Ecological zone-based volume estimation of Calycophyllum spruceanum and Cedrelinga cateniformis in the Northeastern Peruvian Amazon

• Robust volumetric models were developed for two Amazonian forest species. • The performance of the models varies according to the life zones. • The proposed equations are low-cost tools to improve sustainable timber management. Forest volume modeling plays a crucial role in forest inventory, biomass estimation, and the sustainable management of timber resources. In the Amazon region of Peru, native species such as Calycophyllum spruceanum and Cedrelinga cateniformis hold high ecological and commercial value, yet remain understudied in terms of volumetric estimation. This study aimed to develop and evaluate volumetric models for both species across three ecological zones—humid forest, very humid forest, and dry forest—representing the environmental diversity of the northeastern Peruvian Amazon. A total of 18 volumetric models were fitted for each species and site condition using linear regression techniques. Model performance was assessed through adjusted coefficient of determination (Radj 2 ), root mean square error (RMSE), mean absolute error (MAE), Akaike Information Criterion (AIC), and diagnostic analyses including residual plots and relative error histograms. The results revealed that model performance varied by ecological zone, with the dry forest models showing the highest precision and lowest residual dispersion. Models M3 (Spurr), M4 (Schumacher & Hall), and M9 (Meyer) consistently achieved strong predictive accuracy. Prediction errors were higher in small-volume classes, suggesting the need for caution when applying models to young or small-diameter trees. The developed models are statistically reliable models requiring minimal input variables for the accurate estimation of the timber volume of the two species across various Amazonian environments. It is recommended to adopt zone-specific models for operational use and to continue expanding regional forest databases to improve future model calibration and validation.