1 of 31 Relationship between Pacific and Indian Ocean Sea Surface Temperature Variability and Rice Production, Harvesting Area and Yield in Indonesia Rizaldi Boer1,2, Akhmad Faqih1,2 and Rahmi Ariani2 1 Center for Climate Risk and Opportunity Management in Southeast Asia and Pacific (CCROM-SEAP), Bogor Agriculture University, Indonesia 2 Department of Geophysics and Meteorology, Bogor Agriculture University, Indonesia x E-mail : rizaldiboer@gmail.com . Abstract— This study assess the impact of ENSO on anomaly of monthly rainfall and rice production, harvesting area and yield by province and evaluate the potential use of Sea Surface Temperature Index (SST Index) covering Pacific and Indian Ocean for predicting the anomaly of production, area and yield. The SST index representing the dominant variability in the Pacific and Indian Oceans (30°E - 70°W, 30°S - 30°N) was processed using Empirical Orthogonal Function (EOF) analysis. This study found that there was significant relationship between the anomaly of the rice data and the anomaly of SST also exist in Indian Ocean. The relationships were only significant in a number of provinces particularly in the south part of Sumatra (e.g. Bengkulu), Java (West, Central and East Java), and eastern part of Indonesia (South Sulawesi, Southeast Sulawesi, North Sulawesi, Bali and 2016-02-09 2 of 31 Nusa Tenggara). The use of the EOF(1) as predictor for estimating the anomaly of rice production and harvesting is not superior to the one that used only the anomaly of SST NINO-3.4. Considering the simplicity, the use of the anomaly SST at NINO3.4 alone as a predictor for the estimation of anomaly of rice data in Indonesia is preferable. Keywords: Rice, El Nino Southern Oscillation, Sea Surface Temperature Index, Empirical Orthogonal Function Analysis. 1. Introduction Rice is the main food crop commodity cultivated by Indonesian farmers. In 2008, total rice harvesting area reached about 10.7 million ha, while the harvesting areas of other to main food crops, i.e. maize and soybean, were only about 4.0 million ha and 0.5 million ha respectively. Among these crops, rice remains vulnerable to climate variability associated with El Niño/Southern Oscillation (ENSO) events. From historical national data, it was revealed that during El Niño years rice crop was consistently affected by drought and causing significant production loss. While for the other two crops, the impact of El Nino was not consistent. Some studies suggested that this could be due to crop switching since shortage of water during El Nino years may force farmers in some regions to change their dry season rice crop to secondary crops such as maize and soybean (Boer and Subbiah, 2005). Furthermore, the studies also stated that the rice production loss due to drought in El Nino years may be offset by the increase in yield of irrigated rice crop due to higher solar radiation. Many studies on the assessment of the impact of El Nino on crop production used national statistical data (e.g. Falcon et al., 2004; Boer and Subbiah, 2005). It was found that the anomaly of 2016-02-09 3 of 31 national rice production in the next calendar year in Indonesia can be well predicted from the anomaly of August sea surface temperature (ASSTAug) at Region Nino 3.4 of the Pacific Ocean (Falcon et al, 2004). Year-to-year changes in the August ASSTAug explained 94% of the interannual variance of national rice production anomaly. Other studies in Java suggested that the impact of El Nino on rice production occurred due to its effect on the timing of the onset of rainy season. Low rainfall in September–December during El Nino years typically delays plantings until cumulative rainfall is adequate to permit the transplanting of seedlings (Heytens, 1991, Nalyor et al., 2001, 2007). Therefore, this seasonal rainfall could explain about 84% of the variance in area planted in September–December and 81% of the variance in area harvested in January–April. Further study by D’Arrigo and Wilson (2008) suggested that the use of the Nino-3.4 alone in predicting Sept-Dec rainfall over Java