Arctic tundra functional types by classification of single-date and AVHRR bi-weekly NDVI composite datasets

Maps of functional vegetation types are often used to extrapolate in situ measurements of biogeochemical fluxes to make regional-scale or biome-scale estimates. The objective of this study was to determine the optimal spectral-radiometric and temporal features derived from single-date and seasonal time series NOAA AVHRR imagery for classifying three Arctic tundra functional types. A single-date, three-band (VIS, NIR and NDVI) input yielded a map with the highest agreement (86.1% for supervised and 87.8% for unsupervised approaches) compared to a reference map. The map generated with the conventional maximum value NDVI for 10 semi-monthly periods through the growing season was in 82.7% agreement.