@PhDThesis{Oliveira:2018:OtMoMe,
author = "Oliveira, Bruno Silva",
title = "Otimiza{\c{c}}{\~a}o do modelo metric para estimativas de
evapotranspira{\c{c}}{\~a}o no cerrado brasileiro",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2018",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2018-08-27",
keywords = "Cerrado, evapotranspira{\c{c}}{\~a}o, balan{\c{c}}o de energia
na superf{\'{\i}}cie, MODIS, GLDAS, cerrado, evapotranspiration,
surface energy balance, MODIS, GLDAS.",
abstract = "O bioma Cerrado tem sido foco da aten{\c{c}}{\~a}o mundial,
devido {\`a} sua grande diversidade biol{\'o}gica e {\`a}s
mudan{\c{c}}as de uso e cobertura da terra que sofre com a
expans{\~a}o da agricultura. Essas mudan{\c{c}}as causam
impactos no balan{\c{c}}o h{\'{\i}}drico e de energia e,
consequentemente, na evapotranspira{\c{c}}{\~a}o (ET),
contribuindo para as mudan{\c{c}}as clim{\'a}ticas local e
global. Os modelos de ET com dados de sensoriamento remoto
s{\~a}o atualmente a forma mais adequada para representar a
din{\^a}mica espacial e temporal dessa vari{\'a}vel, no entanto
sua aplica{\c{c}}{\~a}o {\'e} limitada pela disponibilidade de
dados meteorol{\'o}gicos de superf{\'{\i}}cie. Neste estudo o
modelo METRIC (Mapping EvapoTranspiration at high Resolution with
Internalized Calibration) foi aplicado em quatro regi{\~o}es
heterog{\^e}neas no Cerrado brasileiro, avaliando-se sua
performance em estimar o balan{\c{c}}o de energia e a ET. As
estimativas foram avaliadas por compara{\c{c}}{\~a}o com fluxos
observados por sistemas eddy covariance instalados sobre
{\'a}reas de cana-de-a{\c{c}}{\'u}car (s{\'{\i}}tio USR),
Cerrado denso (s{\'{\i}}tio PDG) e Cerrado sensu-stricto
(s{\'{\i}}tio RECOR). O submodelo de albedo da
superf{\'{\i}}cie foi ajustado utilizando simula{\c{c}}{\~o}es
de irradi{\^a}ncia solar do modelo SMARTS2 e dados de radiometria
coletados em campo. O METRIC foi aplicado inicialmente em sua
forma padr{\~a}o, com imagens MODIS e dados meteorol{\'o}gicos
observados nos s{\'{\i}}tios experimentais. No segundo momento,
para tornar o modelo independente de dados meteorol{\'o}gicos de
superf{\'{\i}}cie, melhorar a resolu{\c{c}}{\~a}o espacial dos
resultados e reduzir o tempo de processamento, as
simula{\c{c}}{\~o}es foram realizadas aplicando-se somente dados
MODIS e dados meteorol{\'o}gicos de rean{\'a}lise (GLDAS),
desagrega{\c{c}}{\~a}o da banda termal do MOD11 (de 1km para 250
metros) e automatiza{\c{c}}{\~a}o da sele{\c{c}}{\~a}o de
pixels {\^a}ncora para a calibra{\c{c}}{\~a}o interna do
modelo. O ajuste no submodelo de albedo levou a desvios entre 10-
21% e m{\'a}ximo RMSE=0,03. O algoritmo de
desagrega{\c{c}}{\~a}o da banda termal permitiu simular a
temperatura da superf{\'{\i}}cie com R²>0,68 e RMSE < 2,1ºC. O
METRIC em sua forma padr{\~a}o obteve resultados consistentes
para o saldo de radia{\c{c}}{\~a}o (Rn) (R²>0,94 e RMSE<75 W
m-2), por{\'e}m subestimou esse fluxo em 8 e 9% na
cana-de-a{\c{c}}{\'u}car e no Cerrado denso, respectivamente. O
fluxo de calor latente (LE) foi estimado com precis{\~a}o nos
s{\'{\i}}tios PDG e RECOR, por{\'e}m subestimado em 5% no
s{\'{\i}}tio USR (R²=0,94 e RMSE=42,88 W m-2). A ET real foi
estimada com precis{\~a}o nos s{\'{\i}}tios USR e PDG,
por{\'e}m foi superestimada em 14% no s{\'{\i}}tio RECOR
(R²=0,96 e RMSE=0,75 mm dia-1). O modelo ajustado e alimentado por
dados meteorol{\'o}gicos do GLDAS obteve estimativas de Rn
similares {\`a}s da primeira aplica{\c{c}}{\~a}o, por{\'e}m o
esse fluxo foi superestimado em 7% nos s{\'{\i}}tios USR e PDG
(R²>0,93 e RMSE<80 W m-2). Com essa configura{\c{c}}{\~a}o o
modelo superestimou a ET em 25% e 26% na
cana-de-a{\c{c}}{\'u}car e no Cerrado sensu-stricto,
respectivamente. As m{\'e}dias zonais extra{\'{\i}}das para
