Land use planning studies represents, in general, the use of a large amount of information which leads into rather complex interpretation process similarly to the understand of the involved socioeconomic and environmental systems, as a result the decision-making process becomes slower, demanding spatial (scale) and temporal (period) adjustments. The use of inappropriate methods or insufficient data generates fragile instruments that should act as decision support systems. In this scenario, this work aims to contribute to develop a methodology for land use planning processes along the lines of the Ecological-Economic Zoning (EEZ) through Artificial Neural Networks (ANN) - Self Organizing Map (SOM) to subsidize a regionalization process and support land use planning in a more dynamic and faster manner, based upon data collection procedures (socioeconomic and environmental), data adjustment, network implementation and finally interpretation leading into the regional planning synthesis map. These steps allows parameterization and selection of the network that gives the better results accordingly to training and interpreter analysis over the generated data, taking into considerations statistical analysis such as U-matrix, component plans, per class graphs, Cluster analysis and map generation. The area was separated in the 12 units, that were grouped by similarity in 4 categories, which represent the main axes of sustainability proposed in the ZEE. the method adopted can defined distinct zones, that were not identified by ZEE, mainly due to the increased possibility offered by SOM to combine and integrate a large number of physical, social and economic variables.
|
