@PhDThesis{Wiederkehr:2022:UsDaSa,
author = "Wiederkehr, Natalia Cristina",
title = "Uso de dados dos sat{\'e}lites ALOS/PALSAR-2 e Sentinel-1A para
detec{\c{c}}{\~a}o de perdas de volume florestal por processo de
corte seletivo em uma por{\c{c}}{\~a}o da Floresta Nacional do
Tapaj{\'o}s",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2022",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2022-08-19",
keywords = "degrada{\c{c}}{\~a}o florestal, Amaz{\^o}nia, dados SAR,
ALOS/PALSAR-2, sentinel-1A, forest degradation, Amazon, SAR
data.",
abstract = "O desmatamento e a degrada{\c{c}}{\~a}o florestal induzida pelas
a{\c{c}}{\~o}es antr{\'o}picas, sobretudo pela
extra{\c{c}}{\~a}o seletiva de madeira em regime n{\~a}o
sustent{\'a}vel, pelos inc{\^e}ndios florestais e pelo efeito de
bordas, contribuem significativamente para as emiss{\~o}es de
di{\'o}xido de carbono (CO2), degrada{\c{c}}{\~o}es e perdas de
florestas na Amaz{\^o}nia Brasileira. Identificar os impactos
causados ao ambiente florestal torna-se extremamente importante,
pois possibilita compreender as rela{\c{c}}{\~o}es de causa e
efeito da perda de florestas, sobretudo, os ocasionados pela
degrada{\c{c}}{\~a}o florestal, uma vez que ainda n{\~a}o
existe um consenso sobre qual o limiar de perda da biomassa e a
persist{\^e}ncia nessa perda que constitui a
degrada{\c{c}}{\~a}o. Neste sentido, o sensoriamento remoto
torna-se uma ferramenta muito importante, pois {\'e} uma maneira
eficaz e economicamente vi{\'a}vel de monitorar as
mudan{\c{c}}as na cobertura florestal, principalmente de grandes
regi{\~o}es florestais como a Amaz{\^o}nia. O presente estudo
teve como objetivo avaliar individualmente a capacidade das
imagens polarim{\'e}tricas dos sat{\'e}lites ALOS/PALSAR-2
(ALOS2) e Sentinel-1A (S1A) para detectar perdas de volume
florestal por processo de extra{\c{c}}{\~a}o seletiva de madeira
em regime sustent{\'a}vel. Quatro {\'a}reas com alta intensidade
de explora{\c{c}}{\~a}o madeireira, entre 27 m³ ha-1 e 29 m³
ha-1, denominadas de Unidades de Produ{\c{c}}{\~a}o Anual (UPA),
inseridas na Floresta Nacional do Tapaj{\'o}s, foram
selecionadas. Para cada UPA, a explora{\c{c}}{\~a}o ocorreu em
um ano espec{\'{\i}}fico: UPA 2015, UPA 2016, UPA 2017 e UPA
2018. Os atributos extra{\'{\i}}dos a partir dos coeficientes de
retroespalhamento ({\'a}lgebras, raz{\~o}es de bandas,
{\'{\i}}ndices SAR de vegeta{\c{c}}{\~a}o, medidas de
texturas) e de informa{\c{c}}{\~a}o de fase
(decomposi{\c{c}}{\~a}o polarim{\'e}trica entropia-{\^a}ngulo
alfa) das imagens ALOS2 e S1A, foram utilizados para detectar as
perdas de volume florestal. A detec{\c{c}}{\~a}o foi realizada a
partir das diferen{\c{c}}as entre os valores dos pixels das
{\'a}reas exploradas e n{\~a}o exploradas. O teste
n{\~a}o-param{\'e}trico de Wilcoxon, com um grau de
confian{\c{c}}a de 95%, foi empregado para avaliar se as
diferen{\c{c}}as encontradas eram significativas ou n{\~a}o. Os
resultados obtidos pelos dados ALOS2 demonstraram um desempenho
superior aos do S1A. Dentre os atributos associados ao ALOS2,
destaca-se o Radar Normalized Difference Vegetation Index (RNDVI)
que exibiu sensibilidade em detectar perdas de volume florestal
devido {\`a} degrada{\c{c}}{\~a}o por processo de corte
seletivo em todas as {\'a}reas investigadas. A
polariza{\c{c}}{\~a}o HV, a raz{\~a}o de
polariza{\c{c}}{\~a}o cruzada, a medida de textura denominada
Contraste na polariza{\c{c}}{\~a}o HV e os atributos
polarim{\'e}tricos entropia e {\^a}ngulo alfa tamb{\'e}m
apresentaram potencial quando associados {\`a}s UPAs de 2015 e
2016. Os atributos associados ao S1A apresentaram uma discreta
diferen{\c{c}}a entre os valores de retroespalhamento, mesmo
