Spatiotemporal cropland dynamics and sustainable pathways in the Yangtze river economic belt

Abstract

Despite cropland changes’ critical role in ensuring food security, fostering socio-economic development, and balancing ecological protection, the regional variability in cropland dynamics within large-scale watershed economic belts remains understudied. To address this gap, we mapped land use change intensity to analyze cropland change patterns and stability in the Yangtze River Economic Belt (YREB). Through statistical analysis, we identified key driving factors and predicted cropland changes for 2030 using the FLUS model. We also proposed region-specific sustainable use scenarios. The results reveal a systematic trend of cropland conversion to built-up land and water transforming into cropland in the YREB. Despite a steady decline in cropland area over the past two decades, the intensity of cropland inflows and outflows has remained balanced. The cropland center of gravity follows an east-north trajectory, with local meandering, isotropic shifts, and transverse displacements. Transport networks significantly influence cropland changes in the upper YREB, while climate impacts on cropland in the lower YREB are increasing. Future projections suggest the cropland center of gravity will shift toward southern mountainous regions in the upper and middle YREB and toward northern plains in the lower YREB. Among the proposed scenarios, only the Cropland Protection Scenarios (CPS) align closely with the cropland coverage targets of the Sustainable Pathways in Shared Socio-Economic Development Pathways (SPP1). Urban expansion encroaching on agricultural land is evident in Deyang and Mianyang, while competition with ecological space is pronounced in the middle and lower reaches and in riverine and coastal areas. This study provides a novel perspective on cropland changes in the YREB, emphasizing regional variability and the need for tailored policy interventions, thereby offering a scientific foundation for cropland management and planning.

Introduction

Cropland is a fundamental resource widely utilized to support human development¹. The 2019 IPCC Special Report on Climate Change and Land emphasized that sustainable cropland use is crucial for reducing soil erosion, eradicating hunger, and combating climate change. However, global population growth, urbanization, and industrialization impose unprecedented pressures on cropland^2–4. Soil erosion, urban and infrastructure development, cropland abandonment, and forest restoration have resulted in significant encroachment and loss^5,6. Particularly, the global trend of cropland conversion to construction land creates severe challenges for agriculture, as it often entails extensive soil sealing, destroying soil functions⁷ and occupying the most fertile lands⁸. Furthermore, urbanization is projected to cause approximately 2% of global cropland loss by 2030, predominantly in Africa and Asia⁹, with China accounting for a quarter of the total loss. Competing demands for land¹⁰ and the ecological impacts of land use further exacerbate pressures on cropland. In this context, cropland conservation and sustainable use represent major global challenges for scientists and policymakers.

With socio-economic development, dynamic changes and cropland conversion become inevitable^11–13. Various aspects of cropland change drivers have been studied, confirming this phenomenon. Cropland change is influenced by natural and anthropogenic factors, particularly population growth, which directly increases food demand, promotes cropland expansion¹⁴ and drives intensive land use¹⁵. Conversely, urbanization-driven rural-to-urban migration has led to the abandonment or conversion of some cropland to other uses. Economic factors, policy frameworks, and management practices also play critical roles in shaping cropland use¹⁶. Cropland changes exhibit significant temporal and spatial diversity, including expansion, degradation^17,18, shifts in cropping structure¹⁹, and diversified land-use conversions²⁰. The ecological impacts of cropland change include effects on biodiversity^21,22, soil quality^23,24, water resources^25,26, deforestation^27,28, greenhouse gas emissions^29,30 and ecosystem services³¹. Socio-economic impacts include effects on agricultural production³², food security³³, and farmers’ livelihoods³⁴. Research on cropland change has predominantly focused on global-scale phenomena, such as the effects of tropical deforestation³⁵, desertification³⁶ and urbanisation³⁷. Technological advances in remote sensing monitoring³⁸ have revolutionised how cropland data are acquired³⁹ and researched⁴⁰. Despite significant progress, gaps remain in comparative and quantitative research on regional and modal cropland changes, and the variability of drivers at local and micro levels is still poorly understood. In developing countries, land use change in traditional agricultural areas is mainly driven by cropland conversion to non-agricultural land^41,42. This phenomenon usually stems from the population pressure on these areas and the dual need to pursue economic development and ecological protection simultaneously.

China is confronting the significant challenge of a persistent decline in cropland. Since 1979, over 58.5 × 10³ ha of highly productive cropland has been lost annually⁴³. Data from China’s Third National Land Survey indicate that cropland declined by 113 million mu between 2009 and 2019⁴⁴, with a net increase not occurring until 2021. Despite this challenge, China has consistently fed nearly 20% of the global population with less than 10% of the world’s cropland⁴⁵, making significant contributions to the United Nations’ Millennium Development Goals by reducing hunger and poverty⁴⁶. Over the past 20 years, rapid urbanization has consumed substantial high-quality cropland. Combined with over-exploitation and irrational use, this has accelerated cropland change at an unprecedented rate. Non-farm conversion and cropland degradation^47,48 directly impact food security, sustainable cropland use, and ecological security⁴⁹. Additionally, factors such as aspirations for improved living standards, a growing demand for diverse dietary structures⁵⁰, the impacts of the COVID-19 pandemic, the ongoing Russian-Ukrainian conflict, and climate change uncertainties^51–53 have elevated sustainable cropland use as a key priority on the Chinese government’s agenda.