would be less accurate compare to the one using combined index of Indian Ocean SSTs with the Nino-3.4. Therefore, they suggested to use the combined index in developing early warning drought forecast of drought and crop failure risk in Indonesia, particularly in the western part of the country that is most influenced by Indian Ocean climate variability. This study will further assess the impact of ENSO on anomaly of monthly rainfall and rice production, harvesting area and yield by province and evaluate the potential use of SST index covering Pacific and Indian Ocean for predicting the anomaly of production, area and yield. Based on this analysis, we developed models that can be used to predict potential impact of sea surface temperature variability associate with ENSO on rice production by province. 2. Methodology This study used historical data of rice production, harvesting area and yield in all provinces in Indonesia (Figure 1) and sea surface temperature (SST) data from NCDC NOAA called as the Extended Reconstructed Sea Surface Temperature Dataset (ERSST) 2016-02-09 4 of 31 version 3b from 1983-2009 in the Pacific and Indian Oceans (30°E - 70°W, 30°S - 30°N; see Smith et al. 2008; Xue et al. 2003 for explanation of the ERSST dataset). The SST data was downloaded from the IRI Data Library website (link address: http://iridl.ldeo.columbia.edu/SOURCES/.NOAA/.NCDC /.ERSST/.version3b/.sst/). In addition we also used the calculated sea surface temperature anomaly in all four NINO regions (NINO-12 (15°S-5°S, 90°W-80°W), NINO-3 (5°S-5°N, 150°W-90°W), NINO-3.4 (5°S-5°N, 170°W-120°W) and NINO-4 (5°S-5°N, 160°E-150°W)) based on HadISST 1.1 data (Rayner et al. 2003). The long –historical observed rainfall data is taken from the gridded CRU TS2.1 dataset. This data is chosen due to its long-term record and as substitute to unavailability of rain gauge stations data from the provinces. The CRU data is a reconstructed monthly climate observations obtained from meteorological stations that comprises rainfall data and other climate variables (Mitchell & Jones 2005). The grid data covers land area with 0.5°x0.5° horizontal grid resolution for the period of 1901-1998. The analysis was divided into four parts. The first part was to identify relationship between variability of monthly sea surface temperature of the four NINO regions with monthly rainfall variability. This result of this analysis was used to select SST of the NINO regions that can explain most of the variability of monthly rainfall in Indonesia. The second part was to analyze correlation between the single monthly SST of the selected NINO Region and seasonal rice production, area and yield anomalies. This will be used to evaluate the months of the SST at the selected NINO region that significantly explain the variability of the rice data anomalies in each province. The third part was to assess spatial correlation between the SST anomalies of the Pacific and Indian Oceans (30°E - 70°W, 30°S - 30°N) of the significance months and the rice data anomalies. This analysis was to evaluate whether the variability of SST in Indian Ocean also explain significantly the variability of the rice 2016-02-09 5 of 31 data anomaly in the provinces. A seasonal SST index that represents the dominant variability in the Pacific and Indian Oceans was developed if the significant correlation with the Indian Ocean SST existed. The fourth part was to develop models for predicting the anomalies of rice data from the SST of the selected NINO region and from the seasonal SST index and to compare the performance of the prediction models developed from the two predictors. Schematically the step of the analysis is presented in Figure 2. In assessing the relationship between variability of the SST at the four NINO regions with the rainfall variability, we use the monthly data anomaly. The anomaly of monthly rainfall data was calculated using the following formula: ARij Rij R i AR(i) and R(i) were anomaly rainfall and monthly rainfall for month-i respectively, and R (i ) the long term mean of rainfall for month-i. Similar