Sentinel-2-Based monitoring of abandoned croplands and spatial heterogeneity analysis in arid Northwest China: a Case study of the Hexi Corridor, Gansu Province
Мониторинг заброшенных сельскохозяйственных угодий на основе данных Sentinel-2 и анализ пространственной неоднородности в аридном северо-западном Китае: тематическое исследование Хэсийского коридора, провинция Ганьсу
2026-02-03
SCID: 54.1/fck2fvnw
Discuss with AI
Hexi CorridorRandom ForestSentinel-2cropland abandonmentspatial heterogeneity
Figures from the paper
Abstract (AI)
Cropland abandonment poses a significant threat to global food security, especially in ecologically fragile arid regions, where it compromises both agricultural sustainability and ecosystem integrity. Despite its importance, the spatial distribution of abandoned croplands across large geographic areas remains poorly understood. This study focuses on the Hexi Corridor, a representative oasis agricultural zone in northwestern China. Combining multi-dimensional Sentinel-2 features (phenological, spectral, textural, and physiographic) with feature optimisation, we employed a Random Forest-based post-classification change detection approach to map abandoned croplands from 2021 to 2023, achieving high accuracy (overall accuracy: 92.08%; Kappa: 0.84). (1) Landscape pattern indices and kernel density analysis revealed key spatial trends: abandonment rates declined from 7.05% in 2022 to 5.38% in 2023, while persistent abandonment covered 11,170 ha and recultivation reached 34,403 ha. (2) Spatial heterogeneity was pronounced, with core abandonment zones concentrated in Jiuquan (Yumen, Guazhou) and Zhangye (Shandan County), whereas Jinchang (Yongchang County) and Wuwei (Minqin County) underwent significant rehabilitation. (3) Abandoned croplands exhibited increased fragmentation, driven by a 21% rise in patch density, whereas recultivation was characterized by smaller, scattered plots (patch density > 109), indicating a shift from contiguous to fragmented cultivation. Kernel density analysis revealed an ‘oasis-edge aggregation’ pattern in abandonment, characterized by high-density clusters in Shandan County following remediation efforts and a distinct recultivation hotspot in Yongchang County. Our multi-feature RF approach establishes a robust framework for large-scale abandonment monitoring and delivers critical insights to support sustainable land management in arid regions.
Key Findings
1
A multi-feature Sentinel-2 and Random Forest post-classification approach mapped abandoned croplands with 92.08% overall accuracy and a Kappa of 0.84.
2
Abandoned croplands became more fragmented, with patch density increasing by 21%; recultivated areas formed smaller, scattered plots with patch density above 109.
3
Abandonment followed an oasis-edge aggregation pattern, including high-density clusters in Shandan and a distinct recultivation hotspot in Yongchang.
4
Abandonment rates declined from 7.05% in 2022 to 5.38% in 2023; persistent abandonment covered 11,170 ha, while recultivation reached 34,403 ha.
5
Core abandonment zones occurred in Yumen and Guazhou in Jiuquan and Shandan County in Zhangye, whereas Yongchang and Minqin showed substantial rehabilitation.
Research Object
Abandoned croplands in the Hexi Corridor oasis agricultural zone of Gansu Province, Northwest China
Research Subject
The spatial distribution, temporal dynamics, landscape fragmentation, and spatial heterogeneity of cropland abandonment and recultivation from 2021 to 2023
Publication Details
Publication Date
2026-02-03
Journal
International Journal of Remote Sensing
Publisher
Taylor & Francis
ISSN
0143-1161
Cited by
0
Access Type
Author Information
Download PDF
Subscribe to digest