A Multi-Model Coupling Approach to Biodiversity Conservation Strategies for Nationally Important Agricultural Heritage Systems in the Beijing–Tianjin–Hebei Region

Abstract

To address biodiversity degradation in Nationally Important Agricultural Heritage Systems, this study integrates multi-temporal remote sensing data (2000–2023) with the Biodiversity Maintenance Function (BMF) and InVEST Habitat Quality (HQ) models. We assess ecological changes in the Beijing–Tianjin–Hebei (BTH) region and 14 nationally recognized heritage systems. A dual-factor HQ–BMF coupling matrix was developed to trace ecological trajectories shaped by both natural and anthropogenic influences. Results show that (1) regional BMF followed a non-linear trend of increase, decline, and rebound between 2003 and 2023. The mean value rose from 0.1036 in 2003 to 0.1397 in 2023, despite intermediate fluctuations. In contrast, HQ declined steadily from 0.8734 in 2003 to 0.7729 in 2023, reflecting a continuous loss of high-quality habitats. (2) Nearly all heritage systems experienced phased BMF fluctuations—an initial rise, subsequent decline, and eventual recovery. At the same time, HQ showed a continuous decline in 8 of the 14 systems, indicating that more than half of the systems experienced sustained habitat degradation. (3) The HQ–BMF matrix revealed strong spatial heterogeneity. By 2023, only one site remained in a “dual-high” zone, while another had fallen into a “dual-low” condition, suggesting localized ecological degradation. These findings provide quantitative support for conservation strategies, ecological compensation, and land-use regulation in agricultural heritage systems.

1. Introduction

In recent years, intensified human activities have increasingly encroached upon and degraded habitats worldwide. This has posed unprecedented challenges to biodiversity conservation and maintenance [1,2,3,4]. Traditional agriculture, which originates from long-term coadaptation between humans and nature, now faces the dual pressures of ecosystem degradation and biodiversity loss. In agricultural landscapes, these pressures are evident in habitat fragmentation, the loss of traditional plant and animal varieties, and land-use intensification. In response to these challenges, the Food and Agriculture Organization (FAO) of the United Nations launched the Globally Important Agricultural Heritage Systems (GIAHS) initiative in 2002. GIAHS recognizes “remarkable agricultural systems and landscapes shaped by the co-adaptation of people and nature, rich in biodiversity, traditional knowledge, and cultural values.” Its goals are threefold: ecological conservation, cultural heritage preservation, and sustainable development [5,6,7]. Building on this global initiative, China established the Nationally Important Agricultural Heritage Systems (China–NIAHS) framework in 2012 [8,9]. This system aligns with the FAO framework while adapting to China’s diverse ecological and cultural contexts. Since then, 188 systems have been designated, forming a nationwide network that links biodiversity conservation, sustainable resource use, and valorization of ecosystem services [8,10,11]. The Beijing–Tianjin–Hebei (BTH) region is both a major cluster of China–NIAHS and one of the country’s fastest-urbanizing areas. This dual role makes it a representative case for examining how agricultural heritage systems balance conservation with development pressures.

Agricultural biodiversity is the cornerstone of both the GIAHS and China NIAHS evaluation framework. It is explicitly identified as one of the five FAO criteria. The China–NIAHS Recognition Standards likewise highlight the richness and representativeness of agricultural biodiversity [12,13]. Enhancing such biodiversity strengthens food security and ecosystem resilience. It also reduces dependence on external chemical inputs and accelerates the transition to sustainable agriculture [14,15]. Agricultural heritage is more than production areas. It includes mosaic landscapes that integrate farmland, water bodies, forests, settlements and infrastructure [13]. These systems conserve diverse crop varieties and rich agrobiodiversity. They embody ecological wisdom refined through long-term human–nature interaction. However, under ongoing global habitat degradation and increasing anthropogenic pressures, their ecological functions are eroding. This makes biodiversity loss an urgent concern. For example, Liao et al. investigated three agricultural systems in the BTH region. Their results showed that between 1992 and 2022, the mean species abundance in Wangjinzhuang, a designated China–NIAHS system, declined slightly from 0.445 to 0.444, whereas Yuer Mountain Farm dropped more markedly from 0.129 to 0.110 [16].

From a driving-force perspective, biodiversity loss in agricultural heritage systems is shaped by both natural and anthropogenic factors. On the natural side, climate conditions, precipitation, and elevation all play significant roles. Variations in rainfall, temperature, and terrain fragmentation can alter ecosystem structure and productivity. These changes affect ecosystem services and ultimately weaken ecological stability. On the anthropogenic side, pressures stemming from rapid urbanization, land-use change, and intensified land development have emerged as dominant threats to agricultural biodiversity [17,18]. In recent years, several agricultural heritage systems on urban fringes have faced ecological space compression and habitat fragmentation. These processes have accelerated the pace of biodiversity degradation [19,20]. In our framework, both natural and anthropogenic drivers are equally important for explaining biodiversity dynamics in agricultural heritage systems.

As socio-ecological systems, agricultural heritage systems face both natural and anthropogenic pressures. Existing biodiversity assessments often reflect this dual perspective. On the natural side, Net Primary Productivity (NPP) is widely used to quantify ecological capacity. For example, Costanza et al. [21] found a positive correlation between plant species richness and NPP in high-temperature regions, supporting NPP as a proxy for ecological support capacity. Similarly, Siche et al. [22] observed significant multi-scale correlations between NPP and species richness, highlighting the link between biodiversity and primary productivity. In Guangzhou, China, Fu et al. [23] showed that areas with reduced NPP overlapped strongly with biodiversity loss, confirming NPP’s role in tracking ecosystem change.

