# Spatiotemporal Analysis on the Teleconnection of ENSO and IOD to the Stream Flow Regimes in Java, Indonesia

^{1}

^{2}

^{3}

^{*}

## Abstract

**:**

## 1. Introduction

## 2. Data and Methods

#### 2.1. Streamflow Data

^{2}(Cs, Bo, Pr, Ba), >1000 km

^{2}(Cu, Ma), and >10,000 km

^{2}(BS). For further information, the BS River is the largest river basin in Java which flows from Central Java to East Java.

_{50}), high flow (Q

_{10}), and low flow (Q

_{90}). With the definition of the FDC, Q

_{10}(Q

_{90}) is the flow equaled or exceeded 10 (90) percent of all the time. Meanwhile, in data distribution, the definition of Q

_{10}(Q

_{90}) is the 90th (10th) percentile due to the data ranking from lowest to highest. The three flow regimes were calculated using the R package fasstr [14].

#### 2.2. Climate Indices

#### 2.3. Temporal Variation Sets

#### 2.4. Multiple Polynomial Regression

_{50}(medium flow), Q

_{10}(high flow), and Q

_{90}(low flow) for each river. Multiple regression (MR) models were developed using second-order and third-order polynomial functions to minimize the model’s error term as shown in Equations (1) and (2), respectively. All regression analyses and statistics in this study were carried out by using R programming [22,23].

_{0}+ β

_{1}x

_{1}+ β

_{2}x

_{2}+ β

_{3}x

_{1}

^{2}+ β

_{4}x

_{2}

^{2}+ β

_{5}x

_{1}x

_{2}+ ε

_{0}+ β

_{1}x

_{1}+ β

_{2}x

_{2}+ β

_{3}x

_{1}

^{2}+ β

_{4}x

_{2}

^{2}+ β

_{5}x

_{1}x

_{2}+ β

_{6}x

_{1}

^{3}+ β

_{7}x

_{2}

^{3}+ β

_{8}x

_{1}

^{2}x

_{2}+ β

_{9}x

_{1}x

_{2}

^{2}+ ε

_{50}/Q

_{10}/Q

_{90})

_{1}= ENSO index (SOI/Niño 3.4)

_{2}= DMI

#### 2.5. Model Evaluation

^{2}) and the Kling–Gupta Efficiency (KGE) coefficient. The adjusted R

^{2}is commonly used to indicate the goodness of fit in models using multiple predictors. The KGE has been widely used in recent hydrological modeling studies. We used both evaluation metrics hoping that the result of our study can be easily compared to other studies, which use the common metric (R

^{2}) and the recently popular metric (KGE). The two evaluation metrics were calculated using the R package hydroGOF [24].

^{2}is a better evaluation metric than the ordinary R

^{2}because it is adjusted by the number of observations and variables, resolving the bias issue in the ordinary R

^{2}caused by the degree of freedom problem. The adjusted R

^{2}is calculated by Equation (3) (n = number of observations, and p = number of predictors). The value of adjusted R

^{2}closer to 1 indicates better model skill.

## 3. Results

#### 3.1. Temporal Variability

^{2}and KGE coefficient (Figure 3) highlighted several points.

_{a}) set in September–November (SON

_{a}). The seasonal monthly (SM) from September–November (SON) also performed better among the SM sets. The results indicate two things. First, the streamflow predictability by ENSO and IOD was good only in the September–November season, when the transition from dry season to wet season usually occurs in the Java region [29]. Many previous studies have found a high correlation of ENSO and IOD during May–October or September–November to the rainfall [3,4,16,29,30,31] and streamflow [11] in the Java region. Second, in general, the SM

_{a}sets have better model skills than the SM sets. This indicates that even with monthly data series in three-month windows, just ENSO and IOD are not enough to satisfy the need to describe the variance of the streamflows. The comparison between the monthly set (MON) and the six-month moving average set (MA6) also showed the same behavior. MA6 obtained a better model skill than MON due to MA6 carrying over the information of streamflow and climate indices for six months. Thus, it could filter the noises from other factors, such as another climate phenomenon with a shorter timescale, such as Madden–Julian Oscillation.

