Optimizing Landscape Patterns for Tea Plantation Agroecosystems: A Case Study of an Important Agricultural Heritage System in Enshi, China

Abstract

The agroecosystems of tea plantations play a significant role in regional ecosystem services, with some recognized as Important Agricultural Heritage Systems. Despite notable progress in conserving these unique agricultural landscapes, systematic approaches to delineating the core conservation zone and establishing robust ecological networks for agricultural heritage systems remain insufficient. This study employed the Enshi Yulu Tea Agricultural Heritage System as a case study, integrating the MaxEnt model, InVEST model, and circuit theory to quantitatively assess landscape connectivity and prioritize conservation efforts. The analysis delineated a core conservation zone of 718.04 km² for tea plantations, identified 77 ecological corridors, and pinpointed 104 critical ecological nodes. The results indicate 43.96 km² of synergistic areas between tea plantations and ecological sources, demonstrating that the agroecosystems of tea plantations provide higher ESs values compared to monoculture plantations and farmlands. In addition, an ecological optimization framework featuring a “four belts and four zones” spatial configuration was proposed, aimed at enhancing connectivity and promoting the sustainable development of tea plantation agricultural heritage. The proposed framework can provide evidence-based references for future policy formulation, and deliver actionable insights for land-use planning, habitat restoration, and infrastructure mitigation.

1. Introduction

In 2002, the Food and Agriculture Organization of the United Nations (FAO) defined Globally Important Agricultural Heritage Systems (GIAHS) as “Remarkable land use systems and landscapes which are rich in globally significant biological diversity evolving from the co-adaptation of a community with its environment and its needs and aspirations for sustainable development” [1]. Agricultural heritage is an integrative holistic system, with complex connections between various elements within the system. Following FAO, the Ministry of Agriculture and Rural Affairs (MARA) of the People’s Republic of China officially launched the protection work of China Nationally Important Agricultural Heritage Systems (China-NIAHS) in March 2012, describing this agricultural cultural heritage as a dynamic, adaptive, complex, strategic, multifunctional, and endangered system. As of May 2025, China has designated 188 China-NIAHS, among which 23 are tea plantation agricultural heritage systems (TPAHS).

TPAHS are sustainable systems that integrate tea plantations, forests, rivers, and residential areas, representing a harmonious co-evolution of human activities and ecosystems. Research has suggested that tea–forest agroecosystems demonstrate ecological value by sustaining vertebrate populations (e.g., elephants [2], bats [3]) and promoting angiosperm diversity [4]. While agriculturally dominated landscapes have constrained ecological functionality, their extensive distribution in the study area provides important connectivity for wildlife dispersal [5]. However, not all tea plantations demonstrate positive ecological benefits. Research by Biervliet et al. [6] revealed that tea cultivation has degraded stream habitat quality and biodiversity in the East Usambara region. Similarly, He et al. [7] identified moderate ecological risk levels from soil heavy metals and pesticide residues in tea gardens across Tibet, Guangdong, and Fuzhou. Arafat et al. [8] found that continuous tea monocropping reduces soil pH, diminishes beneficial microbial populations, and promotes pathogenic microbe proliferation.

TPAHS demonstrate marked vulnerability and endangerment. Urbanization, accompanied by the transformation of traditional production modes, has significantly impacted the sustainability of TPAHS. Forest conversion to urban and tea plantation land uses has reduced habitat quality, reflecting decreased biological abundance [9]. The loss of rural labor also leads to the abandonment of farmland and damages the integrity of the region’s heritage. The research findings of Xu et al. [10] found that 15.47% of farmland (including tea plantations) in Hubei Province was abandoned in 2014, with labor migration significantly increasing the probability of abandonment. In addition, the modern way of tea planting has also made the connection between tea farmers, villages, and tea plantations less close than traditional tea plantations, and indigenous knowledge and customs associated with tea production have also become scarce [11].

As a highly interdisciplinary field, agricultural heritage systems (AHS) have only been formally recognized for 23 years since their initial conceptualization. Existing research on AHS spans diverse domains, including tourism value assessment [12], food chemistry analysis [13], agro-environmental resource analysis [14,15], and indigenous knowledge systems analysis [16]. However, studies focusing on landscape patterns remain relatively scarce. While existing landscape-related studies have predominantly focused on farmland dynamics [17,18] or landscape characteristics [19], a truly holistic landscape optimization approach must encompass the integrated spatial configuration of the core structural elements defining AHS’s multifunctionality. As for TPAHS, the synergistic interplay of terraced tea plantations, adjacent forests, riparian networks, and villages collectively demonstrate a unique combination of productive value, ecological conservation significance, and aesthetic–cultural value. However, current conservation strategies often overlook these interconnected relationships, failing to address critical spatial synergies.

Furthermore, neither the FAO nor MARA of China provides specific guidance on how to identify the boundaries and core areas of agricultural heritage in their application guidelines. Most local governments applying for important agricultural heritage in China fail to accurately delineate the core conservation zone using required geographic coordinates and mapping methods, instead simplistically substituting administrative boundaries for eco-geographical regions when demarcating protected areas [20]. In contrast, the Italian Register of Historical Rural Landscapes stipulates that its boundaries need not align with administrative divisions, but must encompass all landscape features associated with the proposed landscape type [20]. UNESCO emphasizes that to effectively safeguard the integrity of heritage sites, appropriate boundaries must be delineated, particularly for core zones, which should include all attributes that convey the Outstanding Universal Value (OUV) [21]. To safeguard TPAHS integrity and strengthen agro-biodiversity protection, identifying core heritage elements and optimizing their spatial configuration is essential.

Research on delineating the core conservation zone of agricultural heritage typically considers two key aspects: landscape aggregation and ecological suitability. In terms of landscape aggregation, spatial metrics such as the area index and aggregation degree are commonly employed for quantification [22]. For ecological suitability, studies often adopt methods such as expert scoring and the analytic hierarchy process (AHP) to overlay and compute multiple environmental factors influencing crop growth, thereby identifying suitable zones [23]. However, landscape aggregation analysis is usually confined to the current distribution of crops and lacks dynamic projections for future development, which contradicts the inherently dynamic feature of agricultural heritage systems. In addition, expert scoring and the AHP method yield insignificant differences in weights between different factors, and they have a certain degree of subjectivity.

