Mapping land cover and green vegetation abundance using MODIS-like data: a case study of New England

To reduce the uncertainty of many environmental models it is necessary to both classify land-cover type accurately and to compute dynamic changes in green vegetation abundance of each land-cover type. To meet these needs, a case study is carried out in New England. A univariate decision tree is implemented to classify IGBP land cover using MODIS-like data. Classification results are presented using different combinations of input features including NDVI, NBAR (nadir BRDF-adjusted reflectances), and NDVI+NBAR. It is shown that NBAR is the best input feature for this classification problem. The abundance of green vegetation is calculated using a constrained linear mixture modeling approach based on NBARs. Variation in green vegetation abundance within each land-cover type is analyzed at different phenological stages from July to October 1996.