Introduction

Efficient farmland utilization is critical for China, an agricultural country with a vast population. The Central Document No.1 issued in 2023 and 2024 highlights the urgent need for environmentally sustainable agricultural development, emphasizing the enhancement of high-standard basic farmland and water infrastructure. This strategic direction aims to accelerate the modernization of agriculture and rural areas, providing clear guidelines for managing the farmland spatial transition. Currently, China faces several practical challenges in farmland utilization, including unstable foundations for grain production (Dong et al. 2023); inadequate modern agricultural infrastructure (Wang and Liu, 2019; Zhang, 2016); noticeable trends of non-grain and non-agricultural farmland use (Li et al. 2020; Cao et al. 2022); insufficient disaster prevention and mitigation capabilities (Song et al. 2020); and limited progress in green development (Li et al. 2024; Ke et al. 2021). These challenges arise from several factors, including declining comparative returns in grain production, weakened incentives for agricultural subsidies, escalating resource and environmental constraints, and insufficient scientific and technological research and development capabilities (Deng et al. 2024; Gao and Wang, 2021; Han, 2022). Consequently, promoting farmland spatial transition has become a crucial priority for addressing the current challenges in farmland utilization.

China is actively implementing policies and engineering initiatives related to farmland circulation, the establishment of high-standard basic farmland, and the enhancement of agricultural infrastructure. Notable progress has been achieved in expanding and professionalizing farmland management. According to data released by the Ministry of Agriculture and Rural Affairs of China, by the end of 2021, the total area of farmland circulation exceeded 37 million hectares. Furthermore, per capita farmland area increased from 3866.67 to 5733.33 square meters between 2010 and 2021. Notably, a total of 66.66 million hectares of high-standard basic farmland have been established. Some regions have initially met the objectives of field consolidation, soil improvement, concentrated contiguous land, and enhanced resistance to drought and floods. In terms of spatial morphology, China’s farmland is undergoing transformative changes in management area, shape characteristics, spatial distribution, and cropping patterns. However, significant challenges remain, including farmland fragmentation and dispersion (Zheng et al. 2023; Li et al. 2022), a dispersed industrial spatial layout (Luo et al. 2021), inefficient and wasteful farmland utilization, and pronounced regional disparities that have yet to be fundamentally resolved.

Farmland spatial transition is widely acknowledged as a critical component of farmland use transition and has garnered significant attention from scholars worldwide. It is generally defined as a transformative shift in the spatial utilization patterns of farmland. According to Song and Li (2019), farmland spatial transition is primarily manifested through changes in landscape and management structures. Other scholars, however, interpret farmland spatial transition as changes in quantitative morphological characteristics and spatial structure (Lu et al. 2019; Wang et al. 2024). The driving mechanisms behind farmland spatial transition have been extensively examined, often employing methods such as spatio-temporal geographically weighted regression and spatial econometric models (Ke et al. 2021). Studies indicate that micro-level factors including natural conditions, input structure, output benefits, crop structure, and transportation location (Tian et al. 2020), as well as macro-level factors such as population density, economic development, national policies, and economic stages (Xu and Zhang 2021; Chen et al. 2023) are key drivers of farmland spatial transition. These studies have established a solid foundation and provided valuable insights. However, several limitations remain. First, a unified framework for a comprehensive understanding of farmland spatial transition has yet to be developed. While the spatial attributes such as farmland shape and distribution are recognized, they have not been integrated into existing evaluation systems. Second, most research explores factors influencing farmland spatial transition from a single dimension, despite the process involving complex interactions among multiple drivers. The mechanisms governing these interactions and dominant influences remain unclear. Addressing these limitations, this study aims to develop a comprehensive measurement system for farmland spatial transition, characterize its spatio-temporal patterns in the Yangtze River Economic Belt, and investigate its driving forces. This work is critical for addressing the current challenges in farmland utilization within the region.

The Yangtze River Economic Belt is a key component of China’s central economic zone and a globally significant agricultural production area, serving as a vital center for advancing modern agriculture. Given its strategic importance and the existing disparities in farmland utilization, this study focuses on 897 counties within the region to investigate the spatio-temporal patterns and driving mechanisms of farmland spatial transition. The objective is to develop a theoretical framework for farmland spatial transition, enhance understanding of its spatio-temporal dynamics, and elucidate the underlying drivers.

