The Collection 6 MODIS LC product is still under development with the goal to release the maps sometime in early 2015.  The purpose of this project is to update and improve the methodology used to create the C5 MODIS LC product.  The first major improvement is a new set of daily NBAR inputs that will be splined to fill gaps in the time series.  In addition to ingesting annual metrics derived from these splined inputs, the algorithm will also consider additional spectral indices and phenological information.  The second improvement is the implementation of a new hierarchical classification scheme based on the FAO/UNEP Land Cover Classification System (LCCS) in addition to the five existing land cover legends.  To produce this new legend the training site database (STEP) needed to be redesigned with many existing sites modified and new sites added in areas of geographic or ecological gaps.  A third major improvement is a new multi-year stabilization procedure based on Markov chains developed by Parker Abercrombie here at BU.  A final improvement is to use Random Forests classifiers instead of the C4.5 Decision Tree program.  By producing the maps as a set of nested, hierarchical classifications, the integration of the classifications with a priori information from climate, census data, and other ancillary maps becomes simpler to do and the final maps are more accurate.