The supply of food production at the grid scale was determined based on total grain yield and the NDVI values of cultivated land. The demand for food production at the grid scale was calculated using population density and per capita grain consumption (Whitcraft et al., 2015).

The supply of carbon sequestration was estimated using Net Primary Productivity (NPP) as the core indicator, combined with a CO₂ conversion coefficient. Urban CO₂ emission data served as the basis for assessing demand. Carbon emissions were spatially coupled with population data through spatial disaggregation to generate gridded carbon demand (Yu et al., 2021).

The supply of soil conservation was assessed using the Revised Universal Soil Loss Equation (RUSLE) model by quantifying the difference between potential and actual soil loss. Actual soil loss was used as an indicator of demand (Zhao et al., 2024).

The habitat quality module within the InVEST model was used to assess habitat quality supply, using raster cells as the evaluation units. Based on land-use type data and threat source information, this module quantified the impact of various threat factors on habitat patches to evaluate habitat quality. It comprehensively considered factors such as the distance-decay effect of threat sources, their spatial weighting, and the degree of legal protection, and calculated habitat quality based on the habitat degradation index (Jiang, 2022). For assessing habitat quality demand, with reference to relevant studies, ecosystem degradation is primarily influenced by three major factors: population distribution, land-use development intensity, and economic development level (Wu et al., 2022). Therefore, the proportion of construction land, population density, GDP per unit area, and nighttime light intensity were selected to represent habitat demand.

This study employed the water yield module of the InVEST model to assess the supply of water yield services based on the principle of water balance. The model quantified the water budget by calculating the difference between precipitation and actual evapotranspiration for each raster unit, thereby deriving regional water yield potential (Zhou et al., 2024). The spatial assessment of water yield service demand was achieved by integrating data from water resources bulletins and statistical yearbooks. Specifically, domestic and industrial water use, agricultural irrigation water use, and ecological water demand were spatially linked to corresponding land-use types: construction land for domestic and industrial water use, cultivated land for agricultural water use, and ecological land for ecological water demand. Subsequently, spatial disaggregation was applied to allocate these water demand categories to the grid scale, generating the spatial distribution of water yield service demand (Chen et al., 2020) (Table 1).

In 2020, the total supply of ESs in Guangxi was quantified as follows: food production at 1.37 × 10⁷ t, carbon sequestration at 1.48 × 10¹⁰ t, soil conservation at 5.14 × 10⁸ t, water yield at 2.02 × 10¹¹ m³, with a mean habitat quality index of 0.37. The supply of all services exhibited pronounced spatial heterogeneity (Figure 2).

At the municipal scale, high food supply was concentrated in Nanning, Guilin, Yulin, and Guigang, which together accounted for 51.93% of the total provincial grain output, aligning with key agricultural planning zones. In contrast, cities such as Liuzhou and Laibin showed lower supply levels due to limited cultivated land or an economic structure oriented toward secondary and tertiary industries. Beihai and Fangchenggang exhibited the weakest supply capacity, constrained by limited planting area. Carbon sequestration, soil conservation, and habitat quality displayed highly similar spatial patterns. The core high-value areas were Baise and Hechi within the western and northern ecological barrier zone, which contributed 28.33 and 42.80% of the provincial totals for carbon sequestration and soil conservation, respectively. This underscores the pivotal role of key ecological source areas in regional service supply. Water yield exhibited a clear decreasing gradient from north to south. High-value areas were located along the Guilin-Baise corridor, characterized by abundant precipitation and high vegetation coverage. Guilin and Hechi alone accounted for 35.26% of the total provincial water yield, whereas the southern coastal cities of Fangchenggang and Beihai contributed only 4.23%.

At the county scale, key supply patterns and local variations obscured by municipal-scale averaging became evident. Considerable polarization was observed even within the same municipal administrative unit. For instance, food production remained highly concentrated in major agricultural counties such as Quanzhou County under Guilin City, whereas core urban districts like Xiufeng District in Guilin exhibited extremely weak supply, reaching only 2.1 × 10³ t. High-value areas for carbon sequestration, soil conservation, and habitat quality were mainly distributed in counties in northwestern and northeastern Guangxi, such as Tianlin County in Baise and Huanjiang Maonan Autonomous County in Hechi, representing core supply zones for these ecosystem services. The north–south differentiation in water yield remained evident, with high-value counties such as Rongshui Miao Autonomous County and Quanzhou County located in the northern mountainous areas. Analysis at this scale provides direct spatial targets for implementing differentiated ecological management policies at the county level.