approach was also applied for the SST data of the NINO regions. Since the data used for the analysis was long-term data, i.e. from 1901-1998, significant trends may exist in the data, prior to calculation of the anomaly data, the trend in the data was removed using the regression technique. The anomaly of rice data was calculated using three methods, namely first differencing, five years moving average and polynomial regression. The formulas for the three methods are the following: 1. First Differencing: AD(i) = D(i) – D(i-1) 2. Five years moving average: AD(i) = D(i) – 1/5(D(i-2) + D(i-1) + D(i) + D(i+1) + D(i+2)) 3. Polynomial: AD(i) = D(i) – (a0+a1t + a2t2 + a3t3), where a0, …, a3 are coefficient regression that relate the rice data with time-t. AD(i) and D(i) were anomaly of the seasonal rice data and rice data for year-i. The seasonal rice data were January to April (JFMA), May to August (MJJA), and September to December (SOND). 2016-02-09 6 of 31 In order to develop seasonal SST index representing the dominant variability in the Pacific and Indian Oceans (30°E - 70°W, 30°S - 30°N), the SST data was processed using Empirical Orthogonal Function (EOF) analysis. EOF is a statistical method commonly used to identify patterns of simultaneous variation and had been widely used in many climate studies. Detailed descriptions of the method can be found for example in von Storch & Zwiers (1999), Hannaci (2004) and many others. The EOF analysis is purely based on mathematical approach, but its results represent maximum variability of the data that sometimes can be associated with physical processes (von Storch & Zwiers 1999). 3. Results and Discussions 3.1. Relationship between Monthly SST of the NINO Regions and Rainfall The result of analysis showed that monthly rainfall variability of most of Indonesian region is significantly influenced by the variability of sea surface temperature in Pacific Ocean represented by SST at NINO-12, NINO-3, NINO-3.4 and NINO-4 (Figure 3). With exception in some part of Sumatra, all Indonesian rainfall was negatively affected by the increase of sea surface temperature in the NINO regions. The decrease in monthly rainfall with one degree increase in SST of the NINO regions mostly ranged from 0 to 50 mm/ oC (Figure 4). The impact was stronger in most of Kalimantan, Sulawesi and part of Java and Papua (Figure 4). This is consistent with other study as shown in Figure 2. Figure 4 indicates that the decrease in monthly rainfall with one degree increase in SST in these regions ranged from around 20 and 40 mm, while in other regions were less than 20 mm. Among the four NINO regions, it is apparent that the SST of the NINO3.4 has stronger correlation with the rainfall variability in Indonesia (Figure 3). The correlation in 2016-02-09 7 of 31 part of Sulawesi and Kalimantan increased with NINO3.4 compare to other NINO regions. Based on this result, for subsequent analysis we only used SST of the NINO3.4. 3.2. Relationship between Rice Data Anomaly and SST Anomaly of NINO3.4 As indicated above rainfall variability has strong influenced on the decrease and increase of rice production in Indonesia. Since the rainfall is strongly influenced by the variability of SST in the NINO3.4, the use of the SST anomaly in explaining the rice production variability in Indonesia has been proposed by many studies (e.g. Falcon et al, 2004, Naylor et al., 2001; Naylor et al., 2007; Boer, 2009). In analyzing the relationship between long historical rice data and SST anomaly, the trend in the data is normally removed. In this study, we evaluated three methods for removing the trend as defined above, first differencing (FD), moving average (MA) and polynomial (PN) regression. The results of the analysis from the first differencing method were slightly different from those of the other two methods. Figure 5 shows that using the FD, significant correlation between JFMA rice production and SST with 12, 11 and 10 month lag is observed in many provinces, while using the other two methods the significant correlation is only detected in a few provinces. The