In 1995, the patch–corridor–matrix framework, systematically developed by Forman and Godron [24], established the theoretical foundation for analyzing landscape ecological patterns and structures. Subsequently, with the exploration of numerous scholars, the ecological network construction system of “Identifying source—Constructing resistance surface—Extracting corridor—Constructing ecological network” emerged and became mainstream [25]. Despite achieving methodological maturity, ecological network research still has some limitations. Some scholars have pointed out that ecological connectivity, landscape fragmentation, and the derived concept of ecological networks are affected by semantic ambiguity and are often dogmatically applied, potentially misleading conservation policies [26]. They recommend distinguishing between types of isolation and integrating ecological networks into stricter project cycle management, assessing priorities and cost–benefit trade-offs, and adopting modeling tools such as landscape graphs [27]. In addition, the construction of a resistance surface has two main approaches; one is to employ the reciprocal of habitat quality as resistance values [28] and another is to calculate the weighted sum of human and natural factors [29,30]. Neither of these methods can obtain resistance values well. The first method has limited consideration for resistance factors. The second method assumes a linear relationship between resistance factors and resistance values, without considering the characteristics of wildlife movement.

The Minimum Cumulative Resistance (MCR) model, first developed by Knaapen et al. [31], identifies potential corridors via the Least-Cost Path (LCP) method [32]. While this approach effectively delineates potential corridor distributions, it primarily reflects spatial connectivity rather than quantifying functional differences among corridors. To overcome these limitations, our study employs circuit theory, a transformative methodological advancement for corridor identification. Unlike traditional MCR, circuit theory simulates stochastic charge movement to model species dispersal, capturing multi-path diffusion patterns with ecological widths—better aligning with real-world biological movement dynamics [33]. The use of circuit theory is particularly appropriate for modeling dispersal in species with constrained landscape perception [34]. Implemented through Linkage Mapper and the Circuitscape module, this approach automates the detection of pinch points and barrier points while quantitatively assessing node sensitivity.

The Enshi Yulu Tea Agricultural Heritage System is one of the China-NIAHS, renowned not only for its unique “kill-green” tea processing technique but also for its multi-layered landscape pattern integrating tea plantations as the core element with forests, terraces, and water systems in the unique natural geography of the Wuling Mountains [35]. This study aims to enhance the landscape connectivity of the heritage site and optimize spatial relationships among its elements for regional biodiversity conservation. The study has four objectives: (1) to identify the core conservation zone of tea plantations with the MaxEnt model; (2) to assess ecosystem services (ESs) value and identify ecological sources with MSPA; (3) to generate the resistance surface by applying an exponential transformation to the predicted wildlife habitat suitability derived from MaxEnt; (4) to extract ecological corridors and analyze ecological pinch points and barrier points based on circuit theory, thereby proposing a framework for optimizing the ecological network. The construction of an ecological network and the delineation of the agricultural heritage core zone can provide research-based references for the conservation and management strategies of agricultural heritage.

2. Materials and Methods

2.1. Study Area

Enshi County $({109}^{\circ }{4}^{\prime }{48}^{″}-{109}^{\circ }{58}^{\prime }{42}^{″}\mathrm{E},{29}^{\circ }{50}^{\prime }{33}^{″}-{30}^{\circ }{39}^{\prime }{30}^{″}\mathrm{N})$ is located in southwestern Hubei Province, at the northern edge of the Wuling Mountains and the upper-middle reaches of the Qingjiang River, covering a total area of 3971.58 km² (Figure 1) [36]. It is an important global hot spot for biodiversity [37]. The region exhibits significant vertical climatic variation due to its dramatic elevation changes, with an average annual temperature of 8.8–16.7 °C and annual precipitation of 1400–1500 mm. Coupled with selenium-rich soils and a forest-based ecosystem, these conditions create an ideal environment for tea. This multilevel agroecosystem not only ensures the superior quality of Enshi Yulu tea but also underpins its value as an agricultural heritage system, recognized in 2015 by MARA as a China-NIAHS [38]. Tea is widely cultivated in Enshi, with a total tea plantation area of 258 km², according to statistics in 2024. Traditional Enshi Yulu tea plantations are typically established on forested slopes and valley bottoms, coexisting harmoniously with native flora, fauna, and microorganisms [35].

2.2. Data Sources

The study utilized various data sources (

Table 1. Data description.

Data	Data Type	Resolution	Data Sources
Tea tree data	Raster data	10 m	https://doi.org/10.6084/m9.figshare.25047308 (accessed on 3 February 2025)
Land cover data	Raster data	10 m	https://github.com/LiuGalaxy/CRLC (accessed on 3 February 2025)
Digital elevation model	Raster data	12.5 m	https://nasadaacs.eos.nasa.gov/ (accessed on 18 February 2025)
Soil dataset	Raster data	1 km	Harmonized world soil database v1.2: https://www.fao.org/soils-portal/data-hub/soil-maps-and-databases/harmonized-world-soil-database-v20/en/ (accessed on 21 February 2025)
Monthly precipitation	Raster data	1 km	https://data.tpdc.ac.cn/zh-hans/data/faae7605-a0f2-4d18-b28f-5cee413766a2 (accessed on 4 February 2025)
19 bioclimatic variables	Raster data	30 s	WorldClim: https://www.worldclim.org (accessed on 4 February 2025)
River	Vector data	-	OpenStreetMap: https://www.openstreetmap.org (accessed on 21 February 2025)
Road	Vector data	-	OpenStreetMap: https://www.openstreetmap.org (accessed on 21 February 2025)
Night light index	Raster data	15 arc s	https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/YGIVCD (accessed on 22 February 2025)
NDVI	Raster data	30 m	https://www.nesdc.org.cn/sdo/detail?id=60f68d757e28174f0e7d8d49 (accessed on 20 February 2025)
Population	Raster data	100 m	WorldPop: https://hub.worldpop.org/geodata/summary?id=6524 (accessed on 23 February 2025)
Depth to bedrock	Raster data	1 km	https://doi.org/10.1038/s41597-019-0345-6 (accessed on 23 February 2025)

