The species-area relationship is an important approach for understanding the structure of the plant community, which represents a fundamental issue in ecology, and is an important component of community surveys. The minimum sampling area of a community can be determined by a species-area curve. However, different sampling methods have dramatic effects on the species-area relationship. Previous studies have shown that the minimum sampling area can be determined by species-area relationship models (e.g., logarithmic function, power function, and logistic function), which are constructed from nested plots and quadrate combination methods. However, the forms and parameters of the species-area relationship models vary with sampling method, regional location, and spatial scale. The nested plot method may increase the probability of appearance of rare species, resulting in exaggeration of the number of species, while the quadrate combination method may neglect the aggregation distribution. Therefore, further research is needed to develop methods with decreased uncertainty. Here, based on the inventory data from two 10.4 hm² plots (Shengshan and Liangshui) and five 1.0 hm² plots (Fangzheng, Tangwanghe, Daliangzihe, Dongzhelenghe, and Fenglin) in the broad-leaved Korean pine (Pinus koraiensis) mixed forests in Xiao Xing'an mountains, Heilongjiang Province, the minimum sampling areas were determined by species-area curves with the moving window method. This method overcame the limitations of the nested sampling and quadrate combination methods and prevented the interference of sampling error on the species-area relationship caused by the community spatial heterogeneity. Additionally, the species-area curves of the seven plant communities were fitted with four saturation curve models to simulate the minimum sampling areas, and the suitability of these fits was evaluated. Furthermore, random sampling simulation was carried out with the minimum sampling area to examine the effects of sampling number on the accuracy of the species number estimation. The results showed that the measured minimum sampling areas of the seven forest communities with the moving window method varied from 40 m×40 m to 45 m×45 m, demonstrating that the minimum sampling areas of the broad-leaved Korean pine forest communities were similar. The R² values of the four species-area relationship models were all greater than 0.8 for the seven plots, but the minimum sampling areas determined by the models differed greatly from that measured because of the suitability of the models and the reliability of the curve extrapolation. Moreover, the proportions of the species of minimum sampling areas to that of the entire plot were overestimated by the models. By simulating the process of random sampling, we examined the effects of sampling number on the accuracy of species number estimation and found that the differences in community structures could be reflected in the demands of sampling numbers. Plots at Fenglin and Daliangzihe required only a small number of samples, indicating an even spatial distribution of species. However, the plots at Fangzheng and Shengshan required a large number of samples, indicating high spatial heterogeneity in the species distribution. Further studies are required to determine the mechanisms responsible for this difference.