At the grid scale, micro-scale heterogeneity previously masked by administrative boundaries was revealed. Food supply showed a north-high–south-low spatial trend, contrasting with the southeastern dominance observed at municipal and county scales, which may be related to subtle differences in local climate, field management, and crop structure. High-value areas for carbon sequestration, soil conservation, and habitat quality were concentrated in key ecological patches such as forested land. Water yield generally decreased from north to south, with several high-value units identified in topographically favorable areas such as coastal zones and river valleys, reflecting significant regulation by micro-topography on the water yield process.

In 2020, the total demand for various ecosystem services in Guangxi was as follows: food production at 6.75 × 10⁶ t, carbon sequestration at 4.54 × 10⁹ t, soil conservation at 3.01 × 10⁷ t, water yield at 2.29 × 10¹¹ m³, with a mean habitat quality index of 0.47. The spatial heterogeneity of service demand was significant (Figure 3).

At the municipal scale, the demand for food production was highest in Nanning and Yulin, accounting for 26.36% of the total provincial demand, reflecting the population and economic agglomeration effects of these urban growth poles. Carbon sequestration demand exhibited differentiated distribution patterns, with high-demand areas such as Laibin and Guigang each exceeding 3.55 × 10⁷ t. This pattern was primarily attributable to the concentration of high-energy-consuming industries, notably mining and building materials, within these regions, which had resulted in elevated carbon emissions. Concurrently, elevated carbon sequestration demand was also observed in Hezhou and Baise. Hezhou, as a vital ecological functional zone in central-northeastern Guangxi, and Baise, serving as a crucial ecological barrier in northwestern Guangxi, exhibit elevated carbon sequestration demands that largely reflect their strategic positioning and conservation requirements within the provincial ecological security framework for maintaining and enhancing carbon sink functions. In contrast, Liuzhou, the largest industrial center, showed relatively lower demand at 2.68 × 10⁷ t, attributable to its ongoing green and low-carbon industrial transformation. High demand for soil conservation was concentrated in western and northern Guangxi, such as Baise and Hechi. These areas are characterized by typical karst landscapes with exposed bedrock, shallow soil layers, and consequently high soil erosion risks. Notably, Baise recorded the highest soil conservation demand, accounting for 15.43% of the total provincial demand. The habitat quality demand index exceeded 0.8 in areas such as Baise, Nanning, and Guilin, indicating substantial human-induced pressure on natural habitats. Water yield demand was strongly correlated with economic development and agricultural scale. As the economic and demographic cores, Nanning and Guilin showed the highest demand, whereas Beihai and Fangchenggang, with port- and tourism-oriented industrial structures, registered the lowest demand.

At the county scale, significant disparities within administrative units were revealed. Taking grain demand as an example, municipal-scale assessments indicate Beihai City’s overall demand is relatively low at 20.58 × 10⁴ t. However, county-scale analysis revealed that Hepu County, under its jurisdiction, exhibited notably high demand, reaching 12.23 × 10⁴ t, ranking among the highest across the province. Across the province, food demand is highly concentrated in major plain agricultural production areas such as Guigang under Guiping City and Bobai County underYulin, while regions like Dongxing City, Gangkou District, and Yanshan District exhibit extremely low demand. Soil conservation demand in ecologically fragile mountainous counties, such as Tianlin County, exceeds that in urban core areas like Diecai District under Guilin by more than a thousandfold, highlighting the rigid constraints imposed by ecological conditions on demand. High-value zones of water yield demand are predominantly distributed across agriculture-dominated counties in central and southeastern Guangxi, characterized by plains and hills with dense populations, leading to elevated water requirements. Conversely, low-value zones are concentrated in the western and northern mountainous regions, where development intensity is low and agricultural water use is limited.

At the grid scale, demand exhibited fine-grained spatial clustering that corresponded closely with underlying geographical factors. Food demand formed distinct high-value belts in the urbanized central and southeastern regions, closely aligning with grids of population density. High-value patches of carbon demand were found not only in industrial zones but also widely across agriculturally active hilly areas. The spatial patterns of soil conservation demand and supply were similar overall, showing higher levels in the west and north compared to the central and eastern regions. Habitat quality demand displayed a “high in the center, lower in the periphery” spatial structure, with high-demand grids concentrated in areas undergoing agricultural expansion and urban construction. High water yield demand was not limited to major urban built-up areas but was also distributed across agricultural land, accurately reflecting the spatial distribution of agricultural and domestic water use.

Analysis of supply–demand relationships indicates that the spatial pattern of ES supply–demand matching in Guangxi during 2020 exhibits strong scale dependency.

