The paper describes about rainfall zoning and rainfall prediction modeling and its use for rice availability and vulnerability analysis.  The study used rainfall data from Station Baros (Banten region), Station Karawang and Station Kasomalang Subang (Northern Coastal of West-Java), and Station Tarogong (Garut).  Fuzzy clustering methods, that was applied for rainfall zoning, used the representative data for El-Nino, La-Nina and normal means condition during 1980-2006 periods.  Neural network analysis technique was applied for rainfall prediction modeling.  Training set of the model based on the rainfall data of 1990-2002 periods, and validation model based on data of 2003-2006 periods.  The model were used to predict the rainfall of 2007-2008 periods.  The distibution of equivalence value between rainfall stations was very variative under El-Nino, La-Nina and Normal condition.  On the certain of equivalence level it could be derivated some different rainfall zone under El-Nino, La-Nina and normal condition.  Model training set could explain 88% of Baros rainfall variability, 89% of Karawang rainfall variability, and 72% of Kasomalang rainfall variability.  At Baros, Karawang and Subang, rainfall was predicted to be increased on November 2007-February 2008 period, and to be decreased on the March-June 2008, and to be increased on July-November 2008.  The rainfall decreasing on the March-June would carry a losses of rice production up to 25%.  But, applying the well irrigation management and suitable growing periods could decrease and mitigate the decreasing of paddy production.

 

Key words: rainfall prediction model, fuzzy clustering, neural network analysis, rice vulnerability

Aris Pramudia, Yonny Koesmaryono, Irsal Las, Tania June, I Wayan Astika, Eleonora Runtunuwu

Abstract


The paper describes about rainfall zoning and rainfall prediction modeling and its use for rice availability and vulnerability analysis.  The study used rainfall data from Station Baros (Banten region), Station Karawang and Station Kasomalang Subang (Northern Coastal of West-Java), and Station Tarogong (Garut).  Fuzzy clustering methods, that was applied for rainfall zoning, used the representative data for El-Nino, La-Nina and normal means condition during 1980-2006 periods.  Neural network analysis technique was applied for rainfall prediction modeling.  Training set of the model based on the rainfall data of 1990-2002 periods, and validation model based on data of 2003-2006 periods.  The model were used to predict the rainfall of 2007-2008 periods.  The distibution of equivalence value between rainfall stations was very variative under El-Nino, La-Nina and Normal condition.  On the certain of equivalence level it could be derivated some different rainfall zone under El-Nino, La-Nina and normal condition.  Model training set could explain 88% of Baros rainfall variability, 89% of Karawang rainfall variability, and 72% of Kasomalang rainfall variability.  At Baros, Karawang and Subang, rainfall was predicted to be increased on November 2007-February 2008 period, and to be decreased on the March-June 2008, and to be increased on July-November 2008.  The rainfall decreasing on the March-June would carry a losses of rice production up to 25%.  But, applying the well irrigation management and suitable growing periods could decrease and mitigate the decreasing of paddy production.

 

Key words: rainfall prediction model, fuzzy clustering, neural network analysis, rice vulnerability

Full Text:

PDF