On the anthropogenic side, the Habitat Quality (HQ) module of the InVEST model has been widely adopted to evaluate human pressures on ecosystems and biodiversity. HQ does not measure species richness directly. Instead, it estimates habitat suitability from land-use type and external threat intensity, serving as a proxy for ecological integrity and biodiversity potential. Aznarez et al. [24] applied the model in Vitoria-Gasteiz, Spain, finding strong positive correlations between HQ and urban biodiversity, especially in peri-urban green spaces. In Lazio, Italy, Di et al. [25] used the model to assess future land-use impacts under multiple scenarios, confirming HQ’s efficacy as a biodiversity proxy. Liu et al. [26] integrated the PLUS model with InVEST–HQ to simulate future land-use transitions in the Sanjiangyuan region, while Liao et al. [27] employed HQ to evaluate the co-evolution of agricultural biodiversity and traditional cultural resources in the BTH region.

Although regional biodiversity assessments have made important progress, agricultural heritage systems require more holistic analytical frameworks. Single ecological indicators cannot fully capture the dynamics of biodiversity change. An integrated framework that combines natural ecological potential with anthropogenic pressures is urgently needed. This framework supports accurate evaluation of ecological status and provides a dynamic view of biodiversity in agricultural heritage systems. Preserving agricultural biodiversity is central to heritage protection. It safeguards genetic resources while maintaining the cultural and ecological foundations of traditional farming. To achieve both ecological conservation and cultural continuity, a multidimensional perspective is required—one that examines interactions between natural ecological processes and human activities in cultural landscapes. In these systems, biodiversity supports ecosystem services, farming practices, diets, and rituals. Meanwhile, ecological knowledge and cultural practices accumulated through long-term agriculture help sustain species. Together, biodiversity and cultural traditions co-evolve, forming the foundation of ecological–cultural sustainability in agricultural heritage systems.

In this context, the present study focuses on 14 nationally designated agricultural heritage systems in the Beijing–Tianjin–Hebei (BTH) region. We employ the Biodiversity Maintenance Function (BMF) model, based on NPP, together with the Habitat Quality (HQ) module to evaluate the spatial and temporal dynamics of ecological functions from 2003 to 2023. Furthermore, we construct a HQ–BMF coupling matrix to identify biodiversity hotspots, degradation-prone areas, and other distinct landscape types. Accordingly, the central question of this study is: how have HQ and BMF changed in agricultural heritage systems over the past two decades, and what do these changes mean for their conservation and management?

2. Materials and Methods

2.1. Overview of the Study Area

The Beijing–Tianjin–Hebei (BTH) region is located in the northern part of the North China Plain. As one of the most economically dynamic regions in China, it serves as the core of the national capital economic zone [28]. Geographically, it spans from 36°05′ to 42°10′ N and from 113°27′ to 119°50′ E, covering an area of approximately 210,000 km². The region comprises two municipalities—Beijing and Tianjin—and the surrounding Hebei Province. It is bounded by the Yanshan Mountains to the north, the Taihang Mountains to the west, and Bohai Bay to the east. Its terrain slopes from the northwest to the southeast. Mountainous areas dominate the west and north, while hilly zones occur in the central part and alluvial plains in the east (Figure 1). The region experiences a temperate monsoon climate, with annual precipitation ranging from 400 to 700 mm, mostly concentrated between June and August.

The BTH region represents a typical case of ecological–social–economic tension in China. Beyond its economic prominence, the BTH region plays a crucial ecological role as a transitional zone linking mountains, plains, and wetlands. The region provides essential ecosystem services such as water regulation, soil conservation, and biodiversity maintenance. However, rapid urbanization has significantly disrupted regional ecosystem functions, resulting in declining environmental quality and increasing biodiversity loss. In terms of agricultural heritage conservation, the BTH region hosts several nationally designated China–Nationally Important Agricultural Heritage Systems (China-NIAHS), such as the Jingxi rice culture system in Beijing, the Xiaozhan rice system in Tianjin, and the traditional Xuanhua grape gardens in Hebei. Previous studies have shown that these agricultural heritage systems not only preserve rich biodiversity but also play a vital role in maintaining soil fertility and water conservation. Despite their significant ecological value, they face multiple threats because many are located in rural or peri-urban areas exposed to urban expansion, intensive land-use, and infrastructure development [29,30]. To analyze the spatiotemporal dynamics of ecological functions and biodiversity conservation, this study focuses on 14 officially recognized China-NIAHS systems within the BTH region. Based on information from the Food and Agriculture Organization (FAO) of the United Nations, the Ministry of Agriculture and Rural Affairs of China, and the China Agricultural Museum, we identified the core zones of each heritage system. These were assigned letter codes (A–N) (

Table 1. Codes and core area sizes of 14 China-NIAHS in the Beijing–Tianjin–Hebei Region.

Codename	Name of Heritage System	Main Areas	Area (km2)
A	Beijing Huairou Chestnut Cultivation System	Jiudu River Town, Bohai Town	329.98
B	Beijing Jingxi Rice Cultivation Culture System	Shangzhuang Town, Xibeiwang Town, Sijiqing Town	130.232
C	Beijing Mentougou Jingbai Pear Cultivation System	Junzhuang Town	33.346
D	Beijing Pinggu Sizhu Lou Walnut Production System	Xiong’erzhai Township	58.744
E	Tianjin Jinnan Xiaozhan Rice Cultivation System	Xiaozhan Town, Beizhakou Town, Bailitai Town, Gegu Town, Xinzhuang Town, Shuangqiaohe Town, Xianshuigu Town	305.484
F	Tianjin Binhai Cui Zhuang Ancient Winter Jujube Garden	Taiping Town	194.712
G	Ancient Pear Garden in Zhao County, Hebei Province	Xie Zhuang Township, Fan Zhuang Town	167.892
H	Hebei Qianxi Chestnut Composystem Cultivation System	Han’erzhuang Town, Luanyang Town, Yuhu Zhai Township	299.052
I	Traditional Chestnut Cultivation System in Kuangcheng, Hebei Province	Nanziyu Town, Songling Town, Huajian Town, Boluotai Town, Kuancheng Town, Tashan Township	548.376
J	Ancient Mulberry Forest in Botou, Hebei Province	Yingzi Town	96.108
K	Shixian County Dry Farming Terraced Field System, Hebei Province	Jingdian Town, Gengle Town, Guanfang Township	273.386
L	Xuanhua City Traditional Vineyard	Chunguang Township	18.35
M	Hebei Zhuolu Longyan Grape Cultivation System	Wenquan Town, Wubao Town	172.713
N	Traditional Hawthorn Cultivation System in Xinglong, Hebei Province	Liudaohe Town, Xinglong Town, Beiyingfang Town, Wuling Mountain Town	737.122

), with their geographic distribution shown in Figure 1.

