Individuazione e correzione di outlier nei dati raster LiDAR provenienti dal Geoportale Nazionale e produzione di un CHM idoneo alla stima delle provvigioni legnose della Calabria

Abstract

Abstract: Detection and correction of LiDAR raster data from the Italian national remote sensing programme and production of a suitable CHM to forest volume estimation in Calabria (southern Italy). The AlForLab project, a Public-Private Laboratory which is part of the Cluster MEA (Materials Energy Environment) addressed to the Calabria Region (southern Italy), has gained great benefit by using LiDAR data acquired in the frame of a national remote sensing programme of the Ministry of the Environment and Protection of Land and Sea. This kind of LiDAR data, distributed in raster format and publicly available for research and non-profit purposes, have proved to be a suitable tool to support forest management. Their usage, however, has required the recognition and correction of non-forest elements included in the Digital Surface Model (DSM), like electric powerlines, wind turbines, sub-vertical rocks and viaducts. Such outliers, if remaining into the Canopy height Model (CHM), can generate potential errors in application of LiDAR-based prediction models. This paper proposes some semi-automatic pre-processing procedures, directly applicable on raster data, in order to obtain a CHM without non-forest elements. The methods described here have been developed in open-source environment (R and QGIS). The correction procedures carried out were tested in three municipalities having forest area between 1700 and 5400 ha, and characterized by different types of outliers. The performances of the methods were evaluated by comparing the estimated forest volume obtained before and after their application. Although low total volume changes were observed on the entire study areas (about 0.5%, corresponding to 1500 to 7200 m3), more significant effects, tens to hundreds cubic meters per hectare of overestimation, can occur in stands or forest compartments with a high presence of outliers. In conclusion, the proposed methods have proved to be suitable to achieve a reliable CHM for forest applications.