The Local Characteristics of Indonesian Seas and Its Possible Connection with ENSO and IOD : Ten Years Analysis of Satellite Remote Sensing Data

The overall objective of this study is to investigate the local characteristics of the Indonesian Seas in response to the 6 months, seasonal and inter-annual variations. A moving averages method with 3, 6 and 12 months of lengths has been used to show the temporal variability. Using satellite observed datasets for the 10 year period from December 1999-November 2009, the results of moving averages indicate the 6 months and seasonal variabilities of sea surface temperature (SST), zonal-component of wind speed (U-WS) and rain rate (RR) in most of the Indonesian Seas area. It is also interesting to note that, although in the same latitude with the strong seasonal variability area, a stable condition of SST occurs in some particular areas. The cross-lag correlation analysis shows the relationship among indices with several month time lags. After removing the seasonal variability, the possible connection of indices with the El Nino-Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD) phenomena in the Indonesian Seas can be detected by the 6 month moving averages method.


Introduction
The Indonesian Seas, a semi-closed marginal sea, separates the Pacific Ocean from the Indian Ocean (see Figure 1) with a chain of big and small islands.The bathymetry is generally shallow (<100 m) throughout the western area of the inner Indonesian Seas, while the abyssal sea basin in the east of the inner Indonesian Seas has the depths of 5000 m.The Indonesian Seas lie on and near the Equator including "warm pool" with its high temperature.The region also has an important pathway for water transport (Gordon et al., 2003(Gordon et al., , 2010;;Sprintall et al., 2004;Gordon, 2005), the Indonesian throughflow (ITF).ITF affects both the regional circulation and thermal structure (Godfrey 1996;England & Huang, 2005).
The general characteristic of the Indonesian seas was reported with the satellite remote-sensing dataset (Swardika et al., 2012).It concluded that the ocean characteristics, the sea surface temperature (SST), zonal-component of wind speed (U-WS) and rain rate (RR), are very stable in the Equatorial areas, while their seasonal variability appears as increasing the latitude in both the north and south hemispheres.The two abnormal ocean phenomena have been studied since the 1990s: the El Nino-Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD).ENSO is the anomaly of the wide weather system including ocean, atmosphere and land.Its indicator, or the ENSO index, is the SST in the central tropical Pacific Ocean (Trenberth et al., 1998;Diaz et al., 2001;Li et al., 2010;Weng et al., 2009;Pradhan et al., 2011).IOD is the similar phenomenon as ENSO and its indicator is the SST difference between the eastern and western tropical Indian Ocean (Saji & Yamagata, 2003;Weng et al., 2009;Luo et al., 2010).It is of great interest whether the ENSO and IOD affect the local characteristics of the Indonesian Seas.
The aim of this study is to analyse the ocean characteristics and to show the temporal variability of ocean characters for the local areas in the Indonesian Seas.Then, we compare those temporal variabilities with the ENSO and IOD indices to consider their influence to the Indonesian Seas.We also show the locality of those variabilities and the correlation of indices.
Figure 1.The bottom topography map (scale in meter) of the Indonesian Seas derived from ETOPO 2 minutes (Smith and Sandwell 1997).
Axes indicate longitude 80˚ -180˚ E and latitude 20˚ S -20˚ N. Areas with deep water are shaded in light grey, areas with shallow water are shaded in dark grey and areas with no available data are in white