), including tea tree data, Enshi land cover data, digital elevation model (DEM), soil database, precipitation, and Normalized Difference Vegetation Index (NDVI). The final land cover data were obtained by merging the tea tree data obtained by Peng et al. [39] with the national 10 m resolution land cover data obtained by Liu et al. [40]. The geospatial analysis employed rigorous data standardization, with all layers projected to the WGS_1984_UTM_Zone49N coordinate system and resampled to a consistent 10 m resolution to enhance analytical precision and methodological consistency.

2.3. Methodology

Figure 2 presents a comprehensive framework for delineating the core conservation zone of TPAHS and simulating ecological networks in Enshi.

2.3.1. Identification of Tea Plantation Core Zone

The study used the Maximum Entropy (MaxEnt) model (Version 3.4.1) to identify the highly suitable tea tree distribution areas as the core. The MaxEnt model is one of the Species Distribution Modeling algorithms, which can predict the distribution of suitable habitats for species based on known species distribution coordinates and environmental data [41]. A key advantage of this method lies in its ability to generate reliable predictions despite small sample sizes, making it particularly valuable for conservation planning in data-scarce regions [42]. The MaxEnt jackknife analysis evaluated environmental variable contributions to tea plant growing suitability. This data-driven approach systematically quantifies each factor’s importance through iterative exclusion and model performance evaluation, objectively prioritizing variables based on their actual predictive power. By relying on empirical data rather than subjective judgments, this method effectively eliminates the inherent biases associated with expert scoring techniques, ensuring robust and ecologically meaningful variable selection. According to previous studies [43], 19 bioclimatic variables (

Table A1. Biological climate variable data.

Code	Biological Climate Variable	Code	Biological Climate Variable
Bio 1	Annual Mean Temperature (°C)	Bio 11	Mean Temperature of Coldest Quarter (°C)
Bio 2	Mean Diurnal Range (°C)	Bio 12	Annual Precipitation/mm
Bio 3	Isothermality (bio2/bio7)	Bio 13	Precipitation of Wettest Month/mm
Bio 4	Temperature seasonality	Bio 14	Precipitation of Driest Month/mm
Bio 5	Max Temperature of Warmest Month (°C)	Bio 15	Precipitation Seasonality/mm
Bio 6	Min Temperature of Coldest Month (°C)	Bio 16	Precipitation of Wettest Quarter/mm
Bio 7	Temperature Annual Range (°C)	Bio 17	Precipitation of Driest Quarter/mm
Bio 8	Mean Temperature of Wettest Quarter (°C)	Bio 18	Precipitation of Warmest Quarter/mm
Bio 9	Mean Temperature of Driest Quarter (°C)	Bio 19	Precipitation of Coldest Quarter/mm
Bio 10	Mean Temperature of Warmest Quarter (°C)

), topographic factors including slope, aspect, elevation, and soil conditions, including gravel content, clay content, sand contend, silt content, organic carbon content, and pH, were selected to simulate the habitat suitability for tea. To avoid multicollinearity, pairwise Pearson correlation coefficients were computed for all variable combinations, followed by variance inflation factor (VIF) calculations for each variable [44]. Random points converted from the tea tree raster data and the environmental data were imported into MaxEnt, using 75% and 25% of the points as training and validation sets, respectively, and repeated 10 times.

The jackknife method was employed to assess the importance of each environmental factor. The predictive performance of the model was assessed using receiver operating characteristic (ROC) analysis, with the area under the curve (AUC) serving as the primary evaluation metric. Following established ecological modeling conventions, AUC values were interpreted as follows: <0.7 (poor), 0.7–0.8 (reasonable), 0.8–0.9 (good), and >0.9 (excellent) predictive accuracy [45,46].

2.3.2. Identification of Ecological Sources

Ecological sources function as critical habitat cores with multidimensional ecological attributes including superior landscape permeability and high ESs values. Utilizing land cover data, precipitation data, soil data, etc., the Habitat Quality (HQ), Water Yield (WY), Carbon Storage and Sequestration (CS), and Sediment Delivery Ratio (SDR) were simulated with the InVEST model (Version 3.14.3). The calculation formula for ESs values is as follows.

(1)

Habitat Quality

The InVEST model evaluates HQ based on the premise that different land cover types inherently provide varying levels of suitability for biodiversity. Unlike species-specific approaches, the model assigns relative habitat suitability scores to each land cover type, where higher values indicate a greater potential to support diverse species. Habitat degradation is then calculated by integrating threats with their spatial impacts and specific sensitivity. The final HQ output identifies high-quality patches as areas with both high intrinsic suitability and low degradation. In the relevant literature [47], the relevant parameters for calculating HQ are shown in

Table A2. Maximum distance of threat factors affecting habitat quality and their weights.

Threat Factor	Maximum Impact Distance/km	Weight	Decay Type
Farmland	2.00	0.50	Exponential
Tea plantation	2.00	0.30	Exponential
Impervious surface	6.00	0.80	Exponential
First-level road	3.00	0.50	Linear
Secondary road	1.50	0.40	Linear
Third-level road	0.50	0.20	Exponential

and

Table A3. Habitat suitability of different land cover types and relative sensitivity to threat factors.