As a major agricultural country, China’s cropland distribution and usage exhibit pronounced regional disparities. Geographically, the Northeast Plain features fertile black soil and advanced agricultural mechanisation. Although the cropland area remains stable, it faces threats from black soil erosion and salinisation^54,55. In contrast, cropland in the North China Plain and Huanghuaihai region is concentrated and contiguous, but it is increasingly affected by urbanisation and groundwater over-exploitation⁵⁶. In southern hilly areas, cropland is more dispersed, with significant issues of non-food cropland and ecological fallowing becoming increasingly prominent^57,58. Cropland in the Northwest Arid Region depends on irrigation. Although advancements in irrigation technology have facilitated limited cropland expansion, they have placed considerable strain on already scarce water resources⁵⁹. Although the economically developed eastern region concentrates most of the country’s high-quality cropland, rapid urbanisation and industrialisation have caused it to face a serious problem of cropland encroachment⁶⁰. The central region is rich in cropland resources as the core of agricultural production, but it faces the double threat of cropland degradation and loss due to inefficient use and inappropriate farming practices⁶¹. The under-developed economic regions in the western part of the country are constrained by a fragile ecological environment with sparse cropland and have long faced the challenges of desertification and soil erosion⁶². These regionally distinct characteristics and associated challenges indicate that a universal cropland management strategy is not feasible. Therefore, cropland management must account for regional variability by thoroughly examining changing conditions and trends to formulate precise and effective protection and usage policies.

Human civilisation has historically thrived through the sustenance of large rivers⁶³. Abundant water resources, optimal watercourses, and favourable ecological conditions provide a solid foundation for agricultural development and the creation of watershed economic belts within regional integration frameworks⁶⁴. Since 2014, when the Chinese government proposed the YREB based on the Yangtze River Basin, it has evolved into a globally influential inland river economic belt. It encompasses over 40% of China’s population and economic output⁶⁵ while exhibiting significant regional heterogeneity. The YREB integrates major grain-producing areas, economic growth hubs, and ecological safety barriers, serving as a critical axis for China’s socio-economic sustainable development⁶⁶. Containing 40% of China’s cropland and producing 35% of its grain, the YREB’s cropland changes significantly impact the stability of China’s grain supply and food security strategy⁶⁷. However, the YREB’s strategic development entails intensified economic activities, rapid urbanisation, stringent ecological compliance policies, and ecological restoration measures. This process inevitably results in declining cropland quantity and quality, alterations in land use structure, and intensified land use conflicts⁶⁸. Additionally, it faces threats from uncertainties arising from natural factors such as climate change and soil degradation. Thus, systematically analysing cropland use dynamics—including trends, patterns, and core driving mechanisms—within the YREB’s distinct characteristics is crucial. This analysis aims to offer case-based support for sustainable cropland use in China.

Focusing on the YREB, characterized by significant regional differences, this paper addresses key questions about cropland changes between 2000 and 2020: What are the trends and the spatiotemporal characteristics of cropland changes across different regions? What are the primary drivers of cropland changes? What are the main issues and conflicts impacting the sustainable use of cropland in various regions? What actionable policy insights can be developed to assist policymakers in improving cropland management in China?

Materials and methods

Study area

The YREB spans 11 provinces (Fig. 1) along the Yangtze River, covering an area of approximately 2.04 × 10⁶ km². The upper, middle, and lower sections correspond to three major regions in western, central, and eastern China, with the terrain descending gradually from the plateau of the west to the eastern plains⁶⁹. The YREB hosts six major grain-producing areas characterized by a warm, humid climate and fertile soils suitable for diverse crops, especially rice, oilseed rape, and tea. It serves as a core zone for promoting diversified agriculture and modern technologies, serving as a model for China’s agricultural development. Additionally, the YREB is a significant watershed economic zone, featuring the world’s largest population, industrial scale, and most comprehensive urban system⁷⁰. Its high international profile makes it a quintessential and crucial case study.

Fig. 1.

Study area and cropland distribution in 2020. The administrative boundaries were obtained from the Geographical Information Monitoring Cloud Platform (http://www.dsac.cn). Maps in this figure were created by the authors using ArcGIS 10.8.2 (https://www.esri.com).

Data and processing

The study data consists of the YREB’s land use data for 2000, 2010, and 2020, classified into cropland, forest, grassland, water, and built-up land (BL), which includes urban, rural, industrial, mining, and residential land, as well as unutilised land (UL), based on the national standard ‘Classification of the Current Situation of Land Use GB/T21010-2017’. For spatial conflict analysis, urban-agricultural-ecological functional classification criteria^71,72 were applied, categorising cropland as agricultural space, BL as urban space, and other areas as ecological space. Additionally, terrain data, climate data and ecological data⁷³ were selected to assist the research analysis based on the availability, quantifiability and relevance of the data, and all data were standardised using ArcGIS10.8.2, as shown in Table 1.

Table 1.

Specific data and sources.