Method of Analysis
We reveal the ocean characteristics from the monthly SST, U-WS and RR in 10 years (December 1999 -November 2009) by satellites observation.Those three data show the clear cyclic variability of the ocean characteristics (Swardika et al., 2012;Nurweda & Tanaka, 2012).
To determine the local variability, we calculate the average of the indices over each local area from the monthly grid data, which is downloaded from the satellite data archives.In order to analyse the temporal variability of indices in the Indonesian Seas, we need to take account of the 6 months variability because the solar looking angle becomes at zenith twice a year.
To detect the seasonal variability, we calculate the 6 months moving averages in which no 6 months variability appears.If there exists a longer than 12 months cyclic variability, it appears in the 12 months moving averages.We should know whether those longer cyclic variabilities exist.Thus we calculate the 6 and 12 months moving averages.
To detect the 6 months variability, we calculate the 3 months moving averages.If a seasonal variability exists, it also appears in the 6 months moving averages.But we can distinguish the 6 months variability from the seasonal variability by carefully expecting the average.
The moving averages for the month-i are calculated as follows: ; 3 for the 6 months moving averages ; 6 for the 12 months moving averages (1) ; 1 for the 3 months moving averages where x i is the original monthly data and X i is the monthly data of the moving averages.The value of X i corresponds to the middle of the month.
The ENSO index is defined with the 3 months moving averages of SST in the central tropical Pacific Ocean.If we compare the indices in the Indonesian Seas with the ENSO index, we should take account of both 6 months and seasonal variabilities.As we will show later, all the indices in the ENSO Index area have neither 6 months nor seasonal variabilities.For the comparison, we need the indices free from the 6 months and seasonal variabilities.Thus, the connections of indices with the ENSO signal are observed by the 6 months moving averages of the deseasonalized indices.The deseasonalized (d i ) indices can be calculated as follows: where: (4) W i is the monthly data with no temporal variability longer than 12 months; x i is the original monthly data and X i is calculated by the Equation 1 with k = 6.
The auto-correlation function (R xx ) is expressed as follows: (8) where: To identify the relationship among the indices, the cross-correlation is calculated.The cross-correlation (R xy ) is expressed as follows: (12) where:  (Saji et al., 1999) Positive IOD: The different of WTIO and SETIO is warmer than 30-year (1971-2000) average.
We analyse the whole Indonesian Seas, from 80˚ -180˚ E in longitude and from 20˚ S -20˚ N in latitude (see Figure 1), which spans the tropics of both Indian and Pacific Oceans.In order to analyze the local characteristics, the Indonesian Seas is divided into three regions with twelve areas.These three regions represent the south subtropical region, the near-equatorial region and the north subtropical region.The twelve local areas are (see

Results and Discussion
The 3, 6 and 12 months moving averages for all areas are shown in Figures 2 to 13.The summaries of the seasonal and 6 months variabilities are presented in Tables 2 and 3, respectively.The auto-correlation of indices is presented in Table 4.The correlation coefficients and time lag among indices are presented in Table 5.
The moving averages analysis shows that all indices in area J have neither 6 months nor seasonal variability (see Figures 11a, b and 20).The ENSO event is defined as the SST which stays for longer than 3 months outside the boundaries ± 0.5 (see Figure 21).We summarise the El Nino and La Nina years in Tables 6 and 7.
The 6 months moving averages of DMI show that the index is affected by the 6 months variability (see Figure 22).Thus, in this paper the abnormal year of IOD is considered only for the year of 2007.
From the all 12 months averages, we observed that there is no longer than 12 months variability, or more precisely, the cyclic variability of longer than 12 months in the Indonesian Seas.

The Seasonal Variability
The 6 months moving averages of SST, U-WS and RR identify the signal of seasonal variability.In the High latitude region, all areas show a strong seasonal variability of all indices except for the U-WS in area B. In the Equatorial region, areas D, E and F (the inner Indonesian Seas) shows a strong seasonal variability of all indices except for the SST in area D. The other areas in the Equatorial region show a weak seasonal variability of all indices.It is noted that the ENSO Index area, area J, has no seasonal variability of all indices.