The theoretical explanation of farmland spatial transition

The conceptual implications of farmland spatial transition

Farmland spatial transition refers to the systematic transformation of farmland spatial morphology across multiple dimensions, including operational scale, shape characteristics, spatial distribution, and cropping patterns (Ke et al. 2024; Niu et al. 2025). It targets four key objectives: area scaling, shape regularization, spatial agglomeration, and crop specialization. This transition supports agricultural industrialization, promotes sustainable practices, enhances farmland resilience, and improves farmers’ incomes, thereby embodying the principles of green development and agricultural modernization in the contemporary era (Zhang et al. 2024). Based on morphological components, this study classifies farmland spatial transition into four interrelated types: scale transition, regularization transition, agglomeration transition, and specialization transition. These subsystems form an integrated framework for comprehensively understanding farmland spatial transition (Fig. 1).

Fig. 1
figure 1

Analytical framework of the ‘four-sphere integrated’ measurement system for farmland spatial transition.

Full size image

The core subsystem of farmland spatial transition centers on the complex interplay between the geographic environment and human society, characterized by a continuous exchange of matter and energy. The human–land interaction drives changes in farmland spatial distribution, area, shape, and cropping patterns, collectively constituting the ‘four-sphere integrated’ subsystem of farmland spatial transition. (a) Farmland circulation serves as effective mechanisms for promoting agricultural scale development, providing production and management services for both small-scale and large-scale family operations (Yuan et al. 2024). Scale transition, a key measurement focus, quantifies changes in farmland area with the aim of shifting from small-scale to moderate-scale operations. (b) The construction of high-standard basic farmland is an effective strategy to mitigate land fragmentation, by consolidating and regularizing dispersed plots (Zhang et al. 2023). In this context, regularization transition seeks to simplify and standardize farmland shape, facilitating scale operations and mechanized production. (c) Land consolidation has emerged as a powerful tool for advancing intensive agricultural management (Liu et al. 2019). Agglomeration, emphasizing the transformation of farmland spatial distribution from scattered and fragmented to concentrated and contiguous, plays a pivotal role in this process. (d) The promotion of agricultural industrialization facilitates the transition towards specialization, market orientation, and modernization. Specialization transition focuses on the shift from diversified to specialized crop cultivation, characterized by the concentrated planting of a single crop with regional competitive advantage.

Mechanisms of farmland spatial transition

The scholarly community has extensively examined the driving mechanisms behind farmland use transition. There is broad consensus that micro-level factors, such as natural conditions, input composition, output efficiency, cropping structure, and transportation location, are key drivers of farmland use transition. In addition, macro-level factors, including population density, economic development, national policies, and stages of economic growth, constitute key drivers of farmland use transition (Tian et al. 2020; Xu and Zhang 2021; Chen et al. 2023; Shi et al. 2019). Farmland spatial transition refers to the shift from traditional to modern cultivation methods, influenced by a range of factors including natural environment, institutional environment, economic conditions, and population dynamics. To establish a unified analytical framework, this study adopts the theory of induced factor substitution to elucidate the mechanisms governing farmland spatial transition and evolution (Fig. 2).

Fig. 2
figure 2

Mechanisms of farmland spatial transition and evolution based on the theory of induced factor substitution.

Full size image

Driving factors influencing farmland spatial transition can be broadly categorized into four types: natural, institutional, economic, and population factors. Natural factors encompass the environmental conditions essential for farmland production and management. Regional variations in these conditions lead to differing agricultural practices (Strenge et al. 2023) and crop planting structures (Zhang et al. 2021), which in turn affect the input and substitution dynamics among farmland, capital, technology, and labor.

Institutional factors refer to external governance frameworks that impact farmland production activities. A supportive institutional environment can optimize resource allocation, enhance the scale and specialization of farmland production, and improve agricultural productivity (Men et al. 2024), thereby generating multiple feedback effects on farmland production and management resources.

Economic factors include both internal and external economic variables influencing farmland production and management. Favorable economic conditions promote intensification, scaling, and specialization of farmland operations, strengthen agricultural infrastructure, stimulate diverse market demands, and improve agricultural production efficiency (Tan et al. 2022; Spence 2023; Gong et al. 2024). Collectively, these effects facilitate the optimization of farmland resource allocation and enable factor substitution.

Population factors concern labor dynamics and mobility from a human-land perspective. Areas with high population density experience greater demand for farmland resources (Zhao et al. 2024). Levels of urbanization and mobility further influence the utilization and distribution of farmland resources. Additionally, improvements in population quality can enhance agricultural productivity, reduce land demand, and promote the optimized allocation of farmland resources.