correlation between the JFMA rice production anomaly and SST anomaly in NINO-3.4 increases as the time lag between these two is reduced. It was found that SST anomaly on September, October and November gave highest correlation with JFMA rice production of the following year data compare to other months (Figure 5). Based on SST anomaly data at NINO-3.4 from 1950 to present the onset of ENSO development generally occurred after April and end the cycle in April of the following year (Figure 6). The development of the ENSO will affect the onset of the rainy season in that 2016-02-09 8 of 31 year and this will influence the total area planted in JFMA of the following year as discussed above. It is not surprising that the number of provinces having significant correlation increased after April. In the subsequent analysis, the strength of the relationship between JFMA rice production anomaly and ENSO phenomenon (representing by the SST anomaly) was measured by calculating the mean of the correlation of May to December (8 months). For the MJJA, it was measured by taking the mean of the correlation of May to April (12 months), and for SOND by the mean of the correlation of May-August (4 months). Figure 7 illustrate the time frame for calculating the strength of the relationship. The result of the analysis suggests that there are 16 and 5 provinces in which the anomaly of JFMA rice production is negatively and positively affected by the increase of the SST at NINO-3.4 respectively (Figure 8). Provinces in which their rice production are consistently negatively affected by the increase of the SST irrespective of the methods used in calculating the data anomaly are Bengkulu, DKI Jakarta, West Java (Jabar), Central java (Jateng), East Java (Jatim), DI Yogyakarta, Bali, NTB, East Kalimantan (Kaltim), North Sulawesi (Sulut), South Sulawesi (Sulsel), Southeast Sulawesi (Sultenggara). While the ones that are consistently positively affected by the increase of SST are North Sumatra (Sumut) and Maluku. Referring to Figure 5, it is clear that in general the anomaly of JFMA rice production has significant correlation with the August SST anomaly at NINO-3.4. Loss of JFMA rice production for every one degree increase in August SST anomaly varied across provinces. In some provinces, the production loss could be more than 20% such as South Sulawesi (Figure 9). In other main rice growing areas of Java, the production loss for every one degree increase in SST anomaly at NINO-3.4 were generally between 10% and 20%. 2016-02-09 9 of 31 The pattern of the relationship between the anomaly of MJJA Rice production and the SST is similar to that of JFMA Rice production as shown in Figures 13. However, the direction of the relationship changes from negative to positive for some provinces such as Jambi, Bengkulu and all provinces in Java, and in Sulawesi (Figure 10). As suggested by many findings, the occurrence of El Nino will cause the delay in onset of the season and this will delay the planting time of the first rice crop, thus the total area planted in the wet season will decrease. This decrease is normally compensated by increasing the planting area in the second planting (MJJA) as shown in Figure 6. This partly explains why the rice production of MJJA has positively correlated with the SST. Different from the JFMA and MJJA, the anomaly of SOND rice production has negative correlation with the anomaly of MJJA SST in most of provinces except in Aceh and North Sumatra (Figure 11). The increase in the SST in MJJA may indicate the occurrence of El Nino. During the El Nino, most part of Indonesia experienced below normal rainfall (see Figure 2 and 12). In many cases this condition causes the harvesting failure due to drought (see Figure 7). It is understandable that the increase in SST of MJJA will have direct impact to the decrease of rice production in SOND. The correlation pattern between the anomalies of rice harvesting area and the anomaly of the SST at NINO3.4 were quite similar to those between the anomalies of the rice production and the anomaly of the SST (Figure 12 and 13). This study suggests that, the ENSO phenomena strongly affect the anomaly of rice production area and harvesting area in many Indonesian provinces, particularly in provinces located in monsoonal region (Type A; see Figure 1 and 9). 