@InProceedings{CintraCampCock:2016:MuPeDa,
author = "Cintra, Rosangela Saher Correa and Campos Velho, Haroldo Fraga de
and Cocke, Steven",
affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)} and {Instituto
Nacional de Pesquisas Espaciais (INPE)} and {Florida State
University}",
title = "Multilayer perceptron on data assimilation applied to FSU global
model",
booktitle = "Proceedings...",
year = "2016",
organization = "International Symposium on Uncertainty Quantification and
Stochastic Modeling, 3. (Uncertainties)",
keywords = "data assimilation, artificial neural networks, ensemble Kalman
filter, multilayer perceptron.",
abstract = "Numerical weather prediction (NWP) uses atmospheric general
circulation models (AGCMs) to predict weather based on current
weather conditions. The atmosphere could not be completely
described due to inherent uncertainty. These uncertainties limit
forecast model accuracy to about five or six days into the future.
The process of entering observation data into mathematical model
to generate the accurate initial conditions is called data
assimilation (DA). This paper shows the results of a DA technique
using artificial neural networks (NN) applied to an AGCM used in
Florida State University (FSU) in USA. The Local Ensemble
Transform Kalman filter (LETKF), a version of Kalman filter with
ensembles to represent the model uncertainties, is a traditional
DA scheme. We use Multilayer Perceptron data assimilation (MLP-DA)
with supervised training algorithm where NN receives input vectors
with their corresponding response from LETKF initial conditions.
These DA schemes are applied to FSU Global Spectral Model
(FSUGSM), a multilevel spectral primitive equation model at
resolution T63L27. This data assimilation experiment is based in
synthetic observations: surface pressure and upper-air
temperature. We use a NN self-configuration method to find the
optimal NN parameters to configure the MLP-DA with: four input
vector nodes and one output node for the analysis vector. The NNs
were trained with data from each month of 2001, 2002, and 2003.
The MLP-DA cycle is performed for January 2004. The numerical
results demonstrate the effectiveness of the MLP-DA technique for
atmospheric data assimilation, since the initial conditions have
similar quality to LETKF. The reduced computational cost allows
the inclusion of greater number of observations and new data
sources and the use of high resolution of models.",
conference-location = "Maresias, SP",
conference-year = "15-19 Feb.",
urlaccessdate = "27 abr. 2024"
}