Data type		Data name	Acronym	Data sources
Natural factors	Administrative boundaries	Administrative boundary data for the YREB	-	http://www.dsac.cn/
	Land use	Land use data	-	http://www.resdc.cn
		Nature reserve boundary data	-
	Terrain	DEM	DEM	https://www.webmap.cn/
		Slope	SL
		Slope direction	SD
		Degree of topographic relief	DTR
		Soil	SO	https://gaez.fao.org/pages/hwsd
	Climate	Annual average precipitation	AAP	http://www.worldclim.org/
		Annual average temperature	AAT
	Ecology	NDVI	NDVI	http://www.resdc.cn
		Habitat quality	HQ	https://www.geodata.cn/data/datadetails
Socio-economic data	Human activities	NPP-VIIRS night light data	NV	https://ngdc.noaa.gov/
		Agricultural producer regions	-	Territorial spatial planning
		Distance to city	DC	http://download.geofabrik.de/
		Distance to provincial highway	DPH
		Distance to highway	DH
		Distance to railroad	DR

Methodology

Research framework

We analysed the regional variability of cropland change and its drivers in YREB, modelled cropland change trends, and formulated adaptive cropland development measures. The overall research framework is shown in Fig. 2, including the following steps: (i) integrating multi-source data and pre-processing, (ⅱ) evaluating and screening cropland drivers using binary logistic analysis and Receiver Operating Characteristic (ROC) curves, (ⅲ) predicting cropland change trends using the FLUS model (Liu et al. 2017), and (ⅳ) identifying the primary land use conflicts affecting the sustainable use of cropland through spatial conflict analysis, and proposing recommendations for regional adaptive cropland use.

Fig. 2.

Research framework.

Intensity and stability of cropland use

Estimation of LUCC intensity involves a detailed reinterpretation of land use transfer matrix data. Changes and patterns in land classes are quantitatively analysed using a top-down, hierarchical approach^74,75. This paper explicitly profiles cropland change through the following approaches:

(1) Mapping Analysis: Identifying the systematicity (the existence of regularity and tendency) and stability (the ability to resist perturbation) in the conversion process of multiple land uses is a complex task. In this study, we adopt the LUCC intensity mapping unit⁷⁶ to analyze the transformation patterns of cropland based on the mutual transformation between the initial land class i and the final land class j. As shown in Fig. 3, the x-axis represents the initial land class, while the y-axis represents the final land class. Each cell in the intensity map comprises four elements: ① and②denote the absolute transfer-in and transfer-out intensities, respectively;③and④denote the relative transfer-in and transfer-out intensities, respectively. Within a cell, if ① and ② are in red, it indicates a clear tendency of absolute transfer-in or transfer-out during the land conversion process; if ③and ④are in red, it indicates a relative conversion tendency. If all four elements are red or blue, the conversion process exhibits a strong systematic tendency or inhibition, respectively.

Fig. 3.

Intensity map unit of LUCC.

(2) Intensity analysis: It is a top-down hierarchical, interpretive framework structure for analysing differences in variation between land classes. Based on land-use intensity⁷⁴, we addressed three questions: (i) At the interval level, when was cropland change most rapid or slow in the past 20 years? (ii) At the land type level, was cropland relatively stable or active during a specific period? (iii) At the shift level, was cropland shift strong in a given period?

Cropland centre of gravity trajectories

The shift in the centre of gravity of cropland, along with its standard deviation ellipse⁷⁷, is commonly used to represent the overall characteristics of cropland change⁷⁸. Specifically, the direction and speed of the centre of gravity’s shift over a given period reflect the overall trend in the spatial distribution of cropland⁷⁹. Its mathematical expression is:

1

Here, X and Y represent the horizontal and vertical coordinates of the cropland distribution’s centre of gravity, denotes the cropland area in region , and and are the horizontal and vertical coordinates of the geographical centre of gravity of area , respectively⁸⁰.

Binary logistic regression (BLR) and receiver operating characteristic (ROC) curve

BLR is a non-linear statistical method frequently used in geographic information analysis to model dichotomous dependent variables. It aims to examine the significant relationship between a dependent variable and independent predictors through a regression model⁸¹. This study treats land type (Y) as a binary dependent variable, where Y = 1 represents cropland and Y = 0 represents non-cropland. The potential driver ()(i = 1,2,……,m) is considered as an independent variable, and P represents the probability that land use is cropland (Y = 1). The corresponding BLR model is presented below.

2

3

whereis the regression coefficient. which can reveal the possible impact of each independent variable exerted on the dependent variable, P/(1-P) is the odds ratio of the event⁸¹.

ROC curve is a widely used tool for evaluating the performance of binary classification models and has been instrumental in this regard (Luo et al., 2024; Vakhshoori and Zare, 2018; R. Gil Pontius Jr, 2001). The area under the curve (AUC) is commonly used to assess the discriminative power of predictive models^82–84. When AUC = 0.5, the factor remains unexplained; when AUC < 0.5, it may be unexplained or even counterproductive; and when AUC > 0.5, the model’s effectiveness improves. A higher AUC value, approaching 1, indicates better model performance. We employed the ROC curve to calibrate the suitability drivers further, enhancing Land-Use and Land-Cover Change (LUCC) simulations’ accuracy.

Cropland coverage target for 2030

Ensuring adequate cropland quantity is the core objective and foundation of cropland protection. Considering China’s medium- and long-term development plans and regional development strategy⁸⁵, the food supply of the YREB must not only meet its own demand but also account for the needs of other regions in China. This paper evaluates the food supply requirements of the YREB for 2030, considering future changes in food supply and demand. The methodology employed is as follows:

4

For the YREB, high-quality development is the future growth trajectory, with regional competitiveness as one of the key endogenous drivers. Therefore, we select the population size from both the sustainable (SSP1) and regional competitiveness (SSP3) paths of the common socioeconomic development scenarios⁸⁶. We estimate the national population by averaging the two values from SSP1 and SSP3⁶⁵. The Food and Agriculture Organization (FAO) sets a food security standard of 400 kg/year per capita⁸⁷, while Chinese scholars suggest that per capita food availability should reach 472 kg/year by 2030 based on China’s development level⁸⁸. Additionally, forecasts indicate that per capita food consumption in China may reach 457 kg/year by 2030, accounting for food used for feed, industrial purposes, and seeds⁸⁹. This paper adopts an average of the latter two estimates, 465 kg/year, as China’s per capita grain demand in 2030. In line with the Outline of Medium- and Long-term Planning for National Food Security (2008–2020), which advocates stabilizing China’s food self-sufficiency rate above 95%, by analyzing the trend of YREB food production and its share of national food production from 2000 to 2020 (Table 2), we predict that YREB food production will constitute approximately 36.36% of national food production by 2030.