_{50}and Q

_{10}models achieved better model skills than the Q

_{90}. The overall distribution of Q

_{90}model skills is always lower than the other two regimes in SON and SON

_{a}. It corresponds to the higher flow tending to be more sensitive in the regression with ENSO and IOD indices, as shown in our previous study for Code River [12]. Another previous study on a northwestern Java river showed that low stream flow events are related more to positive IOD based on a probability analysis [11]. However, our result cannot be compared directly to that study due to the different methods. While we used the ENSO and IOD indices together in multiple regression, that study treated them independently in simple probability and correlation analyses.

^{2}and KGE coefficients (Figure 3). It may also be worth noting that, in the averaged seasonal sets (MAM

_{a}, JJA

_{a}, SON

_{a}, DJF

_{a}) of the third-order models, the NINO 3.4 variant showed no KGE outliers. In contrast, the SOI variant has >10 KGE outliers. Meanwhile, there was no significant difference in KGE sets of the second-order models. This may indicate that in our case, the quality of the models developed in third-order MR is better when predicted by NINO 3.4 than SOI.

_{10}, Q

_{50}, Q

_{90}, which Niño 3.4-DMI predicted in the SON

_{a}data set, have the maximum KGE of 0.75 (0.94), 0.74 (0.90), and 0.55 (0.82), respectively. Additionally, in the prediction by Niño 3.4-DMI in SON

_{a}set, the outliers of the adjusted R

^{2}and KGE are found only in the second-order MR models. Overall, the third-order MR improved the model skills after the second-order MR, not only for the maximum but also for the minimum values of the evaluation metrics (Figure 3). The improvement of the predictability by the third-order models is more apparent in the higher flows, as shown by the example of Ma River in Figure 4.

#### 3.2. Spatial Variability

_{a}set. In Figure 2, the boxplots of the SON

_{a}evaluation metrics are generally longer relative to the other sets, indicating a higher spatial variability among the temporal variation sets for all rivers. In this section, we attempted to explain the spatial variability among the rivers by analyzing the goodness of fit metrics (Table 3 and Table 4), the correlation between the predicted (by Niño 3.4-DMI in SON

_{a}) and the observed flow regimes (Table 5). Generally, the west and the central rivers (Cu, Cs, Bo, Pr) have a lower predictability than the eastern rivers (BS, Ma, Ba). The western river models also tend to underestimate the higher flows, while the eastern river models’ estimations are spread evenly (Figure A1).

_{50}and Q

_{90}, even though their correlation coefficients (R) were higher than that of the Q

_{10}. In that case, we can consider that another river in the same region, Ba, has more quality in prediction skill, due to having more consistently lower error distances. Only the highest flow of each flow regime in Ba River showed a relatively too far error distance, which underestimates by about 50% of the observed values. Those error distances were lowered in the third-order MR models (Figure A2).

^{2}(Table 4), and R of predicted–observed values (Table 5) all agree with the conclusion that Pr streamflow regimes cannot sufficiently develop a relationship model to the climate indices by MR analysis. Scatterplots of the second-order MR in Figure A1 showed more details that the predicted–observed correlation of Pr River is poor compared to those of the other rivers. Additionally, among the three flow regime indices of Pr River, only Q

_{10}could achieve a significant correlation to the observed values (Figure A1; the respective p-values for Q

_{50}, Q

_{10}, and Q

_{90}are 0.187, 0.014, and 0.266). In the third-order MR, the Pr River’s KGE (Table 4) and R (Table 5) values are become improved, passing the minimum standards (KGE > −0.41 and R > 0.50). However, the third-order MR models are still poor with the low KGE and adjusted R

^{2}values (Table 3). The scatterplots of predicted–observed values (Figure A2) also showed that the model skill of the third-order MR in the Pr River is more improved than that in the second-order MR (Figure A1). However, the spread of the plots does not make it a strong model.

## 4. Discussion

#### 4.1. Predictability Tendency by the Number of Observations

_{a}and plotted them to the evaluation metrics (adjusted R

^{2}, KGE, and R Pearson). Figure 5 shows the plots for Q

_{50}with the evaluation metrics for second-order MR models. Complete results for all flow regimes on second- and third-order MR models can be seen in Figure A3 in the Appendix B. All results showed similar tendency among the eight rivers. The BS River, which also has a large amount of data, similar to the rivers of Cu and Cs, could achieve a higher skill than the average, nearing the rivers of Ma and Ba (eastern rivers), which have the best skills among all. Meanwhile, Bo River, which has a similarly low number of observations as the Ma and Ba Rivers, had poor model skills. Pr River still had the lowest skills, despite it having more observations than Ma and Ba Rivers.

