Model Error Transparency

The provided emission values are based on multiple factors that can influence the estimation precision like

LULC classification inaccuracies, salt'n'pepper, atmospheric influences, mixed pixels etc. potentially creating random change pixels for unchanged regions
the "unknown" class for which emission estimation is only partly possible (we only know the class at one timestamp or even at none)
other classes for witch no estimation is available (e.g. permanent crops)
inherent uncertainties of the factors (the used values are just estimates themselves)
general LULC diversity and classes being very general (e.g. sparse pine forest vs dense rain forest cannot are both in the "forest" class)
regional specialities not taken into account (e.g. forest in DR Congo different from Germany)

While some of these uncertainties could be reduced through further research, some will always be present.

This Milestone will tackle these problems and strive for more transparency in that area.