Assessment of conservation priorities for agricultural heritage systems in Beijing: a pressure-state-response and fuzzy comprehensive evaluation framework

Abstract

Agricultural heritage is a complex system created by humans during long-term synergistic development with nature. Urbanization and industrialization have resulted in numerous challenges, including the disappearance of local varieties and the degradation of agricultural landscapes. Beijing's 48 municipally recognized agricultural heritage systems (2023 inventory) were selected as the study population, and a 24-indicator protection priority index system was constructed based on the Pressure-State-Response (PSR) model. The fuzzy comprehensive evaluation (FCE) method classified priority levels into red, orange, yellow, and blue groups from high to low. The results indicate a significant variation in risk levels across Beijing's 48 agricultural heritage systems, with two classified as red alerts, and eleven as orange alerts, indicating immediate need for attention. Targeted recommendations for each alert level are presented, encompassing enhanced governmental support and real-time monitoring systems. The findings are crucial for refining conservation practices for Beijing's agricultural heritage and serve as valuable references for structured conservation interventions in other regions. This study provides a decision-support framework for prioritizing agricultural heritage conservation, contributing to sustainable rural development under urban pressure.

1 Introduction

Agricultural heritage systems (AHSs) are complex, diverse, and locally adapted agricultural systems. These systems have been managed with time-tested, ingenious combinations of techniques and practices that have generally led to community food security and the conservation of natural resources and biodiversity (Koohafkan and Altieri, 2011). Owing to the ability of rural communities to adapt to the surrounding environment, AHS has significant social, ecological, economic and cultural value (Bahnasy, 2024; Piras et al., 2024). AHS enables social cohesion, fosters socioeconomic regeneration and poverty reduction, strengthens social wellbeing, improves the appeal and creativity of regions, and enhances long-term tourism benefits (Labadi et al., 2021). In 2002, the United Nations Food and Agriculture Organization (UNFAO) initiated the Globally Important Agricultural Heritage Systems (GIAHS) program and meticulously developed the GIAHS network, which now spans 28 countries worldwide and encompasses 95 heritage sites. The GIAHS initiative has been successful in increasing global awareness of the value of agricultural heritage. Inspired by the GIAHS, and several countries have initiated their own agricultural heritage conservation programs. For instance, China launched its own recognition scheme for China Important Agricultural Heritage Systems (CIAHS) and designated a total of 188 such sites from 2013 to 2025. Concurrently, a surge of enthusiasm has been observed at the regional level, with multiple localities undertaking surveys of agricultural heritage to ascertain the state of its preservation and to advance its sustainable utilization.

Despite the widespread recognition of the significance of agricultural heritage (Koohafkan and Cruz, 2011), its protection and development continue to face a plethora of challenges (Poplawska, 2025) that converge particularly intensely in megacity environments. Modernization and the spread of input-intensive agriculture displace traditional knowledge and varieties (Balick, 2017; Srivastava et al., 2016). Urbanization fragments core areas (Fu et al., 2023), degrades traditional agricultural landscapes (Wei et al., 2025), raises land opportunity costs, exacerbates the risk of agricultural heritage disappearance (Yang et al., 2023), and erodes practitioner cohorts as young rural residents migrate to cities (Nath et al., 2024), leading to the loss of traditional production techniques (Bai et al., 2024) and a decreasing production scale (Du et al., 2024). Climate variability—manifested in more frequent precipitation extremes and shifts in growing seasons—disrupts the ecological niches to which traditional varieties have been locally adapted (Abebaw, 2025). Such spatiotemporal dynamics of ecosystem productivity further influence local agricultural outputs and resource allocation in terrestrial landscapes, highlighting the importance of integrating ecological drivers into agricultural heritage assessments (Li et al., 2025). At the same time, the expansion of global supply chains into previously localized markets diminishes brand competitiveness (Lee et al., 2012). This compounding effect has been documented across diverse contexts, including Spain's historical landscapes (Manrique Anticona et al., 2023) and Slovakia's biocultural plot systems (Izakovičová et al., 2022), and is particularly pronounced in metropolitan regions, where agricultural land directly competes with high-value urban uses.

The literature on AHS protection has matured along four parallel strands. (i) Value assessment (Nahuelhual et al., 2014; Xiang et al., 2024) frameworks quantify cultural, ecological and economic dimensions of AHS to argue for their conservation (Barrena et al., 2014). (ii) Vulnerability and risk-assessment frameworks adopt the cultural-heritage risk index (Ravankhah et al., 2021), the climate-vulnerability function (Ducusin et al., 2019), or the early-warning typology of Cui and Wang (2023) to flag at-risk systems. (iii) Early-warning monitoring systems embed sensor networks and reporting cycles into conservation programmes, with documented benefits in Japan's Noto Peninsula GIAHS (Zhao et al., 2025) and the European agroecological landscape monitoring network (Agbehadji et al., 2023; Petit et al., 2023). (iv) Spatial and management prioritization tools rank sites or land-use types by urgency for intervention (Bennett and Mac Nally, 2004; Akpinar et al., 2005; Zhou et al., 2023). Each strand has methodological strengths but addresses a different question—“how valuable?”, “how vulnerable?”, “is something happening now?”, or “where to act first?”.

