated development among multiple subsystems along two dimensions—coupling degree and coordination degree [35]. It has been widely applied to coupled relationships such as climate change and the ecological environment [36] and urbanization and the environment [37], yet its use in parcel- or city-scale analyses of FRS supply and demand remains limited. Introducing CCDM for the diagnosis of coupling–coordination and typology identification between FRS supply and demand [38] is therefore of great significance for accurately distinguishing supply–demand matching types and formulating scientifically grounded flood-mitigation measures.

This study develops an integrated risk-assessment framework based on FRS supply–demand matching to evaluate flood risk at the urban scale, using Xiamen Island as the study area. The specific objectives are (1) to quantify FRS supply capacity—encompassing ecosystem services of flood prevention and flood mitigation—using the InVEST model; (2) to characterize flood hazard and the exposure and vulnerability of elements at risk under the H-E-V framework, thereby deriving the FRS demand level; (3) to perform parcel-level FRS supply–demand matching via Z-score standardization and quadrant partitioning and to depict the coupling–coordination status with the coupling–coordination degree model (CCDM); and (4) to identify the priority intervention index and propose targeted regulatory measures based on land-use/land-cover (LULC) characteristics and spatial features. This study refines the FRS supply–demand indicator system, reveals the spatial matching pattern and interactions between the two subsystems more systematically, and provides scientific decision support for urban flood-risk management, urban planning, and ecological protection.

The overall structure of this study is as follows: Section 2 describes the materials and methods, presenting the study area and data types. This section also details the research framework, including the selection and calculation of supply and demand indicators, spatial matching, CCDM, and the computation of the priority intervention index. Section 3 presents and analyzes the assessment results for Xiamen Island. Section 4 provides the overall conclusions.

2. Materials and Methods

2.1. Study Area

Xiamen Island (24°23′–24°54′ N, 117°53′–118°26′ E) is the political, economic, and social core of Xiamen City, Fujian Province (Figure 1). The island covers 157.98 km², comprises two administrative districts and 15 subdistricts, and had a resident population of about 2.062 million in 2024—a prototypical high-density built-up area. Influenced by tropical cyclones, the island experiences frequent typhoons and extreme rainfall in summer. Topography is higher in the south and lower in the north, with a maximum elevation of 312 m; surrounded by the sea on all sides, it has small catchments and short, discontinuous natural rivers. Expansion of construction land has sharply reduced natural detention space and intensified ecosystem fragmentation, thereby weakening FRS. During typhoon and rainy seasons, the phenomenon of “external storm surge and internal flooding” often occurs [39], where storm surge and urban flooding overlap, triggering urban flood disasters. Population concentration and intense economic activity further amplify risk. Historical records indicate that Xiamen experiences a major flood approximately every six years, with direct economic losses of 700 million to 1.4 billion USD per event [40]. Consequently, an FRS study from a supply–demand matching perspective on Xiamen Island is crucial for proposing effective flood-mitigation strategies, enhancing FRS capacity, and supporting regional sustainable development.

2.2. Data Sources

This study uses four datasets, primarily sourced from official open data and a series of satellite images (

Table 1. Supply-demand assessment dataset.

Data Type	Data Name	Category	Resolution	Unit	Source
Land use and land cover	LULC	Vector	/	m	Xiamen municipal bureau of natural resources and planning (https://zygh.xm.gov.cn/).
Natural environment	DEM	Grid	12.5 m	m	Geospatial aata cloud platform (https://www.gscloud.cn/).
	Slope	Grid	12.5 m	°	DEM-based extraction.
	Precipitation/evapotranspiration	Grid	30 m	mm	CMADS dataset (https://www.fao.org/soils-portal/en/, accessed on 1 November 2025).
	Soil	Grid	30 m	/	World soil database (https://www.fao.org/home/en, accessed on 1 November 2025).
	Root restriction layer depth	Grid	30 m	m
	Plant available water content	Grid	30 m	%
Socioeconomics	GDP	Grid	30 m	Yuan/km2	Xiamen statistical yearbook (2020) (https://tjj.xm.gov.cn/); AMAP POI open data.
	Population density	Grid	30 m	Per
				/km2
	POI	Vector	/	/
	Building	Vector	/	Piece
Hydrology and hydraulics	Drainage pipe network	Vector	/	m	Xiamen municipal bureau of agriculture and rural affairs; Xiamen city drainage (Rainwater) flood control master plan (2020–2035) (https://sn.xm.gov.cn/) [41].
	Hydrological parameters	Float	/	/
	Historical flood area	Vector	/	Piece
	Contour line	Vector	/	Piece

). (1) Vector land-use data: compiled and redrafted from Xiamen’s Regulatory Detailed Plan, divided into ten land-use categories: residential land, public management-service land, commercial land, industrial land, utilities, transportation, logistics land, green, water, and agriculture and forestry land. (2) Natural-environment data: The DEM was obtained from the Geospatial Data Cloud, and slope was calculated in ArcGIS 10.8. Soil and meteorological data were sourced from the Harmonized World Soil Database and the CMADS dataset. This dataset is used to quantify FRS supply. (3) Socioeconomic data: GDP, population density, POI, and building data were compiled from open sources. (4) Hydrology–hydraulics data: Public data from the Xiamen Municipal Bureau of Agriculture and Rural Affairs, used for flood-process simulation. Datasets (3) and (4) are used to quantify FRS demand.

