Urbanization is a major cause of biodiversity loss and disconnection between people and nature. For this reason, the creation and maintenance of green areas as part of the urban landscape is a common practice in cities around the world. These green areas are generally a result of particular public policies regarding how these areas can be accessed, maintained, and how they are connected to other natural fragments in the city. In an effort to investigate aspects of the ecology of such green areas of the city of Curitiba, southern Brazil, the present study measured distinct parameters of the structure of butterfly assemblages from 19 forest fragments of distinct classes of protection, testing the influence of different policies (protection category) on butterfly assemblages. Comparatively, we also tested the predictive power of landscape variables (at distinct scales) and fragment quality (e.g. diameter at breast height, density of vegetation) using Generalized Linear Models (GLMs). Overall, 14,596 individuals belonging to 412 butterfly species were recorded, but values of species diversity and composition are very heterogeneous, indicating that the protection categories do not significantly interfere with the structure of the butterfly assemblages. More important than the protection category is the forest fragment quality and landscape connectivity. Our models showed that degraded fragments tend to present a specific butterfly composition, whereas the more urbanized landscape have poor butterfly species richness. These results reinforce the importance of implementing public policies that prioritize the conservation of the quality of forest fragments in every class of protection, as well as the conservation of forest fragments throughout the urban landscape.

城市化是生物多样性丧失以及人与自然之间脱节的主要原因。因此，在全球城市中，作为城市景观一部分的绿地的创建和维护是一种普遍的做法。这些绿色区域通常是有关如何访问，维护这些区域以及如何将其与城市中其他自然片段连接的特定公共政策的结果。为了调查巴西南部库里蒂巴市这种绿地的生态方面，本研究从19种不同防护等级的森林碎片中测量了蝴蝶组合结构的不同参数，并测试了不同政策的影响（保护类别）在蝴蝶组合上。比较，我们还使用广义线性模型（GLM）测试了景观变量（在不同尺度上）和片段质量（例如，胸高处的直径，植被密度）的预测能力。总体上，记录了属于412个蝴蝶物种的14,596个人，但是物种多样性和组成的值非常不均一，这表明保护类别并未显着干扰蝴蝶组合的结构。比保护类别更重要的是森林碎片质量和景观连通性。我们的模型表明，退化的碎片往往呈现出特定的蝴蝶组成，而城市化程度更高的景观中蝴蝶物种丰富度较差。