@MastersThesis{Maciel:2019:AbMu,
author = "Maciel, Daniel Andrade",
title = "Quantifica{\c{c}}{\~a}o remota da concentra{\c{c}}{\~a}o de
s{\'o}lidos totais e inorg{\^a}nicos em suspens{\~a}o em lagos
da plan{\'{\i}}cie de inunda{\c{c}}{\~a}o do Baixo Amazonas:
uma abordagem multi-sensor",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2019",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2019-02-25",
keywords = "Landsat-8, Sentinel-2, CBERS-4, corre{\c{c}}{\~a}o
atmosf{\'e}rica, corre{\c{c}}{\~a}o de glnt, Landsat-8,
Sentinel-2,CBERS-4, atmosphere correction, glint correction.",
abstract = "A utiliza{\c{c}}{\~a}o de imagens de sensoriamento remoto {\'e}
de fundamental import{\^a}ncia para aumentar o conhecimento sobre
a din{\^a}mica da troca de sedimentos entre o Rio Amazonas e as
plan{\'{\i}}cies de inunda{\c{c}}{\~a}o j{\'a} que ela pode
ajudar a entender como as mudan{\c{c}}as clim{\'a}ticas e de uso
da terra influenciam esse processo. Neste sentido, este trabalho
investigou a acur{\'a}cia de algoritmos de estimativa de TSS
(Total de S{\'o}lidos em Suspens{\~a}o) e TSI (Total de
S{\'o}lidos Inorganicos em suspens{\~a}o) atrav{\'e}s da
utiliza{\c{c}}{\~a}o de tr{\^e}s sensores de m{\'e}dia
resolu{\c{c}}{\~a}o espacial (Landsat-8/OLI, Sentinel-2A/MSI e
CBERS-4/WFI) em lagos na plan{\'{\i}}cie de
inunda{\c{c}}{\~a}o do Baixo Amazonas. Atrav{\'e}s de
simula{\c{c}}{\~a}o Monte Carlo, foram calibrados e validados
algoritmos emp{\'{\i}}ricos e semi-anal{\'{\i}}ticos a partir
dados de reflet{\^a}ncia de sensoriamento remoto (Rrs) medidos
in-situ e simulados para os tr{\^e}s sensores (Rrs_sim) e dados
de TSS e TSI coletados simultaneamente ao longo de quatro
campanhas de campo em lagos do baixo Amazonas. Para
calibra{\c{c}}{\~a}o dos algoritmos, tr{\^e}s conjuntos de
dados foram avaliados: Conjunto completo, separados por campanhas
e separados por lagos. Ap{\'o}s a calibra{\c{c}}{\~a}o dos
algoritmos, estes foram aplicados a uma cena de agosto de 2017 de
cada sensor para sua valida{\c{c}}{\~a}o com dados de TSS e TSI
in-situ. Al{\'e}m da valida{\c{c}}{\~a}o dos dados de TSS e
TSI, avaliou-se tamb{\'e}m o desempenho de diversos m{\'e}todos
de corre{\c{c}}{\~a}o atmosf{\'e}rica para o OLI (6S, ACOLITE,
L8SR), MSI (6S, ACOLITE, Sen2Cor) e WFI (6S) e tamb{\'e}m de
corre{\c{c}}{\~a}o de glint para o OLI e MSI tomando-se as Rrs
simuladas a partir das medidas de Rrs in-situ como
refer{\^e}ncia. Finalmente, avaliou-se a congru{\^e}ncia entre
os dados de TSS e TSI estimados pelos tr{\^e}s sensores em
imagens adquiridas no mesmo dia da passagem dos tr{\^e}s
sat{\'e}lites afim de avaliar a possibilidade da
cria{\c{c}}{\~a}o de constela{\c{c}}{\~o}es virtuais com estes
sensores. O desempenho dos algoritmos com os dados in-situ mostrou
resultados similares para as faixas espectrais equivalentes dos
tr{\^e}s sensores avaliados e tamb{\'e}m resultados semelhantes
para os algoritmos emp{\'{\i}}ricos e semi-anal{\'{\i}}ticos
que utilizam a mesma faixa espectral. A valida{\c{c}}{\~a}o das
corre{\c{c}}{\~o}es atmosf{\'e}ricas mostrou uma
depend{\^e}ncia da faixa espectral utilizada e melhores
resultados utilizando o 6S. J{\'a} a corre{\c{c}}{\~a}o de
glint se mostrou satisfat{\'o}ria e com grande influ{\^e}ncia
principalmente sobre a acur{\'a}cia do sensor MSI
(Redu{\c{c}}{\~a}o nos valores de MAPE > 100%). Os algoritmos
emp{\'{\i}}ricos e semi-anal{\'{\i}}ticos de estimativa de TSS
e TSI apresentaram melhores resultados de valida{\c{c}}{\~a}o
usando a banda do verde do sensor OLI (561 nm), do red-edge do
sensor MSI (704 nm) do vermelho do sensor WFI (660 nm) quando
aplicado {\`a}s cenas de agosto de 2017 utilizando o os
algoritmos calibrados com o conjunto completo (MAPE < 31%). A
compara{\c{c}}{\~a}o das estimativas de TSS e TSI a partir de
imagens simult{\^a}neas dos tr{\^e}s sensores indicou que eles
permitiram estimar as concentra{\c{c}}{\~o}es de TSS e TSI com
diferen{\c{c}}as entre as medianas das concentra{\c{c}}{\~o}es
