Estimación del volumen individual en Eucalyptus dunnii mediante modelos de regresión y fotogrametría aérea basada en VANT

In Uruguay, Eucalyptus dunnii (E. dunnii) has established itself as one of the most important species for the forestry sector due to its adaptability, frost tolerance, and suitability for pulp production. In this context, accurate estimation of individual tree volume is a fundamental tool for forest planning and management. Traditionally, volume has been determined using destructive methods and regression models based on dasometric variables; however, advances in remote sensing technology have made it possible to incorporate non-destructive alternatives, such as aerial photogrammetry using an unmanned aerial vehicle (UAV).

The objective of this study was to estimate the individual volume of three-year-old E. dunnii using regression models and UAV-based aerial photogrammetry in a plantation located in Quebracho, Paysandú Department, Uruguay. To this end, 10 inventory plots of 300 m² were established, in which dendrometric variables such as diameter at breast height (DBH) and total height (TH) were measured. In addition, 25 trees were selected for destructive sampling to calculate the actual volume using the Smalian method. Subsequently, linear and nonlinear mathematical models were fitted using DBH and TH as predictor variables, and the models were evaluated using adjusted R², root mean square error (RMSE), and residual analysis. The selected model had an adjusted R² of 0.9875 and an RMSE of 0.0035 m³, demonstrating a high predictive capacity for estimating individual volumes.

In addition, an aerial survey was conducted using a DJI Inspire 1 UAV equipped with a multispectral camera. The images were processed using aerial photogrammetry techniques to generate digital terrain models (DTMs), digital surface models (DSMs), and a canopy height model (CHM). Based on the CHM, algorithms for detecting and segmenting individual trees were applied, correctly identifying 97% of the trees present in the study plots.

Finally, a Random Forest-based predictive model was developed to estimate individual volume using metrics derived from the CHM. In cross-validation, the model achieved an R² of 0.59 and an RMSE of 0.0147 m³ per tree, demonstrating moderate predictive ability and a low estimation error relative to the observed volume range. Overall, the results demonstrate that the combination of regression models and UAV-based aerial photogrammetry constitutes a reliable alternative for the non-destructive estimation of individual volume in young E. dunnii plantations, with the potential to complement traditional forest inventories.