Land Use Type	Habitat Suitability	Farmland	Impervious Surface	Tea Plantation	First-Level Road	Secondary Road	Third-Level Road
Farmland	0.50	0.30	0.90	0.10	0.40	0.20	0.10
Forest	1.00	0.50	0.85	0.40	0.60	0.20	0.20
Grassland	0.70	0.50	0.60	0.30	0.50	0.30	0.10
Tea plantation	0.70	0.30	0.70	0.10	0.40	0.30	0.10
Wetland	1.00	0.65	0.75	0.50	0.50	0.30	0.10
Water	0.90	0.65	0.75	0.50	0.50	0.30	0.10
Impervious surface	0	0	0	0	0	0	0

. The formula for calculating the HQ is shown in Formula (1):

$$H{Q}_{xj}=H_j\left(1-{\displaystyle \frac{D_{xj}^z}{D_{xj}^z+k^z}}\right)$$

(1)

In the equation, HQ_xj is the habitat quality of grid cell x in land cover type j; H_j is the habitat suitability for land cover type j; D_xj is the degree of habitat degradation; z is a normalized constant; k is the semi-saturation index.

(2)

Water Conservation

Water conservation (WC) capacity represents nature’s rainwater retention, filtration, and storage functions. The InVEST model calculates WY using parameters including precipitation, plant transpiration, surface evaporation, root depth, and soil depth based on hydrological cycle principles according to the Formula (2). Then, WY is adjusted by topographic index, soil saturated hydraulic conductivity, and flow velocity coefficient to evaluate the water conservation function according to the Formula (3) [48].

$$W{Y}_{xj}=\left(1-{\displaystyle \frac{AE{T}_{xj}}{P_y}}\right)P_y$$

(2)

$$\mathit{Retention}=min\left(1,{\displaystyle \frac{249}{\mathit{Velocity}}}\right)\times min\left(1,{\displaystyle \frac{0.9\times \mathit{TI}}{3}}\right)\times min\left(1,{\displaystyle \frac{\mathit{Ksat}}{300}}\right)\times W{Y}_{xj}$$

(3)

$$TI=lg\left({\displaystyle \frac{Drainage\_Area}{Soil\_Depth\times Percent\_Slope}}\right)$$

(4)

In the equation, WY_xj is the amount of precipitation minus the actual yearly evapotranspiration of a grid cell x, which is the water yield; AET_xj is the annual actual evaporation of grid cell x in land cover type j; P_y is the average yearly precipitation for grid y. Retention is the water source conservation capacity; Ksat is the saturated hydraulic conductivity of soil, calculated using SPAW Hydrology software (Version 6.02.75); Velocity is the coefficient of flow velocity; TI is the terrain index, which is dimensionless and can be obtained by calculating according to the Formula (4). The relevant parameters for calculating WC are shown in

Table A4. Parameters for simulation of water conservation and soil conservaion in Enshi.

Land Cover Type	Root depth/mm	Kc	Velocity Coefficient	Usle_c	Usle_p
Farmland	400	0.70	700	0.22	0.35
Forest	3000	0.95	200	0.05	1
Grassland	500	0.85	500	0.07	1
Tea plantation	1300	0.85	500	0.08	0.35
Wetland	-	0.95	1800	1	0
Water	-	1.00	2012	1	0
Impervious surface	-	0.45	2012	0.20	0

, referring to [47].

(3)

Carbon Storage and Sequestration

CS plays a pivotal role in carbon balance maintenance and climate regulation. Land parcel carbon storage primarily derives from four pools: above-ground biomass, below-ground biomass, soil, and dead organic matter. The formula for calculating the CS is shown in Formula (5).

$$C_{\mathrm{total}}=C_a+C_b+C_s+C_d$$

(5)

In the equation, C_total is the total carbon stock; C_a, C_b, C_s, and C_d represent four carbon pools. The carbon density data was obtained after correction based on the data measured by scholars in Hubei Province (

Table A5. Carbon density of different land cover types in Enshi.

Land Cover Type	C_above	C_below	C_soil	C_dead
Farmland	4.02	0.75	98.13	2.11
Forest	22.62	18.03	126.75	2.78
Grassland	9.05	9.49	97.79	4.89
Tea plantation	14.49	7.27	105.15	2.5
Wetland	2.34	0	70.28	4.62
Water	1.59	0	64.03	3.98
Impervious surface	0.83	0.08	43.71	0

) [49,50].

(4)

Soil Conservation

Soil conservation is vital to maintaining ESs by preventing erosion in its karst landscapes, enhancing water quality and sustaining agricultural productivity that supports both local livelihoods and regional ecological security in Enshi. The formulas for evaluating the soil conservation function are shown below:

$$\mathit{RKLS}=R\times K\times LS$$

(6)

$$\mathit{USLE}=R\times K\times LS\times P\times C$$

(7)

$$\mathit{SD}=RKLS-USLE$$

(8)

In the equation, RKLS represents the potential soil erosion amount under specific geomorphological and climatic conditions in bare land scenarios within the study area. USLE indicates the actual soil erosion amount incorporating management and engineering measures. SD represents the soil retention capacity. R, K, LS, C, and P represent the rainfall erosivity factor, soil erodibility factor, slope length factor, vegetation and crop management factor, and soil and water conservation measure factor. Specific parameter values are shown in Table A4, as referenced in the relevant literature [51].

The four ESs were first standardized and integrated with equal weighting. Using Jenks’ natural breaks classification method, the comprehensive ESs value were reclassified, with the highest-value area assigned as foreground (value = 2) and remaining areas as background (value = 1) in the binary map for MSPA. In MSPA binary maps, the foreground represents target ecological elements (e.g., forests, wetlands, or critical habitats), while the background denotes non-target landscape types (e.g., urban areas, farmland). Through 8-connectivity MSPA processing, the landscape was classified into seven non-overlapping landscape types with core areas identified as key ecological sources. Then, the landscape connectivity was quantitatively assessed using two metrics: the Probability of Connectivity (PC) and the Integral Index of Connectivity (IIC), computed via the Conefor (Version 1.0.218) inputs tool for ArcGIS (Version 10.8.1). In order to determine conservation priorities, each patch’s structural importance for ecosystem stability and biodiversity was quantified by the connectivity index change rate (dM) upon its removal [52,53]. The formulas for calculating the PC, IIC and dM are shown below:

$$PC={\displaystyle \frac{1}{A_L^2}}\sum _{i=1}^n\sum _{j=1}^n{p}_{ij}^{*}{a}_i{a}_j$$

(9)

$$IIC={\displaystyle \frac{{\displaystyle \sum _{i=1}^n}{\displaystyle \sum _{j=1}^n}\frac{a_i\times a_j}{1+n{l}_{ij}}}{A_L^2}}$$

(10)

$$dM(\%)=100\times {\displaystyle \frac{M-M_a}{M}}$$

(11)

In the equation, n is the total number of ecological source patches; a_i, a_j are areas of source patches i and j; P_ij is maximum dispersal probability between patches i and j; A_L is total landscape area; nl_ij is the number of links on the shortest path between patch i and j; M is the connectivity index value (IIC or PC) when all habitat patches are present; M_a is the connectivity index value after removing a specific patch from the landscape.