2.2. Data Sources

The study period spans from 2003 to 2023, with analyses conducted at 5-year intervals. This timeframe was chosen for two main reasons. First, 2003 marks the launch of the Globally Important Agricultural Heritage Systems (GIAHS) initiative by the Food and Agriculture Organization (FAO). It also represents a critical stage in the Beijing–Tianjin–Hebei (BTH) region. During this time, the region experienced rapid socio-economic development alongside the gradual strengthening of ecological protection policies. These conditions allow us to capture biodiversity dynamics of agricultural heritage systems under different development contexts. Second, 2023 is the most recent year for which remote sensing imagery and other relevant datasets were available. This ensures both the timeliness and completeness of the results. To minimize the influence of anomalous years caused by extreme climate events or abrupt land-use changes, Net Primary Productivity (NPP), mean annual precipitation, and mean annual temperature were represented by a three-year average centered on each target year. Meteorological data (spatial resolution: 250 m) were obtained from the National Tibetan Plateau Data Center (https://data.tpdc.ac.cn/, accessed on 22 August 2025). NPP data (250 m resolution) were sourced from the MODIS MOD17A3 dataset (https://www.earthdata.nasa.gov/, accessed on 22 August 2025). Elevation and land-use/land cover data (30 m resolution) were obtained from the Resource and Environmental Science and Data Center (https://www.resdc.cn/, accessed on 22 August 2025). Road network data were sourced from the National Fundamental Geographic Information Database (https://ngcc.cn/, accessed on 22 August 2025), and administrative boundary data were acquired from Tianditu (https://www.tianditu.gov.cn/, accessed on 22 August 2025) under map review number GS(2024)0650.

Due to limitations in data availability, adjacent-year data were used to substitute for missing data where necessary. All spatial analyses were conducted using ArcGIS 10.2 and InVEST 3.14.2 software. Raster datasets were uniformly resampled to a 30 m resolution and projected using the CGCS2000 geographic coordinate system. The CGCS2000_3_Degree_GK_Zone_39 projection coordinate system was added to ensure consistency in model inputs. It should be acknowledged that the original datasets differed in spatial resolution. However, since the research models primarily capture broad-scale spatial patterns rather than fine-scale heterogeneity, such uncertainties are deemed acceptable within the scope of this study.

2.3. Methodology

Based on the above, we developed a coupled framework that integrates the Habitat Quality (HQ) module with the Biodiversity Maintenance Function (BMF) index. The technical roadmap for the HQ-BMF coupled model is shown in Figure 2. This model was applied to conduct a comprehensive multi-model evaluation of agricultural heritage systems in the Beijing–Tianjin–Hebei (BTH) region. It was further used to analyze the key factors affecting biodiversity in these systems and to propose corresponding protection strategies.

2.3.1. Biodiversity Maintenance Function (BMF) Index Model

The Biodiversity Maintenance Function (BMF) index model used in this study is based on the Guidelines for the Delineation of Ecological Conservation Redlines (2017 Edition) issued by the Chinese government [31,32]. The model integrates four key natural factors: net primary productivity (NPP), annual precipitation, mean annual temperature, and elevation. These variables quantify the region’s capacity to sustain biodiversity. The resulting BMF reflects the relative strength of biodiversity–supporting capacity across different areas. The calculation is expressed as follows:

$$BMF={NPP}_{mean}\times F_{pre}\times F_{tem}\times \left(1-F_{alt}\right)$$

(1)

where BMF is the Biodiversity Maintenance Function Index; NPP_mean is the multi-year average Net Primary Productivity; F_pre is the normalized multi-year average precipitation factor; F_tem is the normalized multi-year average temperature factor; F_alt is the normalized elevation factor.

2.3.2. InVEST Habitat Quality (HQ) Model

The Habitat Quality (HQ) module of the Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model was applied in this study. It was used to evaluate the spatial distribution of habitat quality under land-use changes and anthropogenic disturbances. In this model, habitat quality is determined by four main factors: (1) the relative impact of each threat, (2) the sensitivity of each habitat type to each threat, (3) the distance between each grid cell and threat sources, and (4) the level of legal protection afforded to the cell. Given that ecological protection measures have been relatively well implemented in the Beijing–Tianjin–Hebei region, this study focused primarily on the first three factors [33]. The calculation is expressed as follows:

$$D_{xj}={\sum }_{r=1}^R{\sum }_{y=1}^{Y_r}\left({\displaystyle \frac{{\omega }_r}{\sum _{r=1}^R{\omega }_r}}\right)r_y{i}_{rxy}{\beta }_x{S}_{jr}$$

(2)

$$i_{rxy}=1-\left({\displaystyle \frac{d_{xy}}{d_{rmax}}}\right)\left(Linear decay\right)$$

(3)

$$i_{rxy}=exp\left(-\left({\displaystyle \frac{2.99}{d_{rmax}}}\right)d_{xy}\right)\left(Exponential decay\right)$$

(4)

where D_xj is the degree of habitat degradation; R is the total number of threat factors; Y_r is the number of grid cells for threat factor r within the land cover layer; ω_r is the weight assigned to threat factor r; r_y is the intensity of threat at grid cell y; i_rxy is the impact level of threat r from cell y on cell x; β_x is the accessibility level of threat source to grid cell x; S_jr is the sensitivity of habitat type j to threat r; d_xy is the straight-line distance between habitat grid cell x and threat factor grid cell y; d_rmax is the maximum effective distance over which threat r can influence habitat quality.

Habitat quality is calculated based on this, using the following formula:

$$Q_{xj}=H_j\left[1-\left({\displaystyle \frac{D_{xj}^z}{D_{xj}^z+k^z}}\right)\right]$$

(5)

where Q_xj is the habitat quality index; H_j is the habitat suitability index, where 0 ≤ H_j ≤ 1; D_xj is the habitat degradation index; z is the normalization constant, with a value of 2.5; k is the half-saturation constant, usually defined as half the maximum degradation value. In this study, k = 0.5.

