Abstract
Abstract. Carbon losses have enormous impacts on ecosystem and indicative meaning for ecological stability, which are, therefore, necessary to be evaluated. In recent years, with the increasingly number of people settling in Amazon basin, large areas of tropical forests have been exploited, which took the mainly responsible for local deforestation.Carbon losses are hardly to be evaluated directly but can be inferred from the relationship between potential carbon and forest structures referring to previous studies. Remote sensing-based methods are widely used because of the obvious advantages of, such as the repeatability of data collection, high correlations between spectral bands and vegetation parameters etc. However, the selection of the spatial resolution of data is a critical issue that should be taken into consideration for it influences the performance of image texture thereby the discrimination of land covers, especial in complex forest stand structure. The approach for image feature extraction used in the study is explained.Based on the available datasets, this study aims to estimate the potential carbon losses across selected plot in Nova Vida Ranch, Rondônia, Brazil between 1989 and 2001. The results showed that carbon potentials decreased at least 1.1*108 (Kg C) during the 12 years, which would affect all the climate, wildlife habitats and biodiversity. The study suggested that it is significantly important to reduce the uncertainties that are caused by saturation in some of the tropic areas. To pursue this, the research on the integration of optical and radar data is worth to experiment in further study.