Table 2.

YREB grain production and its share in the country, 2000–2020 (tons) from the annual China statistical Yearbook.

Region/Year	2000	2005	2010	2015	2020
Shanghai	174	105. 4	118. 4	112. 1	91. 4
Jiangsu	3106. 6	2834. 6	3235. 1	3561. 3	3729. 1
Zhejiang	1217. 7	814. 7	770. 7	752. 2	605. 7
Anhui	2472. 1	2605. 3	3080. 5	3538. 1	4019. 2
Jiangxi	1614. 6	1757	1954. 7	2148. 7	2163. 9
Hubei	2218. 5	2177. 4	2315. 8	2703. 3	2727. 4
Hunan	2767. 9	2678. 6	2847. 5	3002. 9	3015. 1
Chongqing	1106. 9	1168. 2	1156. 1	1154. 9	1081. 4
Sichuan	3372	3211. 1	3222. 9	3442. 8	3527. 4
Guizhou	1161. 3	1152. 1	1112. 3	1180	1057. 6
Yunnan	1467. 8	1514. 9	1531	1876. 4	1895. 9
Total	20,679. 4	20,019. 3	21,345	23,472. 7	23,914. 1
National Percentage	44. 74%	41. 36%	39. 06%	37. 77%	35. 72%

In summary, we estimate that food production in the YREB must reach 234.88 million tons by 2030. What extent of cropland will be required to meet this demand? Using the standard forecasting methods for cropland coverage⁹⁰, we adopt the scale forecast based on food production security⁹¹, as shown below:

5

Where S represents the required YREB cropland size in 2030 under the food security concept; C is the population size at the end of the projection period; β denotes the food self-sufficiency rate; is the minimum per capita food requirement; P is the grain yield per unit area; Q is the ratio of grain-sown area to total cropland area; and K is the replanting index.

Land use modelling scenarios

LUCC is a complex and multifaceted process⁹². We developed three different land use scenarios to promote the sustainable development of YREB in terms of socio-economic development, food security, and ecological protection.

Natural Development Scenario (NDS): Based on the ‘business as usual’ principle, the Markov model predicted the LUCC state in 2030, drawing on land use trends in the YREB from 2000 to 2020. The catchment’s slow rate of change and its significance to the basin’s economic zone were incorporated as limiting constraints. The NDS serves as the baseline for simulating other scenarios.

Cropland Protection Scenarios (CPS): This scenario adheres to the principle of ‘priority for cropland protection’ and aligns with cropland coverage targets. Major agricultural production areas are set as restrictive conditions, minimizing the probability of cropland conversion to other land types.

Ecological Conservation Scenario (ECS): Based on the YREB’s Ecological Development Report (2019–2020)⁹³, this scenario prioritizes protecting ecological functional areas, strategic water sources, and regions with critical ecosystem functions, using ecological conservation zones as restrictive conditions.

Modelling of GeoSOS-FLUS

The FLUS model is an integrated framework for simulating LUCC scenarios by coupling human activities with natural effects⁹⁴. It addresses the complexity and uncertainty inherent in LUCC processes and has been successfully applied to the dynamic simulation of LUCC in the YREB^95,96. We carried out the following work surrounding the model(Fig. 2 Step3): (i) Training and predicting the adaptive probability of different land use types using an Artificial Neural Network⁹⁷ algorithm based on the natural and socio-economic factors influencing LUCC. (ii) Employ Markov chains⁹⁸ to project the quantitative changes of LUCCs in 2030 based on historical LUCC data spanning 2000 to 2020. (iii) Combining different scenarios with established transformation rules and constraints to simulate spatial changes in LUCC in 2030 using a Cellular Automata (CA) model⁹⁹.

The Kappa and FoM coefficients¹⁰⁰ are widely used to assess model accuracy. The Kappa index quantifies the agreement between modelled and actual land maps, while the FoM evaluates their spatial consistency¹⁰¹. A Kappa value ≥ 0.75 indicates high accuracy, with values between 0.5 and 0.75 reflecting moderate accuracy. The FoM coefficient typically ranges from 1.0 to 0.59, although some studies report lower values around 0.3¹⁰². In this study, the validation results yielded Kappa and FoM values of 0.84 and 0.25, respectively, demonstrating the suitability of the GeoSOS-FLUS model for simulating land use changes in the YREB under consistent transition rules and parameter settings.

