Estimativa de variáveis florestais com perfilamento a laser (Lidar)
Alves, Marcos Vinicius Giongo
MetadadosMostrar registro completo
Lately, data acquisition using Airborne Laser Scanning (ALS) with LiDAR technology (Light Detection and Ranging) is becoming promising in the forest field, especially for estimation of dendrometric variables and to evaluate vertical and horizontal structure of the forest. Topographic and forest coverage information are extremely important to forest and natural resources managers. Accurate information on trees height and density are fundamental for planning, but also hard to obtain by conventional methods. Laser scanning technology, as opposed to satellite images and aerial photographs, can at the same time map the ground and obtain estimates of the trees height. The use of modeling associated with LIDAR data allows the researcher to obtain estimates of several other forest variables, such as basal rea, diameter, volume, biomass and combustible material. It also presents a great potential in planning forest harvesting activities, road construction and maintenance. However, there are still many challenges in developing stronger and more reliable technologies and computational applications for modeling the data acquired with this type of sensor. The objective of this study was to evaluate the potential application of the LiDAR data to estimate forest variables, such as total individual height, average height, canopy base height and number of trees. Different classification methods of airborne laser scan points for thedevelopment of Digital Elevation Model (DEM) were also analyzed. Among the procedures evaluated for the preparation of DEM, the use of the software application TerraScan (TS) showed the best results for the total area, with a standard error of 0,48 m, while for the forest area the standard error was 0,53 m. However, other algorithms used also showed promising results for points classification with laser scanning in forest areas, such as the algorithm of Polynomial Interpolation (PI), which showed a standard error 0.65 m. For automatic recognition of number of individual trees in the study area located in Brazil, the morphologic analysis showed better results when compared with the use of local maximum algorithm, resulting in the recognition of 4848 trees. For study areas located at the Yosemite Park - USA, the different techniques evaluated (local maximum and orphological analysis) showed very similar results. The estimation of the total heights of Araucaria trees using LiDAR data, showed a correlation coefficient of 0.95 and a standard error of 0.91 m, when related to the measurements obtained at the field using a hypsometer. The estimation of the trees base heights with plots of different sizes (10, 15 and 20 meters) showed an standard error of 1.42, 0.95 and 0.82 m, which orrespond to 23.62, 15.70 and 13.84%, respectively.
- Teses