2016-02-09 10 of 31 3.3. Spatial Correlation between Rice Data Anomaly and the SST Anomaly Based on the above finding, we analyze further the spatial correlation between the anomalies of rice data and SST over Pacific and Indian Ocean (30°E - 70°W, 30°S - 30°N). This is to evaluate whether the variability of rice data is also affected by the variability of SST in Indian Ocean. We limit the analysis only for the relationship between the anomaly of rice data and the anomaly of August-October SST. The result of analysis is presented in Figures 14-16. It was shown that the significant relationship between the anomaly of the rice data and the anomaly of SST also exist in Indian Ocean, but the relationship is in different direction compare to those with the Pacific Ocean. Therefore we evaluate the potential use of the seasonal SST index representing the variability of the SST in the Pacific and Indian Ocean. We process the SST data using empirical orthogonal function. It is very apparent that the first mode of EOF is the dominant pattern associated with ENSO (Figure 17), and this first EOF explains most of the variation of the SST over the NINO regions. In the second mode of EOFs, the SST anomalies loading patterns seem to indicate either the warming signals or decadal variations signified by the increasing trends and low-frequency oscillations found in their PC scores. 3.4. Relationship between Rice Data Anomaly and the Anomaly of SON SST at NINO3.4 and the First EOF The result of the analysis shows that the use of the EOF(1) as predictor for estimating the anomaly of rice production and harvesting is not superior to the one used the anomaly of SST NINO-3.4. Figure 18 presents the result of analysis for West Java case. In other 2016-02-09 11 of 31 provinces, the result of the analysis is also the same as that of West Java. The use of the anomaly SST at NINO3.4 as a predictor for the estimation of anomaly of rice data in Indonesia is preferable, since we do not need to process of SST data using the EOF before making prediction. To improve the predictive skill, D’Arrigo and Wilson (2008), proposed to use of the Nino-3.4 together with the index of Indian Ocean SSTs. The use of the first EOF as single predictor is not good enough to accommodate the contribution of SST variation in the Indian Ocean. 4. Conclusion and Recommendation The variability of rice production and harvesting area in Indonesia is strongly related to the variability of the SST at NINO3.4. We can use the anomaly of the SST at NINO3.4 directly for predicting the anomaly of rice production and harvesting area. The use of the EOF that represents the variation of SST over Pacific and Indian Ocean do not improve the skill. To improve the skill of rice anomaly prediction, further study to evaluate the inclusion of SST anomaly over Indian Ocean is required. Acknowledgement The author acknowledges FAO and SEARCA that provided support for this study 2016-02-09 12 of 31 References Boer, R. 2009. Impacts of climate change in agriculture. Country case study submitted to International Food Policy Research Institute, Washington, DC. Unpublished Report. Boer, R., and A.R. Subbiah. 2005. Agricultural droughts in Indonesia. In V.K. Boken, A.P. Cracknell, and R.L. Heathcote. Monitoring and Predicting Agriculture Drought. Oxford University Press, p:330-344. D’Arrigo R, Wilson R. 2008. El Niño and Indian Ocean influences on Indonesian drought: implications for forecasting rainfall and crop productivity. Int J Clim 28:611–616. doi:10.1002/joc.1654 Falcon, W.P., R.L. Naylor, W.L. Smith, M.B. Burke and E.B. McCullough. 