What is missing is a unified, regional-scale workflow that simultaneously captures human pressure, system state, and policy response in a single bounded score, and translates this score into an operationally meaningful four-tier ranking that a municipal conservation agency can act on. Vulnerability frameworks (Hulme et al., 2025) evaluate one site at a time and emphasize sensitivity rather than triage; value frameworks justify protection but do not allocate scarce resources; early-warning systems monitor change but require a baseline ranking to define alert thresholds. Studies that operate at the regional scale tend to focus on inventory or spatial pattern (Xiang et al., 2024) rather than on prioritization. The focus of the present study is on primarily the assessment of value, the transformation of cultural landscapes, and the protection of land use (Dobrovodská et al., 2019). The literature lacks a method that ingests pressure, state, and response indicators concurrently for a heterogeneous (Benton et al., 2003) regional cluster of AHS, derives expert-aggregated weights with verified internal consistency, and outputs a defensible priority tier with quantified uncertainty. This is the gap our study addresses.

We construct a Pressure-State-Response-Fuzzy Comprehensive Evaluation (PSR-FCE) framework for the 48 municipally recognized AHS in Beijing as of the 2023 monitoring update. Twenty-four indicators were retained from a 36-candidate pool through a four-criterion screening protocol applied by a panel of seven sectoral experts. AHP-derived weights, with consistency ratio CR = 0.0039 and leave-one-out stability < = 0.018, are combined with the FCE membership matrix to compute a bounded prioritization score Pi ε [0, 4] for each site. Sites are classified into four tiers (red, orange, yellow, blue) on equidistant intervals, with bootstrap 90 % confidence bands reported alongside. The framework is designed to complement, not replace, existing value-assessment and early-warning approaches: it answers the managerial questions of where to prioritize and how to intervene. We illustrate the framework on the Beijing case, identify two red-tier and eleven orange-tier sites, and propose tier-specific intervention measures. The framework's data requirements are documented as a transferability checklist so it can be applied to other metropolitan AHS clusters with comparable monitoring infrastructure.

2 Study area

Beijing is located in the northwestern part of the North China Plain and has an area of 16,411 km². It is at a relatively high altitude in the northwest and a lower altitude in the southeast, with elevations ranging from 6 to 2294 m a.s.l (Figure 1). Mountainous areas above 100 m a.s.l. account for 62% of the total area, while the remainder are plains and low mountainous areas. Beijing has a monsoon-influenced humid continental climate, with an annual average temperature of approximately 11–13 °C and an annual precipitation of 528 mm. Its favorable climate and landforms, diverse vegetation types, and soil types provide suitable environmental conditions for agricultural development.

Figure 1

Beijing has diverse topographical features and boasts more than 800 years of history as a capital city. Through centuries of agricultural development, it has cultivated a multifaceted agricultural production system including crop cultivation, fruit tree cultivation, and animal husbandry. Agricultural activities, through prolonged interaction with nature, have fostered a rich array of agricultural heritage systems, making the region a concentrated hub for such heritage systems. Beijing is a prime example of where a megacity meets agricultural heritage. It boasts 48 systems recommended by the Beijing Municipal Government (indicated in Figure 1 and summarized in Appendix 1), 4 of which have been recognized as CIAHS sites by the Administrative Measures for Important Agricultural Systems issued by the China Ministry of Agriculture and Rural Affairs (MOARA) in 2015. The 48 sites constitute the complete inventory of municipally recognized AHS in Beijing as of the 2023 monitoring update; no further inclusion threshold beyond municipal recognition was applied.

3 Method

3.1 Research framework

The press-state-response (PSR) model was used to select the indicators for the agricultural heritage conservation priority assessment, and the weights were determined using the AHP model. Agricultural heritage systems were evaluated on the basis of the PSR and AHP procedures, and fuzzy comprehensive evaluation (FCE) was employed to rank all the samples, which were classified into several groups. The flowchart of the research is presented in Figure 2. For the distinct groups of agricultural heritage identified through the ranking process, targeted protection strategies and policy integration have been proposed to promote the preservation and sustainable development of agricultural heritage.

Figure 2

3.2 The PSR model

The PSR framework was developed jointly by the Organization for Economic Cooperation and Development (OECD) and the United Nations Environment Programme (UNEP) and is a commonly used model in multiple disciplines; it performs well in environmental assessments because of its simplicity in identifying and classifying the indicators by fully accounting for causal relationships between the natural environment and human activities (Neri et al., 2016). Hence, the PSR model facilitates comprehensive assessments of agricultural heritage priority. The pressure subsystem (P) refers to the direct pressure factors of natural disasters and human economic and social activities on natural and near-natural ecosystems and reflects the pressures on ecosystems caused by humans or natural disturbances, which mainly include population pressure, economic development pressure (Su et al., 2020) and environmental pressure (Ju et al., 2025). The state subsystem (S) refers to the current state of the agricultural heritage system and the development of the system, including production status, ecological resources, brand influence, and cultural heritage (Zhang et al., 2018). The response indicator (R) characterizes the countermeasures and actions that the human agricultural heritage system can take in the face of environmental development pressures, e.g., sustainable development capacity, which is the quantifiable component of system management actions, mainly including indicators of policy support (Guo et al., 2024) and community participation (Jiang and Zhang, 2023). In other words, to protect agricultural heritage from destruction, it is necessary to consider the strengths of the government and the private sector and to develop a cooperative protection mechanism involving both the government and residents. This model was employed to identify the vulnerability of heritage systems. In addition, it can assess their resilience potential and societal response capacity. The result is a comprehensive support system for decisions regarding prioritization.

