Abstract: Laboratory-simulated rainfall experiments were conducted to quantify the effects of patterns of Artemisia capillaris on runoff and soil loss. The spatial variability of soil erosion and deposition were also preliminarily analyzed using classic statistical and geostatistical methods. Simulated storms (90 mm/h) were applied on a bare plot (CK) and three different plant patterns: a checkerboard pattern (CP), a banded pattern perpendicular to the slope direction (BP), and a single long strip parallel to slope direction (LP). Each patterned plot underwent two sets of test—intact plants and roots only, respectively. All treatments had three replicates. The results showed that A. capillaris of different patterns can effectively reduce runoff and sediment. Compared with CK, the intact plant plots had a 12% to 25% less runoff and a 58% to 92% less sediment. The contributions of roots and canopies of A. capillaris to the reductions in runoff and sediment were different. Roots contributed more (46%–70%) to the sediment reduction, whereas canopies contributed more (57%–81%) to the runoff reduction. BP and CP exhibited a preferable controlling effect on soil erosion than LP. Coefficient of variation indicated that soil erosion, expressed as probe height difference (Δh), was strongly variable. Geostatistical analysis revealed that semivariograms of soil erosion can be fitted with spherical or exponential models and showed a moderate or strong spatial dependence. Interpolation using kriging demonstrated a spatial similarity among the treatments. Erosion (Δh > 0), primarily 0- to 2.67-mm level, spread over the bare plot. For intact plant experiments, the erosion area was relatively smaller than that for CK, and no significant differences were observed between erosion and deposition (Δh < 0) area. The erosion area increased after removing vegetation canopies.