2.3.3. Construction of Comprehensive Resistance Surface

Landscape resistance serves as a fundamental component in connectivity modeling. However, accurately quantifying resistance to movement remains methodologically challenging, primarily due to the scarcity of empirical movement data. In the absence of direct movement or genetic data, habitat suitability indices are commonly employed as proxies to infer resistance patterns across heterogeneous landscapes [54]. Based on the natural environmental characteristics of the study area and existing research, this study selected elevation, slope, land cover type, NDVI, distance to roads (three levels), distance to water, nighttime light index, and population as resistance factors. MaxEnt can be used to predict the distribution of suitable areas for wildlife, while areas unsuitable for species distribution correspond to high-resistance areas unfavorable for wildlife migration. Therefore, this study extracted dominant ESs areas, generated 1000 random points within these regions, and exported their coordinates. The coordinates of these points, along with the resistance factor data, were then imported into MaxEnt for analysis. In the prediction results of the MaxEnt model, areas with low p-values correspond to high-resistance areas.

Research indicates that landscape resistance to species movement typically follows an exponential rather than linear relationship with habitat suitability. Trainor et al. [55] and Mateo-Sánchez et al. [56] demonstrated, using red-cockaded woodpeckers (Picoides borealis) and brown bears (Ursus arctos) as model species, that during long-distance dispersal or pre-dispersal prospecting movements, these species traverse areas of moderate habitat suitability. Their findings reveal that as habitat suitability declines from its peak, resistance increases only marginally. However, when suitability drops below a critical threshold, resistance escalates sharply with further declines in suitability. Keeley et al. [57] suggests that when designing corridors to promote animal dispersal, migration, and other extensive movements, researchers and conservation planners should generally assume a negative exponential relationship when converting habitat suitability into resistance values. To generate the resistance surface, this study applied an exponential transformation to the result of habitat suitability prediction derived from MaxEnt with Formula (12):

$$R={1000}^{(-1\times HS)}$$

(12)

where R is the resistance value, and HS is habitat suitability predicted by MaxEnt. Then, the resistance values were normalized to a 1–100 scale, where HS = 1 maps to 1 and HS = 0 to 100 [58,59].

2.3.4. Construction of Ecological Network

The ecological sources and resistance surface were imported into Linkage Mapper (Version 3.1.0), a spatial analysis tool that synergistically combines the least-cost path (LCP) approaches and circuit theory, to simulate ecological networks. Linkage Mapper first identified adjacent ecological sources and constructed an interconnected network based on spatial proximity. Then it computed cost-weighted distances to determine optimal pathways between ecological sources, ultimately merging all identified corridors into a comprehensive connectivity map. The quality of the linkage was then assessed using two indices: (1) the ratio of cost-weighted distance to Euclidean distance (CWD:EucD), quantifying traversal difficulty relative to straight-line distance between habitat cores, and (2) the ratio of cost-weighted distance to the length of the least-cost path (CWD:LCP), representing mean resistance along optimal ecological corridors. The importance of corridors in maintaining regional connectivity is quantified by centrality values, calculated using Centrality Mapper. Higher centrality values indicate greater ecological significance [60].

2.3.5. Identification of Ecological Nodes

For further prioritizing conservation and rehabilitation efforts, the study identified pinch points, barrier points, and junctions. Pinch points represent critical constrictions within ecological corridors where movement options become severely limited, forcing wildlife to traverse through these obligatory passageways regardless of alternative routes. The study used the Pinchpoint Mapper tool in Linkage Mapper, combining with Circuitscape (Version 4.0.7) to identify pinch points with the cost-weighted width cutoff value set to 1 km [61,62]. The analysis employed an “All to one” computational approach to systematically evaluate connectivity patterns across the ecological network. This methodology generated integrated current density visualizations by sequentially directing current flows from multiple source nodes toward individual ground nodes. The resulting spatial outputs identified pivotal pinch points exhibiting elevated current concentrations, signifying constrained movement options that may serve as connectivity bottlenecks essential for maintaining landscape-scale ecological flows [63].

Barrier point refers to a localized obstacle in a landscape that significantly impedes wildlife movement or gene flow. The study used the Barrier Mapping tool, which employs a moving window method to identify potential obstacles affecting ecological processes. It quantifies the influence on landscape connectivity by expressing connectivity values per unit of distance restored [64]. We established the threshold of the exploration window for testing with an initial value of 50 m, a terminal value of 250 m, and an incremental step size of 50 m. Junctions refers to intersections of ecological corridors. They attract species from multiple corridors, which may result in a high level of biodiversity [65].

3. Result

3.1. Tea Plantation Core Zone

The AUC value to predict the suitable areas for tea is 0.769, giving a reasonable overall predictive accuracy. Ten prediction iterations were averaged to produce the suitability map, delineating the core tea plantation zone in Enshi Yulu Tea China-NIAHS (Figure 3). The prediction map reveals that the spatial distribution of the tea plantation suitability areas is heterogeneous, mainly concentrated in the Grand Canyon in the western part of Enshi, the Bajiao Dong Township and Shengjiaba Township in the southwestern regions, and along the Qingjiang River basin at the border between the Shadi Township and Xintang Township in the eastern regions. MaxEnt variable contribution analysis revealed elevation, isothermality, slope, temperature annual range, and precipitation of the driest quarter to be key determinants of tea plant growing suitability.

