Zone- and stand-level transferability of local and generalized Pinus patula stem-volume equations across three plantation areas in the northern Peruvian Andes

Reliable transfer of individual-tree stem-volume equations requires validation at the spatial and hierarchical levels where they will be applied. We evaluated 221 destructively sampled Pinus patula trees from 30 stands in three sampled plantation areas of the northern Peruvian Andes: Amazonas, Granja Porcón, and Chota. Three common tree-level equations without stand covariates and 33 stand-augmented variants were fitted by joint Gaussian maximum likelihood under constant, DBH-power and predicted-volume-power residual standard-deviation functions. Leave-one-zone-out (LOZO) validation omitted one entire sampled area, whereas leave-one-stand-out (LOSO) validation predicted a new stand within the represented areas. Under apparent LOZO, the density-augmented Spurr equation (M2.N) reduced unit-weighted RMSE% from 25.31% to 21.03%; however, only three areas were available and area-specific bias persisted. Under apparent LOSO, M2.N did not reduce mean RMSE% (14.32% for local M2 versus 14.36% for M2.N), won 15 of 30 folds, and did not show a significant paired improvement. Nested LOSO selected local M2 in all 30 outer folds, whereas M1.HD was selected for the DBH-only family, although uncertainty in its mean gain included zero. Accordingly, local M2 is the parsimonious operational reference within the sampled domain. M2.N is a promising candidate for cross-area transfer when current density is available but requires validation in independent plantation areas before broader use.