@InProceedings{CintraCamp:2012:SaOb,
author = "Cintra, Rosangela Saher Correa and Campos Velho, H. F.",
affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)} and {Instituto
Nacional de Pesquisas Espaciais (INPE)}",
title = "Global Temperature Assimilation using Artificial Neural Networks
in SPEEDY Model: Satellite Observation",
booktitle = "Abstracts...",
year = "2012",
organization = "European Geosciences Union (EGU) General Assembly.",
abstract = "An Artificial Neural Network (ANN) is designed to investigate a
application for data assimilation. This procedure provides an
appropriated initial condition to the atmosphere to numerical
weather prediction (NWP). The NWP incorporates the equations of
atmospheric dynamics with physical process and it can predict the
future state of the atmosphere. Data assimilation procedure
combines information from observations and from a prior short-term
forecast producing an current state estimate. Operational
satellite data are taken and processed in real-time and
distributed around the world. The use of observations from the
earth-orbiting satellites in operational NWP provides large data
volumes and increases the computational effort. The goal here is
to simulate the process for assimilating temperature data computed
from satellite radiances and introduce new technique in analysis
to Weather Forecasting and climate. This performance can be faster
than conventional schemes for data assimilation. The numerical
experiment is carried out with global model: the Simplified
Parameterizations, primitivE-Equation DYnamics (SPEEDY) and the
synthetic observations of temperatures from model plus a random
noise. For the data assimilation technique was applied a
Multilayer Perceptron (MLP-NN) with supervised training, which
observation, local point observation and the Local Ensemble
Transform Kalman Filter (LETKF) analysis are used as input vector.
The global analysis is done in the activation MLP-NN with only,
synthetic observation and its local point. In this experiment, the
MLP-ANN was trained with the first six months considering the
years 1982, 1983, and 1984 data. A hindcasting experiment for data
assimilation performed a cycle for January of 1985 with MLP-NN and
SPEEDY model. LETKF was performed at the same cycle. The results
for MLP-NN analysis are very close with the results obtained from
LETKF. The simulations show that the major advantage of using ANN
is the better computational performance, with similar quality of
analysis. The CPU-time assimilation with MLP-NN is 80% less than
LETKF with the same observations.",
conference-location = "Viena",
conference-year = "22 a 27 de abril de 2012",
language = "en",
targetfile = "cintra_global.pdf",
urlaccessdate = "07 maio 2024"
}