3.2.1 Indicator screening

Candidate indicators were drawn from the Monitoring Report of Beijing Municipal Bureau of Agriculture and Rural Affairs (2023), which inventories 36 indicators across 48 sites. Each candidate was evaluated against four screening criteria: (i) data availability—measurable for at least 80 % of the 48 sites without imputation; (ii) quantifiability or rubric-gradability—a defensible four-level rubric exists; (iii) inter-indicator independence—pairwise Spearman |rho| < 0.7 against already-retained indicators; and (iv) expert endorsement—unanimous agreement by the seven-expert panel that the indicator is material to AHS conservation in a metropolitan context. 24 indicators met all four criteria and were retained (Table 1).

3.3 AHP for determining weights

3.3.1 Expert panel

We convened a panel of seven sectoral experts, including three university professors specializing in agricultural heritage policy research, two senior officers from the Beijing Municipal Bureau of Agriculture and Rural Affairs, one agro-ecologist, and one site manager of a China Nationally Important Agricultural Heritage System (CIAHS), with an average working experience of 10–15 years. Pairwise comparisons were collected one expert at a time, in writing, before any group discussion, so the matrices stay independent.

3.3.2 Qualitative-indicator scoring protocol

For each qualitative indicator, each expert independently assigned a four-level rubric score using a standardized rubric sheet, photographic baseline panels for each site, and the relevant 2023 Monitoring Report annexes. For C11 (Agroecological landscape integrity) the rubric was: 4 = pristine landscape mosaic with intact field-pond-windbreak-settlement structure; 3 = minor fragmentation, 20% or less degraded; 2 = moderate fragmentation, 20–50%; 1 = severe fragmentation, over 50%. Where post-aggregation absolute disagreement was 2 or more levels, a moderated discussion round was held following the Delphi protocol; the median post-discussion score was used to compute the membership degree.

The AHP model is a multilevel weighting analysis method that was proposed by Saaty (2003) and has been widely applied across numerous fields, including systems evaluation, resource allocation, price forecasting, and project selection. It involves dividing a complex research subject into a hierarchical structure, including a target layer, criteria layer, and indicator layer, where factors at the same level hold roughly equal status, while factors across different levels maintain certain interconnections. After expert knowledge, experience, and information are applied, pairwise comparisons are conducted among indicators within each level to construct a judgement matrix, yielding a single-level ranking for that hierarchy. Consistency tests are then performed. Finally, using a layer-by-layer aggregation method, calculations proceed from the highest to the lowest level, and the overall hierarchical ranking value for the top level is derived on the basis of information from all subordinate levels.

The AHP procedure produces indicator weights through five computational steps. Step 1—Pairwise comparison: For each hierarchical level, each expert k constructs an n x n judgement matrix A(k) Step 2—Group aggregation by geometric mean (Equation 1): , with m = 7 experts. Step 3—Weight derivation by the root method (Equation 2): . Step 4—Maximum eigenvalue (Equation 3): λ_max = (1/n) Σ_i (A·W)_i/W_i. Step 5—Consistency check (Equation 4) : . where CR is the consistency ratio, CI is the consistency indicator, and RI is the average random consistency indicator. λ_max is the maximum eigenvalue, and n is the number of indicators. The CR of the judgement matrix was tested according to Equation 4. If the CR is < 0.1, the matrix meets the requirements without modification; otherwise, the expert should be asked to modify the judgement matrix again.

According to Equation 4, the CR value at the project level was 0.0039, which was less than 0.1 and passed the consistency validity test. The scoring matrix, weight values, and consistency tests for the other secondary and below-level indicators are calculated accordingly (Table 2). All the scoring matrices passed the consistency validity test.

3.3.3 Consensus aggregation and sensitivity analysis

We performed two robustness checks. (i) Leave-one-out resampling: Removing each of the seven experts in turn and re-deriving the AHP weights changed any combined weight by at most 0.018 (relative change < = 5%). (ii) Monte Carlo perturbation: We perturbed each AHP ratio entry by +/-15% independently, drew 1,000 replications, and recomputed the priority tier for all 48 sites; 45 of 48 sites retained their original tier in more than 95% of replications, two sites exhibited yellow-orange instability in fewer than 5% of replications, and one site exhibited orange-red instability in 7% of replications.

3.4 Model of the FCE

The FCE method is an application of fuzzy transformation and the maximum degree of membership principle (Zhang and Feng, 2018). First, the fuzzy set is used to represent the various factors related to the evaluation object. Next, it is used to calculate the evaluation matrix and weights of the evaluation factors. Finally, fuzzy linear transformation is used to obtain the evaluation results of fuzzy sets. This method can be applied to solve the problems of multifactor complexity and uncertainty (Hirsenberger et al., 2019). Compared with other multi-criteria decision-making methods (e.g., TOPSIS, entropy weighting, gray relational analysis), FCE offers three distinct advantages for this study. First, it handles fuzzy boundaries naturally. AHS conservation is a dynamic process with vague goals and states, and FCE is specifically designed for such problems, whereas methods like TOPSIS and entropy weighting assume crisp data and linear preferences. Second, FCE directly uses qualitative expert judgments. Our data come from expert panel membership degrees, and FCE incorporates these membership vectors directly. In contrast, entropy weighting requires crisp numerical values and would lose the distribution information inherent in membership degrees. Third, FCE produces a bounded Pi score (0–4) that maps directly to a four-tier action ladder (red to blue), which is readily interpretable for municipal conservation agencies. Other methods do not naturally yield such a policy-friendly output. The FCE method was performed as follows.

