The research analyzed rainfall data from Subang and Karawang as the centers of rice production in West Java.  The objectives of this research were to   (1) develop monthly rainfall prediction model for predicting the next four months rainfall, (2) develop a next three months rice yield prediction model and (3) estimate the availability of rice in Subang and Karawang as a function of monthly rainfall.  Both rainfall and rice yield prediction models were built by ANN technique.  ANN rainfall prediction model was applied at six rainfall stations in Subang and Karawang which are Cigadung, Karawang, Rawamerta, Subang, Sindanglaya and Ciseuti.  It was developed by including 7-8 variables (X) at input layer and 6-10 nodes at a single hidden layer.  Variables at input layer are month code (t) as X1, monthly rainfall values at t, t+1, t+2, and t+3 as X2, X3, X4, and X5 respectively, SOI at t as X6 and SST anomalies at t and t+3 as X7 and X8.  Rice yield model was built to estimate the rice production at t+3 by using four variables at input layer which are t, t+1, t +2 and t+3 as X1, X2, X3 and X4 and also included 6-8 nodes at hidden layer.  The results of this research found that the ANN model could accurately predict the monthly rainfall in all stations with the R2 values ranged from 64-96%, and maximum errors of each month rainfall ranged from 0.4-3.4 mm/month.  Rainfall model predicted that there were trends of Above Normal (AN) rainfall at Karawang and Rawamerta stations in dry season, while at four stations in Subang region would experience Below Normal (BN) rainfall in dry season.  Based on 2009 rainfall prediction, the rice yield model predicted highest rice production to happen during February and March 2009 at values of 299.294 ton and 329.082 ton.

 

Key words: artificial neural network, rainfall prediction, rice production

Magfira Syarifuddin, Yonny Koesmaryono, Aris Pramudia

Abstract


The research analyzed rainfall data from Subang and Karawang as the centers of rice production in West Java.  The objectives of this research were to   (1) develop monthly rainfall prediction model for predicting the next four months rainfall, (2) develop a next three months rice yield prediction model and (3) estimate the availability of rice in Subang and Karawang as a function of monthly rainfall.  Both rainfall and rice yield prediction models were built by ANN technique.  ANN rainfall prediction model was applied at six rainfall stations in Subang and Karawang which are Cigadung, Karawang, Rawamerta, Subang, Sindanglaya and Ciseuti.  It was developed by including 7-8 variables (X) at input layer and 6-10 nodes at a single hidden layer.  Variables at input layer are month code (t) as X1, monthly rainfall values at t, t+1, t+2, and t+3 as X2, X3, X4, and X5 respectively, SOI at t as X6 and SST anomalies at t and t+3 as X7 and X8.  Rice yield model was built to estimate the rice production at t+3 by using four variables at input layer which are t, t+1, t +2 and t+3 as X1, X2, X3 and X4 and also included 6-8 nodes at hidden layer.  The results of this research found that the ANN model could accurately predict the monthly rainfall in all stations with the R2 values ranged from 64-96%, and maximum errors of each month rainfall ranged from 0.4-3.4 mm/month.  Rainfall model predicted that there were trends of Above Normal (AN) rainfall at Karawang and Rawamerta stations in dry season, while at four stations in Subang region would experience Below Normal (BN) rainfall in dry season.  Based on 2009 rainfall prediction, the rice yield model predicted highest rice production to happen during February and March 2009 at values of 299.294 ton and 329.082 ton.

 

Key words: artificial neural network, rainfall prediction, rice production

Full Text:

PDF