Results

LUCC strength and stability

Characteristics of cropland use intensity mapping

According to the LUCC intensity mapping of the YREB in 2000–2020 (Fig. 4), Cropland→BL (conversion of cropland to BL) showed a systematic tendency, indicating a gradual increase in the conversion rate over time. Cropland → forest showed an absolute tendency and stability, reflecting a continuous and substantial rate of change, and stability. In the upper and middle YREB, cropland→forest exhibits absolute tendency and stability. Cropland→UL demonstrates stability in the lower YREB. Additionally, grassland→cropland experienced systematic inhibition from 2010 to 2020. Forest→cropland exhibited absolute tendency in the midstream, whereas grassland→cropland displayed stability in the lower YREB. This period marks accelerated ecological construction in China. Notably, both cropland→water and water→cropland exhibit strong absolute or systematic tendencies. These maps highlight the region-specific patterns of change in the YREB.

Fig. 4.

LUCC intensity mapping of the YREB, 2010–2020.

Regional patterns of change

Over the past twenty years, the intensity of YREB land use change has increased (Fig. 5Ⅰ). Compared to the average growth intensity, changes were slower from 2000 to 2010 but accelerated from 2010 to 2020. Notably, the lower of 0.52% is significantly higher than in other regions. This trend is linked to varying industrialisation and urbanisation processes across regions during different periods. For cropland, the transfer-out area exceeds the transfer-in area, butand are comparable (Fig. 5Ⅱ). Both transformation processes remain active.

Fig. 5.

LUCC interval categories and intensity of cropland change. represents cropland using intensity; represents cropland turning in intensity; represents cropland turning out intensity. Uniform intensity, denotes the uniform line. Should the intensity bar lie to the left of the uniform line, the change in cropland area during that period is relatively slow; should the intensity bar extend to the right of the uniform line, the change in cropland area during that period is relatively rapid.

Regarding transfer levels (Fig. 6), of cropland is larger than Cropland is primarily transferred from forests, grasslands, water, and BL, with water being the most active in transfer activities, showing significantly higher intensity than average. In the upper YREB, cropland is primarily transferred from grasslands and forests to forests, grasslands, and BL. In the middle of YREB, the transfer -in sources of cropland are mainly forests and water, while the transfer -out sources are forests, BL, and water. In the lower YREB, cropland is predominantly transferred from water, forests, and BL, with BL being the primary transfer-out destination. These trends highlight each region’s land use policy orientation and socioeconomic development characteristics, underscoring that water→cropland represents a stable transfer pattern. Cropland→BL remains a stable transfer-out pattern. Additionally, the water domain plays a critical role in ensuring cropland change stability.

Fig. 6.

Intensity of the transfer dimension of cropland use. a, b, c, d are the YREB, Upper YREB, Middle YREB, and Lower YREB, respectively. 1,2 denote the intensity of cropland turn-in and turn-out for 2000–2010, and 3,4 represent the turn-in and turn-out of cropland for 2010–2020. represents the turnout intensity (≠) from the initial land category to the final land category , and represents the turnout intensity (≠) from the initial land category to the final land category .

Spatial variability in cropland use patterns

Spatial pattern variability

Spatial changes in cropland are primarily concentrated near provincial capitals and mountainous areas (Fig. 7) and can be categorized into urban expansion encroachment and ecological fallow encroachment. The urban expansion encroachment category includes cities such as Chengdu, Wuhan, and their surrounding satellite cities. Cropland encroachment is more pronounced in the early stages than in the later stages in Chengdu, Nanchang, and the Yangtze River Delta. In contrast, Chongqing and Hefei exhibit a reversed trend, with more significant encroachment in the later stages. Regional variability exists in the degree of urban expansion’s impact on cropland and its changing trends. The Qianxinan mountainous area in Guizhou exemplifies ecological fallow encroachment alongside the mountainous regions in northern Hubei and Yunnan, representing rocky desertification fallow and sloping cropland fallow, respectively.

Fig. 7.

Spatial change in cropland use in YREB, 2000–2020.

Centres of migration trajectories and patterns

The centre of gravity of cropland in the YREB (Fig. 8) is situated in the Xiangxi Tujia and Miao Autonomous Prefecture, Hunan Province. It demonstrates an overall trend of ‘eastward and northward movement’, with a migration of 26.11 km northeast between 2000 and 2010 and 20.1 km northwest between 2010 and 2020. The cropland distribution shows a tendency to become more centralised rather than decentralised. Significant differences are observed in the cropland centre of gravity trajectory across regions. The upper region shows a ‘westward and northward movement,’ with relatively stable concentration and total cropland area. In the middle region, the centre continues migrating northwest, while in the lower region, it shifts ‘northward and eastward,’ with cropland distribution becoming more concentrated and undergoing greater area changes. These trajectories reflect gradients in natural conditions, urbanisation and land policy.

Fig. 8.

Characteristics of the migration of the centre of gravity of cropland utilization in the YREB, 2000–2020.

Using the dynamic changes in cropland areas across provinces (Table 3), we further analyse the local centre of gravity of cropland (Fig. 8). Three distinct patterns emerge: Meandering Centroidal Shift: In this pattern, the centre of gravity initially moves in a specific direction but partially retraces in the later stage, deviating from its original trajectory. The spatial structure of cropland remains stable. For example, in Anhui, Hunan, Jiangxi, and Sichuan, cropland reduction in the later period was 1 to 6 times greater than in the early period. Uniform Directional Shift: The centre of gravity consistently moves northeast over a relatively short distance, as observed in Hubei, Jiangsu, Zhejiang, and Shanghai. Specifically, the significant early reduction in cropland area in the lower provinces, combined with cropland utilisation intensity analysis, indicates that these regions contributed to the phased development of urbanisation, industrialisation, and agricultural structural adjustment during relatively relaxed land-use policies. In the later period, cropland reduction slowed as policies tightened and awareness of protection increased. Shift in Focus: The centre of gravity moved south-to-north in the early period and westward in the later period, though over relatively short distances. The cropland concentration and shape remained stable, indicating no drastic changes in the overall pattern. The cases of Guizhou, Yunnan, and Chongqing, located in the upper reaches, highlight the profound influence of geographic conditions and policies on cropland changes.

