@InProceedings{GomesJśniorTcheSouzChou:2022:BiCoAn,
author = "Gomes J{\'u}nior, S{\'e}rgio Pinto and Tcheou, Michel Pompeu and
Souza Filho, Jo{\~a}o Baptista de Oliveira and Chou, Sin Chan",
affiliation = "{Universidade Federal do Rio de Janeiro (UFRJ)} and {Universidade
do Estado do Rio de Janeiro (UERJ)} and {Universidade Federal do
Rio de Janeiro (UFRJ)} and {Instituto Nacional de Pesquisas
Espaciais (INPE)}",
title = "A bias correction analysis for SST data from BESM in the tropical
South Atlantic Ocean",
year = "2022",
organization = "Workshop em Modelagem Num{\'e}rica de Tempo, Clima e
Mudan{\c{c}}as Clim{\'a}ticas Usando o Modelo Eta: Aspectos
F{\'{\i}}sicos e Num{\'e}ricos (WorEta), 7.",
publisher = "INPE",
keywords = "bias correction, sea surface temperature, BESM.",
abstract = "Sea surface temperature (SST) is an important variable that drives
climate. Tropical Atlantic Ocean SST variability has a strong
influence on the distribution of precipitation in South America,
including northeastern Brazil and the southwestern Amazon region.
Therefore, to seek for more accurate rainfall forecasts in those
regions, our work aimed at evaluating different bias correction
statistical methods to be applied onto the SST data predicted by
the Brazilian Earth System Model (BESM). The reference
observations were obtained from the ERA5 database. The chosen
region to produce corrected SST is within the coordinates 30W-10E
and 20S-0. The first method evaluated was the basic Quantile
Mapping (QM), considering its wide adoption, ability to deal with
higher order moments, and computationally efficiency. However, it
has some drawbacks, such as requiring the same number of samples
in the time series observed and the one generated by BESM to
ensure an accurate distribution mapping. In addition, QM also
assumes that the error correction function for the modeled and
observed distributions are stationary or time invariant. To
address the first limitation, we evaluate the modified QM method.
Furthermore, since the stationarity assumption of QM models may be
not suitable for long-term observation series, this experimental
study also includes the Scaled Distribution Mapping (SDM) method.
Records are daily based and span the period from November 1980 to
October 2010. The bias correction procedures were individually
applied to each model grid point at the study region. This
investigation indicates the existence of some correlation between
spatially contiguous areas and best performing methods.",
conference-location = "Online",
conference-year = "26-30 set. 2022",
ibi = "8JMKD3MGP3W34T/47MGJMB",
url = "http://urlib.net/ibi/8JMKD3MGP3W34T/47MGJMB",
targetfile = "TO_03_A1_GomesJuniorSP.pdf",
urlaccessdate = "20 maio 2024"
}