Marco Marchetti

Accurate structural information about forests, including canopy heights and diameters, is crucial for quantifying tree volume, biomass, and carbon stocks, enabling effective forest ecosystem management, particularly in response to changing environmental conditions. Since late 2018, NASA’s Global Ecosystem Dynamics Investigation (GEDI) mission has monitored global canopy structure using a satellite Light Detection and Ranging (LiDAR) instrument. While GEDI has collected billions of LiDAR shots across a near-global range (between 51.6°N and >51.6°S), their spatial distribution remains dispersed, posing challenges for achieving complete forest coverage. This study proposes and evaluates an approach that generates high-resolution canopy height maps by integrating GEDI data with Sentinel-1, Sentinel-2, and topographical ancillary data through three machine learning (ML) algorithms: random forests (RF), gradient tree boost (GB), and classification and regression trees (CART). To achieve this, the secondary aims included the following: (1) to assess the performance of three ML algorithms, RF, GB, and CART, in predicting canopy heights, (2) to evaluate the performance of our canopy height maps using reference canopy height from canopy height models (CHMs), and (3) to compare our canopy height maps with other two existing canopy height maps. RF and GB were the top-performing algorithms, achieving the best 13.32% and 16% root mean squared error for broadleaf and coniferous forests, respectively. Validation of the proposed approach revealed that the 100th and 98th percentile, followed by the average of the 75th, 90th, 95th, and 100th percentiles (AVG), were the most accurate GEDI metrics for predicting real canopy heights. Comparisons between predicted and reference CHMs demonstrated accurate predictions for coniferous stands (R-squared = 0.45, RMSE = 29.16%).

Terrestrial laser scanning (TLS) technology characterizes standing trees with millimetric precision. An important step to accurately quantify tree volume and above-ground biomass using TLS point clouds is the discrimination between timber and leaf components. This study evaluates the performance of machine learning (ML)-derived models aimed at discriminating timber and leaf TLS point clouds, focusing on eight Mediterranean tree species datasets. The results show the best accuracies for random forests, gradient boosting machine, stacked ensemble model, and deep learning models with an average F1 score equal to 0.92. The top-performing ML-derived models showed well-balanced average precision and recall rates, ranging from 0.86 to 0.91 and 0.92 to 0.96 for precision and recall, respectively. Our findings show that Italian maple, European beech, hazel, and small-leaf lime tree species have more accurate F1 scores, with the best average F1 score of 0.96. The factors influencing the timber–leaf discrimination include phenotypic factors, such as bark surface (i.e., roughness and smoothness), technical issues (i.e., noise points and misclassification of points), and secondary factors (i.e., bark defects, lianas, and microhabitats). The top-performing ML-derived models report a time computation ranging from 8 to 37 s for processing 2 million points. Future studies are encouraged to calibrate, configure, and validate the potential of top-performing ML-derived models on other tree species and at the plot level.

The heterogeneous ecosystems in the Mediterranean Basin (MB) are becoming sensitive to water stress. To investigate the climatic stress, a water budget study was conducted over the basin using TerraClimate simulations for a long temporal range (1990–2020). According to the budget accounting, forested regions received the highest precipitation (P) on average compared to other land use types (annual mean ≈633 mm yr−1), and even then, they were in a water deficit state (−0.42 mm yr−1). Tree plantations in North Africa (Libya and Morocco) were also in a water deficit state; however, their average P was very low (≈12 mm yr−1) compared to that of northern parts of the MB, and the average Actual Evapotranspiration (AET) was ≈15 mm yr−1. Also, the water balance in other land use systems (rain-fed, irrigated croplands, and rangelands) was either negative or near zero. As a whole, the basin’s average annual P was ≈538 mm yr−1, the annual average AET was ≈415 mm yr−1, and the runoff (Q) was equivalent to 123 mm yr−1, which shows a strong influence of ET over the region. Since runoff was negligible in most dry areas, the AET’s large contribution was notable in the North African base of the Atlas Mountains including the Nile delta region. This indicates that precipitation and evaporation are the principal mechanisms of the water balance in the MB. The result shows strong climate variability over Southern Europe, Turkey, and the western Balkans in the recent years, affecting the AET and making their land use systems more vulnerable to water stress. This benchmark study signifies the consistent need for water storage in the Mediterranean vegetation systems of the basin. It also indicates two distinct climate clusters for water balance modeling.