@MastersThesis{Bandoria:2021:InDaMu,
author = "Bandoria, Marcelo Cardoso da Silva",
title = "Din{\^a}micas de regenera{\c{c}}{\~a}o florestal na Mata
Atl{\^a}ntica: integra{\c{c}}{\~a}o de dados multi-sensores e
medidas de campo",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2021",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2020-04-30",
keywords = "sentinel-2 MSI, Landsat TM, ETM+ e OLI, sensoriamento remoto
multiespectral, DAP, IAF, serrapilheira, regenera{\c{c}}{\~a}o
da Mata Atl{\^a}ntica, multispectral remote sensing, DBH, LAI,
litter, regeneration of the Atlantic Forest.",
abstract = "Desmatamento e degrada{\c{c}}{\~a}o florestal s{\~a}o processos
que impactam diretamente os estoques de carbono, biodiversidade,
reservas h{\'{\i}}dricas e o clima, promovendo ampla perda de
servi{\c{c}}os ecossist{\^e}micos. Na conven{\c{c}}{\~a}o de
Paris, o Brasil assumiu ratificado atrav{\'e}s da
Contribui{\c{c}}{\~a}o Nacionalmente Determinada (NDC), de
restaurar e reflorestar 12 milh{\~o}es de hectares de florestas
at{\'e} 2030. Grande parte dessas a{\c{c}}{\~o}es devem se dar
no bioma Mata Atl{\^a}ntica e metodologias baseadas em imagens
satelitais contribuir{\~a}o para o monitoramento e
verifica{\c{c}}{\~a}o do acordo. Apesar dos grandes
avan{\c{c}}os t{\'e}cnicos sobre metodologias de mapeamento de
desmatamento, a detec{\c{c}}{\~a}o dos diferentes graus de
regenera{\c{c}}{\~a}o florestal, s{\~a}o ainda incipientes.
Nesse sentido, o presente trabalho teve por objetivos identificar,
por meio de invent{\'a}rios de campo (resolu{\c{c}}{\~a}o
CONAMA no 1 de 1994) e sensoriamento remoto orbital as diferentes
idades de fragmentos de Mata Atl{\^a}ntica na Esta{\c{c}}{\~a}o
Experimental Solo Planta Atmosfera de S{\~a}o Francisco Xavier
(EESPA-SFX), localizada no munic{\'{\i}}pio de S{\~a}o
Jos{\'e} dos Campos, SP, administrada pelo Laborat{\'o}rio de
Ecohidrologia (LabEcoh/CCST-INPE). Atrav{\'e}s de produtos de
sistemas sensores multiespectrais (Landsat TM; ETM +; OLI)
analisou-se mudan{\c{c}}as temporais (1984-2019) e investigou-se
processos em escala m{\'e}dia-pequena (Sentinel-2 A e B MSI)
(2019) sobre parcelas em regenera{\c{c}}{\~a}o da
vegeta{\c{c}}{\~a}o da Mata Atl{\^a}ntica. A metodologia
baseou-se em {\'{\I}}ndices de Vegeta{\c{c}}{\~a}o simples e
ajustados (NDVI, EVI e SAVI) e fra{\c{c}}{\~o}es espectrais
(vegeta{\c{c}}{\~a}o, solo e sombra) obtidas pelo Modelo Linear
de Mistura Espectral. Foram associadas medidas de SR e dados de
campo, de crescimento de tronco, {\'{\I}}ndice de {\'A}rea
foliar e produ{\c{c}}{\~a}o de serrapilheira, obtidos em 2019 na
{\'a}rea (EESPA-SFX). Os resultados apontam que as metodologias
combinadas de SR e levantamento in situ permitiram identificar as
idades aproximadas de tr{\^e}s {\'a}reas em diferentes
est{\'a}gios de regenera{\c{c}}{\~a}o, a saber, Ri
(regenera{\c{c}}{\~a}o inicial) com cerca de 7 anos, Rm
(regenera{\c{c}}{\~a}o intermedi{\'a}ria) com cerca de 20 anos
e, Ra (regenera{\c{c}}{\~a}o avan{\c{c}}ada), com > 40 anos. A
an{\'a}lise interanual, EVI e SAVI do Landsat mostraram