Given differences in format, resolution, and coordinate systems across datasets, we first reprojected all spatial layers—LULC, DEM, precipitation, and soils—and resampled them to a 12.5 m × 12.5 m grid. The InVEST model was then run on this unified grid to compute FRS supply. Subsequently, at the parcel-analysis stage, ArcGIS zonal statistics were used to map InVEST raster outputs to parcel vector units, while socioeconomic statistics were area- and land-use-type-weighted to parcels. This ensured compatibility and comparability of multisource data in the FRS supply–demand assessment.

2.3. Methods

2.3.1. Research Framework

This study establishes an FRS supply–demand risk-assessment framework (Figure 2) comprising four components: (1) FRS supply quantification. Based on the InVEST model, three representative ecosystem functions—water retention (WR), soil retention (SR), and nutrient purification (NR)—are selected to measure regional FRS supply capacity. (2) FRS demand quantification. Under the H-E-V risk-assessment framework, specific flood scenarios are simulated to compute hazard, vulnerability, and exposure, thereby characterizing socioeconomic demand for FRS. Subsequently, supply and demand capacities are derived via multi-indicator weighting using the entropy method and TOPSIS, and spatial autocorrelation models are used to identify clustering characteristics of each. (3) Supply–demand matching and analysis. Z-score standardization and quadrant partitioning are employed to identify types of supply–demand matching and their spatial patterns. The CCDM is introduced to diagnose the supply–demand coupling degree and coupling–coordination degree for each subdistrict. (4) Priority-zone identification and pathway formulation. Based on the PRI, zones are leveled from I to V by urgency, and, in combination with LULC characteristics, targeted adjustment measures and planning strategies are proposed for areas with different coordination states.

2.3.2. Selection and Calculation of FRS Supply Indicators

To capture the ecosystem’s substantive contribution to FRS, this study adopts water retention, soil retention, and nutrient retention as core indicators [42] and evaluates them using the InVEST model’s Water Yield, Sediment Delivery Ratio, and Nutrient Delivery Ratio modules, thereby representing runoff regulation capacity, sediment retention capacity, and nutrient reduction capacity.

(1)

Water retention. The Water Yield module, based on the Budyko water-balance principle, computes precipitation retention and infiltration, along with soil–water storage, to determine the study area’s Water Yield. The module’s governing equation is as follows:

$$Y_{xj}=(1-{\displaystyle \frac{AE{T}_{xj}}{P_x}}{)·P}_x$$

(1)

where Y_xj represents the Water Yield supplied by LULC type j on parcel x; AET_xj is the actual evapotranspiration of the parcel, x; and P_x denotes the precipitation on parcel x.

(2)

Soil retention. The Sediment Delivery Ratio module, based on the Universal Soil Loss Equation (USLE), calculates the potential soil erosion and the actual soil erosion for each grid cell; the difference between these represents soil retention. The module equation is as follows:

$$SR=R·K·L·S·(1-C·P)$$

(2)

where R is rainfall erosivity, K is soil erodibility, L is the slope-length factor, S is the slope-steepness factor, C is the cover-management factor, and P is the support-practice factor.

(3)

Nutrient retention. The Nutrient Delivery Ratio module is based on the mechanism by which vegetation and soils in ecosystems reduce or remove nitrogen and phosphorus pollutants in river runoff through storage or transformation. The lower the surface export of TN and TP, the stronger the water-purification capacity and the lower the pollution risk posed by floods. The module equation is as follows:

$$NR={\displaystyle \frac{\mathrm{m}\mathrm{a}\mathrm{x}\left({ALV}_x\right)-{ALV}_x}{\mathrm{m}\mathrm{a}\mathrm{x}\left({ALV}_x\right)-\mathrm{m}\mathrm{i}\mathrm{n}\left({ALV}_x\right)}}$$

(3)

$${ALV}_x={HSS}_x·{pol}_x$$

(4)

where ALV_x is the adjusted loading value of grid cell x, HSS_x is the hydrologic sensitivity score of grid cell x, and pol_x is the export coefficient of grid cell x.

2.3.3. Selection and Calculation of FRS Demand Indicators

Based on the IPCC’s H-E-V framework [43], this study characterizes FRS demand through the interaction of hazard, exposure, and vulnerability. Here, hazard refers to the physical characteristics of flooding—such as the distribution of stormwater runoff, inundation depth, and duration—caused by heavy rainfall; exposure denotes the degree and scale of contact between floodwaters and elements at risk, reflecting potential losses; and vulnerability captures the system’s propensity or potential to experience adverse effects when subjected to stress, typically measured by the extent of damage.