The Auto-Correlation and Cross-Correlation Analyses
We further calculate the auto-correlation of SST, U-WS and RR to find out the temporal variability of indices for those areas with a weak seasonal variability: the areas G, H, I and J in the Equatorial Pacific Ocean (see Figures 14 to 16).These areas have no clear temporal variability of SST and RR, except for areas H.The area H shows combined seasonal and 6 months variability of SST.The temporal variability of U-WS in all areas shows the seasonal variability, except for the ENSO Index area, area J.
We also calculate the cross-correlation among the indices.In the High latitude region, all areas show a strong correlation among indices, except for the area B. The lag time of SST and U-WS for area A is -2 months (it means that the SST comes 2 months later after the U-WS), while those for areas K and L are -1 month.The lag times of U-WS and RR for the areas A, K and L are -2, 0 and 1 months, respectively.The lag times of SST and RR for the areas A, K and L are 2, 1 and 0 months, respectively.The area B only shows a strong correlation between SST and RR with 1 month lag time.
In the Equatorial region, areas D, E and F (the inner Indonesian Seas) show a strong correlation among indices.
The lag time of SST and U-WS for area D is 5 months, while those for areas E and F are 0 month.The lag times of U-WS and RR for areas D, E and F are 0, 2 and 1 months, respectively.The lag time of SST and RR for area D is 0 month, while those for areas E and F are 1 month.The area C only shows a strong correlation between SST and U-WS with 4 months lag times.The ENSO Index area, area J, has no lag time among indices.

The Local Characteristics
The inner Indonesian Seas, areas D, E and F, show the seasonal variability of all indices.However, the area D has more stable SST than the area E, while both areas have a similar U-WS pattern (see Figures 5 and 6).The difference of SST between areas D and E might be caused by the ocean bathymetry (see Figure 1).
The U-WS in the inner Indonesian Seas switches from west to east and vice versa, while it does not change its direction in the outside areas of the inner Indonesian Seas (see Figures 5 to 7).This is the monsoon wind pattern of the inner Indonesian Seas.Saveral studies have described the monsoon pattern of the inner Indonesian Seas (Wyrtki, 1962;Susanto et al., 2006), but the pattern was unclear (Swardika et al., 2012).

Conclusions
The local characteristics of SST, U-WS and RR in the Indonesian Seas are analysed using satellite remote sensing datasets from December 1999 -November 2009.The seasonal variability of the indices is strong in the High latitude region and the inner Indonesian Seas.The 6 months variability of SST is weakly detected in the several areas of the Equatorial region.The inner Indonesian Seas has a strong correlation among indices, but the lag times among indices vary and we can not find out any discernment regarding them.The ENSO signal of SST can be detected in the High latitude Pacific Ocean and the Equatorial region, except for the inner Indonesian Seas.The inner Indonesian Sea only shows the ENSO signal of SST in the year 2007, but it, rather, might be caused by the positive IOD.We show the variabilities and connections of the 12 areas to the ENSO and IOD, but it is difficult to obtain any finding that fits all the areas.One typical locality is the SST difference between the areas D and E.
The 3 months moving averages of SST, U-WS and RR identify the signal of the 6 months variability.In the High latitude region, all areas have no 6 months variability of SST and U-WS, except for the area B. The area B shows a very weak 6 months variability of U-WS.All areas in the High latitude region show a very weak 6 months variability of RR.In the Equatorial region, areas D, E and H show a weak 6 months variability of SST.The other areas in the Equatorial region only show a very weak 6 months variability of SST.The areas H and I show a very weak 6 months variability of U-WS.The other areas in the Equatorial region have no 6 months variability of U-WS.The inner Indonesian Seas, areas D, E and F have no 6 months variability of RR, while the other areas in the Equator show a very weak 6 months variability of RR.It is noted that the ENSO Index area, area J, has no 6 months variability of all indices.

Figure 2 .
Figure 2. Moving averages of SST (upper panel), U-WS (middle panel) and RR (lower panel) in area A The 3 months moving averages are shown in panel (a), (d) and (g); the 6 months moving averages are shown in panel (b), (e) and (h); the 12 months moving averages are shown in panel (c), (f) and (i).The red line indicates the moving averages result, the blue line indicates the original dataset and the green line indicates the average values.Axes indicate averages of indices and year.The positive values of U-WS indicate eastward direction and vice versa.

