@PhDThesis{Petri:2024:EfIlSo,
author = "Petri, Caio Arlanche",
title = "Efeitos de ilumina{\c{c}}{\~a}o solar em atributos
espectro-espaciais da constela{\c{c}}{\~a}o planetscope obtidos
na Amaz{\^o}nia",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2024",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2023-12-15",
keywords = "constela{\c{c}}{\~o}es de sat{\'e}lites, ilumina{\c{c}}{\~a}o
solar, alta resolu{\c{c}}{\~a}o espacial, Amaz{\^o}nia,
satellite constellations, solar illumination, high spatial
resolution, Amazon.",
abstract = "Estudos fenol{\'o}gicos da vegeta{\c{c}}{\~a}o por
sensoriamento remoto orbital na Amaz{\^o}nia s{\~a}o limitados
pela pouca disponibilidade de dados multitemporais, com
observa{\c{c}}{\~o}es de alta resolu{\c{c}}{\~a}o espacial
regularmente espa{\c{c}}adas no tempo. O recente advento das
constela{\c{c}}{\~o}es de sat{\'e}lites poder{\'a} suprir
parte destas limita{\c{c}}{\~o}es. No entanto, quando a alta
resolu{\c{c}}{\~a}o espacial {\'e} utilizada em estudos de
fenologia das florestas tropicais da Amaz{\^o}nia, os efeitos das
condi{\c{c}}{\~o}es de ilumina{\c{c}}{\~a}o solar e do terreno
podem ser cr{\'{\i}}ticos para a determina{\c{c}}{\~a}o de
alguns atributos espaciais e espectrais das imagens. N{\~a}o
h{\'a}, contudo, estudos que abordem os efeitos de
ilumina{\c{c}}{\~a}o solar em dados de constela{\c{c}}{\~o}es
de sat{\'e}lites obtidos ao nadir na Amaz{\^o}nia. Este
trabalho, portanto, teve como objetivo a investiga{\c{c}}{\~a}o
de efeitos da geometria de ilumina{\c{c}}{\~a}o solar em dados
de alta resolu{\c{c}}{\~a}o espacial obtidos pela
constela{\c{c}}{\~a}o PlanetScope (PS) em florestas tropicais da
Amaz{\^o}nia. Para isso, utilizaram-se s{\'e}ries temporais,
compostas por um total de 493 imagens PS, livres de cobertura
nuvens. As imagens foram obtidas entre os anos de 2017 e 2019
sobre 12 {\'a}reas de estudo, abrangendo regi{\~o}es ao norte e
ao sul deste bioma. O processo anal{\'{\i}}tico baseou-se na
observa{\c{c}}{\~a}o da geometria de ilumina{\c{c}}{\~a}o
solar, durante a aquisi{\c{c}}{\~a}o de dados PS na
esta{\c{c}}{\~a}o seca (Junho a Setembro), e das
rela{\c{c}}{\~o}es entre {\^a}ngulos solares com: (i) atributos
espectrais de imagens PS, representados por valores de
reflect{\^a}ncia, {\'{\i}}ndices de vegeta{\c{c}}{\~a}o (IVs)
(Enhanced Vegetation Index - EVI e Normalized Difference
Vegetation Index - NDVI), brilho da vegeta{\c{c}}{\~a}o,
determinado da an{\'a}lise por componentes principais (ACP) e por
fra{\c{c}}{\~o}es-sombra extra{\'{\i}}das de um modelo linear
de mistura espectral {{{(MLME);}}} e (ii) atributos espaciais de
imagens PS, representados por m{\'e}tricas de textura Gray-Level
Co-occurrence Matrix (GLCM). Os resultados mostraram que o aumento
no {\^a}ngulo zenital solar (AZS) e a redu{\c{c}}{\~a}o no
{\^a}ngulo azimutal solar (AAS), em dire{\c{c}}{\~a}o ao final
da esta{\c{c}}{\~a}o seca, produziu um incremento consistente na
reflect{\^a}ncia das quatro bandas do PS, especialmente na banda
do infravermelho pr{\'o}ximo (IVP). De forma concordante, o
brilho da vegeta{\c{c}}{\~a}o aumentou de junho para setembro,
conforme registrado pelos valores da primeira componente principal
(CP1). O EVI, {\'{\i}}ndice mais anisotr{\'o}pico e fortemente
dependente da reflect{\^a}ncia do IVP, apresentou valores mais
altos em dire{\c{c}}{\~a}o a setembro, acompanhando as
varia{\c{c}}{\~o}es nos {\^a}ngulos solares. O NDVI, ao
contr{\'a}rio, foi pouco correlacionado com estes {\^a}ngulos. A
m{\'e}dia da textura GLCM apresentou incremento do
in{\'{\i}}cio para o final da esta{\c{c}}{\~a}o seca, enquanto
a vari{\^a}ncia da textura diminuiu no mesmo per{\'{\i}}odo.
