Abstract: Analysis of gully characteristics and their spatial differentiation across scales provides scientific support for gully erosion control and watershed management in mountainous regions. This study, conducted in a small watershed of the hilly northern Dabie Mountains, employed multi-source data including 5m DEM, 2m Gaofen imagery, and field surveys. Through integrated application of hydrological analysis, ROC curve-Youden’s index method, and spatial statistics, we systematically examined key gully characteristics (gully level, length, gradient, and relief) and their spatial patterns across small-watershed and micro-watershed scales. Results show that: (1) The optimal flow accumulation threshold for gully extraction in the study area is 1.26 hm², extracting a total of 227 gullies with an overall accuracy of 87.22%. The RMSE of gully origins is 24.69m, cosine similarity is 0.9995, and the gully overlay deviation is 1.83%. (2) The number of gullies decreases sharply with increasing gully level, dominated by first-level gullies (80.62%), with an average elevation difference of 6.89m. Gully level shows a significant positive correlation with gully length (r=0.683, p<0.01) and significant negative correlations with gradient (r=-0.358, p<0.01) and gully relief (r=-0.358, p<0.01). Gully relief is significantly negatively correlated with gully length (r=-0.725, p<0.01). (3) Micro-watersheds are predominantly of the complete and interval types (97.12%), with gully densities ranging from 17.13 to 270.17 m/hm². The maximum gully level shows a significant negative correlation with gradient (r=-0.567, p<0.01) and a significant positive correlation with gully density (r=0.340, p<0.01). Spatially, the maximum gully level, mean gradient, and mean relief exhibit significant positive spatial autocorrelation (Moran’s I ≥ 0.135, p<0.01), while gully density shows weak spatial autocorrelation, indicating a random distribution pattern.