2004. Using Climate Models to Improve Indonesian Food Security, Bulletin of Indonesian Economic Studies, 40: 355–77 Hannachi A, 2004, A primer for EOF analysis of climate data, Department of Meteorology, University of Reading, Reading RG6 6BB, U.K., March 24, 2004 Heytens, P.: 1991, ‘Rice Production Systems’, in Pearson, S., Falcon, W., Heytens, P., Monke, E., and Naylor, R. (eds.), Rice Policy in Indonesia, Cornell University Press, Ithaca, NY, pp. 38–57. Mitchell TD, Jones PD (2005) An improved method of constructing a database of monthly climate observations and associated high-resolution grids. International Journal of Climatology 25:693-712 Naylor R, Battisti D, Vimont D, Falcon W, Burke M. 2007. Assessing risks of climate variability and climate change for Indonesian rice agriculture. Proceedings of the National Academy of Sciences of the United States of America 104: 7752–7757. 2016-02-09 13 of 31 Naylor, R.L., W.P. Falcon, D. Rochberg and N. Wada. 2001. Using el niño/southern oscillation climate data to predict rice production in indonesia. Climatic Change 50: 255–265. Rayner NA, Parker DE, Horton EB, Folland CK, Alexander LV, Rowell DP, Kent EC & Kaplan, A, 2003, Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century, J. Geophys. Res., vol. 108, no. D14, doi: 4407 10.1029/2002JD002670. Smith TM, Reynolds RW, Peterson TC & Lawrimore J, 2008, Improvements to NOAA's Historical Merged Land-Ocean Surface Temperature Analysis (1880-2006). J. Climate, 21, 2283-2296. Meinke, H. and R. Boer. 2002. Plant growth and the SOI. In I.J. Partridge and M. Ma’shum (ed) Will It Rain? The effect of the Southern Oscillation and El Niño in Indonesia. Queensland Government, Department of Primary Industry, Australia. p. 25-28. von Storch H & Zwiers F, 1999, Statistical analysis in climate research, Cambridge University Press, Cambridge, United Kingdom, pp. 293-312 Xue Y, Smith TM & Reynolds RW, 2003, Interdecadal changes of 30-yr SST normals during 1871-2000. J. Climate, 16, 1601-1612. 2016-02-09 14 of 31 Figure 1. Name of provinces in Indonesia. Note: Since the rice production data used started from 1979, some of provinces which have been expanded recently have to be united back. These include Papua Tengah, Papua Timur and Papua Barat which were united back as Papua Province, and Maluku Utara and Maluku which were united back as Maluku Province 2016-02-09 15 of 31 Figure 2. Steps of analysis for evaluating the relationship between sea surface temperature variability and rice production, area and yield anomaly. 2016-02-09 16 of 31 Figure 3. Spatial correlation between monthly rainfall anomaly and of SST anomaly of NINO-12, NINO-3, NINO-3.4 and NINO-4 regions (10% of significance Level) 2016-02-09 17 of 31 Figure 4. Change in monthly rainfall with the one degree increase in sea surface temperature of NINO-12, NINO-3, NINO-3.4 and NINO-4 regions 2016-02-09 18 of 31 Rice Production JFMA(t)_FD Rice Production JFMA(t)_FD 0.7 Mean of Correlation Number of Provinces 26 24 22 20 18 16 14 12 10 8 6 4 2 0 0.6 0.5 0.4 0.3 0.2 0.1 0 10% Significance Level 5% Significance Level 10% Significance Level 1% Significance Level 0.7 26 24 22 20 18 16 14 12 10 8 6 4 2 0 Mean of Correlation Number of Provinces 1% Significance Level Rice Production JFMA(t)_PN Rice Production JFMA(t)_PN 0.6 0.5 0.4 0.3 0.2 0.1 0 10% Significance Level 5% Significance Level 1% Significance Level 10% Significance Level Rice Production JFMA(t)_MA 26 24 22 20 18 16 14 12 10 8 6 4 2 0 10% Significance Level 5% Significance Level 1% Significance Level Rice Production JFMA(t)_MA 0.7 Mean of Correlation Number of Provinces 5% Significance Level 0.6 0.5 0.4 0.3 0.2 0.1 0 5% Significance Level 1% Significance Level 10% Significance Level 5% Significance Level 1% Significance Level Figure 5. Number of provinces that show significant correlation between JFMA rice production anomaly and SST anomaly at NINO-3.4 (left) and mean correlations (absolute values) from May-December (right) 2016-02-09 2.00 2.00 1.50 1.50 Anomlay SST at NINO-3.4 Anomlay SST at NINO-3.4 19 of 31 1.00 0.50 0.00 -0.50 -1.00 -1.50 -2.00 1.00 0.50 0.00 -0.50 -1.00 -1.50 -2.00 J FM AM J J A S O N D J FM AM J J A S O N D J FM AM J J A S O N D J FM AM J J A S O N D Figure 6. Pattern of mean monthly of SST Anomaly of El Nino years and La-Nina Years at NINO3-4 (19502010) 2016-02-09 