Throughout the farmland spatial transition process, these driving factors induce substitution effects among labor, capital, technology, and farmland inputs. Such substitution effects drive changes in operational areas, farmland morphology, spatial distribution, and cropping patterns through technological innovation, production mode upgrades, infrastructure improvements, and industrial restructuring, thereby facilitating farmland spatial transition.

Variable selection, research methods and data sources

Variable selection

(1) Farmland spatial transition indicator system: To comprehensively evaluate the extent of farmland spatial transition in the Yangtze River Economic Belt, this study introduces the concept of ‘farmland spatial transition index’ (FTSI) to quantify the transition process. Building on previous analyses, the index encompasses four dimensions: scale transition, regularization transition, agglomeration transition, and specialization transition (Table 1).

Table 1 FSTI system.
Full size table

(a) Scale transition is assessed using indicators such as LF, MPS, and PD, which capture the development level of farmland scale (Ke et al. 2024; Lu et al. 2022). These metrics reflect the scale of farmland and the intensity of land use, providing an effective measure to evaluate the shift from small-scale to moderate-scale operations. (b) Regularization transition is primarily evaluated through ED, SHAPE_MN, and FRAC_MN, which characterize the morphological attributes of farmland (Zhang et al. 2024; Xiao et al. 2024). These indicators describe changes in farmland regularity and geometric shape, reflecting the process of spatial regularization that facilitates agricultural mechanization and optimizes land resource allocation. (c) Agglomeration transition employs indices such as MNN, AI, CONTAG, and COHESION to quantify the spatial clustering and connectivity of farmland. These measures reveal trends in intensive farmland use and spatial concentration of agricultural production (Jing et al. 2023). (d) Specialization transition is captured by indicators including SDI and SHEI, which assess the degree of agricultural specialization and crop diversity (Zhu et al. 2020). These metrics help identify shifts in agricultural production patterns, particularly the move from diversified to specialized crop cultivation.

In Table 1, the symbols used in the formulas are defined as follows: \(A\) is the total area of the farmland landscape; \(L\) is the number of agricultural laborers; \(N\) is the number of farmland plots; \(E\) refers to the total length of landscape edges; \({a}_{ij}\) represents the area of patch \(ij\); \({p}_{ij}\) is the perimeter of patch \(ij\); \(m\) indicates the number of patch types; \(n\) is the number of main crop types under consideration; \({h}_{i}\) represents the sum of the nearest distances from a plot to the nearest patch of the same type; \({N}^{\text{'}}\) indicates the total number of plots with a nearest neighbor calculations; \({P}_{i}\) refers to the proportion of the area of patch type \(i\) in the landscape; \({q}_{k}\) is the proportion of the area occupied by crop type k under investigation; and \(z\) refers to the number of crop types under investigation.

(2) Driving factors: The analysis adopts a human–land interaction perspective, with results summarized in Table 2.

Table 2 Driving factors of the farmland spatial transition.
Full size table

Natural factors refer to environmental elements influencing farmland production and management. These include farmland endowment, commonly measured by farmland area (Lu et al. 2022), topographical and geomorphological features, represented by elevation and slope (Ji et al. 2021; Fu et al. 2020), as well as climatic and water resource conditions, reflected in the cropping system (Ma et al. 2022).

Institutional factors refer to external governance influences on farmland production, including policies related to farmland transfer, designation of major grain-producing areas, subsidies for grain cultivation, and construction of high-standard farmland, with particular emphasis on policies targeting major grain-producing regions (Fu et al. 2020; Ma and Liu 2019). The major crop-producing areas (mcpa) indicator delineates regions with concentrated agricultural output, assigning a value of 1 to counties within these areas and 0 otherwise. This indicator captures zones of significant agricultural production and informs policy measures concerning subsidies, land use, and production incentives, thereby playing a crucial role in understanding the institutional impact on spatial transitions.

Economic factors encompass internal and external economic variables that influence farmland production, including agricultural economic development, fixed asset investment, industrial structure, agricultural technological innovation, financial development, and market demand for agricultural products. This study selects agricultural economic development (Yang et al. 2022; Lyu et al. 2023; Tang et al. 2020), fixed asset investment (Fu et al. 2020), and industrial structure (Qiu and Yang 2016) as representative economic factors.