Table 3.

Dynamics of cropland in YREB provinces, 2000–2020.

Province	Proportion(%)			Dynamic quantity(km²)
	2000	2010	2020	2000–2010	2010–2020	2000–2020
Anhui	57.5	56.73	55.01	−1072.93	−2414.6	−3487.53
Chongqing	47.24	46.39	45.41	−705.81	−806.06	−1511.87
Guizhou	27.95	27.85	27.41	−178.8	−764.88	−943.68
Hubei	37.46	36.79	34.4	−1247.06	−4452.37	−5699.43
Hunan	28.87	28.6	27.91	−569.2	−1457.95	−2027.15
Jiangsu	67.48	61.44	60.2	−5791.03	−1503.39	−7294.42
Jiangxi	27.05	26.96	26.38	−154.4	−962.6	−1117
Shanghai	56.65	42.59	37.7	−817.86	−411.07	−1228.93
Sichuan	24.96	24.65	24.36	−1470.05	−1424.58	−2894.63
Yunnan	17.98	17.81	17.61	−637.59	−765.16	−1402.75
Zhejiang	26.35	24.09	22.97	−2359.17	−1164.67	−3523.84

Identification and assessment of the drivers of cropland change

Driver identification based on BLR

Natural factors such as DTR, AAP, HQ, and SO (Definitions in Table 1) were found to have significant effects (Sig<0.05) on cropland change from 2000 to 2020 (Fig. 9), aligning with the observed trend that cropland change in the YREB decreases as topographic relief increases¹⁰³. The Wald value for DEM increases from 2.45 to 331.73 from 2010 to 2020. This increase is tentatively attributed to the retreat of steep-slope cropland in mountainous areas under the 2014 to 2019 Grain-for-Green (GFG) program, implying that steeper terrain had a negative influence on cropland retention. Additionally, in the upper reaches, the factors DH, AAT, and NV were also significant (Sig < 0.05). Specifically, DH exerted a strong negative influence on cropland area (Wald rose to1003.22), whereas NDVI had a positive influence (Exp(B)>1 during 2010–2020), indicating the positive impact of increased vegetation cover. The construction of transport network facilities emerged as a key driver of cropland change. Moving to the middle reaches, the influence of all factors, except DTR, diminished. In the lower reaches, socio-economic factors dominated cropland change, but the Wald value of AAP rose to 309.76, underscoring the positive effect of precipitation on agricultural development in the region.

Fig. 9.

YREB localized cropland utilization drivers radar map, 2000–2020.

Interpretation of rider confidence based on ROC

Through ROC curves, we further screened factors that demonstrate reliable explanatory power for cropland changes between 2000 and 2020. Their AUC values generally ranged from 0.50 to 0.58(Fig. 10), limited discriminatory ability but still outperforming random classification. Overall, DTR and AAP exhibited relatively high explanatory power. Notably, NV and DH in the upper reaches achieved AUC values approaching 0.7, suggesting high predictive accuracy and generalizability. Additionally, NDVI in the middle reaches shifted from unreliable to reliable explanatory capacity. AAP and AAT provided highly reliable explanations for cropland changes in the lower reaches. Collectively, these factors are critical for improving the accuracy of land-use simulations.

Fig. 10.

Cropland utilization driver ROC curves.

Results of multi-scenario simulations

Cropland coverage in 2030

Based on cereal area and cereal yield data, the average cereal unit area yield in the YREB from 2000 to 2020 is 50,600 − 56,300 tons/10,000 ha. Using regression analysis, incorporating various improvements and influencing factors, P is projected to reach 61,600 tons/10,000 ha by 2030. The mean value analysis predicts Q 60.1% and K 119.02%, accounting for advancements in agricultural technology and cultivation methods. Under the SSP1 path, the required cropland area is projected to exceed 617,740.21 km², representing an increase of 2,174.96 km² compared to the cropland area in 2020. For the SSP3 path, the cropland area is expected to surpass 636,770.84 km², an increase of 21,205.59 km² relative to 2020. Considering the average population projection, the cropland area should exceed 627,255.53 km², reflecting an increase of 11,690.28 km² from 2020.

Trends in cropland change

Based on LUCC intensity mapping under different scenarios (Fig. 11), in the NDS scenario, cropland→BL and BL→cropland exhibit a systematic tendency. At the same time, cropland→forest in the upper and middle regions shows an absolute tendency to align with China’s mainstream land use conversion trends and policy guidance. In the CPS scenario, grassland→cropland the upper region, and BL→cropland in the lower exhibit systematic tendencies based on land suitability. Grassland in the upper region is more likely to be converted into cropland, while the lower region focuses on improving BL use efficiency. In the ECS scenario, cropland→forest exhibits a systematic tendency in upper and lower areas. Cropland→grassland and cropland→water exhibit systematic tendencies in the middle region. This indicates that converting cropland to ecological land remains a common phenomenon under policies promoting the return of cropland to forests, grasslands, and wetlands. This process underscores the critical role of water in balancing regional ecological protection and cropland stability. Only the CPS scenario achieves a total cropland area of 615,714 km², approaching the cropland scale required under the SPP1 pathway.

