ALEPPO PINE ALLOMETRIC MODELING THROUGH INTEGRATING UAV IMAGE-BASED POINT CLOUDS AND GROUND-BASED DATA
Keywords: Mediterranean Forest, Aleppo Pine, UAV, Point Cloud, Machine Learning, Allometric Models, AGB Mapping
Abstract. Effective monitoring of Mediterranean forest is essential to determine the role of forest management in mitigating climate change and ensuring the maintenance of its environmental services. Most of the allometric models to estimate dry above-ground biomass (AGB) at tree level are based on knowing the diameter of the trunk at breast height (DBH). However, it is difficult, if not impossible, to estimate DBH from airborne/spaceborne sensors within the context of a remote sensing-oriented approach, being common to draw upon regression models to relate DBH to remotely sensed dendrometric variables such as total tree height (H) and tree crown diameter (CD). This study uses UAV (unmanned aerial vehicle) image-based data to estimate the dendrometric variables H and CD of the repopulated Aleppo pine (Pinus halepensis Mill.) located in a semiarid continental Mediterranean forest of Almería (southeast of Spain). DBH data were gathered through field work. Both bivariate (DBH = Φ(H)) and multivariate (DBH = Ψ(H, CD)) allometric models were developed by applying least-squares-based regression and machine learning regression methods. The results showed that multivariate allometric models performed better than bivariate at predicting DBH, both in terms of goodness-of-fit and stability against changes in training or testing samples. In addition, least-squares-based regression (linear and potential) provided statistically similar results to those obtained from complex machine learning ensemble algorithms. In this way, the easy-to-apply multivariate linear allometric model DBH = −4.84 + 1.73149H + 3.08114CD (R2 = 89.23%) would be recommended to locally estimate DBH in Aleppo pine from remotely sensed H and CD data.