Population factors primarily comprise population density, urbanization rate, population inflows and outflows, and population quality (Lu et al. 2022; Tang et al. 2020). These factors reflect both the quantity and quality of labor available for farmland production, influencing the intensity and mode of farmland utilization.

Research methods

(1) Entropy method: To ensure a scientifically rigorous assessment of the FSTI, this study applies the entropy method (Ke et al. 2024)—a widely used multi-criteria decision-making approach. The method is particularly effective in evaluating the influence of multiple, interrelated factors within complex systems. In this context, the entropy method is used to determine the relative importance of different factors influencing farmland spatial transition. The resulting weights are then combined through a linear weighting approach to construct a composite index representing the overall degree of transition.

(2) Kernel density estimation and Moran’s I: To explore the spatial and temporal distribution of farmland spatial transition, this study adopts a three-dimensional analytical framework encompassing spatio-temporal, spatial, and temporal perspectives. (a) At the spatio-temporal level, the study generates kernel density estimation curves for the FSTI and its subsystem indices across multiple years. Kernel density estimation (Yang et al. 2024), a non-parametric statistical method, is used to estimate the probability density function of the data based on its distribution. By visualizing these curves over time, the spatial distribution and temporal dynamics of the FSTI can be better understood, providing insights into the evolution of farmland utilization. (b) At the spatial level, the study examines the spatial distribution of farmland spatial transition, identifying areas with high and low index values and exploring underlying spatial heterogeneity. Moran’s I statistic (Roy et al. 2024) is employed to assess spatial autocorrelation among county-level units, thereby revealing spatial clustering patterns within the Yangtze River Economic Belt. (c) At the temporal level, the study systematically compares spatial patterns of the FSTI across multiple years. By analyzing changes over time, it captures the temporal dynamics of farmland spatial transition. By integrating multiple dimensions of analysis, the study provides a comprehensive framework for understanding the mechanisms and underlying drivers of farmland spatial transition.

(3) GeoDetector model: To validate the theoretical mechanisms and underlying drivers of farmland spatial transition across both temporal and spatial dimensions, this study employed the GeoDetector model. The GeoDetector model is a spatially stratified heterogeneity analysis method based on Geographic Information Systems (GIS), designed to identify associations between spatial variables and potential influencing factors. The model consists of four functional modules: the factor detector, interaction detector, risk detector, and ecological detector (Wang and Xu, 2017). Unlike traditional regression-based approaches, the GeoDetector model is capable of quantifying both the explanatory power of individual variables and the interactions between them, particularly in relation to natural, social, and economic determinants. As such, the GeoDetector model provides a robust framework for analyzing the spatial mechanisms and contextual drivers underlying farmland spatial transition. The mathematical formula for the GeoDetector model is as follows:

$$q=1-\frac{\mathop{\sum }\nolimits_{h=1}^{L}{N}_{h}{\sigma }_{h}^{2}}{N{\sigma }^{2}}=1-\frac{{{SSW}}}{{{SST}}}$$
(1)
$${{SSW}}=\mathop{\sum }\limits_{h=1}^{L}{N}_{h}{\sigma }_{h}^{2}$$
(2)
$${{SST}}=N{\sigma }^{2}$$
(3)

In the equation, \(h\)=1,…,\(L\) denotes the strata or categories of variables \(Y\) or the influencing factors \(X\). \({N}_{h}\) and \(N\) represent the number of units in stratum \(h\) and in the entire region, respectively. \({\sigma }_{h}^{2}\) denotes the variance within stratum \(h\), while \({\sigma }^{2}\) represents the overall variance across the entire region. \(SSW\) is the sum of within-stratum variances, and \(SST\) is the total variance of the region. The value of \(q\) ranges from 0 to 1.

Data sources

Data related to farmland spatial transition were primarily obtained from the National Science and Technology Resource Sharing Service Platform. The analysis used the ‘China 30-meter Annual Land Cover Dataset’ (2000–2020) (http://www.ncdc.ac.cn/portal/metadata). Land use remote sensing imagery was processed with the Python-based Pylandstats module, which partitions the land into 897 counties based on formation regions. Subsequently, landscape pattern indices were calculated using Fragstats. Data on specialization transition were derived from multi-temporal, multi-scale remote sensing imagery of agricultural crops.

Natural, social, and economic data defining the driving forces were drawn from corresponding provincial, municipal, and county-level statistical yearbooks, including the China Rural Statistical Yearbook and the China County Statistical Yearbook for the respective years. DEM and slope data were acquired from the ASTER Global Digital Elevation Model. Missing data were interpolated to ensure completeness and consistency.