The MaxEnt model predicts the distribution of suitable areas for tea plantations using the probability p (ranging from 0 to 1), indicating the likelihood of tea tree occurrence. The study area was categorized into four suitability zones based on p-values: high ($p>0.6$), medium ($0.4\le p\le 0.6$), low ($0.2\le p<0.4$), and non-suitable ($p<0.2$) [66]. Then, the Aggregate Polygons tool in ArcGIS was employed to integrate the high-suitability areas while eliminating smaller fragmented patches, resulting in the identification of tea plantation core zones of agricultural heritage. The total area of these core zones is 718.04 km², representing 18.08% of the study area.

3.2. Ecological Sources

As shown in Figure 4e, the comprehensive ESs value is reclassified into four levels, and the area with the highest level of ESs value has high forest coverage, minimal human interference, and high biodiversity. Therefore, it was set as the foreground in MSPA analysis. The MSPA result suggests that the foreground area is 1537.65 km², consisting of seven landscape types (Figure 5). Among them, the core has the largest area at 1015.93 km², comprising 66.07% of the foreground. Core patches predominantly cluster in western Enshi City, comprising protected areas including Xingdou Mountain National Nature Reserve, Tenglongdong Grand Canyon, Tongpenshui Forest Park, Baihuwan Forest Nature Park, Suobuya Forest Nature Park, Camellia polyodonta Forest Park, Fuer Mountain Forest Park, etc. Eastern ecological sources are scarce, primarily located in Fenghuang Mountain Forest Park and Enshi Shuanghe Forest Nature Park. While north–south connectivity exists among sources, the urban core of Enshi fragments east–west linkages due to absent source areas.

The core areas are important habitat patches that play a role in maintaining the integrity of the ecosystem. In general, larger patch areas correlate with enhanced connectivity and superior ecological quality [67]. Consequently, core patches exceeding 3 km² in size were identified as potential ecological sources in Enshi, resulting in 43 potential ecological source patches, and the prioritization for sources was obtained according to the dPC ranking, with $dPC>10.0$ as the first-level sources, $10\ge dPC>5$ as the second-level sources, and $5\ge dPC>0.1$ as the third-level sources. The top 10 important ecological sources are shown in

Table 2. The evaluation result of the importance of ecological sources (top 10).

Rank	Number	Area/${\mathbf{km}}^2$	dPC
1	3	151.20	56.99
2	34	70.13	29.22
3	20	16.39	14.34
4	6	26.48	12.97
5	18	7.28	11.50
6	42	36.88	10.19
7	40	24.45	9.15
8	1	21.97	8.46
9	5	13.62	6.05
10	27	15.31	5.17

. The study analyzed dIIC for evaluation and dPC for validation, with a correlation coefficient of r = 0.92 between these metrics confirming the reliability of the assessment results (Figure 6).

3.3. Comprehensive Resistance Surface

The predicted suitable areas for wildlife distribution are shown in Figure 7a. The jackknife test assessed each environmental factor’s relative influence, with their percent contribution and permutation importance presented in Figure 8. In this study, the main factors affecting the wildlife habitat suitability are mainly the type of land cover, distance to road, NDVI, and slope. The resistance surface is obtained through exponential transformation (Figure 7b). In Enshi, the ecological resistance surface primarily reflects road network patterns, with elevated resistance values clustered in the urban core and radiating along transportation corridors.

3.4. Ecological Corridors

Using circuit theory, this study identified 77 ecological corridors totaling 461 km, averaging 6.00 km (range: 0.10–37.33 km) (Figure 9). L35, L8, L17, and L11 have high CWD:EucD ratios and high CWD:LCP ratios, indicating the high cost for moving along them. Short corridors (e.g., L16, L68, L20) with low CWD:EucD and CWD:LCP ratios demonstrate high movement quality, enhancing species dispersal and genetic exchange within protected areas. Corridors located at the edge of Enshi (e.g., L24, L57, L72) typically exhibit very low centrality values, indicating their relatively less important role in the ecological network. In contrast, corridors near Tongpenshui Forest Park (e.g., L7, L18, L35) demonstrate significantly higher centrality values compared to others. The lengths of L53, L19, L4, L52, L17, and L8 all exceed 16 km, running in an east–west direction, spanning the urban construction area in central Enshi. These pathways effectively link the eastern and western ecological sources, facilitating long-distance resource transfer and significantly contributing to ecosystem stability.

3.5. Ecological Nodes

This study identified 66 pinch points, 25 barrier points, and 13 junctions within the study area (Figure 10). The pinch points are predominantly distributed along east–west-oriented corridors in the southeastern part of Enshi, with some additional nodes observed along east–west-oriented corridors in the northern region. In contrast, few pinch points are detected in the eastern part of Enshi.

The study area’s ecological pinch points fall into two categories: (1) natural topographic constrictions, such as narrow valleys or river confluences, and (2) transition areas between anthropogenic and natural landscapes, including remnant habitat corridors along transportation infrastructure and edges of farmland. Barrier points predominantly cluster within the northern and southwestern corridors of Enshi. This spatial pattern may be attributed to the north–south-oriented roads that traverse the flat central urban area of Enshi. And barrier points predominantly occur at intersections of roads and corridors. Junctions, which represent intersections of multiple corridors, exhibit spatial heterogeneity and typically consist of mixed-habitat transition areas formed by the convergence of diverse ecosystems.

4. Discussion

4.1. Distribution of Tea Plantations

This study predicts the distribution of suitable areas for tea plantations with the MaxEnt model based on tea tree distribution raster data and environmental data. This method accounts for the dynamic feature of agricultural heritage systems, enabling a systematic prediction of suitable areas for tea.

The results of the MaxEnt jackknife test reveal that topographic factors, notably elevation and slope, exert a great influence on tea plant growing suitability (

Table 3. Analysis of variable contributions to tea tree distribution suitability (top 10).