#### 4.2. Change of Predictability over Periods

_{a}data sets from two different periods: 1970–1989 and 1990–2018. We recalculated new data sets which had been divided, then developed new regression models from them. This analysis may reduce the number of observations and become half of the original. Moreover, the number of parameters in the multiple polynomial regressions could further reduce the degree of freedom. Therefore, the second-order MR was used in this reanalysis due to having fewer parameters than third-order MR.

^{2}and KGE values of the second- and third-order MR models in SON

_{a}for the three periods are provided in Table A1, Table A2, Table A3 and Table A4 in the Appendix C.

_{50}. Moreover, the model skill of the original period with all the data from both periods naturally became lower than the first period and higher than the second period. These results indicate that the rivers of Cu and Cs did not always have a poor relationship with the ENSO and IOD. Instead, it suggests that the relationship in the first period was changed in the second period.

#### 4.3. Good Predictability in Code River

_{a}are not. It can be understood that if a river showed high MA6 model skill, its relationship to the climate indices in intra-annual variation is also high. Meanwhile, we knew that, among the four seasons, only the SON season gives the best model skill. Hence, if MA6′s model skill is high, then it must be due to low intra-annual (seasonal) variation, in which the predictability of the other three seasons is not too different from the SON season.

_{50}of Code River is similar in SON

_{a}and DJF

_{a}, while MAM

_{a}and JJA

_{a}were about 0.25 lower. Hence, the intra-annual predictability of Code River is relatively low. Therefore, we concluded that the excellent model skill obtained by Code River in our previous paper seems to be owing to its seasonal characteristic.

#### 4.4. Streamflow Prediction by ENSO and IOD

_{a}) better than the monthly set (SON). It emphasizes that intra-annual variation is crucial to be considered in studying the relationship between climate indices and the streamflow or rainfall in Indonesia. Despite Indonesia having only two seasons (wet and dry), intra-annual variation is still important to be considered, owing to many interrelated climate phenomena around the regions in different periods and timescales.

_{10}) and medium flow index (Q

_{50}) could be predicted better than low flow index (Q

_{90}). While the Q

_{10}predictability level seemed similar among rivers in Java, Q

_{90}was the opposite. The different predictability of Q

_{90}from one river to another is probably caused by the low flow dependency on the physical condition of a river’s watershed. Anthropogenic drought problems, such as land-use change and groundwater exploitation, make Q

_{90}prediction by climate harder. Meanwhile, high flow situations, even in flood, are usually linear to the amount of rainfall (hence climate), making the prediction of Q

_{10}by climate better than Q

_{90}.

## 5. Conclusions

_{a}). The high predictability in the SON season is consistent with the previous studies on ENSO and IOD indices’ connections to rainfall [16,31] and streamflow [11] in Indonesia. Here, we advanced the simple correlation analysis in past studies to multiple polynomial regression and achieved new insights into the prediction of Indonesian streamflow by both ENSO and IOD indices.

## Author Contributions

## Funding

## Institutional Review Board Statement

## Informed Consent Statement

## Data Availability Statement

## Conflicts of Interest

## Appendix A

**Figure A1.**Scatterplots between the predicted (Niño 3.4-DMI, second-order MR, SON

_{a}set) and the observed values (in m

^{3}/s) of the flow regimes (

**a**) Q

_{50}, (

**b**) Q

_{10}, and (

**c**) Q

_{90}for all rivers in this study.

**Figure A2.**Scatterplots between the predicted (Niño 3.4-DMI, third-order MR, SON

_{a}set) and the observed values (in m

^{3}/s) of the flow regimes (

**a**) Q

_{50}, (

**b**) Q

_{10}, and (

**c**) Q

_{90}for all rivers in this study.

## Appendix B

**Figure A3.**The number of daily observations used for developing the second- and third-order MR models of the flow regimes (

**a**) Q

_{50}, (

**b**) Q

_{10}, and (

**c**) Q

_{90}in the SON

_{a}set series. The R Pearson is the coefficient correlation between predicted and observed values.

## Appendix C

**Table A1.**Adjusted R

^{2}of the second-order MR models predicted by Niño 3.4 and DMI in SON

_{a}set in three different periods. The values higher than 0.5 are bolded, indicating good model skill.

