Abstract: Non-Photosynthetic Vegetation (NPV) is not only an important surface cover component during the non-growing season in arid and semi-arid regions but also a key barrier for windbreak and sand fixation in winter and spring. Due to its spectral similarity to bare soil, traditional remote sensing methods have great difficulty in accurately identifying and monitoring NPV at large scales. To address the monitoring challenges caused by fragmented surface landscapes and severe spectral confusion between NPV and bare soil in such regions, this study takes Wuchuan County, located in the agro-pastoral ecotone of northern China, as the study area. A synergistic inversion framework integrating Unmanned Aerial Vehicle (UAV) imagery and Sentinel-2 satellite data was established to carry out NPV inversion during the non-growing season.High-resolution UAV images were first acquired and utilized to screen for optimal classification features. Through comparative analysis, the Blue-to-Near-Infrared ratio index (B/NIR) was identified as the most effective feature for distinguishing NPV from bare soil. Based on this optimal feature, a surface component classification model was constructed using the Support Vector Machine (SVM) algorithm. The model achieved an overall classification accuracy of 91.7%, thereby enabling high-precision extraction of ground truth information for subsequent satellite-based inversion. This step effectively overcame the limitations of traditional ground survey methods in spatially heterogeneous landscapes and provided reliable training and validation samples for large-scale monitoring. Subsequently, the Dead Fuel Index (DFI), which is particularly sensitive to the spectral response of litter and senescent vegetation, was selected from multiple candidate spectral indices to construct a linear regression inversion model for NPV coverage estimation. The model was independently validated using spring field data that were not involved in model training. The validation results demonstrated that the model achieved an inversion accuracy of 77.39% with a Mean Squared Error (MSE) of 0.009250, indicating a satisfactory level of temporal generalization ability. This suggests that the DFI-based linear regression model, calibrated with UAV-derived ground truth, can be effectively transferred to satellite image time series for NPV monitoring across different periods within the non-growing season.Based on the long-term inversion results derived from Sentinel-2 time series data, this study further revealed a distinct seasonal evolution pattern of NPV coverage in the study area. Specifically, NPV coverage exhibited a pattern of "significant accumulation in autumn, relative maintenance in winter, and significant loss in spring." During autumn, the senescence and litterfall of herbaceous and woody vegetation lead to a rapid increase in NPV coverage across various land cover types. Throughout the winter period, NPV coverage remains relatively stable, serving as a persistent surface protection layer against wind erosion. In spring, however, NPV coverage declines markedly due to factors such as wind removal, decomposition, and agricultural tillage activities, which substantially weakens the windbreak and sand fixation function of the surface at a time when strong winds are frequent. The study also found notable differences in NPV dynamics among different land cover types. NPV coverage in forest and grassland areas exhibited large interannual fluctuations, which were closely linked to climate variability and its impact on vegetation growth and litter production in preceding growing seasons. In contrast, cropland areas maintained a persistently low level of NPV coverage throughout the long-term observation period. This consistently low coverage makes cropland a weak link in windbreak and sand fixation during the winter and spring seasons, highlighting the need for targeted conservation measures such as conservation tillage, stubble retention, or cover cropping to enhance surface protection in agricultural areas during the non-growing period.In summary, the "UAV-satellite" collaborative inversion strategy proposed in this study effectively addresses the spectral confusion problem between NPV and bare soil, enables accurate large-scale NPV monitoring in fragmented landscapes, and provides valuable insights into the seasonal and interannual dynamics of surface protective cover. This methodological framework and the findings can provide a scientific basis for the construction of regional ecological security barriers and the implementation of precise supervision and management practices in arid and semi-arid agro-pastoral ecotones.