Results and analysis

Descriptive statistics and indicator weights

Table 3 presents the descriptive statistics of the indicators associated with farmland spatial transition. A noteworthy feature is the substantial variation observed in average plot size and Euclidean nearest neighbor distance. This indicates considerable disparities in farmland endowment and the spatial distribution of farmland patches across different regions within the Yangtze River Economic Belt. These variations serve as a crucial empirical basis for conducting research on spatio-temporal heterogeneity.

Table 3 Descriptive statistics of farmland spatial transition indicators.
Full size table

Table 4 presents the weights of 12 indicators related to farmland spatial transition, calculated using the entropy method, alongside the weights of four transition subsystems. The regularization transition and agglomeration transition subsystems carry relatively high weights of 0.31 and 0.45, respectively, collectively accounting for 0.76—substantially exceeding those of the scale and specialization subsystems. Indicators capturing farmland shape and spatial clustering, including the SHAPE_MN, FRAC_MN, MNN and AI, have the highest weights, ranging from 0.11 to 0.15. These results highlight the paramount role of morphological and spatial distribution changes in driving farmland spatial transition.

Table 4 Weight for farmland spatial transition indicators.
Full size table

In contrast, the scale and specialization transition subsystems display more modest weights of 0.1 and 0.14, respectively, indicating notable room for improvement in farmland scale management and agricultural industrialization. Although the overall degree of farmland specialization in the Yangtze River Economic Belt is not exceptionally pronounced, certain metrics such as the SDI exhibit relatively higher weights (0.1), suggesting that specialization has advanced significantly in specific regions.

Spatio-temporal changes

(1) Spatio-temporal evolution characteristics: To clarify the spatial distribution and temporal trends of the FSTI across different periods, and to gain deeper insights into the evolution of farmland utilization, kernel density curves for the overall FSTI and its four subsystems are presented (Fig. 3).

Fig. 3
figure 3

Kernel density curves of farmland spatial transition and its four subsystems.

Full size image

Analyzing the individual kernel density curve reveals an unimodal distribution with a pronounced peak, reflecting significant regional disparities in the FSTI within the Yangtze River Economic Belt. Comparing the kernel density curves over 5-year intervals shows minimal variation in waveform shape; although the peak height increases slightly, the growth rate remains modest. This indicates that the spatial distribution of the FSTI has remained relatively stable throughout the study period. Such stability suggests that regional disparities in the FSTI persist with no clear evidence of dynamic convergence or divergence.

Moreover, kernel density curves for the four subsystems of farmland spatial transition are presented to elucidate their dynamic evolution characteristics. (1) Scale transition: The kernel density curve is steep, indicating minimal regional variation, with the peak value concentrated near 0.01, reflecting a relatively low level of scale-oriented transition. The distribution lacks multimodality and shows no significant polarization. The left tail is short while the right tail is long, revealing substantial spatial disparity in scale transition within the Yangtze River Economic Belt. (2) Regularization transition: The kernel density curve exhibits an inverted ‘V’ shape with a steep profile, indicating limited regional variation in farmland regularization level. The peak centers around 0.1. The distribution is unimodal and approximately symmetric, with no evident multimodality or pronounced polarization. (3) Agglomeration transition: The kernel density curve is somewhat flattened and broad, with the peak located near 0.25, indicating a relatively high level of agglomeration index. Similar to the previous subsystems, the distribution is unimodal and symmetric, lacking significant multimodality or polarization. (4) Specialization transition: The kernel density curve is steep and right-skewed, mirroring the trend observed in the scale transition subsystem. The peak is concentrated around 0.17. The distribution is unimodal without significant polarization. The right tail is short while the left tail is long. The distribution shows no notable widening trend, suggesting relatively low spatial disparity in the specialization transition across the Yangtze River Economic Belt.

(2) Spatial evolution characteristics: The FSTI ranges from 0.38 to 0.84. Using the natural breaks classification method, the FTSI values were grouped into five categories. Slight adjustments were made to the classification results each year to ensure consistency for both vertical and horizontal comparisons. The classification thresholds are set at 0.53, 0.60, 0.66, and 0.73, alongside the annual maximum and minimum values, establishing the criteria for five categories (Fig. 4). Higher FTSI values correspond to greater farmland concentration, enhanced spatial connectivity, and more advanced scale and specialization transitions. This reflects a progressed stage of farmland spatial transition, consistent with the demands of modern agricultural development.

