| Montane forests are susceptible to various stressors such as land use and climate change. Consequently, research on characterizing montane forest ecosystems should be conducted on a continuous basis for sustainable forest management. In this research, forest type mapping and change analysis, and biomass/carbon stock quantification were performed over a mountainous forest located in the central Adirondack Park, NY, by employing machine learning techniques at the plot level. Multi-temporal Landsat TM data were used to classify forest type cover and to detect forest cover changes for the past 20 years. Forest biomass and carbon stock quantification was then performed using full waveform LiDAR data collected in September 2011. Accuracies from the two case studies were in support of the versatility of machine learning approaches for forest and ecological investigation. Topographic characteristics affected the classification accuracy as well as the forest type change for the past 20 years. LiDAR-derived metrics, especially height-based ones, proved useful for quantifying biomass/carbon stock.;Keywords: Landsat TM, full waveform LiDAR, forest classification, forest change analysis, biomass, carbon stock, machine learning... |
