Abstract: Unplanted cropland, as an important manifestation of uncultivated farmland, exerts complex influences on food production and ecological processes through its quantity, spatial pattern, occurrence frequency, and utilization mode. However, the spatial–temporal characteristics of unplanted cropland and its implications for cropland conditions remain insufficiently understood, particularly in areas characterized by a single-cropping system and strong inter-annual fluctuations in agricultural production.Taking the four Bashang counties of Zhangjiakou City as a case study, this study employed Sentinel-2A imagery in combination with linear spectral mixture analysis and intra-annual CPSEM model to identify unplanted cropland at the pixel scale and delineate its occurrence frequency and temporal trajectories. Linear spectral mixture analysis was used to characterize the subpixel composition of complex land surfaces, while the intra-annual CPSEM model was introduced to capture the seasonal variation in GV and SL components. These spectral and temporal features were further integrated with a random forest classifier to produce annual land-cover maps and identify unplanted cropland. Based on the annual identification results, the frequency and trajectories of unplanting were quantitatively characterized.Furthermore, a Kohonen Self-Organizing Map (SOM) was applied to classify cropland states, and the responses of different unplanting trajectories to soil organic matter (SOM) were analyzed.The results indicated that: 1) the annual land -cover classification accuracy from 2019 to 2024 exceeded 87%, providing reliable support for unplanted cropland identification; 2) the area of unplanted cropland exhibited an increasing-then-decreasing trend, reaching a maximum of 986.34 km² in 2023 and a minimum of 241.78 km² in 2020. Spatially, it showed a pattern of“central aggregation with outward expansion”with Zhangbei County as the core concentration area; 3) during the five-year study period, a total of 1,768.95 km² of cropland experienced unplanted conditions, among which a frequency of one occurrence was the most common, accounting for 50.45% of the total unplanted area, while cropland with multiple unplanted occurrences was mainly concentrated in Guyuan and Zhangbei counties; 4) at the regional scale, cropland was primarily dominated by continuous cultivation and disordered unplanted cropland trajectories. Significant inter-county differences were observed: Guyuan County was mainly characterized by disordered and late-stage unplanted cropland, whereas the other three counties were dominated by disordered and regular unplanted cropland trajectories; 5) the relationship between cropland state and unplanted cropland frequency exhibited a clear nonlinear pattern. A moderate frequency corresponded to higher cropland quality, whereas both lower and higher frequencies were associated with lower state levels. Regular and disordered unplanted cropland trajectories showed continuous SOM accumulation, whereas continuous unplanted cropland trajectories suppressed SOM accumulation, indicating that periodic unplanted conditions were associated with improved cropland conditions. This study provides a scientific basis for understanding the spatial and temporal dynamics of unplanted cropland and optimizing protective fallow–cultivation practices. By considering unplanting frequency and temporal trajectories, the proposed framework supports the optimization of adaptive fallow–cultivation management strategies, including decisions concerning where to fallow, how to fallow, and how long to fallow. It thereby provides practical insights for the protective utilization and sustainable management of cropland and cropland planning oriented toward food security.