Detecting Forest Degradation in Tropical Forests using Earth Observation Satellites

摘要

Deforestation is a process which has attracted considerable scientific interest in remote sensing and successful paradigms of detecting and monitoring have been presented; however, forest degradation in general is a more complicated case the detection of which presents significant challenges (Herold et al. 2011). Countries have been measuring current rates of degradation with field data and remote sensing imagery. Despite the fact that it is well evident that a combination of these two types of data provides the strongest capabilities (Herold et al. 2011), data on rates and processes of degradation are currently not available for many forested systems and information about factors influencing forest degradation is still limited in developing countries. Therefore, in these cases when assessing degradation they are forced to rely strongly on remote sensing approaches supported by any available field assessments of forest degradation. This research investigates the potential of TerraSAR-X remote sensing satellite images combines with field observations to detect and analyse forest degradation in a tropical forested area in Central Kalimantan, Indonesia. Several speckle noise filters have been tested and the Gamma MAP filter technique with a 7x7 window size is proposed as the best for this case study. Processes leading to classification, TerraSAR-X data (HH/HV and VV/VH dual polarizations) pre-processed with the proposed filter are used to classify logging tracks using feature extraction techniques. The results show that Example-Based classification method is a powerful technique to detect and map logging trails clearly while the same task is not addressed equally well by the Rule Based Feature extraction method. \udFurthermore, degraded areas, such as non-woody vegetation, logging trails, burned or logged forest were mapped using a combinations of classification algorithms applied on fused datasets from Landsat 5 TM and dual polarized (HH/HV and VV/VH) TerraSAR-X images; it was found that the fusion of SAR data with TerraSAR-X performed better in degraded areas than only TerraSAR-X dual polarization. Based on the backscatter coefficient and in-situ data the potential of SAR data to estimate biomass is evaluated. TerraSAR-X backscatter at HV polarization (R² = 0.413) outperforms in the study area when classifying logged areas, non-woody vegetation and intact classes. Comprehensively, this study demonstrates the potential of short-wavelength satellite radar to detect and characterise the processes of forest degradation from space.