inferior a 1 mgL-1. Estes resultados permitiram, pela primeira
vez, a calibra{\c{c}}{\~a}o e valida{\c{c}}{\~a}o de
algoritmos emp{\'{\i}}ricos e semi-anal{\'{\i}}ticos de TSS e
TSI em lagos da plan{\'{\i}}cie de inunda{\c{c}}{\~a}o do
Baixo Amazonas utilizando sensores de m{\'e}dia
resolu{\c{c}}{\~a}o espacial. ABSTRACT: Remote sensing (RS) is a
key tool for deepening the knowledge on the spatial and temporal
dynamics of sediment exchange between Amazon River and their
floodplains. Moreover, RS image can help to understand how both
climate change and land use and land cover changes influence the
sediment exchange between the Amazon River and floodplain lakes.
In that sense, this study investigates the accuracy of Total
Suspended Solids (TSS) and Total Inorganic Suspended Solids (TSI)
estimates of Amazon floodplain lakes derived from medium
resolution sensors (Landsat-8/OLI, Sentinel-2A/MSI and CBERS-
4/WFI). Empirical and semi-analytical algorithms were calibrated
and validated through a robust Monte Carlo simulation using both
in-situ simulated remote sensing reflectance (Rrs_sim) and
simultaneous TSS/TSI dataset collected over four field campaigns
in the lower Amazon floodplain lakes. For algorithm calibration,
three different datasets were evaluated: Complete dataset;
Campaign dataset and Lake dataset. After the calibration process,
calibrated algorithms were applied to an august/2017 scene of each
sensor for validation using in-situ TSS and TSI concentration
measurements. Despite TSS and TSI validation, the performance of
several atmosphere correction methodologies for OLI (L8SR, 6S,
ACOLITE), MSI (6S, ACOLITE, Sen2Cor) and WFI (6S) in Rrs retrieval
were evaluated using in-situ Rrs,sim as a reference. Furthermore,
the impacts of glint correction on OLI and MSI Rrs retrieval were
also evaluated. Finally, the consistency between TSS and TSI
estimates by each sensor was accessed using near-simultaneous
imagery aiming to create a virtualconstellation based on those
three sensors to support the generation of sediment products. The
performance of in-situ algorithms demonstrates similar estimates
for similar spectral bands disregarding the sensor and the type of
algorithm (empirical or semi-analytical). Atmosphere correction
validation presented a dependency on the spectral bands used and
better results were obtained using 6S, although satisfactory
results were also observed with other methods. Moreover, glint
correction presented good results and being fundamental to the
accuracy of the algorithms based on MSI imagery, reducing MAPE
values higher beyond 100%. Empirical and semi-analytical TSS and
TSI algorithms best results varied for each sensor when applied to
August/2017 scenes: for OLI the best result was for the green band
(561 nm) while for MSI the best result was for the red-edge band
(704 nm) and for WFI the red band (660 nm) presented best results
(MAPE values lower than 31% for both TSS and TSI) using algorithms
calibrated with the Complete dataset. The comparison between TSS
and TSI estimates using the near-simultaneous overpass indicated
that they allowed sediment estimates with median difference values
lower than 1 mgL-1. These results demonstrated, for the first
time, the calibration and validation of empirical and
semi-analytical algorithms for TSS and TSI retrieval over lower
Amazon Floodplain Lakes using medium-resolution sensors.",
committee = "Kampel, Milton (presidente) and Novo, Evlyn M{\'a}rcia Le{\~a}o
de Moraes (orientadora) and Carvalho, Lino Augusto Sander de
(orientador) and Oliveira, Nat{\'a}lia Rudorff and Montanher,
Ot{\'a}vio Cristiano and Costa, Maycira",
englishtitle = "Remote quantification of inorganic and total suspended solids over
Lower Amazon floodplain lakes: a multisensor aproach",
language = "pt",
pages = "194",
ibi = "8JMKD3MGP3W34R/3SLFNB5",
url = "http://urlib.net/ibi/8JMKD3MGP3W34R/3SLFNB5",
targetfile = "publicacao.pdf",
urlaccessdate = "28 mar. 2024"
}