@InProceedings{JorgeShimSantGasp:2017:InTrDe,
author = "Jorge, Anderson Alex and Shimabukuro, Yosio Edemir and Santos,
Erone Ghizoni and Gasparini, Kaio",
affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)} and {Instituto
Nacional de Pesquisas Espaciais (INPE)} and {Instituto Nacional de
Pesquisas Espaciais (INPE)} and {Instituto Nacional de Pesquisas
Espaciais (INPE)}",
title = "Individual tree detection in intact forest and degraded forest
areas in the north region of Mato Grosso State, Brazilian Amazon",
booktitle = "Proceedings...",
year = "2017",
organization = "AGU Fall Meeting",
abstract = "The State of Mato Grosso - MT has the second largest area with
degraded forest among the states of the Brazilian Legal Amazon.
Land use and land cover change processes that occur in this region
cause the loss of forest biomass, releasing greenhouse gases that
contribute to the increase of temperature on earth. These degraded
forest areas lose biomass according to the intensity and magnitude
of the degradation type. The estimate of forest biomass, commonly
performed by forest inventory through sample plots, shows high
variance in degraded forest areas. Due to this variance and
complexity of tropical forests, the aim of this work was to
estimate forest biomass using LiDAR point clouds in three distinct
forest areas: one degraded by fire, another by selective logging
and one area of intact forest. The approach applied in these areas
was the Individual Tree Detection (ITD). To isolate the trees, we
generated Canopy Height Models (CHM) images, which are obtained by
subtracting the Digital Elevation Model (MDE) and the Digital
Terrain Model (MDT), created by the cloud of LiDAR points. The
trees in the CHM images are isolated by an algorithm provided by
the Quantitative Ecology research group at the School of Forestry
at Northern Arizona University (SILVA, 2015). With these points,
metrics were calculated for some areas, which were used in the
model of biomass estimation. The methodology used in this work was
expected to reduce the error in biomass estimate in the study
area. The cloud points of the most representative trees were
analyzed, and thus field data was correlated with the individual
trees found by the proposed algorithm. In a pilot study, the
proposed methodology was applied generating the individual tree
metrics: total height and area of the crown. When correlating 339
isolated trees, an unsatisfactory Rē was obtained, as heights
found by the algorithm were lower than those obtained in the
field, with an average difference of 2.43 m. This shows that the
algorithm used to isolate trees in temperate areas did not
obtained satisfactory results in the tropical forest of Mato
Grosso State. Due to this, in future works two algorithms, one
developed by Dalponte et al. (2015) and another by Li et al.
(2012) will be used.",
conference-location = "New Orleans",
conference-year = "11-15 Dec.",
language = "en",
targetfile = "jorge_individual.pdf",
urlaccessdate = "11 maio 2024"
}