In this study, five threat factors were selected, including urban construction land, rural residential land, and railways. Their maximum impact distance, weight, and decay type were determined with reference to previous studies in similar regions [34,35,36,37,38,39,40] and the InVEST user guide. These values were further adjusted according to the ecological and geographical characteristics of the study area (

Table 2. Treats and their maximum distance of influence, weights and decay type.

Threat	Max_Dist	Weight	Decay
Urban construction Land	10	1	exponential
Rural Residents’ Construction Land	6	0.7	exponential
Other Construction Land	8	0.8	exponential
Railway	6	0.5	linear
National Highway	4	0.6	linear

). Corresponding sensitivity indices were also established (

Table 3. Habitat Suitability and Relative Sensitivity to Threat Sources.

LULC Code	LULC Type	Habitat Suitability	Relative Sensitivity to Threat Sources
			Urban Construction Land	Rural Residents’ Construction Land	Other Construction Land	Railway	National Highway
1	Cultivated Land	1	0.8	0.5	0.7	0.5	0.4
2	Forest	1	1	0.7	1	0.7	0.7
3	Grassland	1	1	0.7	1	0.8	0.7
4	Water	1	0.8	0.7	0.8	0.5	0.5
5	Urban construction Land	0	0	0	0	0	0
6	Rural Residents’ Construction Land	0	0	0	0	0	0
7	Other Construction Land	0	0	0	0	0	0
8	Unused Land	1	0.3	0.2	0.3	0.2	0.1
9	Ocean	1	0.2	0.1	0.1	0.1	0.1

).

It is important to note that the InVEST–HQ module does not directly measure actual biodiversity. Instead, it serves as a proxy by estimating habitat suitability and potential biodiversity patterns from land-use and threat factors. This limitation should be carefully considered when interpreting the results.

2.3.3. Coefficient of Variation (CV)

The coefficient of variation (CV) is defined as the ratio of the standard deviation to the mean, and it reflects the degree of dispersion of a variable [41,42]. In this study, CV was used to evaluate the relative stability and fluctuation of the Biodiversity Maintenance Function Index (BMF) and Habitat Quality (HQ) over time. A higher CV indicates greater differences in biodiversity among spatial units, reflecting a more complex ecosystem structure and uneven functional distribution. Conversely, a lower CV suggests smaller differences in biodiversity among units, implying a more homogeneous and stable ecosystem. Unlike standard deviation, CV removes the influence of measurement units. This makes it suitable for comparing variables with different units or magnitudes. The formula for calculating CV is as follows:

$$CV={\displaystyle \frac{\sigma }{\mu }}\times 100\%$$

(6)

where CV is the coefficient of variation; σ is the sample standard deviation; μ is the sample mean. A higher CV value indicates greater temporal variability of the sample, while a lower CV suggests more stability over time.

2.3.4. Chi-Square Test

The chi-square test is a widely used non-parametric statistical method suitable for analyzing relationships between categorical variables. The chi-square statistic reflects the degree of deviation between observed and expected frequencies. In this study, the test was employed to assess whether there is a statistically significant relationship between the distribution of heritage systems and elevation, thereby providing insights into the spatial distribution patterns of agricultural heritage systems. This analysis offers additional scientific support for biodiversity conservation in these areas. The chi-square formula is as follows:

$$X^2=\sum {\displaystyle \frac{{\left(O_i-E_i\right)}^2}{E_i}}$$

(7)

where X² is the Chi-square test statistic; O_i is the observed frequency of category i; E_i is the expected frequency of category i. A larger Chi-square value indicates a greater deviation between the observed and expected frequencies, suggesting a stronger discrepancy. Conversely, a smaller Chi-square value implies a closer match between the observed and expected distributions.

Based on these definitions, the hypotheses were formulated as follows. The null hypothesis (H₀) assumes that there is no significant association between distribution agricultural heritage systems and elevation, meaning that the systems are randomly distributed across elevation categories. The alternative hypothesis (H₁) assumes that a significant association exists, indicating that the system distribution is influenced by elevation.

3. Results

3.1. Spatial Distribution and Temporal Dynamics of Biodiversity Maintenance Function (BMF) and Habitat Quality (HQ) in the Beijing–Tianjin–Hebei (BTH) Region

From a temporal perspective, both the Biodiversity Maintenance Function (BMF) and Habitat Quality (HQ) in the Beijing–Tianjin–Hebei (BTH) region showed clear changes between 2003 and 2023. For comparative analysis, BMF values were grouped into three categories: low quality (Lq), medium quality (Mq), and high quality (Hq). The classification followed the Guidelines for the Delineation of Ecological Conservation Redlines (2017 edition), using 50% and 80% cumulative distribution thresholds and considering the characteristics of heritage systems. Similarly, HQ values were categorized using the natural breaks (Jenks) method, with cut-off points set at the 33.3% and 66.6% quantiles. The average BMF values for 2003, 2008, 2013, 2018, and 2023 were 0.1036, 0.1261, 0.1041, 0.1248, and 0.1397, respectively. This indicates a non-linear pattern of increase, decline, and rebound (Figure 3).

Over the 20-year period, the proportion of Lq areas remained relatively stable at approximately 50%, while Mq areas fluctuated around 30%. Both Lq and Mq areas decreased by about 600 km² between 2003 and 2023. Meanwhile, the area of Hq increased slightly, from 43,906 km² to 45,102 km² (Figure 4).

In contrast, the mean HQ values for 2003, 2008, 2013, 2018, and 2023 were 0.8734, 0.8517, 0.8042, 0.7867, and 0.7729, respectively, showing a gradual decline in habitat quality over the past two decades (Figure 5).

The proportion of Hq areas steadily declined by nearly 8% during this period. Concurrently, Lq areas expanded by almost 60%, and Mq areas also increased significantly in the region (Figure 6).