(1)

Hazard. The flood physical characteristics in this study were obtained using the InfoWorks ICM model under a 50-year return-period rainfall scenario (in Supplement Section S1). This study selects maximum inundation depth [44] and inundation duration [45] and defines the hazard index as follows:

$$H=D_{max}·T_{inu}$$

(5)

where H is the hazard index, D_max is the maximum inundation depth (m) experienced by each parcel unit, and T_inu is the inundation duration (h).

(2)

Exposure. This study selects population, medical facilities, government agencies, and social organizations as the elements at risk and counts the number of such elements within each parcel [46]. According to the GB 51222—2017 Technical Code for Urban Waterlogging Prevention [47], a ponding depth exceeding 0.15 m is considered urban waterlogging. Based on inundation depth, five classes are defined: 0–0.15 m, 0.15–0.30 m, 0.30–0.45 m, 0.45–0.60 m, and >0.60 m [48]. Depth-weighting is used to capture the differential impacts across inundation classes, with weights assigned from low to high as 0, 0.25, 0.5, 0.75, and 1. The calculation formula is as follows [49]:

$$E=\sum _{i=1}^n(P_i·W_i)+\sum _{j=1}^m(F_j·W_j)$$

(6)

where E is the exposure index; P_i is the number of affected people of category i within the parcel; F_j is the number of affected facilities of category j within the parcel (medical facilities, government agencies, and social organizations); W_i and W_j are the weights corresponding to the inundation-depth class, and n and m are the numbers of population and facility categories, respectively.

(3)

Vulnerability. Vulnerability is assessed along two dimensions—structural damage to buildings and indoor property and economic losses [50]—using widely applied water depth–damage ratio curves [51], locally calibrated and adjusted to the study area. The calculation formula [52] is as follows:

$$V=V_s·V_i$$

(7)

$$V_s=0.041 + 0.183x-0.059x^2 + 0.003x^3$$

(8)

$$V_i=0.209 + 0.0152x-1.32 \times {10}^{-4}{x}^2+3.786 \times {10}^{-7}{x}^3$$

(9)

where V_s. is the structural damage ratio of buildings, V_i is the indoor property damage ratio of buildings, and x is the depth of waterlogging. The product of structural vulnerability and indoor property vulnerability constitutes the vulnerability index V. For different land-use types, the indoor property damage ratio is determined by substituting the corresponding inundation depth (in Supplement Section S2).

2.3.4. Supply-Demand Calculation

(1)

Supply calculation

FRS supply is computed from the combined contributions of WR, SR, and NR. First, the WR, SR, and NR values are normalized. Given that these three indicators are considered equally important to FRS, equal weights are assigned, and a weighted summation is performed to obtain the FRS supply. The FRS supply derived from the InVEST model is raster data. Using parcels as the basic unit, we apply ArcGIS zonal statistics to map the raster outputs to parcel units, thereby obtaining the parcel-scale spatial distribution of FRS supply. The calculation formula for FRS supply [53] is as follows:

$${FRS}_s={\mu }_1·WR + {\mu }_2·SR + {\mu }_3·NR$$

(10)

where FRS_s is the FRS supply, WR is water retention, SR is soil retention, and NR is nutrient retention. μ denotes the weighting coefficient, which is set to 0.333 because the three components are considered equally important.

(2)

Demand calculation

The FRS demand level is computed from the combined contributions of the H, E, and V indices. To ensure scientific rigor, a composite weighting method is used to determine the weights of the three indicators, which are then aggregated accordingly.

The subjective weight a is calculated using the Analytic Hierarchy Process (AHP), and the objective weight b is calculated using the entropy-weight method. By introducing a priority coefficient μ, the composite weight w is obtained as follows:

$$w =\mu ·a + (1 - \mu )·b$$

(11)

In determining the priority coefficient, the subjective weight and the objective weight are of equal importance; therefore, μ = 0.5.

After normalizing H, E, and V, the formula [54] is as follows:

$$\begin{array}{ll}{FRS}_d\hfill & =\int (H, E, V)\hfill \\ & =w_i·{\sum }_{\mathrm{i}=1}^m{H}_i+w_j·{\sum }_{j=1}^n{E}_j+w_l·{\sum }_{k=1}^p{V}_k\hfill \end{array}$$

(12)

where FRS_d is the FRS demand; H_i is the hazard index of parcel i; E_j is the exposure index of parcel j; V_k is the hazard index of parcel k; and w_i, w_j, and w_l are the composite weights of H, E, and V, respectively: 0.341, 0.358, and 0.301 (in Supplement Section S3).

2.3.5. FRS Supply–Demand Matching

The Z-score standardization method is used to compare FRS supply and demand, enabling a concise and intuitive identification of matching types. For each parcel unit, the FRS supply and demand are standardized using Z-scores and classified by quadrants: the x-axis denotes the standardized FRS supply capacity, and the y-axis denotes the standardized FRS demand level, thereby partitioning the plane into four quadrants. Quadrant I represents high supply-high demand, Quadrant II low supply-high demand, Quadrant III low supply-low demand, and Quadrant IV high supply-low demand.