Rank	Variable	Percent Contribution	Permutation Importance
1	Altitude	28.9	36.8
2	Bio 3 (Isothermality)	15.9	1.6
3	Slope	12.3	7.7
4	Bio 7 (Temperature annual range)	11.3	7.4
5	Bio 12 (Annual precipitation)	6.9	4.8
6	Bio 17 (Precipitation of driest quarter)	5.7	2.3
7	Bio 6 (Minimum temperature of coldest month)	3.4	3.1
8	Bio 2 (Mean diurnal temperature range)	3.0	4.5
9	Bio 18 (Precipitation of warmest quarter)	2.0	8.7
10	Bio 4 (Temperature seasonality)	2.0	6.1

). Furthermore, among the 19 bioclimatic variables examined, Bio 3, Bio 7, Bio 12, and Bio 17 demonstrate particularly high contribution rates. These findings suggest that adequate and evenly distributed annual rainfall, typically between 1460 and 1480 mm in the study area, combined with optimal temperature conditions, constitute essential requirements for successful tea cultivation [68].

However, soil properties such as pH, organic carbon content, clay content, sand content, and silt content show relatively low contribution rates in suitability predictions. This finding contrasts with prior studies that emphasize soil properties as a key determinant of tea plantation productivity [68]. The discrepancy may be attributed to two factors: (1) the relatively small study area with limited soil variability and (2) the predominance of inherently tea-suitable soils across most of the study area.

4.2. Ecosystem Service Value and Resistance Surface

Previous studies have suggested that land cover such as farmland and construction land are key determinants of ESs value [69], and this study confirms this viewpoint. The study further reveals that roads substantially influence ES valuation [70,71], as they are considered as one of the threat factors when calculating HQ, one of the ESs. Then, based on the MaxEnt model, this study obtained the wildlife habitat suitability with various resistance factor variables and high ESs value sample point data. Through exponential transformation and linear interpolation, the resistance surface was obtained with wildlife habitat suitability. This method can eliminate the subjectivity caused by using AHP to determine the weights of the resistance factors, which are still determined based on expert knowledge. NDVI substantially influences ES valuation, thereby shaping resistance surfaces and corridor networks. These findings align with established research linking NDVI to faunal movement, species distributions, and population dynamics [72].

4.3. Ecological Benefits of Tea Fields

This study reveals a 43.96 km² overlap between tea plantation core zones and ecological sources (Figure 11a). The overlapping regions are primarily distributed in the southwestern part of Enshi, which contains the most extensive tea plantations within the study area. According to zonal statistical calculations, these areas exhibit relatively high ESs values, with mean comprehensive ESs values of 0.61 (range: 0.56–0.70), particularly where tea plantation patches are embedded within woodland, forming a landscape of significant ecological importance. This evidence may suggest that tea agroforestry systems provide certain ecological benefits. Research indicates that in anthropogenically modified environments where fragmented forest patches intersect with agricultural systems, recolonization dynamics mitigate species loss while supporting both biodiversity conservation and sustainable agricultural productivity [73].

Nevertheless, tea plantations only contribute effectively when integrated into an appropriate landscape configuration with forests. Research confirms that traditional forest-shaded tea systems outperform monoculture plantations across multiple ecological metrics, including biodiversity conservation, soil and water retention, pest control, climate regulation, and carbon sequestration potential in woody biomass [74,75]. Large-scale monoculture tea plantations, however, do not match the ESs value of forests and may instead cause ecological degradation [76]. Multiple studies have examined how tea plantation microbiomes change over time, consistently reporting the depletion of beneficial bacteria under prolonged cultivation [77,78]. This may explain why not all tea plantation core zones qualify as ecological sources.

Compared to farmlands and urban areas, tea plantations exhibit lower resistance values [79]. Corridors like L52, L53, L42, L31, L77, and L36 traverse extensive tea plantation core zones. However, these pathways predominantly traverse the most constricted sections and gaps in the tea plantation landscape to connect ecological sources (Figure 11b,c). Forests occupy these gaps, demonstrating that tea plantations, despite their partial ecological value, provide poorer habitat suitability than forests, increasing movement resistance for wildlife. This study conducted a buffer analysis of ecological corridors at widths of 50 m, 100 m, 150 m, 250 m, and 300 m to assess land cover composition within corridors. The results demonstrate that forest coverage consistently exceeded 70% across all corridor widths (50–300 m), significantly surpassing tea plantation coverage. However, as corridor width increased, forest coverage exhibited a slight decline from 76.4% to 71.8%, while the proportions of tea plantations and farmland showed a corresponding increase.

4.4. Landscape Pattern Optimization Approach

Based on the mapped ecological sources, corridors, and nodes, this study developed an ecological optimization framework featuring a “four belts and four zones” spatial configuration (Figure 12). The “four zones” refers to the Ecological Conservation Zone (ECZ), Sustainable Tea Cultivation Zone (STCZ), Ecological Barrier Restoration Zone (EBRZ), and Synergistic Tea Forest Rehabilitation Zone (STFRZ).

Establishing an ECZ around ecological sources is essential for safeguarding critical habitats and promoting regional biodiversity conservation. For first-level ecological sources, strict protection measures are recommended, including the prohibition of agricultural and tourism activities as well as the establishment of ecological monitoring stations. Second-level sources should maintain their current ecological conditions while enhancing their functional advantages, whereas fragmented third-level sources require regular quality monitoring to prevent degradation.

To maximize the ecological, social, and economic benefits of the STCZ, an agroforestry-based cultivation model should be implemented. Tea plantation intercropping, as a form of agroforestry, integrates trees and crops with tea plants in a multi-layered system that maximizes spatial and soil resource utilization [80]. This vertically structured ecosystem enhances land productivity while improving soil fertility, increasing biodiversity and sustaining ecological equilibrium through synergistic plant interactions [81]. This approach ensures sustainable yields while enhancing ESs and reducing the vulnerability to climate risk [82], ultimately facilitating the transition into the STFRZ where human disturbances are minimized through regulated agricultural practices and controlled access to preserve ecological integrity. In the STCZ around tourist sites like Tenglong Grand Canyon National Geological Park and Denglongba Village in Bajiao Township, it is essential to coordinate multiple demands including tourist sightseeing, tea production, and heritage conservation, and fully leverage the aesthetic value of tea agricultural heritage to generate economic benefits.

