Monitoring abandoned cropland in the hilly and gully regions of the Loess Plateau using Landsat time series images

Cropland abandonment has become a global issue that poses significant threats to sustainable cropland management, national food security, and the ecological environment. Remote sensing technology is crucial for identifying and monitoring abandoned cropland in large-scale areas. However, limited information is available on the effective identification methods and spatial distribution patterns of abandoned cropland in the hilly and gully regions. This study introduced two methods - the land-use trajectory and normalized difference vegetation index (NDVI) time series - for monitoring abandoned cropland and evaluating its spatial distribution in Yanhe River Basin using Landsat-8 images from 2019 to 2021. The results showed that using a random forest algorithm, high-precision annual land-use classifications were achieved with the generation of reliable land-cover samples and an optimized feature dataset. The overall accuracy (OA) and Kappa coefficient of the land-use maps exceeded 90% and 0.88, respectively, demonstrating the effectiveness of the classification over three years. These two distinct change detection methods were used to identify abandoned cropland in the study area, and their accuracy and effectiveness were evaluated. The land-use trajectory method performed better than the NDVI time series method for extracting abandoned cropland, with an OA of 83.5% and an F1 score of 84.7%. According to the land-use trajectory detection results, the study area had 164.6 km² of abandoned cropland area in 2021, with an abandonment rate of 16.3%. Furthermore, cropland abandonment mainly occurred in the northwestern part of the region, which has harsh natural conditions, while abandonment was rare in the southern and eastern regions. Topography and landforms significantly influenced the spatial distribution of abandoned cropland, with most abandoned cropland located in mountainous regions with higher elevations and steeper slopes. The abandonment rate generally increased with the elevation and slope. These findings provide valuable references and guidance for selecting appropriate methods to identify abandoned cropland and analyze its spatial distribution in the hilly and gully regions. Our proposed methods offer robust solutions for monitoring abandoned cropland and optimizing land-use change detection in similar regions with complex landforms.