3.3 Model summaries and evaluation
Descriptive statistics for each forest-type predicted tree height and density map are presented in Table 1. The mean, maximum and minimum values of predictions generally aligned with those of the plots. Tree height models with the highest r2 values were coniferous-broadleaf forest, broadleaf forest, and mangroves (Table 1) but appeared to have the best performance for tropical dry forests as it had relatively small RMSE and a correlation of ~0.65 (Fig 3a).
Table 1. Summary statistics of tree height and tree density predictions for all forest ecosystems in Mexico. *r2: coefficient of determination, RMSE: root mean square error.