TL;DR
This study investigates cropland abandonment (CA) in the rapidly urbanizing Chengdu metropolitan suburbs, revealing an average annual abandonment rate of 8% with significant spatial clustering. Surprisingly, proximity to roads, rather than increasing abandonment, was linked to higher rates, suggesting policy-driven land speculation as a key factor.
What Problem Does This Paper Address?
Global food security is increasingly threatened by factors like climate change, land degradation, and, crucially, cropland abandonment (CA). While CA is a growing concern worldwide, its dynamics in rapidly urbanizing metropolitan suburbs, where urban expansion and agricultural land preservation clash, are poorly understood. These areas face unique pressures from development, making it vital to identify where and why farmland is being abandoned to inform sustainable land management and urban planning. This paper addresses this gap by focusing on the Chengdu Directly Administered Zone (CDAZ) in China, a rapidly developing new area, to understand the spatiotemporal patterns of CA and its driving factors.
Methods
The researchers employed a multi-pronged approach:
- Land Use Change Trajectory Analysis: Using high-precision vector data from China's National Land Survey (2019-2023), they identified abandoned croplands by tracking land use changes over time. Abandonment was defined as land left uncultivated for two consecutive years and converted to grassland, forest, or bare land.
- Exploratory Spatial Data Analysis (ESDA): This involved using global and local Moran's I indices to assess spatial autocorrelation, revealing whether abandoned croplands were clustered or dispersed.
- Kernel Density Estimation (KDE): KDE was used to visualize the spatial density and distribution of CA hotspots, highlighting areas with high concentrations of abandoned land.
- Geographically Weighted Regression (GWR): This advanced regression technique was crucial for understanding the spatial heterogeneity of driving factors. Unlike traditional regression, GWR allows coefficients to vary across space, capturing how the influence of different factors changes from one location to another.
Datasets
The study utilized several datasets:
- Land Use Data (2019-2023): High-resolution vector data from China's National Land Survey, providing parcel-level information.
- Cropland Quality Data (2021): Vector data classifying cropland quality based on various factors.
- Road Networks (2021): Vector data of road infrastructure.
- Digital Elevation Model (DEM) (2009): Raster data for topographic analysis.
- Gross Domestic Product (GDP) Spatial Distribution (2020): Raster data indicating economic activity.
- Demographic Data (2020): Raster data on population distribution.
- Subdistrict Administrative Centers (2019): Vector data for locational analysis.
Key Insights and Ideas
- Urbanization's Unique Impact: The study highlights that CA in metropolitan suburbs operates differently than in traditional rural or mountainous areas. The "push-pull" dynamics are more about urban encroachment and land speculation rather than solely labor migration.
- Policy-Driven Abandonment: The most surprising finding is the negative correlation between road proximity and abandonment. This suggests that anticipated land expropriation and development potential along transportation corridors, rather than accessibility, drive abandonment.
- High-Quality Land at Risk: Even high-quality croplands are being abandoned, particularly in peri-urban areas facing intense development pressure. This points to a conflict between economic development incentives and agricultural sustainability.
- Spatial Heterogeneity is Key: The GWR model's success underscores that driving factors for CA are not uniform. Understanding these localized variations is essential for targeted policy interventions.
Key Results
- Average Abandonment Rate: Approximately 8% of cropland was abandoned annually between 2021 and 2023.
- Spatial Clustering: Abandonment was not random but showed significant spatial clustering, with hotspots concentrated in peri-urban areas.
- Key Drivers:
- Restraining Factors: Cropland quality (CQ) and aggregation index (AI) were found to reduce abandonment risk.
- Elevating Factors: Increasing slope (SLO) and land development intensity (LDI) increased abandonment risk.
- Counter-intuitive Factor: Distance to roads (DisR) showed a negative effect, meaning closer proximity to roads was associated with higher abandonment.
- GWR Model Performance: The GWR model demonstrated excellent fit (Adjusted R² = 0.743), outperforming traditional OLS regression in explaining the spatial variability of CA.
Limitations and Future Work
- Temporal Scope: The study's timeframe (2019-2023) is relatively short, limiting insights into long-term trends. While the data is the most up-to-date available, comparing it with older land surveys could introduce biases due to methodological differences.
- Natural Succession vs. Afforestation: The detected cropland-to-forest conversions might be more indicative of policy-driven afforestation than spontaneous natural succession, especially within the short study period.
- Underexplored Drivers: Institutional drivers (e.g., land transfer policies, subsidies) and household characteristics (e.g., age, income) were not deeply explored.
- Factor Interactions: The GWR model does not capture interactions between explanatory variables.
- Future Research Directions:
- Extend the time series analysis and incorporate high-resolution remote sensing and UAV data.
- Integrate household surveys and policy analysis to understand institutional and household-level drivers.
- Explore machine learning approaches to capture complex factor relationships and temporal dynamics.
- Prioritize research on optimal reuse strategies for abandoned croplands, considering regional contexts and dominant drivers.
Summary updated