@MastersThesis{Mariano:2015:DeAvSe,
author = "Mariano, Denis Araujo",
title = "Detec{\c{c}}{\~a}o e avalia{\c{c}}{\~a}o de seca
agron{\^o}mica atrav{\'e}s da an{\'a}lise de s{\'e}ries
temporais de dados MODIS e PERSIANN",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2015",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2015-03-16",
keywords = "seca agron{\^o}mica, MODIS, PERSIANN, s{\'e}ries temporais,
transformada wavelet, agricultural drought, MODIS, PERSIANN,
time-series, wavelet transform.",
abstract = "Eventos de seca causam grande preju{\'{\i}}zo para a agricultura
brasileira, sendo a regi{\~a}o Sul frequentemente castigada por
esse fen{\^o}meno. Essas ocorr{\^e}ncias prejudicam gravemente a
cadeia agr{\'{\i}}cola nacional, o que causa
oscila{\c{c}}{\~a}o de pre{\c{c}}os, onera produtores e
empresas. Tais ocorr{\^e}ncias evidenciam a necessidade de
sistemas para monitorar e quantificar a seca com
informa{\c{c}}{\~o}es precisas e espacializadas. O presente
trabalho objetivou mensurar em termos de intensidade e
abrang{\^e}ncia os eventos de seca agron{\^o}mica ocorridos no
Paran{\'a} para as safras de ver{\~a}o de 2002 a 2013,
considerando milho e soja. Foram utilizados dados dos espectros
vis{\'{\i}}vel, infravermelho e termal do \emph{Moderate
Resolution Imaging Spectroradiometer} (MODIS) e de
precipita{\c{c}}{\~a}o estimada por sensoriamento remoto (SR)
oriundos do produto \emph{Precipitation Estimation from Remotely
Sensed Information Using Artificial Neural Networks} (PERSIANN).
V{\'a}rios {\'{\i}}ndices de vegeta{\c{c}}{\~a}o (IV) foram
avaliados, tendo estes em geral apresentado maior
correla{\c{c}}{\~a}o com a precipita{\c{c}}{\~a}o de outubro
para os anos secos e janeiro para os normais, indicando que, para
o desempenho da vegeta{\c{c}}{\~a}o, estes meses foram os mais
relevantes para os respectivos tipos de anos. Foram criadas linhas
de base para cada IV, considerando a mediana hist{\'o}rica para
cada data fenol{\'o}gica. Os IVs foram ent{\~a}o avaliados para
caracteriza{\c{c}}{\~a}o da seca, tendo o \emph{Land Surface
Water Index} (LSWI) se destacado, apresentando maior
correla{\c{c}}{\~a}o com m{\'e}tricas de
precipita{\c{c}}{\~a}o e produtividade agr{\'{\i}}cola. Foram
estudadas as rela{\c{c}}{\~o}es entre o LSWI e \emph{Land
surfasse temperature} (LST), sendo detectada uma
rela{\c{c}}{\~a}o inversa entre as vari{\'a}veis. Foram
analisadas as transformadas de \emph{wavelet [wavelet transform}
(WT)] para cada vari{\'a}vel e utilizado os m{\'e}todos da WT
cruzada (XWT) e coer{\^e}ncia de WT (WCT). N{\~a}o foi
verificada rela{\c{c}}{\~a}o de causa-efeito entre as
vari{\'a}veis, mas sim, uma rela{\c{c}}{\~a}o de espelho, ou
seja, outros fatores governam o comportamento de LSWI e LST.
Segundo a literatura, o principal fator {\'e} a umidade do solo,
a qual {\'e} bem correlacionada com o LSWI. Por fim, as
diferen{\c{c}}as acumuladas entre LSWI de cada safra e sua linha
de base (LSWI-dif) foram espacializadas na forma de mapas para
cada ano-safra, o que visualmente foi bastante coerente com os
mapas de precipita{\c{c}}{\~a}o acumulada. O m{\'e}todo de
mensura{\c{c}}{\~a}o de seca agron{\^o}mica proposto se mostrou
eficiente e potencialmente aplic{\'a}vel para fins de
monitoramento agr{\'{\i}}cola, tendo como maior
limita{\c{c}}{\~a}o a resolu{\c{c}}{\~a}o espacial dos dados
utilizados. ABSTRACT: Drought events strike Brazilian agriculture
causing yield losses, being the south region often stricken by
this phenomenon. These occurrences lead to negative impacts in the
agricultural chain by causing commodities prices fluctuation and
hampering farmers and companies finances condition. A need for a
system for monitoring and retrieving drought on time and
spatialized information regarding the agriculture emerges in this
context. The present study aimed at measuring and quantify the
intensity and geographical spreading of the agricultural drought
phenomena occurred in Paran{\'a} state between the 2002 and 2013
summer seasons, considering maize and soybean crops. Remote sensed
reflectance and thermal data from Moderate Resolution Imaging
Spectroradiometer (MODIS) sensors and precipitation from the
Precipitation Estimation from Remotely Sensed Information Using
Artificial Neural Networks (PERSIANN) were used as input for the
methods. Several vegetation indices (VI) were tested and
generally, they were better correlated to accumulated
precipitation in October for dry years and January for normal
years, showing that these months are crucial for the vegetation
condition regarding these kind of years. Using phenology and
historical data, baselines for each index were created in a median
basis. The VIs were then analyzed in order to better assess the
agricultural drought occurrences being the Land Surface Water
Index (LSWI) the most well suited for this task. LSWI showed
better correlation to precipitation metrics and estimated yield.
The relationships between LSWI and Land Surface Temperature (LST)
were studied as well, showing an inverse correlation between then.
The wavelet transform (WT) were used in each variable and the
cross WT (XWT) and WT coherence (WCT) methods were applied for
testing these relations. Cause-effect relationships were not
found, instead, LST and LSWI simply mirrored each other, this
means that other variables govern the LST and LSWI behavior,
according to the literature, the main factor is the soil moisture
which is also well correlated to LSWI. Finally, the accumulated
differences between LSWI for each season and the baseline
(LSWI-dif ) were spatialized and mapped being possible to
visualize the coherence between precipitation and LSWI-dif maps.
The proposed method proved itself on being well suitable for
agricultural monitoring needs, however, the main hurdle was the
spatial resolution of the input data.",
committee = "Moreira, Maur{\'{\i}}cio Alves (presidente/orientador) and
Formaggio, Ant{\^o}nio Roberto and Sanches, Ieda Del'Arco and
Galv{\~a}o, L{\^e}nio Soares and Zullo J{\'u}nior, Jurandir",
copyholder = "SID/SCD",
englishtitle = "Agricultural drought detection and assessment through MODIS and
PERSIANN time-series data analysis",
language = "pt",
pages = "110",
ibi = "8JMKD3MGP3W34P/3HU5BP5",
url = "http://urlib.net/ibi/8JMKD3MGP3W34P/3HU5BP5",
targetfile = "publicacao.pdf",
urlaccessdate = "15 jun. 2024"
}