Figure 3 .
Figure 3. Moving averages of SST (upper panel), U-WS (middle panel) and RR (lower panel) in area B The 3 months moving averages are shown in panel (a), (d) and (g); the 6 months moving averages are shown in panel (b), (e) and (h); the 12 months moving averages are shown in panel (c), (f) and (i).The red line indicates the moving averages result, the blue line indicates the original dataset and the green line indicates the average values.Axes indicate averages of indices and year.The positive values of U-WS indicate eastward direction and vice versa.

Figure 4 .
Figure 4. Moving averages of SST (upper panel), U-WS (middle panel) and RR (lower panel) in area C The 3 months moving averages are shown in panel (a), (d) and (g); the 6 months moving averages are shown in panel (b), (e) and (h); the 12 months moving averages are shown in panel (c), (f) and (i).The red line indicates the moving averages result, the blue line indicates the original dataset and the green line indicates the average values.Axes indicate averages of indices and year.The positive values of U-WS indicate eastward direction and vice versa.

Figure 5 .
Figure 5. Moving averages of SST (upper panel), U-WS (middle panel) and RR (lower panel) in area D The 3 months moving averages are shown in panel (a), (d) and (g); the 6 months moving averages are shown in panel (b), (e) and (h); the 12 months moving averages are shown in panel (c), (f) and (i).The red line indicates the moving averages result, the blue line indicates the original dataset and the green line indicates the average values.Axes indicate averages of indices and year.The positive values of U-WS indicate eastward direction and vice versa.

Figure 6 .
Figure 6.Moving averages of SST (upper panel), U-WS (middle panel) and RR (lower panel) in area E The 3 months moving averages are shown in panel (a), (d) and (g); the 6 months moving averages are shown in panel (b), (e) and (h); the 12 months moving averages are shown in panel (c), (f) and (i).The red line indicates the moving averages result, the blue line indicates the original dataset and the green line indicates the average values.Axes indicate averages of indices and year.The positive values of U-WS indicate eastward direction and vice versa.

Figure 7 .
Figure 7. Moving averages of SST (upper panel), U-WS (middle panel) and RR (lower panel) in area F The 3 months moving averages are shown in panel (a), (d) and (g); the 6 months moving averages are shown in panel (b), (e) and (h); the 12 months moving averages are shown in panel (c), (f) and (i).The red line indicates the moving averages result, the blue line indicates the original dataset and the green line indicates the average values.Axes indicate averages of indices and year.The positive values of U-WS indicate eastward direction and vice versa.

Figure 8 .
Figure 8. Moving averages of SST (upper panel), U-WS (middle panel) and RR (lower panel) in area G The 3 months moving averages are shown in panel (a), (d) and (g); the 6 months moving averages are shown in panel (b), (e) and (h); the 12 months moving averages are shown in panel (c), (f) and (i).The red line indicates the moving averages result, the blue line indicates the original dataset and the green line indicates the average values.Axes indicate averages of indices and year.The positive values of U-WS indicate eastward direction and vice versa.

Figure 9 .
Figure 9. Moving averages of SST (upper panel), U-WS (middle panel) and RR (lower panel) in area H The 3 months moving averages are shown in panel (a), (d) and (g); the 6 months moving averages are shown in panel (b), (e) and (h); the 12 months moving averages are shown in panel (c), (f) and (i).The red line indicates the moving averages result, the blue line indicates the original dataset and the green line indicates the average values.Axes indicate averages of indices and year.The positive values of U-WS indicate eastward direction and vice versa.