diferentes classes de cobertura da terra do mapa do TerraClass
Cerrado mostraram que entre as quatro {\'a}reas de estudo
avaliadas os fluxos de energia e a ET s{\~a}o similares quanto
{\`a} magnitude e varia{\c{c}}{\~a}o sazonal, mostrando-se
fortemente influenciados pela energia dispon{\'{\i}}vel,
precipita{\c{c}}{\~a}o, umidade no solo e NDVI. ABSTRACT: The
Brazilian Cerrado has been the focus of world attention due to its
great biological diversity and land cover/land use changes, caused
by the expansion of agricultural activities. These changes cause
impacts on water and energy balance, and, consequently, on
evapotranspiration (ET), contributing to local and global climate
change. The ET models that use remote sensing data as inputs are
currently considered the best method to represent the spatial and
temporal dynamic of this variable, however, its application is
limited by the availability of surface meteorological data. In
this study we evaluated the potential of METRIC (Mapping
EvapoTranspiration at high Resolution with Internalized
Calibration) model to estimate the energy balance fluxes and ET
over four heterogeneous Cerrado regions. The estimates were
evaluated by comparing them to flux tower data collected over
sugarcane (USR site), woody cerrado (PDG site) and stricto-sensu
cerrado (RECOR site) areas. The broadband albedo submodel was
adjusted by applying solar radiation simulations with the SMARTS2
model and radiometry data collected during fieldworks. METRIC was
initially applied in its standard form, with MODIS images and
meteorological data observed at the experimental sites. In the
second phase, in order to the model become independent of surface
meteorological data, to improve the spatial resolution of the
outputs, and to reduce processing time, the simulations were
performed by applying only MODIS and reanalysis (GLDAS) data;
spatial sharpening of MOD11 thermal band (1km to 250 meters) and
automation of anchor pixels selection for the internal
calibration. The surface albedo model adjustment led to deviations
between 10-21% and maximum RMSE=0.03. The global algorithm for
thermal disaggregation allowed to simulate the land surface
temperature with R²>0.68 and RMSE<2.1ºC. METRIC in its standard
form presented consistent results for net radiation (Rn)(R²>0.94
and RMSE<75 W m- 2), however it underestimated this flux by 8 and
9% in sugarcane and woody cerrado, respectively. Latent heat flux
(LE) was accurately estimated at PDG and RECOR sites but
underestimated by 5% at USR site (R²=0.94 and RMSE=42.88 W m-2).
METRICderived ET estimates showed agreement with ground data at
USR and PDG sites, but at RECOR site ET was overestimated by 14%
(R²=0.96 and RMSE=0.75 mm day-1). METRIC model adjusted and fed by
GLDAS simulated obtained Rn results close to those obtained in the
first simulation but overestimated this flux by 7% at USR and PDG
sites (R²>0.93 and RMSE<80 W m-2). With this configuration, the
model overestimated ET by 25% and 26% in sugarcane and
stricto-sensu savanna, respectively. The zonal means extracted for
different land cover types (TerraClass Cerrado map) showed that,
among the four study areas evaluated, the surface energy fluxes
and ET are similar in magnitude and seasonal variation, with
direct dependence on available energy, precipitation, soil
moisture and NDVI.",
committee = "Shimabukuro, Yosio Edemir (presidente) and Moraes, Elisabete Caria
(orientadora) and Sanches, Ieda Del'Arco and Zeri, Luis Marcelo de
Mattos and Oliveira, Gabriel de",
englishtitle = "Optimization of metric model for evapotranspiration estimates in
the brazilian cerrado",
language = "pt",
pages = "188",
ibi = "8JMKD3MGP3W34R/3RKS7A2",
url = "http://urlib.net/ibi/8JMKD3MGP3W34R/3RKS7A2",
targetfile = "publicacao.pdf",
urlaccessdate = "27 abr. 2024"
}