Both BMF and HQ exhibited notable spatial heterogeneity, but with largely inverse spatial patterns. BMF was generally higher in the southeast and lower in the northwest, with high-value clusters located in areas such as Chengde–Qinhuangdao–Tangshan, southeastern Cangzhou, Handan–Xingtai, and parts of Baoding. In contrast, HQ was higher in the northwest and lower in the southeast. High HQ values were concentrated in mountainous zones like the Yanshan–Taihang range, while low HQ values were primarily observed in highly urbanized areas such as Beijing, Tianjin, and Shijiazhuang, where human activities are most intensive.

To assess temporal variability, 1600 raster pixels were randomly extracted from the study area. After removing outliers, the coefficient of variation (CV) was calculated for each indicator across the five time points. The maximum CV for BMF was 0.5000 (Figure 2), and for HQ it was 0.3524 (Figure 4), both falling within the moderate variability range of 0.25–0.50 [43]. The findings indicate that changes in BMF and HQ occurred gradually rather than abruptly. This suggests that ecological functions in the BTH region remained relatively stable over the study period. These regional-scale patterns provide a basis for testing whether similar dynamics are also evident within the nationally recognized agricultural heritage systems, discussed in the next section.

3.2. Spatial and Temporal Dynamics of Biodiversity Maintenance Function (BMF) and Habitat Quality (HQ) in the 14 Heritage Systems of the Beijing–Tianjin–Hebei (BTH) Region

Building on the regional analysis in Section 3.1, this section focuses on the spatiotemporal dynamics of Biodiversity Maintenance Function (BMF) and Habitat Quality (HQ) within the 14 agricultural heritage systems of the Beijing–Tianjin–Hebei (BTH) region.

3.2.1. Spatial Distribution Characteristics of Agricultural Heritage Systems

A total of 14 nationally designated agricultural heritage systems have been identified in the BTH region. To examine their elevation characteristics, the terrain was classified into three elevation zones: low elevation (<200 m), medium elevation (200–500 m), and high elevation (>500 m). The low-elevation areas are mainly located in the Haihe Plain and coastal alluvial plains along the Bohai Bay, the medium-elevation zone corresponds to transitional areas between hills and low mountains, and the high-elevation zone comprises typical mountainous terrains. The proportions of the total area covered by these three elevation levels are 48.4%, 9.9%, and 41.7%, respectively.

Against this topographical backdrop, the 14 heritage systems exhibit pronounced spatial clustering: 43.0% of their total area lies in high-elevation zones, 27.1% in medium-elevation zones, and only 29.9% in low-elevation zones. Although the medium and high elevation zones account for just 58.3% of the total land area, they encompass 70.1% of the agricultural heritage land area, indicating a clear preference for mountainous and hilly areas. Further chi-square testing confirmed a highly significant spatial association between heritage site distribution and topography (χ² = 124,683, df = 2, p < 0.001), highlighting a marked preference for mid to high-elevation slopes over lowland plains (

Table 4. Expected and actual values for different heritage systems at different altitudes.

	Proportion of Each Terrain Type	The Heritage System Is Located at Different Altitudes	Total Area of Heritage System (km2)	Expected Value (km2)	Actual Value (km2)
low altitude area	48.4%	29.9%		162,887	100,505
Medium-altitude area	9.9%	27.1%	336,543	33,318	91,233
High altitude area	41.7%	43.0%		140,339	144,806

).

3.2.2. Temporal Changes in BMF of the 14 Agricultural Heritage Systems

The study assessed the BMF of 14 agricultural heritage systems in the BTH region across five time points (Figure 7, illustrated using 2023 as an example). In accordance with the established approach, BMF data were categorized into low quality (Lq), medium quality (Mq), and high quality (Hq) areas. Mean BMF and coefficients of variation (CVs) were computed for each system (

Table 5. Average BMF values and CV for 14 heritage systems in the Beijing–Tianjin–Hebei region for each year.

	Year	2003		2008		2013		2018		2023
Code		Average	CV	Average	CV	Average	CV	Average	CV	Average	CV
A		0.1228	0.2340	0.1664	0.2638	0.1530	0.2494	0.1670	0.2816	0.1559	0.2773
B		0.0759	0.8956	0.1038	0.8966	0.0919	0.8991	0.1074	0.9002	0.1164	0.9028
C		0.1077	0.3918	0.1439	0.4213	0.1355	0.3798	0.1549	0.4304	0.1687	0.4315
D		0.1043	0.2406	0.1483	0.2548	0.1321	0.2627	0.1381	0.2974	0.1552	0.2554
E		0.1432	0.4092	0.1628	0.4110	0.1381	0.4168	0.1843	0.4107	0.2225	0.4065
F		0.1410	0.4636	0.1710	0.4677	0.1400	0.4770	0.1897	0.4775	0.2287	0.4757
G		0.0830	0.3125	0.1324	0.3095	0.0831	0.3162	0.0965	0.3246	0.1243	0.3181
H		0.1593	0.1739	0.2058	0.1736	0.1987	0.1828	0.2169	0.1748	0.2325	0.1743
I		0.1441	0.1925	0.1692	0.2047	0.1670	0.2143	0.1833	0.2122	0.1898	0.2144
J		0.0893	0.1154	0.1213	0.1060	0.0634	0.1505	0.0995	0.1274	0.1366	0.1164
K		0.2452	0.2757	0.2143	0.2843	0.1735	0.2828	0.2235	0.2788	0.2731	0.2819
L		0.0245	0.4111	0.0377	0.4126	0.0267	0.4187	0.0271	0.4104	0.0171	0.4092
M		0.0501	0.2130	0.0839	0.2401	0.0580	0.2075	0.0671	0.2021	0.0470	0.1991
N		0.1104	0.2310	0.1489	0.2367	0.1309	0.2378	0.1451	0.2493	0.1413	0.2563

). The percentage change in BMF (ΔBMF%) from 2003 to 2023 (Figure 8) followed a three-phase pattern: initial increase (2003–2008), subsequent decline (2008–2013), and recovery (2013–2023). During 2003–2008, most systems experienced significant BMF increases. This was followed by widespread decline during 2008–2013, and a gradual rebound between 2013 and 2023, with several systems recovering previous losses. For instance, System B is located in Haidian District. Its BMF rose by +36.8% during 2003–2008, then dropped by −11.5% in 2008–2013. After that, it rebounded sharply, with increases of +16.9% (2013–2018) and +8.4% (2018–2023).