Figure 10 .
Figure 10.Moving averages of SST (upper panel), U-WS (middle panel) and RR (lower panel) in area I The 3 months moving averages are shown in panel (a), (d) and (g); the 6 months moving averages are shown in panel (b), (e) and (h); the 12 months moving averages are shown in panel (c), (f) and (i).The red line indicates the moving averages result, the blue line indicates the original dataset and the green line indicates the average values.Axes indicate averages of indices and year.The positive values of U-WS indicate eastward direction and vice versa.

Figure 11 .
Figure 11.Moving averages of SST (upper panel), U-WS (middle panel) and RR (lower panel) in area J The 3 months moving averages are shown in panel (a), (d) and (g); the 6 months moving averages are shown in panel (b), (e) and (h); the 12 months moving averages are shown in panel (c), (f) and (i).The red line indicates the moving averages result, the blue line indicates the original dataset and the green line indicates the average values.Axes indicate averages of indices and year.The positive values of U-WS indicate eastward direction and vice versa.

Figure 12 .
Figure 12.Moving averages of SST (upper panel), U-WS (middle panel) and RR (lower panel) in area K The 3 months moving averages are shown in panel (a), (d) and (g); the 6 months moving averages are shown in panel (b), (e) and (h); the 12 months moving averages are shown in panel (c), (f) and (i).The red line indicates the moving averages result, the blue line indicates the original dataset and the green line indicates the average values.Axes indicate averages of indices and year.The positive values of U-WS indicate eastward direction and vice versa.

Figure 13 .
Figure 13.Moving averages result of SST (upper panel), U-WS (middle panel) and RR (lower panel) in area L The 3 months moving averages are shown in panel (a), (d) and (g); the 6 months moving averages are shown in panel (b), (e) and (h); the 12 months moving averages are shown in panel (c), (f) and (i).The red line indicates the moving averages result, the blue line indicates the original dataset and the green line indicates the average values.Axes indicate averages of indices and year.The positive values of U-WS indicate eastward direction and vice versa

Figure 14 .Figure 15 .
Figure 14.The auto-correlation of SST within 4 years in area G (panel a), area H (panel b), area I (panel c) and area J (panel d) Axes indicate the correlation values and time lag.

Figure 16 .
Figure 16.The auto-correlation of RR within 4 years in area G (panel a), area H (panel b), area I (panel c) and area J (panel d) Axes indicate the correlation values and time lag.
indicates the signal can be detected, while NC indicates the signal is not clear.
indicates the signal can be detected, while NC indicates the signal is not clear.

Figure 17 .
Figure 17.The 6 months moving averages of the deseasonalized dataset in area A (left panel), area B (middle panel) and area C (right panel) The upper panel shows the SST; the middle panel shows the U-WS; and the lower panel shows the RR.The red line indicates the moving averages result and the green line indicates the average values.

Figure 18 .
Figure 18.The 6 months moving averages of the deseasonalized dataset in area D (left panel), area E (middle panel) and area F (right panel) The upper panel shows the SST; the middle panel shows the U-WS; and the lower panel shows the RR.The red line indicates the moving averages result and the green line indicates the average values.

Figure 19 .
Figure 19.The 6 months moving averages of the deseasonalized dataset in area G (left panel), area H (middle panel) and area I (right panel) The upper panel shows the SST; the middle panel shows the U-WS; and the lower panel shows the RR.The red line indicates the moving averages result and the green line indicates the average values.

Figure 20 .
Figure 20.The 6 months moving averages of the deseasonalized dataset in area J (left panel), area K (middle panel) and area L (right panel) The upper panel shows the SST; the middle panel shows the U-WS; and the lower panel shows the RR.The red line indicates the moving averages result and the green line indicates the average values.

Figure 21 .
Figure 21.ENSO years based on ONI dataset.SST is at least 0.5 warmer than 30-year average indicate the El Nino.SST is at least 0.5 cooler than 30-year average indicate the La Nina

Table 2 .
Seasonal variability of indices

Table 4 .
Auto-correlation of indices

Table 7 .
La Nina signal from the 6 months moving averages of the deseasonalized dataset