• (1) Establishing the index set of the evaluation object

  where the elements are the evaluation indices for a certain object. U refers to the indicator selected for the PSR system. The weight (W_i) of each indicator is calculated with the AHP method.

• (2) Establishing the evaluation categories sets

  where V is the evaluation set corresponding to the evaluation indices in U. The set of rubrics for the indicators was assigned as V = {high, moderately high, medium, and low}. Each rating level corresponds to a score of {4, 3, 2, 1}, and the median of the seven expert scores is used to determine the degree of belonging to each rubric level—a robust choice for ordinal Likert-type ratings under small panels (Table 3).

3.4.1 Convention

Throughout the FCE, a higher rubric level (e.g., ‘high') denotes a more favorable state of the heritage system on that indicator, irrespective of whether the underlying variable is a pressure, state, or response measure. For pressure indicators (P1–P3) the rubric is therefore inverted with respect to the raw variable: a ‘high' rating on a pressure indicator means low actual pressure on the AHS.

(3) Creation of a fuzzy evaluation matrix

The degree of affiliation (F) refers to the i-th factor to the j-th rating and is the ratio of the number of experts who rate the j-th factor to the total number of experts who participate in the rating process. The membership function is defined as F_ij = n_ij/N, where N = 7 is the total number of experts, and n_ij is the number of experts who rate indicator i as rubric level j. Thus, a fuzzy relationship matrix is created as follows:

(4)Fuzzy Comprehensive Evaluation

The fuzzy evaluation result was calculated by multiplying the weight W_i by the fuzzy matrix F_ij (Equation 4). The process was carried out sequentially to obtain fuzzy evaluation results at the indicator layer, criteria layer, and target layer. The maximum degree of affiliation is used to obtain the maximum value of the U row vector, which is the final evaluation result.

(5) FCE result of AHS

The FCE result for each AHS was calculated by the following equation. P_i represents the final score of the i-th agricultural heritage. Z_j represents the evaluation grade score of each rating level (Table 3).

Pi represents the prioritization score of each agricultural heritage site, which ranges from 0 to 4. AHSs with different P_i values were classified into 4 alarm status groups (Table 4). A lower P_i value corresponds to more urgent conservation status and higher intervention priority.

3.5 Threshold rationale and uncertainty quantification

The four-tier classification uses equidistant intervals on the Pi axis: red [0, 1), orange [1, 2), yellow [2, 3), blue [3, 4]. The choice is justified by three arguments: (i) the rubric scores Z = (3.5, 2.5, 1.5, 0.5) are equally spaced by construction, so equidistant intervals on Pi preserve the natural ordering; (ii) equidistant thresholds avoid layering an additional analyst-chosen weighting on top of the AHP-derived weights, preserving the interpretability of P_i as a calibrated PSR-weighted state score; (iii) a quantile-based robustness check using the 25th, 50th, and 75th percentiles of the empirical P_i distribution reassigned only 4 of 48 sites, none of which crossed the red-orange boundary, supporting the equidistant choice for the present sample.

To further address the concern that FCE results depend heavily on expert scoring, we performed a bootstrap uncertainty quantification. For each of the 48 sites, we resampled the seven-expert membership vector with replacement, drew 1,000 bootstrap samples, and computed a 90% confidence band on Pi. The mean half-width of the band was 0.23 score points, indicating that the tier assignments are stable.

3.6 Data acquisition

In this study, 48 agricultural heritage systems in Beijing were selected as samples, and the list of the systems was derived from the Census Report of Beijing Municipal Bureau of Agriculture and Rural Affairs (2022). The indicators used in the PSR research came from the Monitoring Report of Beijing Municipal Bureau of Agriculture and Rural Affairs (2023), which provides 36 indicators for 48 heritage sites, including basic information on AHS, economic development, ecological conditions, social maintenance, scientific research, popularization of science, and management services of the heritage sites. After careful consideration and consultation with relevant experts, we selected 24 indicators to incorporate into the PSR system. We consulted 7 experts specializing in agricultural heritage who were familiar with Beijing agricultural heritage conservation to conduct pairwise comparisons for the AHP process. During the FCE phase, they assessed the degrees of affiliation of indicators for each agricultural heritage system.

4 Results

4.1 Weights of the PSR indicator system

A series of expert assessments were collated for all the indicators of each agricultural heritage system. These data were then analyzed to determine the proportion of each indicator receiving a certain grade in a four-grade evaluation, as well as their respective scores (Table 5). 80% of the evaluators assigned a moderate rating to Indicator C1, indicating relatively low pressure from population density on agricultural heritage; 84% assigned a low rating to C12, suggesting a continued focus on the status of traditional varieties. Across the 24 indicators, four were rated as high (C7, C11, C13, C20), ten as moderately high, seven as medium, and three as low (C12, C22, C24). C13 (Agrobiodiversity) recorded the highest score (3.88); C12 (Endangerment of traditional varieties) recorded the lowest among state indicators (1.30). C22 (Professional management) and C24 (Infrastructure) recorded the lowest response-subsystem scores. Subsystem-level interpretation.