All but one system (System B, CV = 0.9028) maintained moderate to low variability, suggesting generally stable internal dynamics. Sankey diagrams (Figure 9) revealed that approximately 60% of BMF changes occurred within the same quality class. The period 2008–2013 represented the key phase of improvement. During this time, upward transitions increased by 62%, with Mq→Hq changes accounting for more than 68% of the total gains. Downgrades mainly occurred as Mq→Lq, while Hq→Lq losses were negligible. From 2003 to 2008, most systems exhibited intra-class transitions, with some Lq→Mq (e.g., System G: 103.97 km²) and Mq→Hq (System A: 91.30 km²; System B: 23.72 km²). The 2008 to 2013 period featured active transitions from Lq to Mq and Mq to Hq. Meanwhile, the total area of Hq continued to expand. System I showed the highest single-phase gain (Mq→Hq: 222.89 km²). Systems E, F, and K performed well in retaining Hq areas. During 2013–2018, Hq areas remained mostly stable. Only a few systems, such as A and N, experienced localized Mq→Lq downgrades (28.29 km² and 23.51 km², respectively), which did not alter the general upward trajectory. From 2018 to 2023, systems generally retained Mq and Hq areas while seeing local rebounds in Lq areas. Systems I, N, and A maintained strong Mq→Mq retention (354.69 km², 480.23 km², 139.44 km², respectively). Systems E, F, and K continued to preserve large Hq zones. However, System H exhibited some degradation (Hq→Mq: 71.37 km², 24% of its total area). Overall, while BMF recovery was evident, localized volatility and degradation risks warrant continued monitoring.

3.2.3. Temporal Changes in HQ of the 14 Agricultural Heritage Systems

HQ was similarly assessed over five time points (Figure 10, using 2023 as an example). Consistent with the previous method, HQ data were categorized into low quality (Lq), medium (Mq), and high (Hq) quality areas. Each system’s mean HQ, CV, and ΔHQ% (Figure 11) values were calculated (

Table 6. Average HQ values and CV for each year at 14 heritage systems in the Beijing–Tianjin–Hebei region.

	Year	2003		2008		2013		2018		2023
Code		Average	CV	Average	CV	Average	CV	Average	CV	Average	CV
A		0.9656	0.1268	0.9349	0.1397	0.9323	0.1750	0.9169	0.1862	0.9154	0.1867
B		0.5094	0.7829	0.3972	0.9817	0.2898	1.3318	0.3229	1.2180	0.2356	1.4480
C		0.6691	0.3990	0.6330	0.4218	0.5568	0.6136	0.5965	0.4458	0.5802	0.4624
D		0.9701	0.1124	0.9685	0.1128	0.9397	0.1347	0.9401	0.1335	0.9400	0.1339
E		0.7276	0.4361	0.5975	0.5579	0.5853	0.6513	0.4634	0.8454	0.4163	0.9042
F		0.8480	0.2954	0.8434	0.2999	0.8747	0.2791	0.7186	0.5253	0.8523	0.2721
G		0.8629	0.3608	0.8616	0.3617	0.8363	0.4114	0.7966	0.4220	0.8313	0.4206
H		0.9799	0.0767	0.9774	0.0773	0.8696	0.2099	0.8602	0.2249	0.8658	0.2219
I		0.9679	0.0936	0.9466	0.1146	0.8268	0.2405	0.7985	0.2933	0.7953	0.3071
J		0.8305	0.4342	0.8273	0.4362	0.8720	0.3430	0.8693	0.3442	0.8680	0.3438
K		0.9023	0.2069	0.8358	0.3099	0.7863	0.3769	0.7666	0.4023	0.7476	0.4231
L		0.7879	0.3083	0.6022	0.3982	0.5106	0.5989	0.4203	0.6681	0.3637	0.7691
M		0.9066	0.2101	0.8776	0.2173	0.8294	0.2490	0.8261	0.2510	0.7890	0.2613
N		0.8990	0.1766	0.8974	0.1772	0.8347	0.2517	0.8262	0.2628	0.7988	0.2826

). Results showed a continuous decline in HQ for 8 of the 14 systems, indicating that more than half of the systems experienced sustained habitat degradation. For System L, HQ showed a continuous decline across all periods. Specifically, HQ decreased by −23.6% between 2003 and 2008, followed by further reductions of −15.2% in 2008–2013, −17.7% in 2013–2018, and −13.5% in 2018–2023.

Across all periods, most systems showed varying degrees of HQ degradation, reflecting a steady decline in ecological quality. Systems B, E, and L exhibited the largest CV (1.448, 0.904, and 0.769, respectively), while most other systems maintained low to moderate variation. HQ Sankey diagrams (Figure 12) showed that around 80% of transitions involved the retention of Hq areas. The critical period for degradation was 2008–2013, when downgrades (Hq→Mq, Hq→Lq, Mq→Lq) surged by 175% compared to 2003–2008. From 2003 to 2008, most systems exhibited stable Hq→Hq retention. Systems A, H, I, and N retained over 90% of their Hq areas. Upward transitions were rare (e.g., E: 0.79 km² Lq→Hq; C and N: 3.04 and 5.91 km² Mq→Hq). From 2008 to 2013, Hq retention remained dominant (82% of changes), with minor downgrades in some systems. During 2013–2018, Hq area generally declined while Lq expanded. For instance, System I’s Hq shrank from 487.19 to 465.75 km²; System N dropped from 615.01 to 606.51 km². Concurrently, Lq grew: System B from 53.07 to 63.87 km²; System I from 4.36 to 18.37 km². The 2018–2023 period continued this trend, with further Hq degradation and Lq expansion. Systems I and N saw notable Hq→Mq downgrades (14.13 km² and 34.97 km², respectively). Lq areas reached peak levels: System B (63.87→73.69 km²), System E (74.51→113.81 km²). Only System F showed notable improvement (Lq→Hq: 5.57→30.49 km²). Overall, HQ in these systems is still on a slow downward trajectory.