Foram observadas correla{\c{c}}{\~o}es estatisticamente
significantes entre as m{\'e}tricas de textura GLCM e os
{\^a}ngulos solares. O EVI apresentou alta correla{\c{c}}{\~a}o
com a m{\'e}dia GLCM da textura da banda do IVP do PS, ao
contr{\'a}rio do observado para o NDVI. As
fra{\c{c}}{\~o}es-sombra, resultantes das varia{\c{c}}{\~o}es
de ilumina{\c{c}}{\~a}o entre as copas das {\'a}rvores e no
terreno, diminu{\'{\i}}ram na dire{\c{c}}{\~a}o do final da
esta{\c{c}}{\~a}o seca. Estas fra{\c{c}}{\~o}es foram
correlacionadas com o EVI, mas n{\~a}o com o NDVI. Em geral, os
resultados destacaram a import{\^a}ncia da corre{\c{c}}{\~a}o
dos efeitos de ilumina{\c{c}}{\~a}o solar e, em alguns casos, de
ilumina{\c{c}}{\~a}o do terreno, se IVs anisotr{\'o}picos como
o EVI forem utilizados em estudos de fenologia da
vegeta{\c{c}}{\~a}o com dados de alta resolu{\c{c}}{\~a}o
espacial do PS. ABSTRACT: Vegetation phenological studies using
orbital remote sensing in the Amazon are hampered by the limited
availability of high spatial resolution data and the analysis of
regularly spaced time series observations. The recent advent of
satellite constellations may overcome some of these limitations.
However, when high spatial resolution is used in phenology studies
of Amazonian tropical forests, the effects of solar and terrain
illumination conditions can be important in determining spatial
and spectral attributes from images. In this scenario, there are
currently no studies addressing the effects of solar illumination
on data from satellite constellations obtained at nadir in the
Amazon. This work aimed to investigate the effects of solar
illumination geometry on high spatial resolution data obtained by
the PlanetScope (PS) constellation in tropical forests of the
Amazon. A time series from a total of 493 PS cloud-free images was
used. The images were obtained between 2017 and 2019 over 12 study
areas, covering regions from the north and south of this biome.
The analytical process was based on the observation of solar
illumination geometry during the acquisition of PS data in the dry
season (June to September), and the relationships between solar
angles with: (i) spectral attributes of PS images, represented by
reflectance values, vegetation indices (VIs) (Enhanced Vegetation
Index - EVI and Normalized Difference Vegetation Index - NDVI),
vegetation brightness, determined from a principal component
analysis (PCA), and shade fractions extracted from a linear
spectral mixture model {{{(LSMM);}}} and (ii) spatial attributes
of PS images, represented by Gray-Level Co-occurrence Matrix
(GLCM) texture metrics. The results showed that the increase in
the solar zenith angle (SZA) and the reduction in the solar
azimuth angle (SAA), towards the end of the dry season, produced a
consistent increase in the reflectance of the four PS bands,
especially in the near-infrared band (NIR). Concordantly,
vegetation brightness increased from June to September, as
recorded by the scores of the first principal component (PC1). The
EVI, a more anisotropic index and strongly dependent on NIR
reflectance, showed higher values towards September, following
variations in solar angles. NDVI, on the contrary, was poorly
correlated with these angles. The average GLCM texture increased
from the beginning to the end of the dry season, while the texture
variance decreased in the same period. Statistically significant
correlations were observed between GLCM texture metrics and solar
angles. The EVI showed a high correlation with the GLCM average of
the PS NIR band texture, contrary to what was observed for the
NDVI. Shade fractions, resulting from variations in illumination
of tree canopies and terrain, decreased towards the end of the dry
season. These fractions were correlated with EVI, but not with
NDVI. Overall, the results highlighted the importance of
correcting for solar illumination effects, and in some cases
terrain illumination, if anisotropic VIs such as EVI are to be
used in vegetation phenology studies with high spatial resolution
data of PS.",
committee = "K{\"o}rting, Thales Sehn (presidente) and Galv{\~a}o, Lenio
Soares (orientador) and Arag{\~a}o, Luiz Eduardo de Oliveira e
Cruz de (orientador) and Moura, Yhasmin Mendes and Breunig,
F{\'a}bio Marcelo",
englishtitle = "Solar illumination effects on spectral-spatial attributes of the
planetscope constellation obtained in the Amazon",
language = "pt",
pages = "96",
ibi = "8JMKD3MGP3W34T/4AH9P78",
url = "http://urlib.net/ibi/8JMKD3MGP3W34T/4AH9P78",
targetfile = "publicacao.pdf",
urlaccessdate = "07 maio 2024"
}