One of the most important tasks in community ecology is identifying the co-occurrence patterns of the community and identifying the underlying processes that control such patterns. Null model analysis,which has been facilitated by the processes of biodiversity maintenance, has been used to identify the co-occurrence patterns in communities. However, few studies have focused on co-occurrence pattern dynamics of belowground soil communities. In this study, we identify the co-occurrence pattern dynamics of a Carabidae community in a Pinus koraiensis planted forest on a small scale, based on five investigations from June 2013 to October 2014. A plot (20 m×20 m) was established and further divided into 100 squares of 2 m×2 m at the Maoer Mountain Ecosystem Research Station of the Northeast Forestry University in Heilongjiang Province, Northeast China. One hundred pitfall traps were set to collect the Carabidae community for each investigation. To determine non-random patterns of species co-existence and significant species pairs in Carabidae communities, the null model was used to generate expected patterns in the absence of species interactions. In total, 20 species and 2278 individuals were captured. Carabus billergi maoershanensis was the most abundant and most widely distributed species among all communities. Carabidae community composition exhibited significant temporal variability among individuals during the experimental period. With the model algorithm of fixed row and column sums, results of the C-score and V-ratio showed that co-occurrence of species in June 2013 was aggregated, a finding that was inconsistent with Diamond's assembly rules. Moreover, with the model algorithm of fixed row sums and probability column sums and considering both algorithms, results of the C-score and V-ratio indicated that co-occurrence patterns of Carabidae communities in August and October 2014 were segregated, which was consistent with Diamond's assembly rules. Although the results of the C-score and V-ratio suggest that the co-occurrence patterns of the Carabidae communities in July and September 2013 were aggregated or segregated based on analysis of the null model, it can be inferred that non-random co-occurrence patterns were detected in all Carabidae communities. Furthermore,based on our observations of the co-occurrence patterns in the Carabidae communities, the Diamond's assembly rules were partly rejected. Few significant species pairs were detected among all Carabidae communities. A greater number of aggregated species pairs were observed in comparison to segregated species pairs among all communities. The species comprising those significant species pairs were dominant or common species, and were comparatively more widely distributed throughout the plot in each season. In addition, the Pianka (O_ik) indices between those dominant and common species were relatively greater than that observed between other species pairs. The results suggest that those significant species pairs might also make important contributions to community composition. The present study demonstrated that a non-random co-occurrence pattern might be common in a small-scale Carabidae community in a Pinus koraiensis planted forest. This non-random co-occurrence pattern also showed short-term stability during the study period;however, the properties of the sepatterns vary among the different seasons. The significant species pairs possibly contribute directly to observed community co-occurrence patterns, and more focus should be placed on the contributions of those dominant and common Carabidae species in the future.