At the municipal scale, to align with macro-management requirements, this study employs actual supply–demand differentials for assessment. Since habitat quality is a dimensionless index, its supply–demand ratio is used for characterisation. Results indicate that food production exhibits a surplus across all prefecture-level cities, with particularly pronounced surpluses in basin-plain agricultural hubs such as Nanning, Guilin, Yulin, and Guigang. In contrast, Fangchenggang and Beihai exhibit smaller supply–demand surpluses, primarily due to their location within the Beibu Gulf coastal ecological functional zone. These cities have limited arable land resources, smaller-scale agricultural production, and industrial structures dominated by non-agricultural sectors such as ports and tourism, resulting in relatively weaker grain supply capacity. The high-value zones for carbon sequestration, soil conservation, and habitat quality supply–demand surpluses largely coincide with their respective supply surplus areas. Carbon sequestration services exhibited deficits solely in Beihai City. Soil conservation services demonstrated surpluses across all prefecture-level cities. Habitat quality services showed deficits in multiple cities, particularly in rapidly developing economic centres such as Nanning, Liuzhou, and Yulin, where the supply–demand ratio ranged between −0.25 ~ 0. This indicates that accelerated urbanization has led to significant degradation of natural ecological functions. Water yield services exhibited deficits in nine central and southern prefecture-level cities including Nanning, Guigang, and Yulin, highlighting water resource pressures in urbanized and agricultural areas with high water consumption (Figure 4).

Analysis of the supply–demand ratio at the county scale reveals localised contradictions obscured by the municipal-scale conclusion of an overall surplus. Grain production exhibits a supply surplus in Wuming District under Nanning, Guiping City and Pingnan County under Guigang, and Bobai County under Yulin, whereas widespread deficits are evident in Qingxiu District under Nanning, Yuzhou District under Yulin, and Diecai and Qixing Districts under Guilin. This reflects the displacement effect of urbanization on cultivated land resources. Deficits in carbon sequestration services are concentrated in centrally located counties with higher industrialisation and urbanization levels, as well as in certain coastal regions. Deficits in soil conservation, habitat quality, and water yield services are particularly pronounced in central and southern areas, whereas western and northern mountainous regions exhibit surpluses due to relatively stable ecological conditions.

Grid-scale analysis reveals micro-patterns of supply–demand relationships across continuous spatial domains. Grain production demonstrates surpluses across most regions, though deficits emerge in certain southern and central grid cells. Soil conservation services exhibit relative equilibrium in central and northwestern Guangxi, whilst deficits occur in grids with high soil erosion risk in the south and southeast. Habitat quality and water yield services face widespread deficits in southeastern and southern grids, highly coinciding with areas of intensive agricultural development and urban land expansion.

Previous researches have revealed the macro-level patterns of ES supply and demand in Guangxi. For instance, Xie et al. (2020) concluded that overall ES supply trend in Guangxi is relatively higher in the north than in the south, while ES demand shows the opposite pattern, being lower in the north than in the south, with high ES demand values clustered in the southeast. The findings of this study align with this conclusion, as the supply and demand patterns of the five key ES indicators selected all exhibited similar trends. Concurrently, Xie Yuchu et al. found that at the county scale, H-L and L-L ES supply–demand clusters were distributed in northern Guangxi, while H-H and L-H areas tended to be concentrated in the southeast. In this study, apart from soil conservation, the other ESs exhibited similar spatial clustering characteristics at the county level. This further indicates that our findings reliably reflect the ES supply and demand patterns in Guangxi. Compared to existing research, this study innovates by overcoming the limitations of single-scale analysis, establishing a nested multi-scale analytical framework encompassing “municipal-county-grid” scales. The results indicate that the matching patterns of ecosystem service supply and demand exhibit significant scale dependency. At the municipal level, overall surpluses or deficits in ecosystem services may mask severe localised imbalances. For instance, while Nanning and Guilin exhibit overall surpluses in grain production supply and demand, county-scale districts such as Qingxiu in Nanning and Diecai and Qixing in Guilin face deficits. At the county and grid scales, micro-level heterogeneity gradually emerges. Taking water yield services as an example, supply–demand matching exhibits a significant positive correlation at the city level, whereas it shifts to a significant negative correlation at the county level. This underscores that relying solely on a single-scale analysis may lead to biases in ecological management decisions. Adopting a multi-scale perspective facilitates a more comprehensive identification of spatial targets for supply–demand mismatches, providing a more layered and targeted scientific basis for macro-level planning and micro-level governance.

