3D Remote Sensing technologies for Precision Forestry

Precision forestry is a new approach for more sustainable forest management. Modern technologies are important to the development of new tools and applications to conduct site-specific management practices. 3D remote sensing technologies are new tools and have new applications useful for improving the data collection, work efficiency and quality of forest information that can be used to take better management decisions. This thesis is aimed at assessing the use of 3D data to develop new tools and procedures useful for forest inventories and for the estimation of soil disturbances caused by forest operations. In so doing, this study attempts to close the gaps underlined by previous studies. The thesis is divided into two main sections. The first one comprises the studies I, II, and III related to forest inventory optimization, while the second section comprises the studies IV and V related to estimation of soil disturbances caused by forest operations. Study I demonstrates how a 3D point cloud acquired by a Terrestrial Laser Scanner (TLS) and a Hand-Held Mobile Laser Scanner (HMLS) can be used to automatically derive forest single tree variables such as diameter at breast height (DBH) and tree position (TP). Moreover, the study underlines how the integration of TLS with Airborne Laser Scanner (ALS) point clouds improves the estimation of tree top height (H) and crown base projection (CPA). In study II a novel approach is presented for the extraction of explanatory variables from unmanned aerial vehicle (UAV) 3D photogrammetric data for predicting forest biophysical properties without relying on a digital terrain model. This study assesses the use of DTM-independent variables to predict forest biophysical proprieties using as a benchmark two more traditional sets of variables: (i) height and density variables from UAV photogrammetric data normalized using a DTM acquired using airborne laser scanning (ALS) (Image-DTMALS variables), and (ii) height and density variables extracted from normalized ALS data (ALS variables). We obtained comparable results between the models developed with DTM-independent models and the ones obtained with the other two types of variables (i.e. Image-DTMALS and ALS) to predict: Growing Stock Volume (V), Basal Area (G), Number of trees (N), Dominant Height (Hdom) and Lory’s height (Hl). Study III used the new set of DTM-independent variables developed in study II to predict area-based (ABA) forest structure variables (e.g. V, G, Mean Diameter (DBHmean), Gini coefficient of DBH (Gini), standard deviation of DBH(σdbh), Hdom, Hl and standard deviation of H (σh)) using as benchmarks the variables from ALS. The results underline comparable results between the two types of metrics in the estimation of forest structure variables. Moreover, the models developed with DTM-independent metrics were used to create two maps of two forest structure indices. In study IV and V we tested the utility of multi-temporal high resolution DTM derived by Personal Laser Scanner (PLS) (IV) and by close range photogrammetry (V) to measure and quantify soil disturbances caused by forest operation. These studies underline how multi-temporal high resolution (DTM) can be used to quantify rut deep, bulges, and soil volume changes. In conclusion, 3D RS data appears useful in the development of new methods to collect and measure forest ecosystem components such as vegetation and soils