Detecting cropland abandonment and intentional afforestation using MODIS and landsat across agroclimatic zones

Intentional afforestation on abandoned croplands represents a viable strategy for terrestrial carbon sequestration without competing with food production. However, few studies have addressed how to distinguish intentional afforestation from spectrally similar processes such as natural succession following cropland abandonment or fallow. Focusing on large-scale environmental engineering projects under China's Grain-to-Green initiative, this study employed a fused Moderate Resolution Imaging Spectroradiometer (MODIS)-Landsat normalized difference vegetation index (NDVI) dataset, generated using the Enhanced Spatio-temporal Adaptive Reflectance Fusion Model (ESTARFM). Vegetation change rates during key crop growth stages were analyzed in combination with the coefficient of variation (CV) method to identify cropland abandonment hotspots. Subsequently, the Landsat-based trend detection of disturbance and recovery (LandTrendr) algorithm was applied to the 2000–2022 NDVI time series to detect disturbance and recovery patterns and identify afforestation turning points and greening trends within these hotspots. The results show that integrating NDVI trajectories with regional phenological parameters effectively differentiates intentionally afforested areas from natural succession on abandoned croplands, achieving an overall accuracy of 90.5 % (95 % CI: 86.4 %–94.6 %). This study demonstrates that NDVI trend analysis serves as a robust indicator for identifying both intentional afforestation and natural succession on abandoned croplands, providing an effective monitoring framework and valuable data for evaluating regional ecological restoration efforts. • A gradual increase in vegetation cover served as a key indicator of cropland-to-forest conversion and provided a reliable basis for distinguishing intentional afforestation. • Afforestation on abandoned cropland was detected based on interannual variations in vegetation phenological information. • The proposed method mitigated the effects of landscape heterogeneity on afforestation mapping accuracy. • The temporal and spatial patterns of afforestation on abandoned cropland were revealed across different agroclimatic zones.