Banca de QUALIFICAÇÃO: RICARDO BRUNO DE ARAÚJO TENÓRIO

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
DISCENTE : RICARDO BRUNO DE ARAÚJO TENÓRIO
DATA : 20/03/2019
HORA: 08:30
LOCAL: Auditório do CCET
TÍTULO:

Sea ice modeling with Deep Learning in the Antarctic region


PALAVRAS-CHAVES:

Machine Learning. Long-Short Term Memmory. Sea-Ice Concentration. Southern Ocean.


PÁGINAS: 54
GRANDE ÁREA: Ciências Exatas e da Terra
ÁREA: Geociências
SUBÁREA: Meteorologia
ESPECIALIDADE: Climatologia
RESUMO:

Sea ice, covering approximately 7% of the Earth's ocean surface, is a key climate component for studying the climate in polar regions, primarily because it works as a natural barrier - a thin floating layer - at the ocean-atmosphere interface that restricts heat, mass and momentum exchanges between the sea and the air, in addition to reflecting much of the incident solar radiation. Satellite observations, since the 1970s, shows an increasingly thinner and younger sea ice Arctic, followed by a decline in its extent. While some areas of the Antarctic (e.g the Ross Sea and the Weddell Sea) have shown a slight increase in the extent of this climatic component for the same period. In contrast to these observations are simulations of Global Climate Models (GCMs), which exhibt an average decrease in ice extent for both hemispheres. In this sense, the main goal of this research is to develop a two-dimensional model of sea ice concentration that is able to simulate the observed increase of this variable in real data over Antarctica. This model is being developed with the use of Deep Learning (or Deep Neural Networks) with the Long-Short Term Memmory (LSTM) architecture. Data used as predictors and target for training the Artificial Neural Networks  belong to Era-Interim  (for the period 1979 - 2014) and  the Coupled Model Intercomparison Project Phase 6 (CMIP6, 1850 - 2014 and 2015 - 2100) databases. The first tests performed for the Weddell Sea presented a good performance for monthly and seasonal simulations of the model for 2018, with the best results obtained in the austral winter with a Mean Square Error of (0.006 +/- 0.090) %. Maps with the ice edges also showed promising results, for example, superposition of these curves (simulated and observed) in the austral summer of 2018 that presented a negative concentration anomaly in relation to the 1981-2010 average.


MEMBROS DA BANCA:
Presidente - 1808127 - JOSE HENRIQUE FERNANDEZ
Interno - 1858120 - DAVID MENDES
Interno - 2411669 - JONATHAN MOTA DA SILVA
Externo ao Programa - 2276514 - MARCIO MACHADO CINTRA
Notícia cadastrada em: 12/03/2019 10:37
SIGAA | Superintendência de Tecnologia da Informação - (84) 3342 2210 | Copyright © 2006-2024 - UFRN - sigaa08-producao.info.ufrn.br.sigaa08-producao