One of the most important topics in community ecology is identifying the relative contributions of dispersal limitation and environmental filtering in community construction. However, the relative roles of these processes in above- and below-ground soil organism communities are still not well known. In order to reveal the relative contributions of environmental filtering and dispersal limitation in above- and below-ground soil mite community compositions, a small-scale plot (with a spatial extent of 50 m×50 m and a spatial resolution of 5 m×5 m) was established in a farmland of the integrated experimental field of a wetland, the Sanjiang Mire Wetland Experimental Station, Chinese Academy of Sciences. The plot was equally divided into 100 subplots, and samples for investigating soil mite communities and environmental factors were collected from the bottom left-hand corner of each subplot in August and October in 2011, respectively. Soil corner samplings and pitfall traps were used to capture soil animals in above- and below-ground communities, respectively. Moran's eigenvector maps (MEMs) were calculated to model spatial components in both months. Variation partitioning (based on partial redundancy analysis, pRDA) was used to estimate and test the proportion of total variation explained purely by environmental factors and spatial variables in soil mite community compositions. Furthermore, the partial Mantel test was selected to analyze the relationships between community dissimilarity and environmental and spatial distances. Finally, redundancy analysis (RDA) was performed to unveil the relative importance of the environmental variables on soil mite community composition. In total, 12 soil mite species were captured in the experiment. In August, 1, 6, and 6 significantly positive MEMs were selected for above-ground, below-ground, and aboveground-belowground (a combination of above- and below-ground) soil mite communities, respectively; 3, 8, and 7 significantly positive MEMs were selected for these communities in October. The results of variance partitioning showed that 3.8%, 12.8%, and 11.1% of the total variation was explained by spatial and environmental variables in the above-ground, below-ground, and aboveground-belowground soil mite communities in August, respectively; these values were 6.6%, 18.4%, and 16.9% in October. The relatively large and significant variation was attributed to purely spatial variables in all three types of soil mite communities in both months, whereas the contributions of purely environmental variables and spatially structured environmental variation were relatively low and non-significant. The partial Mantel test showed no obvious contribution of spatial or environmental distances for all soil mite communities. Based on the results of RDA analysis, soil pH and the average height of soybean explained a significant portion of the variance in August for all communities. In October, soil water content and soil pH explained a significant portion of the variance for the above-ground and aboveground-belowground soil mite communities. These results suggested that dispersal limitation is an important regulator in the composition of above- and below-ground soil mite communities in the farmland of the Sanjiang Plain at a small scale (50 m). However, the relative contribution of environmental filtering should not be overlooked. Collectively, construction of the above- and below-ground soil mite communities appears to be governed by niche- and neutral-based theories simultaneously at the small scale.