diferen{\c{c}}as significativas (p < 0,05) entre essas
{\'a}reas. Atrav{\'e}s da reamostragem simples e
itera{\c{c}}{\~a}o de 10000 vezes, o discernimento
alcan{\c{c}}ou 95% de credibilidade. Esse resultado sugere que os
dados Landsat foram sens{\'{\i}}veis na detec{\c{c}}{\~a}o de
processos de regenera{\c{c}}{\~a}o florestal em paisagens
heterog{\^e}neas, na escala do presente trabalho. Foram
selecionadas um conjunto de imagens Sentinel-2 MSI sobre o ano de
2019 e extra{\'{\i}}dos os IVs, ajustada uma biblioteca
espectral para as {\'a}reas estudadas e extra{\'{\i}}das as
fra{\c{c}}{\~o}es espectrais de cada parcela, que foram
comparados aos resultados das medidas de campo. Na an{\'a}lise
intra-anual, o NDVI do Sentinel-2 se mostrou mais
sens{\'{\i}}vel para detec{\c{c}}{\~a}o dos diferentes
est{\'a}gios (p < 0,05). A vari{\'a}vel incremento acumulado
explicou o EVI em 17% para Rm e o NDVI em 34% para Ra. A
vari{\'a}vel IAF explicou o EVI em 21% para Rm, o NDVI em 20%
para Rm e o SAVI em 18% para Rm. Em rela{\c{c}}{\~a}o {\`a}s
fra{\c{c}}{\~o}es espectrais a parcela P apresentou
separabilidade em todas as fra{\c{c}}{\~o}es. Destacamos que a
fra{\c{c}}{\~a}o solo apenas n{\~a}o separou as parcelas Ri e
Ra. Em termos de caracter{\'{\i}}sticas e evolu{\c{c}}{\~a}o
dos processos nos 3 est{\'a}gios sucessionais as medidas de
tronco apontam que a taxa de crescimento da
regenera{\c{c}}{\~a}o inicial {\'e} maior em
rela{\c{c}}{\~a}o {\`a}s outras parcelas. A Biomassa Acima do
Solo (BAS) calculada foi maior na Ra em rela{\c{c}}{\~a}o aos
outros est{\'a}gios, e apresentou diferen{\c{c}}a entre Rm e Ra.
Ainda sob reamostragem simples, as vari{\'a}veis {\'{\I}}ndice
de {\'A}rea Foliar (IAF) e serrapilheira diferenciam as parcelas
Ri e Ra, sob 95% de credibilidade. Verificou-se sincronicidade
entre as vari{\'a}veis de incremento acumulado e IAF sobre a
parcela Ri, o que sugere sincronismo entre crescimento de tronco e
folhas sobre a {\'a}rea. O cruzamento com dados de campo apontou
que na parcela Ri, a BAS e incremento acumulado, explicam
respectivamente 22% da varia{\c{c}}{\~a}o da fra{\c{c}}{\~a}o
sombra, o IAF e a serrapilheira, explicam respectivamente 22% e
17% a fra{\c{c}}{\~a}o solo (valor-p < 0,05). J{\'a} na parcela
Rm a BAS e IAF, explicaram respectivamente 18% e 23% da
varia{\c{c}}{\~a}o das fra{\c{c}}{\~o}es solo e sombra
(valor-p < 0,05), o incremento acumulado e serrapilheira n{\~a}o
apresentaram valores significativos para nenhuma
fra{\c{c}}{\~a}o espectral. Na parcela Ra o incremento acumulado
e IAF, explicam respectivamente 76% e 13% as fra{\c{c}}{\~o}es
solo e vegeta{\c{c}}{\~a}o (valor-p < 0,05), IVs e
fra{\c{c}}{\~o}es sob intervalo de credibilidade de 95%. Em
linhas gerais, esse estudo mostrou que, apesar das
limita{\c{c}}{\~o}es, foi poss{\'{\i}}vel driblar o desafio de
associar medidas por SR e medidas in situ em fragmentos de
florestas em diferentes est{\'a}gios de regenera{\c{c}}{\~a}o e
usar SR orbital, por meio de IVs tanto do Landsat quanto do
Sentinel-2, quando tratamentos apropriados s{\~a}o aplicados.
