Table 2.

Coefficient of determination (R²) and root mean square error (RMSE) after excluding individual biomes from the ´best´ linear regression model.

Biome excluded	n	R²	RMSE
Temperate conifer forest	4	0.90	0.62
Temperate broadleaf forest	6	0.91	0.59
(Sub-)Tropical moist broadleaf forest	5	0.89	0.58
(Sub-)Tropical savannas and woodlands	3	0.82	0.64
Boreal forest	2	0.91	0.58

Each linear regression sub-model was significant at p < 0.0001. ‘n‘ refers to the number of sites in the respective biome that was excluded from the model.