Percent Tree Cover Estimation Using Regression Tree Method - A case study of Africa with very-high resolution QuickBird images as training data

摘要

Different percent tree covers store different amounts of carbon. Regression tree is more robust than linear regression method, primarily due to its capability of approximating complex non-linear relationships using a set of linear equations. To cover large area with daily acquisition, the coarse resolution MODIS data were used. As the training data for this regression tree method, 11 scenes of very-high resolution QuickBird satellite were employed The explanatory variables acquired from MODIS data such as surface reflectance, NDVI, EVI, NDSI and thermal data were used as predictor. The tree proportions from QuickBird and MODIS-derived variables were then used to produce tree cover percentage layer for Africa. From prediction error measurement, the results were in good agreement with ground truth data.