2.3.6. Coupling-Coordination Degree Evaluation

The degree of coupling-coordination reflects the tendency toward order or disorder between systems. The CCDM comprises three components: the comprehensive system index, the coordination degree, the coupling degree, and the coupling-coordination degree. The specific model equations are as follows:

$$D=\sqrt{C·T}$$

(13)

where D denotes the coupling-coordination degree, C denotes the coupling degree, and T denotes the comprehensive coordination level.

2.3.7. Spatial Priority Method for Planning Intervention

PRI is a scientific tool for the orderly regulation of the FRS supply–demand relationship, thereby enhancing system detention capacity and guiding spatial-planning implementation. The intervention sequence is established based on the demand-to-supply ratio; the underlying logic is that the higher the demand and the lower the supply, the larger the PRI, and the parcel thus requires higher-priority intervention. In this study, PRI is introduced to prioritize planning interventions for imbalance hotspots characterized by “low supply–high demand.”

$$PRI={\displaystyle \frac{FRSs+p}{{FRS}_d+q}}$$

(14)

where FRS_s and FRS_d are the Z-score standardized FRS supply and demand, respectively, and p and q are adjustment coefficients. To prevent abnormal inflation of the PRI caused by a negative or near-zero denominator, the standardized demand and supply are translated to a positive interval away from zero. In this study, p = 1 and q = 2.5.

3. Results

3.1. Results of Supply and Demand Indicators

3.1.1. Results of Supply Indicators

The FRS supply indicators based on ecosystem services are shown in Figure 3. Using the equal-interval method, the standardized indices of WR, SR, and NR are divided into five classes: very low (0–0.2), low (0.2–0.4), medium (0.4–0.6), high (0.6–0.8), and very high (0.8–1). WR and SR display similar spatial patterns—generally low in the central and northern zones and gradually increasing toward the south as well as to the east and west. Parcels with very high WR and SR values are mainly located in Xiagang, Binhai, Lianqian, Jinshan, and Huli subdistricts. These areas are typically adjacent to water bodies, green spaces, and agriculture and forestry land; have good landscape connectivity and pronounced relief; and possess favorable soil infiltration and detention conditions. By contrast, very low values are concentrated in the contiguous built-up areas of the central and northern subdistricts—Jialian, Jiangtou, and Heshan—which are dominated by impervious surfaces. The NR values are higher in the north and northwest and lower in the central-southern part of the island; high to very high NR is distributed continuously along the coastal belt from the Xiangyu Bonded Area to Xiagang. In the mid-northern and southern subdistricts—Wucun, Jiangtou, Dianqian, and Binhai—parcel-level NR values are concentrated within 0–0.6.

Box plots intuitively reveal dispersion and skewness, making them well-suited for presenting the distribution of supply indicators. The results show marked differences between WR, SR, and NR. NR tends to cluster toward lower values. SR is more evenly distributed. WR is generally higher, primarily due to the strong retention capacity of ecological land, such as agricultural and forestry land and water bodies, which elevates WR levels locally and in surrounding areas. The large share of contiguous built-up areas on Xiamen Island exerts a pronounced adverse effect on NR. In contrast, ecological land underpins WR and SR, yielding a more robust overall distribution.

3.1.2. Results of Demand Indicators

Based on the H-E-V framework, the FRS demand indicators are shown in Figure 4. Using the equal-interval method, each indicator is classified into five levels: very low (0–0.2), low (0.2–0.4), medium (0.4–0.6), high (0.6–0.8), and very high (0.8–1). Parcels with high and very high H values exhibit spot-like clusters, mainly in Jiangtou, Jialian, and Dianqian. These areas have a high proportion of impervious surfaces, short flow paths, and low-lying terrain; combined with a dense drainage network that is prone to overload during heavy rainfall—leading to frequent overflows at nodes and outfalls—and limited direct discharge when distant from water bodies, these conditions cause surface ponding and backflow, thereby elevating overall hazard. Parcels with E values > 0.6 are concentrated in infrastructure-dense, socioeconomically active subdistricts such as Yundang, Huli, Wucun, Zhonghua, and Lianqian, where approximately 40% of the island’s hospitals and 42% of government and social organizations are located. Heshan and Lianqian also show higher exposure indices due to large resident populations. The spatial pattern of V is more dispersed. Owing to older building stock, dense indoor property, and earlier construction vintages in these areas, resistance and recovery capacity are relatively limited, resulting in higher overall vulnerability with pronounced internal variability.

The distributional trends of the demand indicators show that the indices for H and E are concentrated mainly within extremely low levels, with most parcels falling within 0–0.1. By contrast, the indices for V value are significantly higher than those for H and E and are concentrated mainly in the low-to-mid-range. When confronted with flood risk, the sensitivity to potential loss is high, which can lead, over the long term, to sustained outflows of population and economic activity.