4.2 PSR subsystem analysis

We determined the weights of the indicators at each level through the AHP method (Table 6). The weights for P, S and R were 0.20, 0.46 and 0.34, respectively. The pressure subsystem indicators demonstrated moderate to high levels, although the impact of urbanization in the surrounding regions necessitates particular consideration. The status subsystem indicators predominantly indicated moderate to high threat levels, with a particular emphasis on agrobiodiversity and the management of agricultural product branding. Within the response subsystem, overall ratings were predominantly low to moderate, indicating deficiencies in institutional development and management. This suggests the need to prioritize infrastructure improvements and increase the professional management level.

4.3 Priority analysis of AHS conservation

After the fuzzy comprehensive evaluation, the P_i scores of the 48 agricultural heritage systems were calculated, and the priority rankings are summarized in Table 7. There are 2, 11, 26 and 9 items of red, orange, yellow and blue priority order, respectively, with approximately 27.1% of the total in the red and orange classes; additionally, the priority status in Fengtai District and Changping District is below that in other districts (Figure 3). The blue priority order indicates relatively good conditions, the yellow priority order indicates potential for development, and the orange priority order indicates that conservation should be explored. It is imperative that the areas with red priority orders receive timely intervention and real-time monitoring. The analysis identifies two red-tier and eleven orange-tier sites that warrant immediate attention. The Five-color Chives Cultivation System (Daxing) and Crabapple Cultivation System (Changping) are the two red-tier sites; both have core areas below 10 hectares, severe practitioner shortage, and weak community-side response. Targeted interventions—informed by interviews with local administrators and practitioners—are summarized in Table 8.

Figure 3

5 Discussion

The FCE results reveal the stratified conservation priorities of Beijing agricultural heritage systems. Significant differences emerge in terms of urgency and intervention strategies across the four priority tiers. Thus, appropriate governance approaches that are tailored to their differing PSR statuses are needed. Among the agricultural heritage conservation strategies of varying priorities, we analyzed the principal characteristics of the three dimensions of PSR on the basis of preliminary field investigations and expert consultations and proposed key intervention measures in Table 8 drawing on the recommendations gathered during our research process through interviews with local administrators and practitioners, as well as expert consultations.

5.1 Indicator-system limitations and transferability

To facilitate operation and application, all possible indices cannot be exhaustively considered. Two further dimensions (Salpina, 2020)—climate resilience (Zhao et al., 2025) and intergenerational transmission (He et al., 2020)—were excluded for the data-availability reasons. Site-level data for both were unavailable across the 48 sites in the 2023 monitoring update. Their omission is conservative—incorporation would likely raise, not lower, the priority of two types of sites: Mountain rain-shadow sites in Fangshan and Mentougou, where CMIP6-downscaled projections indicate a 12–18% increase in precipitation variability by 2050, and Aging-practitioner sites, where the majority of practitioners are over 60 years old (e.g., Five-color Chives, Watermelon, Goldfish systems documented in our field investigation). These indicators should be incorporated into priority assessment systems in future research.

The transferability of the framework to other regions is constrained primarily by data availability rather than by the framework structure itself: indicators C1–C3 are obtainable from standard statistical yearbooks; C6–C7, C10–C15, C17, C20–C22, C24 require either a one-off field survey or an expert Delphi process. The remaining eight indicators can either be obtained by contacting relevant government departments or be replaced by qualitative expert scoring, so the data requirement is a small expert panel plus county-level statistical yearbooks, not a city-level monitoring system.

5.2 Methodological limitations and uncertainty

Three methodological caveats accompany the analysis. (i) The expert panel of seven, while supported by leave-one-out checks, is at the lower bound of common AHP practice; replication with a larger panel would tighten weights. (ii) The bootstrap confidence intervals assume that the seven expert ratings are exchangeable for a given indicator and site; if expert specializations introduce systematic between-expert variance, the bands underestimate uncertainty. (iii) The four-tier classification uses equidistant thresholds on Pi; while supported by the quantile robustness check (Methods), the choice of threshold scheme is ultimately a policy parameter that should be calibrated to the conservation budget.

5.3 Policy implications and management inspirations

The importance of government investment in addressing these issues is reflected in the AHP weights. This is particularly evident for C19 and C20, which carry high weights. The seven-expert panel converged on the judgement that, in the Beijing metropolitan context, the binding constraint on AHS conservation outcomes is the level and continuity of public investment. Three contextual factors drove this judgement: (i) Beijing's AHS sites are in or near a megacity where land-opportunity costs are extreme, so without targeted public finance to offset opportunity cost, abandonment is rapid; (ii) the institutional architecture is centralized, so policy-management strength has high marginal effect compared with less-centralized regions; (iii) the 2021 Beijing 14th Five-Year Plan for AHS Protection and Development established a fiscal envelope that varies considerably across districts, making C19 a tight binding constraint at the cross-site level. The C19 and C20 combined weight should therefore be read as Beijing-specific rather than universal; the framework can accept different weights derived for other regions without modification.

This pattern is clearly reflected in the empirical results. Fengtai-district sites are spatially constrained: core areas of 2–47 ha is encroached by the urban footprint, amplifying pressure indicators (C1, C7). Changping-district sites face large built-up area expansion (2010–2023 land-use record), eroding state-subsystem indicators (C10, C11). Both districts receive below-municipal-mean fiscal allocation per AHS site, compounding the response-subsystem deficit (C19, C22). The policy implication from district-level patterns in Figure 3 is concrete: re-balancing the 2026–2030 fiscal envelope toward Fengtai and Changping would yield the largest expected upgrade in tier composition.

