Prediction of the Onset of Rainy Seasons in South Sulawesi Province Using Sea Surface Temperature Anomaly Data.
Prediksi Awal Musim Hujan di Provinsi Sulawesi Selatan Menggunakan Data Anomali Suhu Muka Laut
Abstract
The Province of South Sulawesi is one of the centers for national paddy production with 7% contribution from total paddy production in national level. The increasing of paddy production depends on the water availability. Moreover, the water availability is always associated with the onset of rainy seasons, because if an area has the forward or backward of onset of rainy season, it will influence the water availability and crop calendar. In order to onset of rainy seasons, to reduce the farmer’s loss and to protect the food stability level in Indonesia, it needs to identified the areas which have the sensitivity for the onset of rainy season’s variability to climate control factors like ENSO and IOD. This research has been conducted to study the correlation between onset of rainy seasons (AMH) in South Sulawesi and sea surface temperature anomaly (AnoSML) in Pacific Ocean and Sulawesi Sea, to determine the domain of sea surface temperature predictor which has influence the onset of rainy seasons in South Sulawesi and to develop the onset of rainy season model prediction in South Sulawesi based on the sea surface temperature anomaly in Pacific Ocean and Sulawesi Sea. This research has been conducted using daily rainfall data during 30 years (1980-2010) from rainfall stations in South Sulawesi, sea surface temperature anomaly data for 4 domains i.e. Nino 3 region, Nino 3.4, Nino 4 and Sulawesi Sea on June, July, August and September (JJAS) during 30 years. The first analysis of data was conducted with the examination of observation daily rainfall data and selection of rainfall stations which are used for the development of the rainy season onset prediction model. Further analysis was conducted to determine the onset of rainy seasons using Moron’s method (five days of daily rainfall data/pentad analysis). The result of pentad analysis was used to calculate Principle Component Analysis (PCA) which was used to determine the cluster of rainfall stations. The correlation analysis was then calculated between the onset rainy seasons on each cluster and AnoSML data on JJAS. The last analysis was conducted for development of onset rainfall season model prediction and verification of the model. Based on the checking of observation daily rainfall data with the data availability >70%, there were 45 chosen stations for onset of the rainy season analysis. The result of AMH analysis showed that the average of AMH for whole stations was on the 348th Julian Date (JD) or on 14th December. The highest AMH was obtained at Stasiun BPP. Panincong (PNCG) in Soppeng District with 473th Julian Date or on 18th April. Otherwise, the lowest AMH was obtained at Stasiun BPP. Keera (KERA) in Wajo District with 278th Julian date or on 5th October. The result of PCA analysis showed that there were six PCs which were explained variability of data ± 80%. Based on the PC clustering, there were 3 clusters of rainfall stations. Cluster 1 has 13 rainfall stations which are spread in 9 districts i.e. Bantaeng District, Bulukumba, Enrekang, Gowa, Luwu, Maros, Pinrang, Soppeng and Wajo District. On Cluster 2, there are 10 rainfall stations which are spread in Barru District, Bulukumba, Enrekang, Gowa, Luwu, Makassar, Maros, Soppeng, Takalar and Wajo District. On Cluster 3, there are 12 rainfall stations in Bulukumba District, Jeneponto, Gowa, Luwu, Maros, Pangkep, Pinrang, Soppeng, Takalar. The average of AMH at each cluster in South Sulawesi has different value. On cluster 1, the average of AMH is on 344th Julian Date (JD) or on 10 December with the standard deviation is 30. On Cluster 2, the average of AMH is on 318th JD or on 14 November with standard deviation value is smaller than in Cluster 1 (standard deviation=16). On Cluster 3, the standard deviation value is the higher than Cluster 1 and Cluster 2. This result showed that Cluster 3 has a high sensitivity to ENSO events. The spatial distribution of correlation on Cluster 1 and 2 showed the significant correlation at 95% confidence level between AMH in South Sulawesi and AnoSML in Pacific Ocean and Sulawesi Sea on June, July, August and September (JJAS). On Cluster 3, the significant correlation only found on June which is concentrated on East Indonesia, whereas there was no significant correlation with Pacific Ocean and Sulawesi Sea on July, August and September. The domain predictor in Pacific Ocean (coordinate: 148oW, 12oS) has more significant influence at 95% confidence level to AMH in South Sulawesi than other domains. This result was showed with the highest correlation value (0.82). The domain of Sulawesi Sea has the influence to AMH in South Sulawesi. This statement is strengthened by grid which has influence AMH on Cluster 2 and 3 i.e. grid 28oBT, 2oLU and 126oBT, 2oLU (grid of Sulawesi Sea). Moreover, AMH on each cluster in South Sulawesi has different correlation level and sensitivity to AnoSML in Pacific Ocean and Sulawesi Sea. Based on the verification of onset prediction model using four years data (2006, 2007, 2008, and 2009) on each cluster, the result showed that the smallest error (RMSE) found on the Cluster 3 with RMSE=3, while RMSE for Cluster 1 and Cluster 2 respectively is 16 and 29. This showed that linear regression model is good enough to used for predict the AMH on Cluster 3 which has a small standard deviation.