Soil coring was used to obtain live fine roots (D≤2 mm, D is root diameter) sampled from each site taken randomly in the northern part of the Development and Management Research Center in the Mu Us Sandland. Root samples were collected five times from April to October of 2005 using 8cm diameter cores. Roots were separated into D≤1 mm and 1 mm＜D≤2 mm diameter classes and dry weights were determined. A well-established asymptotic equation Y=1-βd \[β= root persistence coefficient, d = soil depth (cm), and Y = percentage of total root weight from soil surface to depth d\] was used to describe the relationship between the fine root vertical distribution and soil depth for the Sabina vulgaris community, the Sabina vulgaris cluster and the Artemisia ordosica community. The average root diameter was defined as the median of the root diameter distribution, and the root length density (RLD) and root area index (RAI) were calculated according to the regression equations obtained by a pilot experiment. The results are the following: (1) The means of biomass by month for live fine roots (D≤2 mm) reached a maximum at the 0-15 cm depth interval, but decreased with the increase in soil depth. About 50% of the fine roots occurred in the 0-20 cm depth and 90% occurred in the 0-597 (or -73.7) cm depth. The fine root distribution pattern was similar in the S. vulgaris community, the S. vulgaris cluster and the A. ordosica community. However, differences were found in the average biomass for the same depth, in which the value for the S. vulgaris community was 6.7-14.6 times that of the A. ordosica community (P<0.05), and the value for the S. vulgaris cluster was 14.0-19.2 times that of the A. ordosica community (P<0.05). (2) In both S. vulgaris (community or cluster) and A. ordosica communities, the dry mass percentages of the roots of diameter classes of both D≤1 mm and 1 mm＜D≤2 mm had a logarithmic relationship with soil depth, though the correlation was stronger for the former class (with a correlation coefficient of R2＞0.97 and 0.81＜R2＜0.94, respectively). In the S. vulgaris community, the weight percentage for the fine root diameter class of 1 mm＜D≤2 mm reached a maximum in the 15-30 cm depth range and then decreased with increasing soil depth, however, for both the S. vulgaris cluster and the A. ordosica community, the maximum occurred in the 0-15 cm depth range. (3) The monthly average of RLD for the D≤1mm diameter class decreased with the soil depth. In April, the A. ordosica community had the maximum RLD and RAI in the 15-30 cm depth, however, the maximum for S. vulgaris cluster (also in the 15-30 cm depth) occurred in October. The remaining maxima lie in the 0-15cm depth range for the D≤1 diameter class of fine roots. In terms of the monthly average of RLD, the value for the S. vulgaris community was 11.2 times (P＜0.05) and for the S. vulgaris cluster, 14.5 times (P＜0.05) that of A. ordosica community. The distributions of RAI across soil layers were similar for all three community types. The monthly means of the sums of the RAI within the 0-90 cm depth for the diameter classes of D≤2mm, D≤1mm and 1mm＜D≤2mm decreased in this order: S. vulgaris cluster＞S. vulgaris community＞A. ordosica community.