Fig. 4
figure 4

Spatial distribution pattern of the FSTI from 2000 to 2020.

Full size image

Overall characteristics: The spatial distribution pattern of the FTSI values reveals notable unevenness, characterized by ‘strong in the east, weak in the middle’ pattern. High-value areas (Levels IV and V) are predominantly clustered in the eastern Yangtze River Economic Belt, including Jiangsu Province, Anhui Province, and Shanghai. These high-value zones extend into the western Yangtze River Economic Belt, encompassing central and western Sichuan Province as well as western Yunnan Province. Conversely, low-value areas (Levels I and II) dominate most of Chongqing Municipality, Guizhou Province, Hunan Province, Jiangxi Province, Hubei Province, and Zhejiang Province. From a geographical perspective, high-value regions are mainly located in the middle and lower reaches of the Yangtze River Plain, the Chengdu Plain, central and northern Sichuan Plateau, and western Yunnan Province, whereas low-value areas concentrate in the hilly and mountainous zones of the Yangtze River Economic Belt’s central region. This distribution suggests a potential correlation between farmland spatial transition and regional topography.

(3) Temporal distribution characteristics: The overall spatial distribution of FTSI values remained relatively stable between 2000 and 2020, consistently following the ‘higher in the east, lower in the middle’ trend. High-value areas in the eastern Yangtze River Economic Belt are concentrated and contiguous, whereas those in the western part are more dispersed. Over time, the spatial extent of Level II regions noticeably expanded, reflecting a modest increase in FTSI values in some areas and indicating an overall enhancement in farmland spatial transition. This improvement is primarily observed in the Dongting Lake Plain and Poyang Lake Plain in Hubei Province, the eastern part of Zhejiang Province, and the northeastern part of Guizhou Province. Additionally, several eastern and western regions experienced a subtle ‘rise-fall-rise’ pattern, notably in southwestern Sichuan and the Huai Bei region.

Subsystem contributions

To elucidate the temporal variations in the contribution of various subsystems to farmland spatial transition, Table 5 presents the percentage scores of the four subsystems from 2000 to 2020. The results reveal that within the Yangtze River Economic Belt, the regularization transition and agglomeration transition subsystems have played the most prominent roles in driving farmland spatial changes. The specialization transition follows, while the scale transition consistently shows the lowest contribution. These findings suggest that the farmland spatial transition in the region has been predominantly characterized by enhanced spatial agglomeration, morphological reshaping of farmland parcels, and adjustments in production structures. The primary drivers behind these trends include the construction of high-standard basic farmland, the expansion of agricultural production cooperatives, and the promotion of specialized agricultural production. In contrast, the relatively limited impact of the scale transition highlights persistent challenges in achieving large-scale farmland operations. This points to the need for further institutional innovation and policy support to accelerate farmland transfer and consolidate farmland management at scale.

Table 5 Proportion contribution of farmland spatial transition subsystems.
Full size table

Spatial correlation

To examine whether discernible spatial correlations exist in the farmland spatial transition across different regions of the Yangtze River Economic Belt, we employed GeoDa to calculate the univariate global Moran’s I index for the FSTI in 2000, 2005, 2010, 2015, and 2020 (Fig. 5). Across all observed years, the Moran’s I values exceed 0.6, indicating a strong and stable positive spatial autocorrelation in the FSTI throughout the region.

Fig. 5
figure 5

Global Moran’s I scatter plots for FSTI from 2000 to 2020.

Full size image

Driving mechanisms

The spatial distribution of farmland spatial transition reveals substantial heterogeneity, with notable intra-subregional similarities and pronounced inter-subregional differences. To identify the underlying driving forces behind these patterns, we applied the Geographical Detector model, a spatial statistical method designed to identify spatial stratified heterogeneity and quantify the explanatory power of potential driving factors (Wang and Xu, 2017). In this study, we conducted the analysis using the GD package in R. Explanatory variables were discretized using the natural breaks, divided into five levels, labeled 0 to 4. The results of the factor detection module are summarized in Table 6.

Table 6 Results of factor detection analysis for the driving mechanisms of farmland spatial transition.
Full size table

The table above demonstrates that the natural, institutional, economic, and population factors all exert significant and consistent influence on farmland spatial transition throughout the study period.