River | 1970–2018 | 1970–1989 | 1990–2018 | ||||||
---|---|---|---|---|---|---|---|---|---|

Q_{50} | Q_{10} | Q_{90} | Q_{50} | Q_{10} | Q_{90} | Q_{50} | Q_{10} | Q_{90} | |

Cu | 0.277 | 0.309 | 0.170 | 0.618 | 0.478 | 0.568 | 0.090 | 0.100 | −0.001 |

Cs | 0.142 | 0.078 | 0.085 | 0.615 | 0.027 | 0.579 | 0.029 | 0.055 | 0.011 |

Bo | 0.181 | 0.260 | 0.162 | 0.533 | 0.411 | 0.626 | −0.037 | 0.022 | −0.009 |

Pr | −0.101 | 0.025 | −0.117 | 0.231 | 0.329 | 0.189 | −0.149 | −0.184 | −0.122 |

BS | 0.431 | 0.415 | 0.248 | 0.653 | 0.738 | 0.361 | 0.362 | 0.398 | 0.147 |

Ma | 0.590 | 0.537 | 0.606 | 0.568 | 0.508 | 0.188 | 0.727 | 0.771 | 0.741 |

Ba | 0.502 | 0.576 | 0.372 | 0.711 | 0.653 | 0.527 | 0.502 | 0.548 | 0.389 |

Co | - | - | - | - | - | - | 0.479 | 0.575 | 0.078 |

**Table A2.**Adjusted R

^{2}of the third-order models predicted by Niño 3.4 and DMI in SON

_{a}set in three different periods. The values higher than 0.5 are bolded, indicating good model skill.

River | 1970–2018 | 1970–1989 | 1990–2018 | ||||||
---|---|---|---|---|---|---|---|---|---|

Q_{50} | Q_{10} | Q_{90} | Q_{50} | Q_{10} | Q_{90} | Q_{50} | Q_{10} | Q_{90} | |

Cu | 0.192 | 0.257 | 0.099 | 0.563 | 0.294 | 0.505 | 0.160 | 0.284 | −0.048 |

Cs | 0.093 | −0.002 | 0.030 | 0.564 | −0.326 | 0.560 | −0.178 | −0.148 | −0.216 |

Bo | 0.212 | 0.328 | 0.255 | 0.525 | 0.580 | 0.645 | −0.222 | −0.297 | −0.060 |

Pr | 0.065 | 0.188 | 0.091 | 0.540 | 0.648 | 0.318 | −0.189 | −0.253 | 0.040 |

BS | 0.526 | 0.433 | 0.383 | 0.887 | 0.876 | 0.655 | 0.361 | 0.353 | 0.086 |

Ma | 0.670 | 0.571 | 0.652 | 0.643 | 0.419 | −0.254 | 0.863 | 0.893 | 0.839 |

Ba | 0.530 | 0.596 | 0.401 | 0.625 | 0.514 | 0.300 | 0.708 | 0.674 | 0.653 |

Co | - | - | - | - | - | - | 0.741 | 0.841 | −0.058 |

**Table A3.**KGE of the second-order MR models predicted by Niño 3.4 and DMI in SON

_{a}set in three different periods. The values higher than 0.5 are bolded, indicating good model skill.

River | 1970–2018 | 1970–1989 | 1990–2018 | ||||||
---|---|---|---|---|---|---|---|---|---|

Q_{50} | Q_{10} | Q_{90} | Q_{50} | Q_{10} | Q_{90} | Q_{50} | Q_{10} | Q_{90} | |

Cu | 0.441 | 0.323 | 0.473 | 0.800 | 0.771 | 0.718 | 0.334 | 0.230 | 0.345 |

Cs | 0.281 | 0.205 | 0.194 | 0.792 | 0.771 | 0.377 | 0.254 | 0.232 | 0.284 |

Bo | 0.383 | 0.362 | 0.462 | 0.769 | 0.818 | 0.701 | 0.371 | 0.395 | 0.420 |

Pr | −0.105 | −0.152 | 0.146 | 0.573 | 0.544 | 0.637 | 0.085 | 0.125 | 0.028 |

BS | 0.587 | 0.410 | 0.572 | 0.826 | 0.657 | 0.870 | 0.575 | 0.383 | 0.604 |

Ma | 0.735 | 0.747 | 0.696 | 0.797 | 0.586 | 0.767 | 0.865 | 0.872 | 0.887 |

Ba | 0.663 | 0.555 | 0.719 | 0.875 | 0.789 | 0.849 | 0.711 | 0.634 | 0.741 |

Co | - | - | - | - | - | - | 0.689 | 0.371 | 0.753 |

**Table A4.**KGE of the third-order MR models predicted by Niño 3.4 and DMI in SON

_{a}set in three different periods. The values higher than 0.5 are bolded, indicating good model skill.