For EBRZs, ecological restoration should be implemented to facilitate wildlife migration. Since these areas often intersect with transportation infrastructure, adopting ecological engineering measures, particularly wildlife movement facilitation structures such as overpasses and underpasses, is crucial for minimizing anthropogenic disturbances to species movement [83,84]. Additionally, it is recommended to implement acoustic warning signals alongside buffer forest belts. These methods can effectively reduce noise disturbance and minimize wildlife–vehicle collisions by alerting animals to approaching traffic [85,86]. When combined with vegetative buffers, this approach provides a non-invasive solution that enhances habitat connectivity while addressing sensory pollution.

Among the four belts, the two north–south ecological conservation corridors enhance short-distance material exchange between the ecological sources in the western and eastern regions. Meanwhile, the Qingjiang River Ecological Restoration Belt and the East–West Ecological Construction Belt address the fragmentation caused by Enshi’s central urban zone, thereby enhancing connectivity. The east–west corridors that traverse the central urban area, particularly those crossing the core tea plantation zone which contains multiple ecological nodes, should incorporate stepping stone habitats at junctions. These habitats can improve landscape connectivity by linking isolated habitat patches, thereby facilitating species movement and genetic exchange across the fragmented tea-dominated landscape [87]. To protect water quality, maintain aquatic habitats, and filter pollutants, priority should be given to eco-friendly tea cultivation practices in riparian buffers of over 30 m in the Qingjiang River Ecological Restoration Belt [88].

The proposed “four belts and four zones” optimization framework, while theoretically constructing an ideal ecological pattern, necessitates the balanced integration of multi-stakeholder interests in practical implementation. To balance ecological conservation and economic growth, it is essential to leverage governmental coordination through scientific planning, policy guidance, and public services to align diverse stakeholder interests. A cost–benefit prioritization strategy should be implemented by systematically evaluating and comparing the ecological effectiveness versus implementation costs of potential corridors, with priority given to establishing those demonstrating the highest ecological returns per unit investment. Particular attention must be paid to reconciling short-term economic benefits with long-term ecological sustainability by establishing equitable ecological compensation mechanisms to incentivize local community participation in conservation efforts while ensuring the selected corridors optimize both biodiversity outcomes and socioeconomic feasibility through multi-criteria decision analysis. Furthermore, the delineated sustainable tea cultivation zones and ecological protection areas should be harmonized with existing policies such as ecological redlines and tea plantation expansion restrictions set by natural resource authorities while allowing for adaptive adjustments based on field conditions. Given the dynamic feature of agricultural heritage systems, it is recommended to implement a comprehensive monitoring and evaluation system to periodically revise functional zone boundaries, ensuring continuous alignment between the conservation framework and the evolving landscape pattern conditions. This adaptive management approach safeguards ecological security while maintaining flexibility for regional development.

4.5. Other Limitations and Future Research Priorities

This study proposes a landscape pattern optimization strategy for the Enshi Yulu Tea Agricultural Heritage site. Nevertheless, several limitations were identified in the study. (1) The MaxEnt model’s moderate prediction accuracy (AUC=0.769) indicates room for improvement. This limited accuracy may stem from unconsidered social and cultural factors affecting tea plant growing suitability, potentially resulting in the incomplete delineation of the tea plantation core zone. (2) The study applied equal weighting to four ecosystem services, failing to account for their differential impacts on habitat suitability. (3) Generalized exponential functions were applied to habitat suitability and resistance transformation for all organisms instead of species-specific data. (4) The corridors constructed primarily enhance structural connectivity, while functional connectivity receives relatively less attention [89]. (5) The lack of continuous ecological monitoring data (e.g., animal migration routes, biodiversity records) prevented the rigorous validation of corridor connectivity, centrality ranking reliability, and ecological benefits of tea fields.

Future research priorities can focus on the following: (1) conducting sensitivity analyses of MaxEnt parameters and alternative ES weighting schemes to reduce model uncertainty; (2) developing species-specific resistance surfaces through field surveys of key fauna movement patterns; and (3) establishing long-term ecological monitoring to validate corridor functionality using camera traps and GPS tracking.

5. Conclusions

This study established a framework for delineating the core conservation zone of the tea plantation agricultural heritage system and optimizing landscape patterns through an integrated approach combining MaxEnt model, MSPA, and circuit theory. This framework can be effectively transferred to other agricultural heritage sites when the following conditions are met: (1) availability of high-resolution land cover data that specifically includes the distribution of major agricultural commodities, which can be obtained either from existing datasets or through high-precision remote sensing classification; (2) the proper adaptation of environmental factors and resistance factors according to local biophysical conditions; and (3) the localized calibration of InVEST-based ecosystem service assessments using field survey data collected by regional researchers to ensure ecological relevance.

This study delineated a 718.04 km² core conservation zone for tea plantations, identified 77 ecological corridors, and pinpointed 104 critical ecological nodes. The results indicate 43.96 km² of synergistic areas between tea plantations and ecological sources, demonstrating that agroecosystems of tea plantations provide higher ESs value compared to monoculture plantations and farmlands. In addition, the suitability of tea plantations is significantly influenced by topographic and climatic factors, while the resistance surface is primarily affected by land cover type, road networks, and NDVI. These findings were utilized to develop an ecological optimization framework featuring a “four belts and four zones” spatial configuration, aimed at enhancing connectivity and promoting the sustainable development of TPAHS. The framework can provide evidence-based references for future policy formulation through collaboration with local agricultural authorities, and deliver actionable insights for land-use planning, habitat restoration, and infrastructure mitigation.