3.1.3. Spatial Distribution and Clustering Characteristics of Supply–Demand

The FRS supply and demand are derived from the supply and demand indices, respectively. Using the natural breaks method, the FRS supply is classified into five levels, very low (0–0.322), low (0.322–0.405), medium (0.405–0.499), high (0.499–0.598), and very high (0.598–0.778), with the spatial distribution shown in Figure 5a. Areas with high and very high values are mainly concentrated in the southern, northwestern, and northeastern parts of the island, exhibiting a clear clustering pattern. Such clusters of relatively high supply form several core hotspots in Jinshan, Huli, Binhai, and Lianqian subdistricts (Moran’s I = 0.986, z-score = 82.8263, p < 0.001), as shown in Figure 5c. These subdistricts contain extensive water bodies and mountainous forest land, with strong ecological baselines and high landscape connectivity; consequently, supply values range from 0.47 to 0.62, significantly exceeding the island-wide mean (0.448). In contrast, areas with very low supply are concentrated primarily in the contiguous central-to-northern subdistricts of Wucun, Jialian, Jiangtou, Heshan, and Dianqian; additionally, parts of the southeastern sector also exhibit low supply. This pattern is attributable to high levels of land development in these areas, which has led to varying degrees of ecological degradation and weakened ecosystem supply capacity.

Similarly, the FRS demand is classified into five levels—very low (0–0.047), low (0.047–0.124), medium (0.124–0.236), high (0.236–0.402), and very high (0.402–0.661)—with the spatial distribution shown in Figure 5b. Compared with the clustered spatial characteristics of FRS supply, demand hotspots are more dispersed (Moran’s I = 0.852, z-score = 72.9512, p < 0.001) (Figure 5d). This is mainly because demand evaluation depends strongly on the distributions of floods, population, critical infrastructure, and buildings, which are particularly scattered in high-density urban areas, thereby producing the observed differences in matching outcomes. High and very high demand values are locally clustered in the eastern, southeastern, western, and central sectors. Topographically, these areas are relatively low-lying. Moreover, Xiamen Island’s main transportation hubs are concentrated in the Dianqian, Heshan, Jiangtou, and Jialian subdistricts, where population density is high. Consequently, when flood events occur, these areas incur greater population and economic losses. By contrast, zones near water bodies and foothills can discharge floodwaters promptly or absorb them in situ, resulting in minimal losses and generally low demand levels.

3.2. Results of Supply–Demand Matching and Coupling–Coordination

3.2.1. Results of Supply–Demand Matching

Based on Z-score standardization, the spatial distribution of urban FRS supply–demand matching in the spatial quadrants is shown in Figure 6a. Quadrants I–IV correspond, in order, to high supply–high demand, low supply–high demand, low supply–low demand, and high supply–low demand. Among these, low supply–high demand parcels indicate severe supply–demand mismatch and are deemed to have higher planning urgency; there are 1815 such parcels, accounting for 11.90% of the area. High supply–high demand parcels indicate that although demand is elevated, supply can still maintain a present balance; however, this balance is relatively fragile and may readily tip into mismatch, making these areas priorities for preventive control and structural optimization. There are 1567 such parcels, accounting for 11.87% of the total area. By contrast, low supply–low demand and high supply–low demand parcels, to some extent, help maintain the balance between FRS supply and demand and display stronger adaptive and responsive capacity to urban flood risk.

The supply–demand matching results are spatially mapped in ArcGIS (Figure 6b). By subdistrict, the area of each class and its share of the study area are summarized to reveal fine-scale spatial differences (

Table 2. Calculations of the area and proportion of each supply and demand matching type according to the street statistics.

Subdistrict Name	Low Supply- Low Demand		Low Supply- Low Demand		High Supply- Low Demand		High Supply- High Demand
	Area (km2)	P (%)	Area (km2)	P (%)	Area (km2)	P (%)	Area (km2)	P (%)
Binhai	4.094	2.86	9.093	6.35	1.584	1.11	1.022	0.71
Heshan	8.388	5.85	2.365	1.65	0.714	0.50	4.888	3.41
Zhonghua	0.348	0.24	0.748	0.52	0.414	0.29	0.015	0.01
Dianqian	11.686	8.16	2.820	1.97	0.898	0.63	2.114	1.48
Lianqian	5.994	4.18	13.929	9.72	3.063	2.14	2.231	1.56
Jinshan	3.699	2.58	4.299	3.00	5.299	3.70	0.819	0.57
Huli	2.679	1.87	4.858	3.39	1.791	1.25	0.051	0.04
Jiangtou	4.052	2.83	0.327	0.23	0.385	0.27	1.400	0.98
Jialian	3.441	2.40	0.000	0.00	0.000	0.00	1.135	0.79
Kaiyuan	2.350	1.64	3.591	2.51	0.541	0.38	0.529	0.37
Yundang	3.895	2.72	3.607	2.52	0.951	0.66	1.043	0.73
Xiangyu	1.277	0.89	4.526	3.16	0.325	0.23	0.007	0.00
Xiagang	0.425	0.30	0.895	0.62	0.285	0.20	0.000	0.00
Wucun	2.563	1.79	1.715	1.20	0.016	0.01	1.628	1.14
Lujiang	0.000	0.00	0.000	0.00	0.000	0.00	0.021	0.01
Total	56.114	39.17	53.090	37.06	17.011	11.87	17.054	11.90