GWR models reveal significant spatial non-stationarity and scale dependence in the influence of natural and anthropogenic factors on the supply–demand relationship of ES. GWR results indicate that the intensity and direction of the same influencing factor vary considerably across different geographical locations, reflecting the regional specificity of spatial coupling mechanisms between natural and human systems. For instance, the positive effect of forest cover ratio on balancing soil conservation service supply and demand is most pronounced in the western ecological barrier zone. This stems primarily from the region’s high vegetation cover, where forests enhance soil and water conservation functions through ecological processes such as canopy interception and root system stabilization. Conversely, in the urbanized southeastern regions, intensive human activities have led to landscape fragmentation. Forest patches become isolated by developed land, disrupting ecological processes and compromising their functions. Consequently, their regulatory role in ES supply and demand is significantly weakened or even negligible. This spatial heterogeneity indicates that a one-size-fits-all management policy may yield limited results, necessitating regionalised strategies tailored to local driving mechanisms. Previous studies have demonstrated that different factors influence ESs at varying scales (Cui et al., 2019). This study further reveals that natural baseline factors (such as precipitation and forest cover proportion) exhibit relatively prominent explanatory power at the municipal scale, reflecting that at macro-management units, ES supply capacity remains heavily constrained by regional natural conditions. At the county level, driving mechanisms diverge. For supply-oriented services like food production and soil conservation, natural geographical factors such as DEM play a decisive role; whereas for demand-driven services like carbon sequestration and habitat quality, factors indicating human activity—such as night-time light indices and proportion of built-up land—become dominant. This confirms that at the county level, as the fundamental unit for China’s socio-economic planning and implementation, human activities directly reshape land cover and ecological processes through land use planning and industrial layout, thereby becoming key variables influencing the ES supply–demand relationship. Human activity factors such as population density and night-time light intensity exert a more pronounced influence at the grid scale, even becoming dominant in water yield services. Concurrently, the decisive role of cultivated land proportion in shaping grain production supply–demand dynamics becomes more pronounced at the grid scale. This demonstrates that high-resolution analysis can reveal spatial heterogeneity within administrative units, simultaneously capturing micro-pattern variations in natural elements and spatial clustering characteristics of human activities. Consequently, it more accurately reflects localised processes of human-land interaction.

This study offers significant insights for spatial governance and ecological civilisation development in the new era: regarding governance philosophy, it is imperative to transcend the homogenised management paradigm and implement differentiated strategies based on the coupling characteristics of regional natural endowments and human activity intensity. Specifically, in the ecologically dominant northwestern mountainous regions, priority should be given to maintaining and restoring functions such as water conservation and soil preservation; In the densely populated urban areas of the southeast, where human demands are concentrated, strict controls on the expansion of construction land are required, alongside efforts to enhance the quality and connectivity of ecological spaces. Regarding governance methods, a multi-tiered, nested monitoring and evaluation system spanning “municipal–county–grid” scales should be established. Research indicates significant variations in the dominant contradictions and driving mechanisms of ES supply–demand relationships across different scales. Consequently, policy formulation must correspond to the appropriate scale: the municipal scale is suitable for ecological functional zoning and cross-regional coordination; the county scale can identify internally imbalanced areas and delineate key ecological restoration zones; while the grid scale provides direct targets for plot-level ecological engineering deployment. In practical implementation, spatial explicit models such as geographically weighted regression can be employed to spatially visualise the intensity and direction of driver impacts. This provides evidence for optimizing ecological conservation redlines, cross-regional ecological compensation, and refined land use regulation, thereby facilitating a transition from “empirical” planning to a data-driven, spatially intelligent ecological governance paradigm.

Whilst this study systematically assessed multi-scale ES supply–demand relationships, certain limitations remain. Firstly, although models such as InVEST and RUSLE are widely applied, their parameters primarily rely on literature and generic datasets. Integrating locally measured data from Guangxi could enhance research precision in future work. Secondly, this study integrates diverse datasets including remote sensing, meteorological, soil, and socio-economic information. Differences in spatial resolution and temporal consistency among these datasets, particularly during spatial downscaling and aggregation processes, may impact analysis outcomes at finer scales (e.g., grid cells). Thirdly, this study constitutes a static cross-sectional analysis, failing to reveal the spatiotemporal dynamics of ES supply and demand. More critically, it focuses on quantitative matching and spatial co-location of supply and demand within individual scale units, without quantifying the actual flows and feedback mechanisms of ES (e.g., water yield, carbon sequestration) between different scale units. This is essential for understanding regional ES networks and formulating cross-regional ecological compensation mechanisms. Finally, the geographically weighted regression revealed heterogeneous influences of factors, necessitating particular attention to localised landscape contexts, environmental resources, socio-economic characteristics, and other drivers to formulate efficient management measures tailored to local conditions. Future work could couple ecological models (such as InVEST) with socio-economic scenarios to simulate ecosystem service dynamics under different development pathways, incorporating local knowledge and governance preferences into ecosystem service assessments to enhance the practical relevance of spatial planning recommendations.