Contribuições metodológicas ao Planejamento Urbano: comparação de métodos de análise multivariada e modelos de mudança da paisagem urbana - aplicações na Regional Pampulha - Belo Horizonte, Minas Gerais - Brasil

Abstract

The Chaos Theory or Complex Systems defines a larger system is composed of smaller subsystems that operate interconnected, and that the attempt to understand and interfere in a subsystem excluding the interaction between other subsystems takes the larger system to chaos. Based on this principle, we present the urban system as a complex system and therefore should be studied and assessed in the light of complex methodologies. Thus, this work presents the general objective to compare different methodologies - simple and complex - variable selection and urban modeling in order to understand the behavior of this object in the light of landscape change models. For urban planning, change models are presented as support tools in the decision process as they can assist in the understanding of the dynamics that involve certain territory, as well as simulation scenarios according to the interests of the groups that has the power of decision in the strategic planning sector. Such models allow to simulate and verify in advance the life of certain public policies and projects as well as technically to base such decisions. This paper presents four articles of results which compare and evaluate variable selection methods to compose the model change of the urban landscape of Pampulha, Belo Horizonte - Brazil, by the methods of multivariate analysis: factor analysis, Monte Carlo, Crámer-Von -Mises and Logistic Regression. Once identified the most efficient variable selection method for this process, and identified the variables that make up the model, went to the second part where it makes comparisons between methods of modeling the dynamics of cover / land use Pampulha, among them: Exogenous Quantity and Linear Regression (GEOMOD); Markov chain and Cellular Automata (Ca-Markov); and Markov Chain and Neural Networks (LCM), all dynamic modeling module in Idrisi / Jungle platform. On the evaluation of the results, we conclude that the biggest challenge for the change models become a handy tool for managers is the database. The models showed better validation indices required at least 3 maps of use / land cover different times and scale compatible with urban dynamics (much detail). It is noted however that there are still no investment in the organization and collection of geographical information in a format compatible with GIS in most Brazilian municipalities. It is advised even the presence of a professional who is able to understand the spatial dynamics in the territory studied for this structure and identify more quickly the transition rules that make up the model change