). The results indicate that low supply–high demand parcels are concentrated mainly in the central and northern zones, with local clusters in the southeast. Among these, Heshan subdistrict covers 4.888 km² (3.41%). Next are the western Lianqian, Dianqian, and Wucun subdistricts, accounting for 1.56%, 1.48%, and 1.14%, respectively. These subdistricts encompass large areas of high-intensity built-up land and functionally dense clusters, reflecting systemic shortcomings in basic infiltration and detention, as well as ecological carrying capacity. High-supply, high-demand parcels are concentrated primarily in Jinshan subdistrict (approximately 5.299 km²). Benefiting from Lakeside Reservoir, Huzai Mountain, and Wuyuan Bay Wetland Park, Jinshan maintains a strong supply base alongside high demand. Low supply–low demand and high supply–low demand parcels are broadly distributed across Xiamen Island. Overall, the spatial matching of FRS supply and demand exhibits a pattern of “pronounced mismatch in the central and northern core, with sufficient supply along the periphery.”

3.2.2. Supply–Demand Coupling–Coordination Degree

This study employs the coupling degree model and the coupling–coordination degree model to evaluate the interaction between the supply and demand subsystems in the FRS assessment. The coupling degree and coupling–coordination degree between subsystems are calculated as C* and D*, respectively. Using the natural breaks method, C* and D* are divided into five intervals (

Table 3. Classification of coupling degree and coupling–coordination degree.

Classification of Coupling Types		Classification of Coupling–Coordination Degree
Coupling Degree C* Value	Coupling Type	Coupling–Coordination Degree D* Value	Coupling–Coordination Type
C* ∈ [0, 0.55]	Low coupling	D* ∈ [0, 0.20]	Seriously discoordination
C* ∈ (0.55, 0.77]	Medium-low coupling	D* ∈ (0.20, 0.40]	Slightly discoordination
C* ∈ (0.77, 0.85]	Medium coupling	D* ∈ (0.40, 0.60]	Basically coordinated
C* ∈ (0.85, 0.90]	Medium-high coupling	D* ∈ (0.60, 0.80]	Well-coordinated
C* ∈ (0.90, 1.00]	High coupling	D* ∈ (0.80, 1.00]	Highly coordinated

). Based on these intervals, a type-distribution matrix of coupling degree and coupling–coordination degree is constructed to determine the category of each subdistrict (Figure 7).

The results indicate that the 15 subdistricts of Xiamen Island are distributed across five matrix levels, resulting in seven basic types. Low values of coupling degree and coupling–coordination degree are concentrated in Jialian, Jiangtou, and Heshan subdistricts, with coupling degrees of 0.334, 0.348, and 0.380 (average 0.354), indicating a low-coupling state; the corresponding coupling–coordination degrees are 0.104, 0.144, and 0.146 (average 0.131), reflecting serious discoordination. By contrast, high values of coupling and coupling–coordination are mainly found in Binhai, Jinshan, Yundang, and Lianqian. Among them, Binhai exhibits a coupling degree of 0.938 and a coupling–coordination degree of 0.810, indicating high coupling and high coordination. In Binhai, the supply and demand of FRS form a positive feedback relationship, in which ecosystem service supply effectively supports socioeconomic flood-prevention needs. Yundang, Lianqian, and Jinshan also show high coupling and good coordination, with mean values of 0.913 and 0.757, respectively. Overall, the spatial pattern of the FRS supply–demand system on Xiamen Island is characterized by serious discoordination in the central area and high coordination in the east and south, consistent with the spatial trends of supply–demand balance.

3.3. Spatial Prioritization of Planning Interventions

In this study, the PRI values of “low supply–high demand” areas are classified into five levels using the natural breaks method and, in descending order of value, are denoted as level I–level V, where level I represents the highest intervention priority and level V represents the lowest. The spatial distribution pattern of each level is shown in Figure 8.

The intervention priority overall shows a spatial pattern that decreases from the central part of Xiamen Island toward the periphery. Parcels with level I priority intervention are mainly concentrated in Jialian Subdistrict, reaching 0.312 km², which is the core area where the current supply–demand contradiction is most prominent. Level II–III priority intervention zones are mainly distributed in Yundang, Wucun, and Heshan subdistricts, among which the total area of level II and level III priority intervention zones in Heshan Subdistrict reaches 1.298 km², significantly higher than in other subdistricts. Level IV–V priority intervention zones include the Heshan, Lianqian, and Dianqian subdistricts, with total areas of 3.590 km², 1.948 km², and 1.942 km², respectively. Overall, as the area with the highest concentration of level I priority intervention parcels, Jialian Subdistrict should be the first to be allocated more intensive and time-sensitive comprehensive governance measures to quickly alleviate the core supply–demand contradiction and play a demonstrative role; meanwhile, Heshan Subdistrict has the largest total area of priority intervention zones across all levels, reaching 4.888 km², and implementing graded and categorized governance there is expected to optimize the overall flood-risk pattern on a larger scale.