considerando a alta intensidade de corte seletivo nas UPAs. Os
melhores resultados foram obtidos pelos atributos \σ°VH
associado {\`a} UPA 2015, Radar Ratio Vegetation Index (RRVI) na
UPA 2016, medidas de texturas denominadas de Energia e M{\'a}xima
Probabilidade na polariza{\c{c}}{\~a}o VH associados {\`a} UPA
2017, {\^a}ngulo alfa, Contraste na polariza{\c{c}}{\~a}o VH,
al{\'e}m do RRVI para a UPA 2018. ABSTRACT: Human-induced
deforestation and forest degradation, especially by illegal
logging, fires, and edge effects contribute significantly to
carbon dioxide (CO2) emissions, degradation, and loss of forests
in the Brazilian Amazon. Identifying the impacts caused to the
forest environment becomes extremely important, as it makes
possible to understand the cause and effect relationships of
forest loss especially those caused by forest degradation, since
there is still no consensus on the threshold of biomass loss and
the persistence of this loss that constitutes degradation. In this
sense, remote sensing becomes a very important tool, as it is an
effective and economically viable way to monitor changes in forest
cover, especially in large forest regions, such as the Amazon. In
this context, this work aimed to evaluate the ALOS/PALSAR-2
(ALOS2) and Sentinel-1A (S1A) polarimetric images to detect losses
in forest volume through the process of selective logging in a
sustainable regime. Four areas with high timber exploration
(between 27 m³ ha-1 and 29 m³ ha-1), called Annual Production
Units (APU), inserted in the Tapaj{\'o}s National Forest, were
selected. For each APU, the exploration took place in a specific
year: APU 2015, APU 2016, APU 2017, and APU 2018. The attributes
derived from radar backscatter (algebras, band ratios, SAR
vegetation indexes, texture measures) and phase information
(entropyalpha angle) from ALOS2 and S1A images were used to detect
forest volume losses. The detection was performed from the
differences between the pixel values of the explored and
unexplored areas. Wilcoxon's nonparametric test, with a confidence
level of 95%, was used to assess whether the differences found
were statistically significant or not. The results obtained by the
ALOS2 data showed a superior performance in relation to the S1A
datasets. Among the ALOS2 attributes, Radar Normalized Difference
Vegetation Index (RNDVI) showed best potential to detect forest
volume losses due to degradation by the selective logging process
in all investigated areas. The HV polarization, the cross
polarization ratio, the Contrast texture measure, and the entropy
and alpha angle attributes also showed potential when associated
with the 2015 and 2016 APUs. The S1A attributes showed a slight
difference between the backscatter values, even considering the
high intensity of selective logging in the APUs. The best results
were obtained by the \σ°VH associated with the 2015 APU,
Radar Ratio Vegetation Index (RRVI) in the APU 2016, Energy and
Maximum Probability textural measures in the VH polarization
associated with the 2017 APU, alpha angle, Contrast in the VH
polarization, as well as by RRVI attribute for the APU 2018.",
committee = "Shimabukuro, Yosio Edemir (presidente) and Gama, F{\'a}bio Furlan
(orientador) and Bispo, Polyanna da Concei{\c{c}}{\~a}o
(orientadora) and Mura, Jos{\'e} Cl{\'a}udio and Sano, Edson
Eyji and Araujo, Luciana Spinelli",
englishtitle = "Use of ALOS/PALSAR-2 and Sentinel-1A satellite data for the
detection of forest volume losses by selective logging process in
a portion of the Tapaj{\'o}s National Forest",
language = "pt",
pages = "353",
ibi = "8JMKD3MGP3W34T/47PQBTS",
url = "http://urlib.net/ibi/8JMKD3MGP3W34T/47PQBTS",
targetfile = "publicacao.pdf",
urlaccessdate = "24 abr. 2024"
}