5.4 Comparison with existing frameworks and future research directions

Established frameworks—the FAO GIAHS criteria, cultural-ecosystem-service valuations, and vulnerability-only risk frameworks—answer “why conserve?” Our PSR-FCE framework is a complement, not a replacement, that answers “where to prioritize?” and “how to intervene?” The complementarity is illustrated by the case where a high cultural-value site simultaneously faces extreme population pressure (P1 low) and weak policy response (R1 low): a value-only assessment flags importance but our framework assigns a high-priority tier and recommends concrete intervention.

One limitation of this study is that the complete 48 × 4 membership degree matrix for all 48 agricultural heritage systems is not systematically reported. Although Supplementary material provides a detailed step-by-step calculation of the membership vectors using the Chaoyang Langjiayuan jujube system as an example, the membership vectors for the remaining sites are not presented in the main text. This simplification is primarily due to space constraints, and because the main focus of this study is the final priority score (Pi) and the four-tier warning classification, which are sufficient to support the core conclusions. Nevertheless, membership degrees can reveal patterns, clusters, and borderline cases that the final scores alone cannot convey. For example, a more detailed analysis of the membership degrees for the Daxing Five-color Chives system (red alert) and the Fengtai Changxindian white jujube system (orange alert) could provide additional nuance. Therefore, future research should systematically report the full membership degree matrix and explore its potential for identifying borderline cases and clustering patterns, in order to further improve the early-warning system for agricultural heritage protection.

Furthermore, based on the findings of this study, the following future research directions are warranted:(i) extension to climate resilience and intergenerational transmission indicators once the 2026+ Beijing monitoring update is released; (ii) coupling of the PSR-FCE workflow with a real-time monitoring layer so that Pi can be updated annually, transforming the framework from a one-shot prioritization to a true early-warning system; (iii) cross-regional replication in other megacity AHS clusters (e.g., the Yangtze River Delta and Greater Bay Area), which will allow assessment of whether the high C19/C20 weights observed here are general or Beijing-specific.

References

• 1

  AbebawS. E. A. (2025). Global review of the impacts of climate change and variability on agricultural productivity and farmers' adaptation strategies. Food Sci. Nutr.13:e70260. doi: 10.1002/fsn3.70260

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  AgbehadjiI. E.MabhaudhiT.BotaiJ.MasindeM. A. (2023). Systematic review of existing early warning systems' challenges and opportunities in cloud computing early warning systems. Climate. 11:188. doi: 10.3390/cli11090188

  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  AkpinarN.TalayI.GunS. (2005). Priority setting in agricultural land-use types for sustainable development. Renew. Agric. Food Syst. 20, 136–147. doi: 10.1079/RAF200483

  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  BahnasyN. A. (2024). A conceptual framework for characterization of agricultural heritage in desert-prone areas. J. Humanit. Appl. Soc. Sci.7, 236–255. doi: 10.1108/JHASS-11-2023-0163

  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  BaiY.LiX.FengY.LiuM.ChenC. (2024). Preserving traditional systems: identification of agricultural heritage areas based on agro-biodiversity. Plants People Planet. 6, 670–682. doi: 10.1002/ppp3.10479

  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  BalickM. J. (2017). “Traditional knowledge: lessons from the past, lessons for the future,” in Routledge Handbook of Biodiversity and the Law, eds. McManisC. R.OngB. (London: Routledge), 239–253. doi: 10.4324/9781315530857-16

  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  BarrenaJ.NahuelhualL.BáezA.SchiappacasseI.CerdaC. (2014). Valuing cultural ecosystem services: agricultural heritage in Chiloé Island, southern Chile. Ecosyst. Serv. 7, 66–75. doi: 10.1016/j.ecoser.2013.12.005

  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  Beijing Municipal Bureau of Agriculture and Rural Affairs. (2022). Census Report of Beijing Agricultural Heritage.

  • Google Scholar
  • View reference in article

• 9

  Beijing Municipal Bureau of Agriculture and Rural Affairs. (2023). Monitoring Report of Beijing Agricultural Heritage.

  • Google Scholar

• 10

  BennettA. F.Mac NallyR. (2004). Identifying priority areas for conservation action in agricultural landscapes. Pac. Conserv. Biol.10, 106–123. doi: 10.1071/PC040106

  • CrossRef
  • Google Scholar

• 11

  BentonT. G.VickeryJ. A.WilsonJ. D. (2003). Farmland biodiversity: is habitat heterogeneity the key?Trends Ecol. Evol.18, 182–188. doi: 10.1016/S0169-5347(03)00011-9

  • CrossRef
  • Google Scholar

• 12

  CuiF.WangZ. (2023). Construction and method of early-warning evaluation system of agricultural heritage protection. J. Nat. Resour. 38, 1119–1134. doi: 10.31497/zrzyxb.20230501

  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  DobrovodskáMKankaR.DavidS.KollárJ.SpulerovaJ.StefunkovaD.et al. (2019). Assessment of the biocultural value of traditional agricultural landscape on a plot-by-plot level: case studies from Slovakia. Biodivers. Conserv. 28, 2615–2645. doi: 10.1007/s10531-019-01784-x