Natural factors. The q-values of farmland endowment (farml) remain relatively stable, ranging from 0.54 to 0.56 across all observed years, with p-values consistently at 0, indicating robust statistical significance. Elevation (dem) exhibits the highest q-values, ranging from 0.61 to 0.64, also with p-values of 0 in each year. This suggests that elevation is the most influential individual factor explaining farmland spatial transition. In contrast, slope and cropping present q-values close to 0.5 with relatively higher p-values, reflecting weaker explanatory power and limited statistical significance.

Institutional factors. The q-values associated with major crop-producing area policies (mcpa) range from 0.52 to 0.53, with p-values at 0 throughout, underscoring their statistically significant role in shaping farmland spatial transition.

Economic factors. Indicators such as per capita agricultural GDP (pcaov) and per capita fixed asset investment (fai) show significant effects in most years (60%). The q-value of the proportion of the primary sector (vaa) ranges from 0.53 to 0.55, with consistently significant p-values (all 0), highlighting the importance of agricultural industry development as a key driver. However, agricultural yield per unit (ypna) shows no significant influence.

Population factors. Both the population influx rate (rip) and the student-teacher ratio in primary education (tprfe) significantly affect farmland spatial transition in 80% of the study years. Average population density (apd) exhibits q-values ranging from 0.56 to 0.59 with p-values of 0 in all years, confirming its strong explanatory power in the model.

Even in the absence of p-value testingFootnote 1, the four factors with the strongest explanatory power for farmland spatial transition are elevation (dem), average population density (apd), farmland endowment (farml), and the proportion of the primary sector (vaa), highlighting the interactive dynamics between human activity and land resources in driving spatial transitions.

Accordingly, Table 7 presents the results of the interaction analysis among the primary influencing factors, including farml, dem, slope, cropping, mcpa, pcaov, ypna, fai, vaa, apd, up, rip, and tprfe. The findings reveal that elevation (dem) consistently emerges as the dominant interaction factor across all years. This indicates that elevation not only has a strong independent effect but also significantly enhances the influence of other variables on farmland spatial transition. Moreover, the q-values for most interaction terms exceed those of individual factors, suggesting widespread nonlinear enhancement. This further corroborates the conclusion from factor detection that elevation plays a central role in shaping farmland spatial transition.

Table 7 Interaction effects analysis of driving factors in farmland spatial transition.
Full size table

Additionally, the interactions between elevation (dem) and the proportion of the primary sector (vaa), average population density (apd), and population influx rate (rip) exhibit particularly high q-values, indicating strong synergistic effects. These results imply that the combined influence of topography, agricultural industrial structure, and labor dynamics exerts a substantial impact on farmland spatial transition.

Ecological detection reveals significant variation in the spatial distribution of farmland spatial transition, particularly in relation to changes in elevation (dem) (Table 8). However, no statistically significant differences are observed in the spatial distribution of other variables relative to one another. This suggests that the influence of these other factors on farmland spatial transition may be relatively limited, or that their effects are interrelated or mutually offsetting. These findings further underscore the dominant role of elevation (dem) in farmland spatial transition.

Table 8 Ecological detection results analysis of farmland spatial transition mechanisms.
Full size table

Additionally, similar conclusions can be drawn from the risk detection results. The risk detection results indicate that elevation (dem) stands out as the most significant driving factor influencing farmland spatial transition, with the highest explanatory power among all factors and statistically significant differences across multiple categories (p < 0.05). This highlights the crucial role of elevation in shaping farmland dynamics, likely through its influence on terrain and ecological conditions. Additionally, farmland endowment (farml) exhibits strong spatial heterogeneity, and factors such as average population density (apd) and the proportion of the primary sector (vaa) also demonstrate notable significance, suggesting their meaningful impact on farmland changes. In contrast, other factors show lower explanatory power and mostly insignificant differences between categories, indicating a relatively weaker influence on farmland transition.

In summary, elevation (dem) stands out as the most influential factor driving farmland spatial transition, exerting a consistent and statistically significant impact. This highlights the crucial role of topographic variation in influencing farmland distribution and utilization, potentially due to its association with hydrological conditions, soil properties, and land suitability.

Farmland endowment (farml) also stands out as a stable and statistically significant factor, with strong explanatory power for farmland spatial transition. This suggests that regional differences in farmland resource availability substantially shape farmland use and transformation processes, making farmland endowment a key driver of spatial transition.

In addition, major grain-producing area policies (mcpa), the proportion of the primary sector (vaa), and average population density (apd) exhibit strong explanatory power. These findings imply that agricultural policies, industrial structure, and population dynamics significantly contribute to farmland spatial transition, with government policy and population changes serving as critical forces in this process.

