Fig. 6

Rank of importance (mean decrease Gini) of each variable for predicting each site burn severity class (unburned, low, moderate, high) as determined by post-fire archaeological assessments. The Gini coefficient represents node impurity. Variables that lead to more homogenous nodes give greater decreases in mean Gini. Variable importance was tested as part of an analysis to determine the best environmental predictors of burn severity for 858 sites burned during five wildfires in the Jemez Mountains, New Mexico, USA, from 2000 to 2011. Numbers represent variable importance rank, with 1 being most important. Predictors are presented in order of importance (greatest to least) for predicting sites with a high severity designation. See Tables 1 and 2 for predictor definitions. Additional file 1 reports overall importance ranks for variables as measured by accuracy and mean decrease in Gini