The arid and semi-arid regions of the Eurasian continent (the Eurasian arid region) are home to extensive shrublands and these shrublands are critical components of the regional biomes. However, identifying shrublands and distinguishing them from grasslands via remote sensing remains challenging due to their spectral and phenological similarity, as evidenced by the varying distribution areas and patterns of shrublands across each common land cover product. To address this, we combined a multi-strategy sampling framework with a phenology-pixel-based method to map shrubland distribution in the arid Eurasian region. We achieved a mean F1 score of 67.05%, notably higher than that of existing global land cover datasets. The total mapped shrubland area is approximately 427,643 km2, exhibiting a clear latitudinal peak (30 degrees-50 degrees N) and multimodal longitudinal distribution pattern. In different biomes, climatic conditions-moisture or temperature-regulate the optimal phenological windows by driving functional trait divergence between shrubland and grassland. Background shrubland cover strongly determines identification accuracy. When shrubland coverage exceeds 60%, mapping accuracy shifts from "detectable" to "reliable". Below this threshold, classification accuracy is limited by phenological noise and topography interference, whereas above it, canopy continuity enhances spectral separability and reduces terrain-related noise. This study demonstrates that shrub coverage, phenological timing, and terrain collectively regulate mapping accuracy. The phenology-optimized classification framework significantly outperforms existing common shrubland distribution datasets, especially in complex microtopography and transition zones. Improved shrubland mapping is critical for evaluating ecosystem health and investigating carbon cycling in arid regions.