After overlaying the PRI classification results with the LULC of Xiamen Island, marked differences in land-use structure across priority levels become evident (Figure 9). In level I priority-intervention zones, industrial and transportation land together account for about 91%, while ecological land is only about 6%. Highly hardened underlying surfaces and limited detention space are key factors driving the pronounced increase in flood risk for these parcels. In level II–III zones, the shares of industrial and transportation land decline, whereas residential, public service, and commercial land increase notably—37%, 14%, and 16% in level II, and a combined ~46% in level III. The land-use composition of levels II and III reflects broad exposure risks arising from population and urban–functional agglomeration. Compared with level I, these two levels show a clear rise in ecological land—roughly 25%–29%—but the concentration of high-density population and urban functions keeps exposure and vulnerability elevated, so the risk-intervention level remains high. In levels IV–V, residential and public-service land still dominate, but the overall share of ecological land remains relatively high, with a notably increased proportion of water bodies—4% in level V. Overall, when the internal share of ecological land rises from the extremely low level of level I to moderate or higher, it can partially offset the runoff-concentration effect generated by construction land, thereby providing some buffering of flood risk.

4. Discussion

4.1. Advantages of the Research Method

This study takes FRS supply–demand matching as the entry point. By improving the indicator systems for ecosystem-function supply and socioeconomic demand, harmonizing indicators of different types and units, and strengthening the depiction of flood processes, it constructs an FRS supply–demand assessment framework suitable for the urban scale, providing a more systematic technical pathway for the identification and regulation of urban flood risk.

On the FRS supply side, numerous studies have demonstrated the pivotal role of multiple ecosystem services in flood regulation and water provision processes [55]. Compared with prior work that relied on distributed hydrological models emphasizing flood-mitigation capacity, this study adopts the InVEST model to integrate three ecosystem functions closely related to flood processes—WR, SR, and NR—thereby encompassing both flood-prevention and flood-mitigation services of urban ecosystems and producing estimates closer to the actual FRS supply capacity. On the FRS demand side, the H-E-V framework has been widely applied to complex flood-risk assessment. Considering risk characteristics at the urban scale, inundation depth, extent, and duration are commonly used to characterize flood hazard [56]. Accounting for the risk profiles of population and critical infrastructure under flood impacts captures potential losses to exposed elements, while post-event building structural integrity and property damage inform potential economic losses [57]. Integrating indicators of flood hazard, population and critical-infrastructure exposure, and the vulnerability of social assets to quantify the magnitude and spatial distribution of FRS demand enables a more comprehensive understanding of socioeconomic needs for FRS.

At the urban scale, FRS supply–demand assessment requires higher spatial precision and finer process representation [58]. Prior studies commonly use grid [59], subdistrict [60], or functional-zone [61] units to summarize and calculate potential surface characteristics and socioeconomic densities; however, larger analytical units struggle to capture gradients in risk accurately, thereby affecting the precision of FRS demand assessment. By contrast, the parcel, as the smallest management unit in urban planning and construction [62,63], can depict LULC spatial patterns and functional differences in detail and aligns with the way statistics on population, social services, and assets are organized [64]. Choosing parcels as the basic analytical unit can substantially remedy the above limitations, enabling differentiated identification of flood risk [65]. When comparing FRS supply and demand, there are large differences in the dimensions of ecosystem function and socioeconomic indicators, and incompatibilities exist among their spatial data units. To address this, the raster-based FRS supply is spatially mapped to the parcel scale, enabling supply-demand comparison within a unified parcel unit. In the matching stage, Z-score standardization and quadrant partitioning methods are introduced [66], allowing indicators of different types and magnitudes to be compared on the same scale, thereby identifying multiple supply–demand combinations and mismatch units. This makes the assessment results directly actionable for urban planning and regulatory schemes and enhances the method’s operability and practical value.

4.2. FRS Supply-Demand Risk Assessment and Policy Recommendations

Xiamen Island exhibits pronounced spatial heterogeneity in FRS supply–demand patterns, characterized overall by “marked mismatch in the central and northern areas, with sufficient supply along the periphery.” This configuration results from the combined effects of supply and demand. On the supply side, LULC types—through WR, SR, and NR—significantly influence ecosystems’ detention and purification capacity: areas with more green space, forest land, and water bodies show stronger supply. The southern and southwestern parts of the island are dominated by forest, and the northeastern sector contains extensive water bodies; these regions have higher shares of ecological land and thus superior supply capacity. By contrast, in the central–northern area, impervious surfaces extensively encroach on ecological space, constraining supply. On the demand side, patterns are driven mainly by inundation depth, population density, points of interest, and building distributions. Compared with natural factors, socioeconomic factors are more complex and dispersed in high-density built-up zones, yet hotspots of elements at risk still show a center-to-periphery declining trend. Accordingly, the spatial distribution of FRS demand mirrors this pattern.

