Delimitação e estimativa da estrutura da mata ribeirinha a partir de imagens de alta resolução

Abstract

Riparian forests play an important role in the ecological balance of river ecosystems. The knowledge of riparian vegetation biophysical parameters is essential for developing strategies aimed at their conservation. Medium resolution sensors, such as Landsat-TM, have limited use in delineating and obtaining quantitative information of the riparian forests, given their narrow cross section in the landscape. An alternative to this fact consists in using high-spatial resolution data (<5m). The objective of this study is to describe a methodology for delineating riparian forests and extracting their biophysical parameters from field data and IKONOS images. The riparian forest delineation process is divided into two steps. In the first step, the riparian zone was defined on Depth to Water (DTW) algorithm application in the digital surface models. Then, the IKONOS images were classified into two classes: arboreal and non-arboreal using the 5m DTW threshold as a processing mask. The softwares e- Cognition, MAGIC (Map Guide Image Classification) and SPRING were compared in the riparian forest unsupervised classification. Allometric measurements were made in 280 plots (10x10m) of riparian area from both sides of the Pandeiros River, located in Northern Minas Gerais. These data were analyzed together with spectral and textural parameters of IKONOS images in order to develop predictive models of riparian forest biophysical structure. The texture image data were acquired from co-occurrence gray levels matrix features and semivariograms analysis. The classifiers e-Cognition, MAGIC and SPRING showed results more than 85% of accuracy, when compared with the validation data developed for this research. In general, predictive models results showed low determination coefficients. The best results for biophysical structure modeling were obtained, respectively, for the leaf area index and volume in the Pântano site, with R2 = 0.83 and R2 = 0.55.