Discussions

Compared to previous studies, this study offers additional insights into the drivers of farmland spatial transition. The identification of elevation (dem) as the dominant factor aligns with earlier research indicating that topographical variations influence farmland utilization through hydrological conditions and soil types (Ji et al. 2021; Fu et al. 2020). However, this study further reinforces the critical role of elevation by applying ecological and risk zone detection methods, providing new perspectives and evidence for ongoing investigation.

Similarly, farmland endowment (farml) is confirmed as a stable and statistically significant factor, supporting prior findings (Lu et al. 2018) and highlighting the considerable impact of resource availability on farmland spatial transition. Consequently, regional variations in farmland endowment should be carefully considered in agricultural policy formulation to promote efficient use and protection of farmland resources.

Furthermore, this study underscores the significant explanatory power of policies related to major crop-producing area policies (mcpa), the proportion of the primary sector (vaa), and average population density (apd) in farmland spatial transition. These results emphasize the influence of economic policies, industrial structure, and demographic factors, providing quantifiable analyses that deepen understanding of their roles (Fu et al. 2020; Ma and Liu 2019; Yang et al. 2022; Lyu et al. 2023; Tang et al. 2020; Qiu and Yang 2016; Lu et al. 2018).

However, several limitations should be noted. The temporal scope of the data may exclude relevant changes and influencing factors beyond the study period. The analytical models and methods incorporate only a subset of possible drivers. And the effects of other variables warrant further exploration. Finally, given that the findings are derived from specific regions and time periods, caution is advised in extrapolating results to other contexts or periods.

Conclusions and policy implications

This study analyzed the spatio-temporal patterns of the FSTI and its driving factors, yielding several key conclusions. Between 2000 and 2020, the spatial transition index within the Yangtze River Economic Belt showed no significant dynamic convergence or divergence, with temporal variations in spatial disparities evolving gradually. The scale index remained relatively low, whereas the agglomeration index consistently exhibited higher values. The FSTI ranged from 0.38 to 0.84, displaying an uneven spatial distribution characterized by higher values in the eastern and western regions and lower values in the central areas. Over time, the overall FSTI steadily increased and demonstrated strong positive spatial autocorrelation.

Farmland spatial transition results from the interplay of natural and socio-economic factors, with natural factors dominating during the study period. In particular, elevation (dem) exerted the strongest influence, with q-values ranging from 0.61 to 0.64. Findings from ecological and risk zone detection further corroborate the predominant role of elevation in driving farmland spatial transition, offering new insights and evidence for further research. Additionally, farmland endowment (farml) exhibited considerable explanatory power. Furthermore, policies related to major crop-producing area policies (mcpa), the proportion of the primary sector (vaa), and average population density (apd) all played significant roles in explaining farmland spatial transition.

These findings carry important implications for farmland development and policy-making. First, the influence of terrain elevation on farmland utilization must be given paramount consideration. Given the predominant role of elevation (dem) in farmland spatial transition, policymakers should explicitly consider terrain elevation in farmland management strategies. In high-altitude or steep terrain regions, targeted interventions—such as terrace construction and measures to prevent excessive land development—are essential to protect farmland.

Second, effective regulation of farmland resource allocation is crucial. Considering the significant influence of farmland endowment (farml), policymakers should adopt rational and context-specific approaches to allocating farmland resources. In resource-rich areas, policies and incentives should encourage agricultural development, while in resource-scarce regions, efforts should focus on promoting structural adjustment and reducing dependence on farmland.

Third, regionally differentiated agricultural policies are recommended. Considering the effects of policies related to major crop-producing area policies (mcpa), the proportion of the primary sector (vaa), and average population density (apd) on farmland spatial transition, agricultural policies should be tailored to regional characteristics. In major crop-producing areas, supportive policies and incentives can enhance agricultural capacity and sustainability. Regions with a high proportion of primary industry should prioritize diversification and modernization of agricultural practices. Furthermore, in densely populated areas, policies should emphasize farmland protection to prevent overdevelopment and excessive exploitation.

Finally, integrating ecological environment protection with farmland utilization is imperative. Agricultural development policies must carefully balance ecological conservation with farmland use, particularly in high-elevation or ecologically sensitive areas. Specific environmental protection measures should be implemented to prevent excessive farmland expansion, thereby safeguarding ecosystem stability and sustainability while promoting agricultural productivity.