Soil aggregates reflect the stability of soil structure, soil fertility and quality, and they are closely related to the environmental quality, soil and water loss, and soil erosion. Although vegetation types, soil depths, and altitude likely have important effects on soil properties and soil aggregates, few studies have concentrated on the relationships between their interactive effects and soil aggregate stability on the regional scale. In order to study the variation characteristics of soil aggregates in the Qilian Mountains. In this study, 4 vegetation type zones (desert, steppe, meadow, and shrub) were selected in the Qilian Mountains (northwest China) as the subjects. The stability and size distribution of soil aggregates was measured by the method of wet-sieving. The characteristics of soil aggregate index were analyzed at different soil depths and different altitudes, including the percentage of water-stable aggregate (WSA), mean weight diameter (MWD), geometric mean diameter (GMD), mean weight soil specific area (MWSSA), aggregate processing damage rate ( PAD) and fractal dimension (D). The results showed that the coefficients of variation of soil aggregate index under the 4 vegetation types were high (ranging from 10.91% to 62.50%), which indicated a moderate spatial variability. Among the 4 grassland types, WSA, MWD, GMD and MWSSA all showed the same increased in order: desert < steppe or meadow 30-40 cm depth, no significant differences in aggregate stability index were found among different vegetation types (P>0.05). The reason why the aggregate stability of desert was lowest could be explained by the fact that the lower soil organic carbon (SOC) and biomass of desert, and the small aggregates couldn’t be formed. Along the vertical direction of the soil profiles from aboveground to underground, with the increase of soil depth (0-40 cm), WSA, MWD, GMD, and MWSSA gradually all decreased. Whereas PAD and D behaved opposite to them, and with the increase in soil depth, they gradually increased, which indicated that the stability and aggregation degree of soil aggregates decreased with soil depth. This could be explained as that the SOC appeared enrichment phenomenon in the 0-20 cm layer and decreased with soil depth. The soil aggregate index all tended to be significantly higher in surface layer than in lower layers only at the meadow (P<0.05), but no significant difference in MWSSA (P>0.05) were found among all soil depths. With increasing altitude, the aggregate stability (WSA, MWD, and GMD) increased gradually from 1 692 m, reached a peak at 2 800 m, and then decreased quickly. But the trends were reverse for the PAD and D, and there was no significant correlation between MWSSA and altitude under different vegetation types (P>0.05), which suggested that their distribution clearly showed unimodal patterns across all the altitude. This indicated that the stability and aggregation degree of soil aggregates first increased and then decreased with increasing altitude, and the maximum point appeared at elevation of 2 800 m. In addition, correlation analysis showed that WSA, MWD, and GMD were mainly affected by 1-4 mm, D was mainly affected by 0.038-0.25 mm, and MWSSA couldn't accurately express the characteristics of water-stable aggregates. This research will guide the practice of reducing soil erosion for the different conditions and different vegetation types, and results have great significance for controlling grassland degradation, promoting soil structure stability and the sustainable development of grassland animal husbandry. [ABSTRACT FROM AUTHOR]