Conclui-se que os dados derivados do sat{\'e}lite Sentinel-2
apresentam o potencial para serem utilizados para
verifica{\c{c}}{\~a}o do crescimento de florestas
secund{\'a}rias, e podem ser utilizados para
caracteriza{\c{c}}{\~a}o das diferentes idades e estruturas
destas florestas. ABSTRACT: Deforestation and forest degradation
are processes that directly impact carbon stocks, biodiversity,
water reserves and the climate, promoting a wide loss of ecosystem
services. At the Paris Convention, Brazil was ratified through the
Nationally Determined Contribution (NDC), to restore and reforest
12 million hectares of forests by 2030. Most of these actions
should take place in the Atlantic Forest biome and methodologies
based on satellite images will contribute to monitoring and
verification of the agreement. Despite major technical advances in
deforestation mapping methodologies, the detection of different
degrees of forest regeneration is still incipient. In this sense,
the objective of the present work was to identify, by means of
field inventories (CONAMA resolution 1 of 1994) and remote orbital
sensing, the different ages of fragments of the Atlantic Forest in
the Experimental Soil Atmosphere Station of S{\~a}o Francisco
Xavier (EESPA- SFX), located in the city of S{\~a}o Jos{\'e} dos
Campos, SP, managed by the Ecohydrology Laboratory (LabEcoh /
CCST-INPE). Through multispectral sensor system products (Landsat
TM; ETM +; OLI), temporal changes (1984-2019) were analyzed and
processes were investigated on a small-medium scale (Sentinel-2 A
and B MSI) (2019) on regenerating parcels in Atlantic Forest
vegetation. The methodology was based on simple and adjusted
Vegetation Indices (NDVI, EVI and SAVI) and spectral fractions
(vegetation, soil and shade) obtained by the Linear Spectral
Mixture Model. Remote sensing measurements and field data, stem
growth, leaf area index and litter production, obtained in 2019 in
the (EESPA-SFX) were associated. The results show that the
combined methodologies of SR and in situ survey allowed to
identify the approximate ages of three areas in different stages
of regeneration, namely, Ri (initial regeneration) with about 7
years, Rm (intermediate regeneration) with about 20 years and, Ra
(advanced regeneration), > 40 years. Landsat's interannual
analysis, EVI and SAVI showed significant differences (pvalue
<0.05) between these areas. Through simple resampling in 10000
interactions, the confidence reached 95% credibility. This result
suggests that the Landsat data were sensitive in the detection of
forest regeneration processes in heterogeneous landscapes, on the
scale of the present work. A set of Sentinel-2 MSI images for the
year 2019 were selected, obtained by Google Earth Engine (GEE) and
the IVs were extracted, a spectral library was adjusted for the
studied areas and the spectral fractions of each plot were
extracted, which were compared results of field measurements. In
the intra-annual analysis, the Sentinel-2 NDVI was more sensitive
for detecting the different stages (pvalue <0.05). The accumulated
increment variable explained EVI in 17% for Rm and NDVI in 34% for
Ra. The LAI variable explained EVI in 21% for Rm, NDVI in 20% for
Rm and SAVI in 18% for Rm. In relation to spectral fractions, the
P portion presented description in all fractions. We emphasize
that the soil fraction did not separate the Ri and Ra plots. In
terms of characteristics and evolution of the processes in the 3
successional stages, the accumulated wood growth measures indicate
that the growth rate of the initial regeneration is higher in
relation to the other plots. The Above Ground Biomass (AGB)
calculated was higher in Ra compared to the other stages, and
showed a difference between Rm and Ra. Still under simple
resampling, the Leaf Area Index (LAI) and litter variables
differentiate the Ri and Ra plots, under 95% credibility. There
was synchronicity between the accumulated increment and LAI
variables on the Ri plot, which suggests synchrony between trunk
and leaf growth over the area. The crossing with field data showed
that in the Ri plot, the AGB and accumulated increment,
respectively explain 22% of the variation of the shadow fraction,
the LAI and the litter, respectively explain 22% and 17% of the
soil fraction (p-value <0.05). In the Rm plot, AGB and LAI
explained 18% and 23%, respectively, of the variation of soil and
shade fractions (p-value <0.05), the accumulated wood growth and
litter did not show significant values for any spectral fraction.
In the Ra plot, the accumulated wood growth and LAI, respectively,
account for 76% and 13% of the soil and vegetation fractions
(p-value <0.05), IVs and fractions under a 95% credibility
interval. In general, this study showed that, despite the
limitations, it was possible to circumvent the challenge of
associating measures by SR and measures in situ in forest
fragments in different stages of regeneration and using orbital
SR, through IVs - both from Landsat and of Sentinel-2, when
appropriate treatments are applied. It is concluded that the data
derived from the Sentinel-2 satellite have the potential to be
used to verify the growth of secondary forests, and can be used to
characterize the different ages and structures of these forests.",
committee = "Galv{\~a}o, L{\^e}nio Soares (presidente) and Borma, Laura de
Simone (orientadora) and Anderson, Liana Oighenstein (orientadora)
and Monteiro Junior, Mauro Brum",
englishtitle = "Dynamics of forest regeneration in the Atlantic Forest:
integration of multi-sensor data and field measurements",
language = "pt",
pages = "179",
ibi = "8JMKD3MGP3W34R/42EKG7L",
url = "http://urlib.net/ibi/8JMKD3MGP3W34R/42EKG7L",
targetfile = "publicacao.pdf",
urlaccessdate = "27 abr. 2024"
}