Caracterização de solos e avaliação da vulnerabilidade de ambientes no Parque Nacional de Itatiaia, Brasil

Abstract

Knowledge of soils and their properties is essential for environmental planning in natural systems especially in a conservation unit such as the Itatiaia National Park (INP). The INP, despite the importance of ecology and preservation, does not have information on its soils in detail that can support research and management plan. Aiming to understand the process involving the genesis and soils distribution in the mountainous environment of the INP and factors that involve the environmental vulnerability in this region the present study was developed. The objectives were to develop a database in a GIS environment with information on soils (classes and attributes), vegetation, relief, geology and geomorphology and to produce (environmental covariates) to support interdisciplinary research actions, environmental education programs and plan of park management. To further evaluate environmental vulnerability by integrating information from the physical environment with expert knowledge to reconcile public use demand with ecosystem conservation. In order to do so, sampling, collection, description, characterization, classification and mapping of soils was prepared and a database was prepared with all the environmental variables of data ownership, robust methods of digital soil mapping were tested in order to optimize the performance of the algorithms for the prediction of soil attributes and uncertainty evaluation. Finally, data from the literature review, participatory approach and specialized knowledge and biophysical variables produced in the previous steps were incorporated into a Bayesian belief network (BBN) to predict environmental vulnerability as well as to produce associated uncertainty. The results produced were sufficient to fill the gap in the lack of information on soils in the INP and to understand the factors related to the landscape soil relationship of the INP and are useful for several purposes. Generalized Additive Model Algorithms (GAM) with covariates selection based on the scorpan model are efficient in predicting attributes of the same using a limited number of points. And despite the complexity of the study area, the BBN was able to produce a significant result of the spatial distribution of environmental vulnerability and proved to be an alternative approach less subjective than conventional methods of assessing environmental vulnerability.