3.3. BMF–HQ Coupling Analysis of the 14 Agricultural Heritage Systems

Biodiversity Maintenance Function (BMF) reflects the natural potential of heritage systems, while Habitat Quality (HQ) measures anthropogenic pressures. Using GIS-based zonal statistics, annual mean values for both indicators were calculated. These values were categorized into high, medium, and low quality, and cross-tabulated into a 3 × 3 matrix to visualize the ecological state of each system over time (Figure 13). A “dual -high” cell—where high habitat quality combined with high ecological potential—denotes a well preserved state. Conversely, a “dual-low” cell—low quality coupled with low ecology—indicates severe degradation.

Between 2003 and 2023, HQ–BMF coupling patterns shifted significantly. The number of “dual-high” system (high HQ and high BMF) increased from one in 2003 to three in 2008–2013, declined to two in 2018, and dropped to one in 2023. This trend indicates that systems with strong ecological potential and low disturbance became fewer after reaching a peak. In contrast, “dual-low” systems were absent until 2013, after which one system consistently fell into this category, signaling emerging ecological degradation. Some systems transitioned from “high HQ–high/medium BMF” to “high HQ–medium/low BMF” or “medium HQ–medium BMF,” suggesting declining biodiversity despite intact habitats. The most common pattern in 2023 was “high HQ–medium BMF” (6 systems), indicating good habitat quality but only moderate biodiversity levels. “High HQ–low BMF” remained present, while “medium HQ–medium BMF” stabilized at two systems from 2008 onward.

4. Discussion

This study employed the Biodiversity Maintenance Function (BMF) model and the Habitat Quality (HQ) model to systematically assess the spatiotemporal dynamics of biodiversity and habitat quality in the Beijing–Tianjin–Hebei (BTH) region and its 14 agricultural heritage systems between 2003 and 2023. The results revealed the ecological service evolution and potential risks at both regional and system scales. By integrating multi-temporal remote sensing data and a dual-factor coupling matrix of HQ and BMF, the study captured pronounced spatial differences in ecosystem status across heritage systems, effectively identifying typical categories such as “dual-high” and “dual-low” zones. These findings provide a scientific basis for refined management strategies, including system-specific protection, ecological compensation, and development control.

At the regional level, the overall BMF in the BTH region exhibited a slight upward trend from 2003 to 2023, with the area of high-BMF zones increasing from 43,906 km² to 45,102 km², indicating the positive impact of ecological projects such as reforestation and green barrier initiatives in enhancing ecological potential [44,45,46]. In contrast, HQ continued to decline, marked by a significant reduction in high-quality habitats and a rapid expansion of low-quality areas. Since 2003, rapid urbanization and industrialization in the region have accelerated land-use changes, particularly the conversion of cultivated land into construction land, which directly contributed to the observed decline in HQ (Figure 14). Such a decline is consistent with the empirical findings of Li et al. and Deng et al. [47,48], who identified urban expansion and human activities as key drivers of HQ degradation. The contrasting pattern of rising BMF alongside steep HQ decline suggests that enhancing natural potential alone is insufficient to counter habitat degradation. Ecological governance must also address human-induced pressures.

Beyond the regional perspective, the spatial distribution of heritage systems provides further insight into ecological vulnerability. In the spatial distribution of the 14 agricultural heritage systems, there is a clear preference for mid- to high-elevation areas, especially hilly and mountainous zones. In contrast, low-elevation plains are sparsely represented. This pattern closely relates to the stable ecological conditions of mid- to high-elevation areas. These zones often serve as water conservation regions and ecological barriers, with abundant natural resources that support the long-term maintenance of traditional agricultural systems. Limited accessibility has also helped preserve traditional farming practices [49]. As a result, distinctive agricultural cultures and landscapes have developed in these areas. In addition, high-elevation regions face less pressure from urban expansion and experience lower levels of land-use change. These factors further contribute to the conservation and continuity of heritage systems. Since China launched its Globally Important Agricultural Heritage Systems (GIAHS) initiative in 2005, a national and global heritage identification and management framework has gradually been established. However, dynamic monitoring of HQ and BMF in this study indicates that the ecological status of most heritage systems has not sustained improvement.

The BMF of most heritage systems showed a three-stage fluctuation: an initial increase, a decline, and a recovery. This pattern mirrored the regional BMF trend and reflected the responsiveness of agricultural heritage systems to ecological changes at the regional scale. Among all systems, System B had the highest coefficient of variation (CV). The system is located in Haidian District, the core area of Beijing’s urban development. Under the pressures of rapid urban expansion and intensive land-use, it experienced severe habitat degradation. Later, the city carried out large-scale greening and ecological restoration projects [50,51,52]. These measures improved local ecological functions and caused pronounced fluctuations at System B. In contrast, HQ declined consistently. The deterioration exceeded the regional average, especially during 2008–2013, when several heritage systems suffered substantial habitat loss. Although partial recovery occurred in some areas, the overall ecological condition remains fragile. The vulnerability is due to the high sensitivity of HQ to land-use types and intensity. Agricultural heritage systems usually include not only cropland but also built-up areas, forests, grassland, and transportation land. For example, the CV of HQ in System B and System L reached 1.448 and 0.904, respectively, indicating high spatial variability.

Land-use transition matrices (Figure 15) reveal that both systems underwent continuous expansion of built-up areas at the expense of cropland and ecological land. In System B, built-up land increased by 39.40 km² over 20 years, with approximately 77% converted from cropland. The peak period for this transition was 2008–2013, and about 10 km² of forestland was also lost. Compared to the initial year, cropland, forest, and water bodies decreased by 36.11 km², 4.84 km², and 0.44 km², respectively. In System L, built-up land increased by 4.33 km², mainly from cropland (60%) and grassland (1.38 km²). Cropland and water areas decreased by 1.97 km² and 1.30 km², respectively. This expansion of built-up areas has significantly impacted habitat quality. Therefore, protecting agricultural heritage systems requires not only reinforcing boundary control but also acknowledging that their ecological vulnerability stems not only from internal spatial changes but also from broader natural and anthropogenic factors. Future strategies must go beyond the traditional ‘core agricultural’ perspective. They should move toward integrated, landscape-based ecological management and land-use optimization.

