Gully morphology and area–volume relationships across gully types under different topographic conditions in the Mollisol region of Northeast China

Gully morphology and area–volume relationships vary with geomorphic position and environmental setting, yet how they differ across gully types and topographic conditions remains poorly resolved. We investigated 110 gullies in three agricultural catchments with different topographic conditions in the Mollisol region of Northeast China using UAV imagery, field cross-sections, contributing-area topographic analysis, bootstrap resampling, repeated five-fold cross-validation, ANCOVA, and nested-model comparisons. Valley-bottom gullies (BGs) had morphometric values approximately 1.8–11.2 times those of hillslope gullies (HGs), with contributing area and relative relief approximately 11.2 and 1.9 times those of HGs, respectively. Although the three catchments had different topographic conditions, only gully depth showed a significant catchment-level morphometric difference. Planimetric area was most strongly correlated with volume across the full dataset, both gully types, and all three catchments. The pooled model, V = 0.1445A 1.2326 , had R 2 = 0.873, while repeated five-fold cross-validation yielded R 2 = 0.868 ± 0.002. V–A slopes did not differ significantly between HGs and BGs, and type-specific modelling did not significantly improve the pooled model, although the estimated volume of BGs was approximately 1.53 times that of HGs at the same planimetric area. In contrast, catchment-specific models showed high goodness of fit (R 2 = 0.903–0.931) and significantly improved the pooled model ( P < 0.001); the V–A slopes also differed significantly among catchments overall. Thus, gully-type differences were expressed primarily in morphometric scale and in volume level at a given planimetric area, whereas V–A scaling showed clearer catchment dependence. The pooled model supports rapid volume estimation without local cross-sectional data, while catchment-scale calibration can improve local applicability when sufficient field data are available.