Identification of Short-Rotation Eucalyptus Plantation at Large Scale Using Multi-Satellite Imageries and Cloud Computing Platform

摘要: A new method to identify short-rotation eucalyptus plantations by exploring both the changing pattern of vegetation indices due tree crop rotation and spectral characteristics in red-edge region is presented. It can be adopted produce maps high spatial resolution (30 m) at large scales, with use open remote sensing images from Landsat 8 Operational Land Imager (OLI), MODerate Imaging Spectroradiometer (MODIS), Sentinel-2 MultiSpectral Instrument (MSI), as well a free cloud computing platform, Google Earth Engine (GEE). The composed three main steps. First, time series Enhanced Vegetation Index (EVI) constructed data for each pixel, statistical hypothesis testing followed determine whether pixel belongs plantation or not based on idea that crops should harvested specific period. Then, broadleaf/needleleaf classification applied distinguish coniferous trees such pine fir using bands data. Refinements superpixel are performed last remove salt-and-pepper effects resulted per-pixel detection. proposed allows gaps very common tropical subtropical regions employing segmentation testing, could capture forest disturbances conversion natural agricultural lands emerged recent years short observing time. experiment Guangxi province China demonstrated had an overall accuracy 87.97%, producer’s 63.85% user’s 66.89% plantations.