Reconhecimento de padrões com técnicas de agrupamento aplicadas ao inventário da arborização urbana do município de Jataizinho - PR

Abstract

Urban tree planting provides various social and environmental benefits when well planned, so as to avoid damage and problems related to plant health, tree health and structure, as well as conflicts with infrastructure. Through the inventory, it was possible to obtain information about the characteristics of the trees present in the municipality of Jataizinho, including their health and geographic distribution. Based on this information, an efficient Urban Tree Planting Master Plan can be developed, taking into account the needs of each region and best practices for preserving and developing urban tree planting. However, the process of analyzing inventory data can be complex and lengthy due to the large dimensionality of the data. In this context, this study proposes the application of clustering techniques to recognize patterns using the VIS-STAMP computational tool. The goal is to identify clustering patterns in the urban tree planting inventory dataset for the municipality of Jataizinho in order to assist in diagnosing for the development of PDAU and new regulations, aiming to optimize the extraction of more relevant information, reducing the amount of data volume to be analyzed and facilitating decision-making. The initial process was to prepare a geographic database, adopting two approaches: by blocks and by individuals. Through clustering techniques using Kohonen’s unsupervised neural network model, it was possible to identify areas with plant health problems, conflicts with urban infrastructure and pedestrians, and tree health and structure problems, making it possible to propose solutions for managing urban tree planting. The results revealed that 13.38% of occurrences are related to plant health problems, with ants being the main one. In relation to tree health and structure, a percentage of 68% was found, with absence of free area being the most prominent attribute. In relation to conflicts, a percentage of 66.06% occurrence was observed, with root breaking sidewalk being the attribute with the highest percentage. In addition, it was possible to identify clusters that were considered most important for maintaining tree planting. Cluster A2 showed a high value for conflicts, corresponding to a quantity of 3.92% of the total trees analyzed. Cluster Q1 showed high values for most attributes evaluated, representing about 9.78% of total blocks. This technique allowed identifying trends that could go unnoticed in a simple analysis, providing more detailed information for decision-making in managing urban tree planting. It was found that this method of interpreting multivariate data can be a valuable tool for optimizing and speeding up the process of analysis and urban planning, being able to provide a healthier, safer and more sustainable urban tree planting.