Desarrollo de modelos de estimación de variables forestales en sistema silvopastoril utilizando métricas LiDAR

Silvopastoral systems (SPS) are a productive alternative that integrates trees, pastures, and livestock, generating both productive and environmental benefits. In this context, the accurate estimation of forest variables such as total height (Ht) and diameter at breast height (DBH) is essential for forest management and monitoring. The use of remote sensing technologies, particularly airborne LiDAR, enables rapid and accurate estimation of these variables at the individual tree level. The study was conducted at the Prof. Bernardo Rosengurtt Experimental Station in Cerro Largo, Uruguay, on clonal and seed-origin trials established under a silvopastoral regime with Eucalyptus grandis W. Hill and Eucalyptus dunnii Maiden plantations. A total of 96 trees were selected and measured in the field, recording Ht and DBH. Subsequently, a LiDAR survey was performed using a Zenmuse L2 sensor (DJI, Shenzhen, China) mounted on a Matrice 350 drone (DJI, Shenzhen, China). From the point cloud, a digital vegetation model (DVM) was generated, from which the maximum height value (maximum DVM) of each segmented tree was extracted. Linear models were then fitted to estimate Ht and DBH, and their performance was evaluated using the coefficient of determination (R²), root mean square error (RMSE), and leave-one-out cross-validation (LOOCV). The developed models showed high predictive capacity. For Ht estimation, R² values of 0.93 and 0.92 were obtained for the clonal and seed-origin trials, respectively, with an RMSE of 0.40 m in both cases. For DBH estimation, the models achieved R² values of 0.73 (clonal) and 0.78 (seed-origin), with RMSE values of 0.73 cm and 0.59 cm, respectively. Significant differences were also detected among silvopastoral treatments, demonstrating the influence of spacing and row arrangement on tree growth. Treatments with narrower alley widths (9 m) and triple-row arrangements showed the highest DBH and Ht values in both clonal and seed-origin trials. Significant differences were also observed between treatments with different alley widths. In the clonal trial, maximum values of 8.28 cm DBH and 8.94 m Ht were recorded under 9 m alley widths, while minimum values of 5.05 cm DBH and 4.93 m Ht were observed under 15 m alley widths. In the seed-origin trial, maximum values reached 9.54 cm DBH and 9.36 m Ht under 9 m alley widths, whereas minimum values of 7.61 cm DBH and 6.88 m Ht were recorded under 15 m alley widths. In conclusion, airborne LiDAR proved to be an efficient and accurate tool for estimating forest variables at the individual tree level in SPS. The developed models reduce the time and costs associated with traditional forest inventories, contributing to improved planning and decision-making.