Skip to main content
Silvilaser 2019

Silvilaser 2019 - Poster Presentations »

Qualification of emergent trees extracted from LiDAR point cloud in the Amazon forest

Before any logging operation in Brazilian native forests, managers are required to provide location and volume for all commercial trees above 50 cm of DBH. A census and a map geolocating all these trees is also required. The forest census is time consuming and expensive. This research aimed to qualify how many commercial trees are able to be detected by airborne laser scanning (ALS). To validate the trees extracted from LiDAR point cloud, we used an inventory dataset from 22 one-hectare plots. The study area is located in Paragominas, state of Pará, Brazil. Every tree within the plots with DBH larger or equal to 35 cm had its diameter, height and canopy radius measured and its geographic coordinate recorded. Additionally, the stem quality and canopy position were also determined. We used in our analysis a subset of the collection of measurements, as we only considered trees with DBH ≥ 50 cm. An ALS assessment flown in 2012 coincided with the inventory. We applied the maximum local (ML) algorithm to individualize the emergent trees, extracting also height and canopy radius. Key points indicated the presence of a tree if located by the ML algorithm and validated by the uncertainty index threshold. The uncertainty index was a function of (i) the horizontal distance to neighbor trees, (ii) the difference between canopy size estimated by LiDAR and measured in the field, and (iii) difference between LiDAR and field recorded height. From the 504 trees (DBH ≥ 50 cm) coming from 96 different species found in field plots, LiDAR was capable to locate 217 individuals coming from 69 species (considering an uncertainty index threshold of 9). From the 69 species, 25 were considered as commercial species. According to the structural analysis, 12 species listed as highest importance values (IV) from LiDAR plots were included between the 15 highest IV species from field plots. LiDAR sampling missed 57% of total trees (DBH ≥ 50 cm), however, maintaining a similar forest structure. Our results demonstrated that the total emergent trees located by LiDAR are enough to plan the logging, and could be an alternative to the forest census.

Eduardo Pelli
Universidade Federal dos Vales do Jequitinhonha e Mucuri
Brazil

Cristiano Rodrigues Reis
Universidade de São Paulo
Brazil

Gilciano Saraiva Nogueira
Universidade Federal dos Vales do Jequitinhonha e Mucuri
Brazil

Cristiano Christófaro Matosinhos
Universidade Federal dos Vales do Jequitinhonha e Mucuri
Brazil

Alessandro Vivas Andrade
Universidade Federal dos Vales do Jequitinhonha e Mucuri
Brazil

Luiz Carlos Estraviz Rodriguez
Universidade de São Paulo
Brazil

Eric Bastos Görgens
Universidade Federal dos Vales do Jequitinhonha e Mucuri
Brazil

 


Powered by OpenConf®
Copyright ©2002-2018 Zakon Group LLC