Coarser-Level (Hierarchical) Classification of MODIS Land Cover 2021.

[shiweda_markus]

I am downscaling future land use change cover projections for Southern Africa using MODIS Land Cover's 17 classes as baseline data. However, Land Cover classes are slightly many (17), leading to some classes missing in some countries: Each missing one or more classes. Grouping them into broader, ecologically relevant categories (e.g., Forest, Shrubland, Grassland, Wetland, Urban, Water) naturally sidesteps the problem of certain fine classes appearing or disappearing in each country.

Is there a confusion matrix table or a better alternative I can use to combine country-specific classes?

Your advice will be appreciated.




Additional information:

Additional info for my downscaling implementation approach:
I am downscaling few classes (7) projections, matched to MODIS 500m LC classes. I then need distribution Proportions between MODIS LC and projections. Due to the class availability difference per country, I need dynamic percentage proportions for the classes.

[LP DAACx - lien]

Hi,
I am assuming you are referring to the IGBP Landcover classification in the MCD12Q1. I am not exactly sure what you are trying to. As far as Land Cover classification types, we have 5 and the BGC has just 8 classes as compared to the IGBP. Also, if you are looking for coarser data, we have the MCD12C: https://doi.org/10.5067/MODIS/MCD12C1.061
If you are attempting to create a different data set perhaps some folks in the user community may be able to make recommendations.
Thanks,
LPDAAC