Song et al. (2025) Analysis of cultivated land changes and driving factors in the Alar Reclamation Area (1990–2019) based on multi-temporal Landsat data and machine learning algorithms

Identification

• Journal: Scientific Reports
• Year: 2025
• Date: 2025-12-06
• Authors: Qi Song, Wanming Zhang
• DOI: 10.1038/s41598-025-29175-z

Research Groups

• School of Environment and Resource, Xichang University, Xichang 615013, China

Short Summary

This study analyzed the spatiotemporal dynamics of cultivated land and its driving factors in the Alar Reclamation Area, southern Xinjiang, China, from 1990 to 2019 using multi-temporal Landsat data and machine learning. It found a significant increase of 729.97 km² in cultivated land, primarily driven by anthropogenic factors such as population growth, agricultural output value, and fixed asset investment, with GDP showing a negative direct effect.

Objective

• To analyze the dynamic changes in cultivated land area and its spatial distribution in the Alar Reclamation Area between 1990 and 2019.
• To compare and identify the optimal machine learning algorithm for land use/cover classification in an arid oasis environment.
• To investigate and quantify the direct and indirect effects of socio-economic and natural driving factors on cultivated land change using path analysis.
• To provide a scientific basis for cultivated land monitoring, rational utilization, and sustainable land management in arid oasis regions.

Study Configuration

• Spatial Scale: Alar Reclamation Area, southern Xinjiang, China (80°30′–81°58′E, 40°22′–40°57′N), covering a total area of 4,197.58 km².
• Temporal Scale: 1990–2019, with data from seven key years (1990, 1994, 2000, 2006, 2010, 2015, and 2019).

Methodology and Data

• Models used:
  • Classification Algorithms: Spectral Angle Mapper (SAM), Artificial Neural Network (ANN), Minimum Distance Classification (MDC), Maximum Likelihood Classification (MLC), Support Vector Machine (SVM), Support Vector Machine–Conditional Random Field (SVM–CRF).
  • Atmospheric Correction: LEDAPS (Landsat 5/7), LaSRC (Landsat 8).
  • Cloud/Shadow Masking: Fmask 4.0.
  • Statistical Analysis: Path analysis, correlation coefficient matrix, Variance Inflation Factor (VIF), Durbin-Watson statistic, McNemar’s test, bootstrapping (n=1000).
• Data sources:
  • Remote Sensing: Multi-temporal Landsat satellite images (Landsat 5 TM, Landsat 7 ETM+, Landsat 8 OLI/TIRS) from the U.S. Geological Survey (USGS).
  • Socio-economic Data: Statistical Yearbook of the Xinjiang Production and Construction Corps (1990–2019) for the First Division, including total population, non-agricultural population, GDP, total fixed asset investment, primary industry, total agricultural output value, and cotton price.
  • Meteorological Data: Monthly Surface Meteorological Records for the Alar meteorological station (station number 51730), including mean annual temperature and mean annual precipitation.
  • Validation Data: Visually interpreted regions based on field observations and Google Earth imagery, with 119,075 independent validation samples.

Main Results

• The SVM–CRF algorithm achieved the highest land use/cover classification accuracy with an Overall Accuracy (OA) of 0.95 and a Kappa coefficient of 0.94, significantly outperforming other tested methods.
• Cultivated land area in the Alar Reclamation Area increased by 729.97 km² (312.21%) from 233.81 km² in 1990 to 963.76 km² in 2019, showing a continuous outward expansion trend.
• The most rapid cultivated land expansion occurred in the early 1990s, with growth continuing steadily through the 2000s and moderately after 2010.
• Cultivated land expansion primarily occurred in the southeastern and northwestern parts of the area, mainly through the reclamation of unused land and saline–alkali land.
• Reductions in cultivated land were observed mainly along the Tarim River, where it was converted into orchards and forest–grassland.
• Path analysis identified total population, GDP, total fixed asset investment, total agricultural output value, and cotton price as the main anthropogenic driving factors influencing cultivated land changes.
• GDP exhibited the largest direct negative effect on cultivated land area (path coefficient = –1.51), indicating that economic growth is often accompanied by a reduction in cultivated land due to industrialization and urban expansion.
• Total fixed asset investment (1.12), total agricultural output value (0.63), total population (0.57), and cotton price (0.26) showed positive direct effects on cultivated land area.
• The cumulative effect of these driving factors was positive, leading to a net increase in cultivated land area over the three decades.

Contributions

• Systematically compared multiple machine learning algorithms for land use/cover classification in an arid oasis environment, demonstrating the superior accuracy of the SVM–CRF approach (OA = 0.95, Kappa = 0.94) in handling spectral confusion and mixed pixels.
• Integrated multi-temporal Landsat imagery with both socio-economic and natural driving factors using path analysis to quantify their direct and indirect effects on cultivated land change, a methodological framework rarely applied to oasis reclamation areas in southern Xinjiang.
• Provided new insights into human–land interactions under extreme arid conditions, highlighting the dominant role of anthropogenic factors in cultivated land expansion in the Alar Reclamation Area.
• Offered a scientific basis for cultivated land monitoring, rational utilization, and sustainable land management strategies in the Alar area and similar arid regions in northwest China.

Funding

• Key Project of Sichuan Provincial Department of Natural Resources, grant number “ZRS202001.”

Citation

@article{Song2025Analysis,
  author = {Song, Qi and Zhang, Wanming},
  title = {Analysis of cultivated land changes and driving factors in the Alar Reclamation Area (1990–2019) based on multi-temporal Landsat data and machine learning algorithms},
  journal = {Scientific Reports},
  year = {2025},
  doi = {10.1038/s41598-025-29175-z},
  url = {https://doi.org/10.1038/s41598-025-29175-z}
}