Fig. 11.

LUCC Intensity Mapping of the YREB different scenarios,2030.

Spatial change in cropland: trends and conflicts

Combining land use conflict characteristics (Fig. 12) with spatial cropland changes under different scenarios (Figs. 13, 14 and 15) reveals that the cropland’s centre of gravity in the NDS scenario shifted southwestward (Fig. 8), decrease in cropland area 9285.83 km². The urban space encroached on agricultural space around core cities and the eastern coastal area, with notable concentrations in Deyang and Mianyang near Chengdu in the upper and northeastern Zhejiang in the lower. Encroachment of urban space on agricultural space and ecological space is more significant in the lower, characterized by flat terrain and dense rivers and lakes. In the CPS scenario, the centre of gravity of cropland shifts southeastward, increased in cropland area 11147.17 km², concentrating in the upper central mountainous areas and along the Yangtze and Huai Rivers in the middle and lower regions. Agricultural space becomes more concentrated while encroachment on ecological space increases, leading to some disturbances in the urban space. In the ECS scenario, the cropland’s centre of gravity shifts northwestward, decrease in cropland area 2275.83 km². Mountainous areas in the lower central and southern regions exhibit a denser concentration of agricultural space, while urban space encroachment on agricultural space becomes more evident in northern regions. The encroachment of the urban space on agricultural space is particularly noticeable. This underscores the complexity and diversity of cropland conversion in rapidly developing economic regions, highlighting the determination and effectiveness of land use managers in protecting cropland space through policy orientation and technological advancements.

Fig. 12.

YREB land use shift shock maps for different scenarios agricultural space (AS), urban space (US), ecological space (ES).

Fig. 13.

Spatial changes in the upper cropland utilization under different scenarios.

Fig. 14.

Spatial changes in the middle cropland utilization under different scenarios.

Fig. 15.

Spatial changes in the lower cropland utilization under different scenarios.

Discussion

4.1. Drivers, Mechanisms, and Spatial Patterns of Cropland Change in Watershed-Based Economic Belts: A Multi-Scaling and Regionalized Perspective.

Cropland change is a complex and dynamic process shaped by multiple factors, levels, and scales, following specific transformation mechanisms^104–106. Endogenous development shapes cropland change in the YREB, while external policy orientations and socio-economic development⁹⁴ dominate its direction, location, and scale¹⁰⁷. Natural conditions govern cropland’s distribution and production potential, while environmental factors influence its stability and sustainability. YREB cropland is primarily distributed along the Yangtze River in plains with flat terrain and favourable soil and water¹⁰⁸. Population and economic development are direct drivers of cropland change. Urbanisation has led to a significant decline in cropland over the past 20 years, especially in areas with sustained population growth, such as the Yangtze River Delta, the central Yangtze River region, and the Chengdu-Chongqing urban agglomeration¹⁰⁹, where there is a large area of cropland→BL. Policies drive and respond to cropland change. As a critical ecological conservation area in China, the GFG and ‘Grain for Lakes and Wetlands’ programs have brought some cropland under ecological protection and restoration¹¹⁰. In addition to improving land use efficiency and safeguarding cropland areas, the ‘occupying and replenishing balance’ promotes the optimal allocation and protection of cropland resources at the macro level. Moreover, some upper and middle regions have established incentive mechanisms for cropland protection and ecological protection compensation systems¹¹¹, contributing to maintaining regional cropland stability. Marketisation and labour mobility levels influence structural changes in cropland use. The intensity of cropland development is higher in downstream areas with greater marketisation. In contrast, in some mountainous regions in the upper and middle areas, phenomena such as ‘non-food’ production, cropland abandonment, and desertification are more pronounced due to the pursuit of economic income and labour loss¹¹². Additionally, the smart stations for agricultural facilities and equipment and upgrades in farming techniques have improved cropland utilisation efficiency¹¹³. However, they may also lead to repurposing some cropland for other land uses.

The sustainable use of cropland: problems and practical challenges

Significant regional variability exists in cropland cross-conversion types, scales, and locations across different scenarios¹¹⁴. The study shows that, in the future, problems and conflicts affecting the sustainable use of cropland in the YREB will continue to be concentrated between agricultural space, urban space, and agricultural space and ecological space. First, fluctuations in cropland change threaten food supply capacity and national food security. Although some mountainous and riverine areas in the CPS have potential for cropland development, these regions are also critical ecological conservation zones with fragile and difficult-to-repair environments. Second, although long-term and large-scale ecological restoration measures have been implemented, the scale of restoration in the upper regions of ecologically fragile areas remains limited⁹⁴. The protection of lakes and wetlands in the middle and lower parts of the CPS still needs strengthening, and the construction of the overall ecological barrier is lagging¹¹⁵. Additionally, climate has always had a significant impact on the productivity and suitability of cropland. According to the Statistical Yearbook of YREB-2021, from 1991 to 2000, the likelihood of converting land to cropland in the 0–250 mm rainfall interval was higher, reflecting that the maintenance and utilisation of cropland remain under greater pressure in areas experiencing drought or insufficient rainfall.

