Skip to main content
Fig. 5 | Carbon Balance and Management

Fig. 5

From: Questioning emissions-based approaches for the definition of REDD+ deforestation baselines in high forest cover/low deforestation countries

Fig. 5

Ranking following Mean Decrease in Gini index (MDG) of spatial variables included within the three deforestation location models. NGMM, GMM, HA stand for no gold-mining model, gold-mining model and historical scenario respectively. Mean Decrease in Gini index (MDG) is used to rank variables by importance. When random forest algorithm is used to classify a sample of pixels within two groups (deforested, not deforested), a very important variable will increase a lot the “purity” within each group and will be associated with a large MDG. On the contrary, a variable of low importance will not contribute to increasing intra-group purity and will have a lower MDG

Back to article page