  • CrossRef
  • Google Scholar
  • View reference in article

• 14

  DuH.XuG.LinY. (2024). Ecological compensation of agricultural heritage conservation: case of the mountainous Juglans hopeiensis planting system in Beijing. Front. Sustain. Food Syst. 8:1425738. doi: 10.3389/fsufs.2024.1425738

  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  DucusinR. J. C.EspaldonM. V. O.RebancosC. M.De GuzmanL. E. P. (2019). Vulnerability assessment of climate change impacts on a globally important agricultural heritage system (GIAHS) in the Philippines: the case of Batad rice terraces, Banaue, Ifugao, Philippines. Clim Change. 153, 395–421. doi: 10.1007/s10584-019-02397-7

  • CrossRef
  • Google Scholar
  • View reference in article

• 16

  FuL.ZhangQ.TangY.PanJ.LiQ. (2023). Assessment of urbanization impact on cultural heritage based on a risk-based cumulative impact assessment method. Herit. Sci. 11:177. doi: 10.1186/s40494-023-01024-0

  • CrossRef
  • Google Scholar
  • View reference in article

• 17

  GuoX.MinQ.JiaoW. (2024). Spatial distribution characteristics and differentiated management strategies of China nationally important agricultural heritage systems. J. Geogr. Sci. 34, 483–498. doi: 10.1007/s11442-024-2214-2

  • CrossRef
  • Google Scholar
  • View reference in article

• 18

  HeS.MinQ.LiH.LiuM.JiaoW.BaiY. (2020). Value typology and evaluation of important agricultural heritage systems. Chin. J. Eco-Agric. 28, 1314–1329. doi: 10.13930/j.cnki.cjea.190881

  • CrossRef
  • Google Scholar
  • View reference in article

• 19

  HirsenbergerH.RanogajecJ.VuceticS.LalicB.GracaninD. (2019). Collaborative projects in cultural heritage conservation: management challenges and risks. J. Cult. Herit.37, 215–224. doi: 10.1016/j.culher.2018.10.006

  • CrossRef
  • Google Scholar
  • View reference in article

• 20

  HulmeP. E.LieuranceD.RichardsonD. M.RobinsonT. B. (2025). Multiple targets of the global biodiversity framework must be addressed to manage invasive alien species in protected areas. NeoBiota. 99, 149–170. doi: 10.3897/neobiota.99.152680

  • CrossRef
  • Google Scholar
  • View reference in article

• 21

  IzakovičováZ.ŠpulerováJ.RaniakA. (2022). The development of the Slovak agricultural landscape in a changing world. Front. Sustain. Food Syst. 6:862451. doi: 10.3389/fsufs.2022.862451

  • CrossRef
  • Google Scholar
  • View reference in article

• 22

  JiangY.ZhangR. (2023). Characteristics of urban agricultural heritage sites: policies and management methods for their conservation in China, Germany, and Italy. Habitat Int.131:102710. doi: 10.1016/j.habitatint.2022.102710

  • CrossRef
  • Google Scholar
  • View reference in article

• 23

  JuF.YangR.YangC. (2025). Analysis of spatiotemporal dynamics and driving factors of China's nationally important agricultural heritage systems. Agriculture. 15:221. doi: 10.3390/agriculture15020221

  • CrossRef
  • Google Scholar
  • View reference in article

• 24

  KoohafkanP.AltieriM. A. (2011). Globally Important Agricultural Heritage Systems: A Legacy for the Future. Rome: Food and Agriculture Organization of the United Nations.

  • Google Scholar
  • View reference in article

• 25

  KoohafkanP.CruzM. J. D. (2011). Conservation and adaptive management of globally important agricultural heritage systems (GIAHS). J. Resour. Ecol. 2, 22–28. doi: 10.3969/j.issn.1674-764x.2011.01.004

  • CrossRef
  • Google Scholar
  • View reference in article

• 26

  LabadiS.GilibertoF.RosettiI.ShetabiL.YildirimE. (2021). Heritage and the Sustainable Development Goals: Policy Guidance for Heritage and Development Actors. Paris: ICOMOS.

  • Google Scholar
  • View reference in article

• 27

  LeeJ.GereffiG.BeauvaisJ. (2012). Global value chains and agrifood standards: challenges and possibilities for smallholders in developing countries. Proc. Natl. Acad. Sci. U.S.A. 109, 12326–12331. doi: 10.1073/pnas.0913714108

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 28

  LiM.LiangL.HuangZ.SongH.WangS.WangQ.et al. (2025). Spatiotemporal dynamics and driving factors of net primary productivity in Asian terrestrial ecosystems. Ecol. Model. 510:111322. doi: 10.1016/j.ecolmodel.2025.111322

  • CrossRef
  • Google Scholar
  • View reference in article

• 29

  Manrique AnticonaC. E.Yagüe BlancoJ. L.Pascual CastañoI. C. (2023). Characterization of potential Spanish territories for creating a national network associated with the globally important agricultural heritage systems. Land Use Pol. 131:106667. doi: 10.1016/j.landusepol.2023.106667

  • CrossRef
  • Google Scholar
  • View reference in article

• 30

  NahuelhualL.CarmonaA.LaterraP.BarrenaJ.AguayoM. A. (2014). Mapping approach to assess intangible cultural ecosystem services: the case of agricultural heritage in southern Chile. Ecol. Indic.40, 90–101. doi: 10.1016/j.ecolind.2014.01.005