20 of 31 M J J A S O N D J F M A M J J A May(t-1)-Dec(t-1) Rice Data(t) May(t-1)-April(t) Rice Data(t) May(t0-Aug(t) S O N D Rice Data(t) Figure 7. Time frame used for calculating the correlation between the anomaly of rice data SST. 2016-02-09 Production JFMA vs SST NINO3.4 Papua Maluku Sultenggara Sulsel Sulteng Sulut Kalsel Kaltim Kalbar Kalteng NTT Bali First Differencing NTB Jatim Jateng Moving Average DI Yogya DKI Jakarta Lampung Sumsel Bengkulu Jambi Riau Sumbar Aceh Polinomial Jabar 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 Sumut Mean Correlation of May(t-1)-Dec (t-1) 21 of 31 Figure 8. Mean of the correlation between the anomaly of JFMA Rice Production and Anomaly of SST at NINO3.4 2016-02-09 22 of 31 Yd<10%: 10%<=Yd<15%: 15%<=Yd<20%: Yd>20%: Figure 9. Percent JFMA Rice Production Loss for every one degree increase in Anomaly of August SST at NINO3.4 2016-02-09 23 of 31 Production MJJA vs SST NINO3.4 Mean Correlation of May(t-1)-Apr(t) 0.8 Polinomial 0.6 Moving Average First Differencing 0.4 0.2 0.0 -0.2 -0.4 -0.6 Papua Maluku Sultenggara Sulsel Sulteng Sulut Kalsel Kaltim Kalbar Kalteng NTT Bali NTB Jatim Jateng DI Yogya Jabar DKI Jakarta Lampung Sumsel Bengkulu Jambi Riau Sumbar Aceh Sumut -0.8 Figure 10. Mean of the correlation between the anomaly of MJJA Rice Production and Anomaly of SST at NINO3.4 2016-02-09 Harvesting Area MJJA vs SST NINO-3.4 Papua Maluku Sultenggara Sulsel Sulut Sulteng Kalsel Kaltim Kalbar Kalteng First Differencing NTT Bali NTB Jatim DI Yogya Moving Average Jabar Lampung DKI Jakarta Bengkulu Jambi Sumsel Riau Sumbar Aceh Polinomial Jateng 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 Sumut Mean Correlation of May(t-1) to Apr(t 24 of 31 Figure 11. Mean of the correlation between the anomaly of SOND Rice Production and Anomaly of SST at NINO3.4 2016-02-09 25 of 31 Papua Maluku Sultenggara Sulsel Sulut Sulteng Kalsel Kaltim Kalbar Kalteng First Differencing NTT Bali NTB Jatim DI Yogya Jabar Moving Average Jateng Lampung DKI Jakarta Bengkulu Jambi Sumsel Riau Sumbar Aceh Polinomial Sumut Mean Correlation of May(t-1) to Dec(t-1) Harvesting Area JFMA vs SST NINO-3.4 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 Figure 12. Mean of the correlation between the anomaly of JFMA Rice Harvesting Area and Anomaly of SST at NINO3.4 2016-02-09 Harvesting Area MJJA vs SST NINO-3.4 Papua Maluku Sultenggara Sulsel Sulut Sulteng Kalsel Kaltim Kalbar Kalteng First Differencing NTT Bali NTB Jatim DI Yogya Moving Average Jabar Lampung DKI Jakarta Bengkulu Jambi Sumsel Riau Sumbar Aceh Polinomial Jateng 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 Sumut Mean Correlation of May(t-1) to Apr(t 26 of 31 Figure 13. Mean of the correlation between the anomaly of JFMA Rice Harvesting Area and Anomaly of SST at NINO3.4 2016-02-09 27 of 31 Figure 14. Spatial correlation between Anomaly of ASO SST (t-1) and the anomaly of JFMA Rice Production (t). 2016-02-09 28 of 31 Figure 15. Spatial correlation between Anomaly of Aug-Oct SST (t-1) and the anomaly of May-Aug Rice Production (t). 2016-02-09 29 of 31 Figure 16. Spatial correlation between Anomaly of Jun-Aug SST (t) and the anomaly of Sep-Dec Rice Production (t). 2016-02-09 30 of 31 Figure 17. Spatio-temporal of seasonal SST loading factors (SON) over the Pacific and Indian Ocean based on EOF analysis 2016-02-09 31 of 31 West Java y = -129.9x + 0.036 R² = 0.362 2.00 3.00 Anomaly Harvested Area Anomaly JFMA Harvested Area West Java 3.00 1.00 0.00 -1.00 y = -0.719x + 0.126 R² = 0.431 2.00 1.00 0.00 -1.00 -2.00 -2.00 -3.00 -0.01 -0.005 0 0.005 0.01 0.015 -3.00 -2 First EOF 2.00 1.00 0.00 -1.00 3.00 y = -0.728x + 0.183 R² = 0.467 2.00 1.00 0.00 -1.00 -2.00 -2.00 -3.00 -0.01 3 West Java y = -137.3x + 0.091 R² = 0.427 Anomaly Harvested Area Anomaly JFMA Production West Java 3.00 -1 0 1 2 Anomaly of Aug-Oct SST NINO3.4 -0.005 0 0.005 First EOF 0.01 0.015 -3.00 -2 -1 0 1 2 Anomaly of Aug-Oct SST NINO3.4 3 Figure 18. Regression equations relating the anomaly of rice production and harvesting area with the first EOF and the anomaly of SST at NINO3.4. 2016-02-09