Against this backdrop, a large number of parcels in the central–northern part of Xiamen Island exhibit a low supply–high demand mismatch, with LULC dominated by impervious surfaces and flood risk clearly higher than in surrounding areas. In these areas, WR, SR, and NR are the weakest across the island, while carriers such as population, points of interest, and buildings are most severely impacted by flooding, resulting in the greatest FRS demand. The combination of the weakest FRS supply and the strongest FRS demand places these areas at the highest flood risk.

For Jialian subdistrict, which has the highest priority within the central area’s FRS supply–demand profile, large-scale reductions in industrial, residential, and public-service land to enhance supply are impractical under the existing high-intensity development pattern; instead, optimization should focus on improving drainage and detention capacity in the foundational drainage system. In low-lying areas, the flood-discharge capacity of the local pipe network should be comprehensively assessed for storm conditions; where systems do not meet design standards, pipe diameters should be appropriately increased or auxiliary pipelines added. Meanwhile, transportation corridors, green squares, and residential neighborhoods should, wherever possible, adopt permeable pavements—such as permeable bricks, permeable asphalt, or permeable concrete—to increase infiltration and avoid the accumulation of surface runoff. Adding rooftop gardens or vertical greening to expand the vegetated system can help store and attenuate floodwater. For Yundang, Wucun, and Heshan subdistricts—where level II–III priority interventions predominate—ecological restoration and structural adjustments should take precedence: during the renewal of mixed residential–commercial precincts and commercial centers, strictly control any new encroachment on ecological land and enhance green-space connectivity; embed vegetated buffer belts and rain gardens within industrial parks and at transportation nodes to compensate for ecological-function deficits. Level IV–V lower-priority subdistricts include Lianqian in the southeast and Dianqian in the north. Given their generally favorable ecological baselines, these areas should strengthen ecological redlines and binding controls and prudently guide eco-agriculture, cultural tourism, and environmentally friendly industries to maintain robust ecological foundations and independent disaster resilience.

To comprehensively resolve the mismatch in FRS supply–demand, it is necessary to integrate Xiamen Island’s specific conditions and spatial pattern and to advance efforts in a coordinated manner from two dimensions: optimizing urban spatial structure and reorganizing functional layouts [67]. This study proposes reallocating part of the core urban functions within the island to surrounding areas to reduce the pressure arising from residential, commercial, industrial, and transportation land in higher-priority intervention zones; to restore the natural functions of fragmented and scattered green spaces; and to offset the stock deficits of agriculture and forestry land caused by compression and encroachment—thereby establishing a balanced development model with mutual coordination. This approach can prevent excessive agglomeration of commercial–residential uses in the urban core, balance resource allocation, and improve overall urban efficiency to achieve FRS supply–demand balance.

4.3. Limitations and Prospects

This study has several limitations that warrant further investigation. The results are unavoidably affected by uncertainties in model assumptions and parameter settings. First, the estimation of FRS supply and demand depends on predefined model structures and simplifying assumptions. For example, we assess ecosystem supply capacity using only WR, SR, and NR, whereas other ecosystem services related to flood processes could also be considered. Climate regulation services [68], for instance, moderate atmospheric CO₂ levels through carbon storage and sequestration, thereby reducing, to some extent, the frequency and intensity of extreme weather events. At the parcel scale, the H-E-V indicators include only a limited set of exposure and vulnerability variables [69]. Second, the 50-year return period rainfall scenario adopted here does not fully capture uncertainties in future precipitation under different climate models, potentially leading to under- or overestimation of future flood risk and FRS demand. Consequently, the spatial patterns identified in this paper should be interpreted as one plausible outcome under the stated assumptions, rather than a precise forecast. Future research should incorporate multiple climate scenarios, model structures, and parameter combinations and conduct sensitivity analysis or uncertainty-propagation analysis to more systematically characterize how modeling assumptions and climate projections influence the assessment results.

5. Conclusions

Centering on balancing FRS supply and demand, this study enhances the comprehensiveness of urban flood-risk indicators and the accuracy of their assessment scope while precisely identifying fine-scale spatial features of supply–demand and pinpointing critical areas. Empirical results for Xiamen Island indicate an overall spatial pattern of “pronounced mismatch in the central and northern areas, with sufficient supply along the periphery.” In the central–northern zone—particularly Heshan, Wucun, Jialian, and Jiangtou—numerous low supply–high demand parcels lead to markedly higher flood risk than elsewhere. The root causes are dense population and commercial activity, a high proportion of impervious surfaces, and insufficient ecological function, which together severely limit the system’s capacity to cope with flood hazards. By contrast, most peripheral areas possess ample water bodies and green space, robust natural detention capacity, and lower levels of socioeconomic activity and industrialization, yielding higher supply and lower demand and thus substantially reduced flood risk. The study elucidates how the natural ecological base and socioeconomic layout jointly shape urban flood risk. Based on risk stratification, we propose differentiated management strategies that emphasize expanding ecological land and green infrastructure, reducing impervious coverage, and optimizing drainage systems—thereby jointly improving urban flood resilience and the quality of spatial development and offering insights for flood management and long-term planning on Xiamen Island.