  • CrossRef
  • Google Scholar
  • View reference in article

• 31

  NathT. K.InoueM.EeW. Y.TakahashiS. (2024). Globally important agricultural heritage systems in Japan: investigating selected agricultural practices and values for farmers. Int. J. Agric Sustain. 22:2355429. doi: 10.1080/14735903.2024.2355429

  • CrossRef
  • Google Scholar
  • View reference in article

• 32

  NeriA. C.DupinP.SánchezL. E. (2016). A pressure–state–response approach to cumulative impact assessment. J. Clean. Prod.126, 288–298. doi: 10.1016/j.jclepro.2016.02.134

  • CrossRef
  • Google Scholar

• 33

  PetitS.AlignierA.AllartR.AvironS.BoussardH.FranckP.et al. (2023). Building capacities for the design of agroecological landscapes: the added value of landscape monitoring networks. Agric. Ecosyst. Environ. 342:108263. doi: 10.1016/j.agee.2022.108263

  • CrossRef
  • Google Scholar

• 34

  PirasF.PanY.SantoroA.FioreB.MinQ.GuoX. (2024). Agro-silvo-pastoral heritage conservation and valorization: a comparative analysis of the Chinese nationally important agricultural heritage systems and the Italian register of historical rural landscapes. Land13:988. doi: 10.3390/land13070988

  • CrossRef
  • Google Scholar

• 35

  PoplawskaM. (2025). Intangible heritage system: challenges of culture preservation in a cross-cultural perspective. Ethnomusicol. Forum. 34, 91–119. doi: 10.1080/17411912.2025.2484707

  • CrossRef
  • Google Scholar

• 36

  RavankhahM.SchmidtM.WillT. (2021). An indicator-based risk assessment framework for world heritage sites in seismic zones: the case of “Bam and its cultural landscape” in Iran. Int. J. Disaster Risk Reduct. 63:102405. doi: 10.1016/j.ijdrr.2021.102405

  • CrossRef
  • Google Scholar

• 37

  SaatyT. L. (2003). Decision-making with the AHP: why is the principal eigenvector necessary?Eur. J. Oper. Res.145, 85–91. doi: 10.1016/S0377-2217(02)00227-8

  • CrossRef
  • Google Scholar

• 38

  SalpinaD. (2020). Cultural Dimension of Agricultural Landscape: The Study on Protection, Management, and Governance of the Multifunctional Heritage. Lucca: IMT School for Advanced Studies.

  • Google Scholar

• 39

  SrivastavaP.SinghR.TripathiS.RaghubanshiA. S. (2016). An urgent need for sustainable thinking in agriculture: an Indian scenario. Ecol. Indic.67, 611–622. doi: 10.1016/j.ecolind.2016.03.015

  • CrossRef
  • Google Scholar

• 40

  SuM. M.SunY.WallG.MinQ. (2020). Agricultural heritage conservation, tourism and community livelihood in the process of urbanization: Xuanhua grape garden, Hebei Province, China. Asia Pac. J. Tour. Res. 25, 205–222. doi: 10.1080/10941665.2019.1688366

  • CrossRef
  • Google Scholar

• 41

  WeiJ.JiY.YangD.LiangF. A. (2025). Multi-model coupling approach to biodiversity conservation strategies for nationally important agricultural heritage systems in the Beijing–Tianjin–Hebei region. Sustainability. 17:7959. doi: 10.3390/su17177959

  • CrossRef
  • Google Scholar

• 42

  XiangH.ZhouC.CuidongSong, H. (2024). High-quality agricultural development in the central China: Empirical analysis based on the Dongting Lake area. Geomatica. 76:100010. doi: 10.1016/j.geomat.2024.100010

  • CrossRef
  • Google Scholar

• 43

  YangJ.YouY.YeX.LinJ. (2023). Cultural heritage sites risk assessment based on RS and GIS: the fortified manors of Yongtai as an example. Int. J. Disaster Risk Reduct. 88:103593. doi: 10.1016/j.ijdrr.2023.103593

  • CrossRef
  • Google Scholar

• 44

  ZhangP.FengG. (2018). Application of fuzzy comprehensive evaluation to evaluate the effect of water flooding development. J. Pet. Explor. Prod. Technol. 8, 1455–1463. doi: 10.1007/s13202-018-0430-y

  • CrossRef
  • Google Scholar

• 45

  ZhangY.LiX.MinQ. (2018). How to balance the relationship between conservation of important agricultural heritage systems (IAHS) and socio-economic development? A theoretical framework of sustainable industrial integration development. J. Clean. Prod. 204, 553–563. doi: 10.1016/j.jclepro.2018.09.035

  • CrossRef
  • Google Scholar

• 46

  ZhaoL.MaR.FanC.LiJ. (2025). Enhancing the spatial resilience of agricultural heritage through constructing a balanced production-living-ecological space protection network. NPJ Herit. Sci. 13:368. doi: 10.1038/s40494-025-01920-7

  • CrossRef
  • Google Scholar

• 47

  ZhouL.ZhaoY.YangX.HeJ.WangH. (2023). Identification of important terraced visual landscapes based on a sensitivity-subjectivity preference matrix for agricultural cultural heritage in southwestern China. Ecol. Indic.154:110573. doi: 10.1016/j.ecolind.2023.110573

  • CrossRef
  • Google Scholar