Cropland expansion in this study was derived from grasslands, forests, and water. Using correlation method¹⁰¹, we calculated that YREB watersheds consistently exhibited a correlation of p > 0.1 with cropland change (Table 4), emphasising the critical role of water in sustainable cropland management within watershed economic belts. On one hand, direct use and pollution of water bodies may be reduced to meet the diverse needs of land use. On the other hand, this could harm wetland resources and ecosystems. Grasslands are typically found in wetter or poorly drained areas, while cropland requires fertile soil and suitable irrigation conditions. Therefore, although grassland→cropland conversion shows a systematic inclination in the CPS, the conversion from upper region grassland→cropland faces systematic inhibition, making it difficult to transform the natural environment. Moreover, under China’s strict land use control system, grassland or forest→cropland conversion is restricted to the areas designated in planning. Additionally, it must comply with national laws and regulations on land management, environmental protection and agricultural production. The destruction of the crop layer and the associated socio-economic costs must also be considered¹¹⁶.

Table 4.

Importance of the YREB water for cropland Change.

Type/Year		2000		2010		2020
		Cropland	Water	Cropland	Water	Cropland	Water
Cropland	Pearson	1	−1.70E-04	1	−3.29E-04	1	−2.85E-04
	p	--	0.96794	--	0.93725	--	0.94486
Water	Pearson	−1.70E-04	1	−3.29E-04	1	−2.85E-04	1
	p	0.96794	--	0.93725	--	0.94486	--

*Two-tailed significant tests were used.

Sustainable cropland management recommendations for YREB

Addressing the core contradictions between cropland utilization and regional natural conditions, we propose strategies for agricultural development and measures to ensure sustainable cropland utilization (Fig. 16). Specifically, our recommendations include the following:

Fig. 16.

Adaptive cropland development recommendations for diverse regions.

Combining the centre-of-mass trajectories with scenario outputs, for cropland loss, optimize land-use planning and urbanization strategies, and clearly delineate city expansion boundaries to prevent the disorderly occupation of cropland. Scientific planning and renovation of urban areas are essential for high-growth regions such as Chengdu, Anhui, and northern Jiangsu. Optimising urban layouts, identifying low-value land, and enforcing the “occupation-compensation balance” policy are critical in slower-urbanising regions like Yunnan and Jiangxi.

ECS simulations reveal that clusters of cropland-to-forest conversion are concentrated in the mountainous upper reaches of watersheds. This spatial overlap corroborates the dual imperative of protecting upper-reaches water sources while sustaining investment in high-standard cropland within the remaining concentrated plains. Ecologically fragile regions should therefore implement enhanced restoration programmes and establish stringent policy safeguards. Along the Yangtze River corridor, anthropogenic disturbance must be curtailed and the construction of ecological security barriers actively promoted. upper-reaches water source protection areas, middle-reaches ecological corridors, riparian wetlands, and lake regions require the establishment of scientific monitoring systems and systematic maintenance mechanisms to halt ecological degradation at source.

Improve cropland quality through targeted interventions. In the concentrated plains, expand ‘High-Standard Farmland Construction’ and ‘Comprehensive Land Improvement’ projects to upgrade soil quality and agricultural infrastructure. Concurrently, refine the policy framework for reclaiming abandoned rural homesteads, appropriate land-renovation measures should be encouraged to convert these parcels into well-functioning cropland, thereby facilitating a healthy, spatially balanced expansion of cropland area.

Limitations and future research

This study employs the GeoSOS-FLUS zoning model to predict land use trends in the YREB. However, the model encounters challenges in quantifying and spatially representing human factors like policy and technological advancements. For instance, the NV indicator, used as a proxy for population and GDP, demonstrated limited interpretability. Additionally, this study primarily examines cropland change at the zonal level, with a limited focus on micro-scale analyses. Potential interactions between drivers and indirect mechanisms influencing cropland change should be explored in future research to enhance understanding and support more targeted policy recommendations.

Conclusions

This study analyzes the patterns, characteristics, drivers, and trends of cropland evolution in the YREB over the past 20 years, based on cropland use intensity and stability, employing graphical and regression analyses along with models like gravity shift and FLUS. The study also proposes regionally differentiated measures for optimizing cropland use. The findings indicate that cropland→BL and water→cropland are the dominant and stable conversion modes in the YREB. Water is critical in maintaining stability and balancing cropland development with ecological protection. Variations in geographic resources and economic development activities have driven the cropland’s centre of gravity to shift eastward and northward. During this process, transport network construction and average annual rainfall have had significant and sustained impacts on cropland changes in the upper and lower regions, respectively. Multi-scenario results reveal that the cropland’s centre of gravity in the upper and middle regions tends to migrate to southern mountainous areas, while in the lower regions, it shifts toward northern plains. In the NDS, cropland→BL is prevalent, accompanied by notable conflicts between agricultural space and the urban space. In contrast, the CPS and ECS scenarios exhibit more complex and diverse cropland transformations involving forest, grassland, and water. Agricultural space and the urban space experienced intense conflicts with a significant spatial overlap in the lower region of YREB. The primary conflict in cropland development remains between urban expansion and ecological protection. Thus, scientific urban planning and enhanced ecological restoration are essential for achieving sustainable cropland development and balancing economic growth with ecological conservation.

Author contributions

M. W. conceptualization, reviewing, editing, and writing. S. H. Methodology and Supervision. M. S. Writing – review & editing. Y. Z. original draft, Conceptualization. D.S.T. Formal analysis, Data curation.

Funding

This study was supported by the Key Project of the National Social Science Foundation of China (Grant No.18ZDA053).

Data availability

All data generated or analysed during this study are included in this published article. The data cannot be publicly shared but are available upon request to the corresponding author, SG H.

Declarations

Competing interests

The authors declare no competing interests.