The HQ–BMF coupling matrix highlights the spatial heterogeneity of ecosystem conditions in agricultural heritage systems. It offers a quantitative basis for differentiated management and intervention. Distinct “dual-high” and “dual-low” spatial patterns were observed. The “dual-high” areas indicate both high habitat integrity and strong biodiversity maintenance functions. These zones suggest stable ecosystems and well-preserved heritage systems. From a management perspective, their priority is to maintain this favorable ecological status. In contrast, the “dual-low” areas reflect ecological fragility and high land-use pressure. These systems should be prioritized for ecological restoration and policy interventions. Overall, the HQ–BMF framework offers policymakers clear priorities and supports a more science-based approach to the protection of agricultural heritage systems.

Among the 14 systems, the “high HQ–medium/low BMF” type was predominant. This suggests that habitat conditions face limited anthropogenic interference, but biodiversity potential remains underutilized. The loss of traditional agricultural elements may explain this. Weakened landscape functions and simplified land-use structures are also important drivers. This pattern is consistent with findings by Volpato et al. [53] who observed that high ecological potential often remains untapped in low-intensity agricultural systems. For such nature-dominated low-BMF systems, ecological restoration measures should be prioritized. Examples include reviving terraces, water channels, and traditional irrigation systems to enhance water regulation. Agroforestry systems and integrated eco-farming can also play a role. In addition, creating microhabitats and ecological buffers improves habitat diversity and resilience. These strategies have shown positive results elsewhere. For instance, Brancalion et al. [54] reported significant biodiversity and ecosystem service gains through forest restoration in tropical agricultural landscapes. Similarly, Assandri et al. [55] found that preserving traditional features (e.g., hedgerows, dry-stone walls, and microhabitats) enhanced both biodiversity and cultural value, particularly for bird species.

In contrast, low HQ areas reflect stronger human disturbances and are typically located on urban fringes or in systems under high development pressure. Examples include the Jingxi rice zone and Xuanhua vineyard in Beijing, which face ongoing urban encroachment, habitat fragmentation, and land conversion pressures [56,57]. These “dual-low” zones urgently require ecological restoration and strict development control. Similar issues have been observed in agricultural heritage systems in Guangzhou, Zhuhai, and the polder (duotian) systems of Jiangsu Province [58,59].

Based on the ecological differentiation and systems distribution, this study proposes a multi-level protection strategy informed by the HQ–BMF coupling framework.

• Dual-high zones (high HQ–high BMF): The focus should be on keeping current ecological advantages. Long-term conservation and monitoring mechanisms are needed to protect against external disturbances.
• Dual-low zones (low HQ–low BMF): These systems are most affected by human activities. They should be placed in strict protection areas with tightly controlled development intensity. Ecological compensation and pilot restoration programs should be prioritized to prevent further loss of core agricultural and cultural values.
• Other zones: The main drivers of degradation must be clarified. If human disturbances dominate, measures should include land-use optimization and restrictions on development. If natural factors dominate, restoration should focus on creating microhabitats, recovering native vegetation, and revitalizing traditional agro-technical systems. These steps can strengthen ecological carrying capacity and biodiversity.

This study highlights the coupling analysis of HQ and BMF as a useful tool for adaptive management of agricultural heritage systems. The HQ–BMF framework helps identify spatial heterogeneity, track ecological changes, and detect potential risks. It also guides managers to adjust conservation actions based on different HQ–BMF types, enabling more precise and differentiated strategies. By doing so, managers can implement more precise and differentiated strategies. This approach supports both habitat quality and biodiversity maintenance, and is essential for the long-term sustainability of agricultural heritage systems.

While the proposed framework offers practical insights, several methodological limitations should be noted. The HQ module of the InVEST model is highly sensitive to the assigned weights and sensitivity scores of threat factors. In this study, these parameters were derived from existing literature rather than field calibration, which may lead to deviations between model outputs and actual conditions. Moreover, due to limited availability of meteorological and road network data for some target years, adjacent-year data were used as substitutes. This approach may introduce certain inaccuracies. In addition, urbanization rates, elevation-linked land management history, and external disturbances (e.g., natural disasters and ecological restoration projects) were not fully considered in the current analysis. These factors may partially explain the spatial and temporal variations observed. Future studies should seek higher-resolution datasets, incorporate broader socio-ecological drivers, and apply more rigorous parameter validation to improve the accuracy and explanatory power of agricultural heritage ecosystem assessments.

5. Conclusions

This study systematically assessed the spatiotemporal dynamics of Biodiversity Maintenance Function (BMF) and Habitat Quality (HQ) across the Beijing–Tianjin–Hebei (BTH) region and 14 nationally designated agricultural heritage systems from 2003 to 2023. While regional BMF showed a slight upward trend, HQ continuously declined, with a notable reduction in high-quality areas and expansion of low-quality areas. At the heritage systems level, BMF experienced a phased trajectory—an initial increase, followed by decline, and subsequent recovery—peaking during 2008–2013. In contrast, HQ steadily deteriorated across most systems, highlighting ecological vulnerability under urban expansion. The HQ–BMF coupling analysis revealed strong spatial heterogeneity. By 2023, only one system remained within a “dual-high” zone, while another had fallen into a “dual-low” condition, suggesting localized ecological degradation. Most systems were categorized as “high HQ–medium BMF”, indicating latent ecological potential in relatively undisturbed areas. Based on these findings, we propose a typology-based intervention strategy. Overall, the proposed HQ–BMF coupling framework provides a valuable decision-support tool for adaptive spatial management of agricultural heritage systems. It helps identify ecological risks, detection of change trajectories, and formulation of precise and differentiated protection strategies. This dual-indicator approach is key to enhancing both habitat quality and biodiversity conservation, supporting the long-term sustainability of agricultural heritage systems.