TY - GEN
T1 - Applying artificial neural network models for ENSO prediction using SOI and Nino3 as onset indicators
AU - Baawain, M. S.
AU - Nour, M. H.
AU - El-Din, M. G.Gamal
PY - 2003
Y1 - 2003
N2 - ENSO, El Nino southern-oscillation, is known to be the strongest climatic variation on seasonal to inter-annual time scales. It causes severe droughts, floods, fires, and hurricanes leading to economical disasters. Hence, ENSO prediction is crucial to help in preserving world economy. This study explores the use of relatively simple inputs in developing an artificial neural networks (ANN) model for predicting ENSO, identifying the best ENSO indicator for neural networks, and assessing the impact of different time steps on the model efficiency. Two types of ENSO indicators, Southern Oscillation Index (SOI) and NiñoS, were used to develop neural networks models in this study. The use of SOI as an ENSO indicator performed much better compared to NiñoS. ANN model using a monthly time span proved to have a high potential of success. The devised monthly model was a three-layer multi-layer perceptron model trained with error backpropagation algorithm using the logistic activation function for hidden and output layers. The model showed high stability due to the similar R2values when the training and testing data patterns in the Input data sets were swapped.
AB - ENSO, El Nino southern-oscillation, is known to be the strongest climatic variation on seasonal to inter-annual time scales. It causes severe droughts, floods, fires, and hurricanes leading to economical disasters. Hence, ENSO prediction is crucial to help in preserving world economy. This study explores the use of relatively simple inputs in developing an artificial neural networks (ANN) model for predicting ENSO, identifying the best ENSO indicator for neural networks, and assessing the impact of different time steps on the model efficiency. Two types of ENSO indicators, Southern Oscillation Index (SOI) and NiñoS, were used to develop neural networks models in this study. The use of SOI as an ENSO indicator performed much better compared to NiñoS. ANN model using a monthly time span proved to have a high potential of success. The devised monthly model was a three-layer multi-layer perceptron model trained with error backpropagation algorithm using the logistic activation function for hidden and output layers. The model showed high stability due to the similar R2values when the training and testing data patterns in the Input data sets were swapped.
UR - http://www.scopus.com/inward/record.url?scp=33645713748&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=33645713748&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:33645713748
SN - 1894662040
SN - 9781894662048
T3 - Proceedings, Annual Conference - Canadian Society for Civil Engineering
SP - 858
EP - 867
BT - CSCE 31st Annual Conf. Proceedings
T2 - Canadian Society for Civil Engineering - 31st Annual Conference: 2003 Building our Civilization
Y2 - 4 June 2003 through 7 June 2003
ER -