Urban forests are very important ecosystems, performing key cultural services and having crucial regulatory functions. However, naturally valuable areas in cities continue to decline or undergo irreversible transformations. It is necessary to determine the extent of human influence on forest ecosystems in highly urbanized environments. Our study examined the indicators of diversity, naturalness, lability, and stability of flora along transects from edge to center of six urban forests. Furthermore, the shape of gradients of changes and the threshold distance and straightness between forest edge and core for these indicators were measured. Finally, the impact of the number of inhabitants, the degree of protection, and the size and shape of the forests on each indicator separately and their aggregation in the form of the Flora Transformation Index were analyzed. The findings revealed that the gradients of changes were rarely linear and showed a logarithmic or polynomial shape. The most regular gradients were observed for lability and naturalness of flora. The regularity of the shape of the thresholds differed between the studied indicators. The threshold distance was relatively far (93–113 m) and depended heavily on human influence, but forest shape and area were also very important. Our results showed that legal protection, even as a reserve, has no effect in the case of solid human penetration of the forest. Our findings are expected to help determine suitable indicators for defining threshold values, i.e., the size of the edge and core zones of urban forests, which is crucial to the protection of the native plant species pool.

城市森林是非常重要的生态系统，具有重要的文化服务和重要的调节功能。然而，城市中具有天然价值的区域继续衰退或经历不可逆转的转变。有必要确定高度城市化环境中人类对森林生态系统的影响程度。我们的研究检查了从六个城市森林边缘到中心的横断面的植物群多样性、自然性、不稳定性和稳定性的指标。此外，测量了这些指标的变化梯度形状以及森林边缘和核心之间的阈值距离和直线度。最后，居民数量、保护程度的影响，分别分析了各指标上森林的大小和形状，并以植物区系转化指数的形式对它们的聚集进行了分析。研究结果表明，变化的梯度很少是线性的，呈对数或多项式形状。对于植物群的不稳定性和自然性，观察到最规则的梯度。阈值形状的规律性在所研究的指标之间有所不同。阈值距离相对较远（93-113 m），很大程度上取决于人类的影响，但森林形状和面积也非常重要。我们的研究结果表明，即使作为保护区，法律保护在人类深入森林的情况下也没有效果。我们的研究结果有望帮助确定定义阈值的合适指标，即