River | 1970–2018 | 1970–1989 | 1990–2018 | ||||||
---|---|---|---|---|---|---|---|---|---|

Q_{50} | Q_{10} | Q_{90} | Q_{50} | Q_{10} | Q_{90} | Q_{50} | Q_{10} | Q_{90} | |

Cu | 0.441 | 0.356 | 0.502 | 0.858 | 0.837 | 0.761 | 0.560 | 0.417 | 0.637 |

Cs | 0.337 | 0.268 | 0.230 | 0.847 | 0.845 | 0.453 | 0.287 | 0.251 | 0.314 |

Bo | 0.533 | 0.564 | 0.615 | 0.875 | 0.907 | 0.890 | 0.597 | 0.659 | 0.567 |

Pr | 0.351 | 0.375 | 0.464 | 0.863 | 0.792 | 0.897 | 0.387 | 0.536 | 0.339 |

BS | 0.708 | 0.604 | 0.642 | 0.968 | 0.899 | 0.964 | 0.673 | 0.497 | 0.668 |

Ma | 0.829 | 0.819 | 0.773 | 0.920 | 0.694 | 0.867 | 0.961 | 0.954 | 0.969 |

Ba | 0.737 | 0.653 | 0.777 | 0.932 | 0.870 | 0.911 | 0.887 | 0.865 | 0.873 |

Co | - | - | - | - | - | - | 0.895 | 0.500 | 0.937 |

## References

- Poff, N.L. Beyond the natural flow regime? Broadening the hydro-ecological foundation to meet environmental flows challenges in a non-stationary world. Freshw. Biol.
**2018**, 63, 1011–1021. [Google Scholar] [CrossRef] - Ropelewski, C.F.; Halpert, M.S. Global and Regional Scale Precipitation Patterns Associated with the El Niño/Southern Oscillation. Mon. Weather Rev.
**1987**, 115, 1606–1626. [Google Scholar] [CrossRef][Green Version] - Hendon, H.H. Indonesian rainfall variability: Impacts of ENSO and local air-sea interaction. J. Clim.
**2003**, 16, 1775–1790. [Google Scholar] [CrossRef] - Aldrian, E.; Susanto, R.D. Identification of three dominant rainfall regions within Indonesia and their relationship to sea surface temperature. Int. J. Climatol.
**2003**, 23, 1435–1452. [Google Scholar] [CrossRef] - Lee, H.S. General Rainfall Patterns in Indonesia and the Potential Impacts of Local Seas on Rainfall Intensity. Water
**2015**, 7, 1751–1768. [Google Scholar] [CrossRef] - Saji, N.; Goswami, B.; Vinayachandran, P.; Yamagata, T. A dipole mode in the Tropical Ocean. Nature
**1999**, 401, 360–363. [Google Scholar] [CrossRef] - Cardoso, A.O.; Silva Dias, P.L. The relationship between ENSO and Paraná River flow. Adv. Geosci.
**2006**, 6, 189–193. [Google Scholar] [CrossRef][Green Version] - Cluis, D.; Laberge, C. Analysis of the El Niño Effect on the Discharge of Selected Rivers in the Asia-Pacific Region. Water Int.
**2002**, 27, 279–293. [Google Scholar] [CrossRef] - Zhang, Q.; Xu, C.Y.; Jiang, T.; Wu, Y. Possible influence of ENSO on annual maximum streamflow of the Yangtze River, China. J. Hydrol.
**2007**, 333, 265–274. [Google Scholar] [CrossRef] - Mohsenipour, M.; Shahid, S.; Nazemosadat, M.J. Effects of El Nino Southern Oscillation on the Discharge of Kor River in Iran. Adv. Meteorol.
**2013**, 2013, 1–7. [Google Scholar] [CrossRef] - Sahu, N.; Behera, S.K.; Yamashiki, Y.; Takara, K.; Yamagata, T. IOD and ENSO impacts on the extreme stream-flows of Citarum River in Indonesia. Clim. Dyn.
**2012**, 39, 1673–1680. [Google Scholar] [CrossRef] - Nugroho, A.R.; Tamagawa, I.; Harada, M. The Relationship between River Flow Regimes and Climate Indices of ENSO and IOD on Code River, Southern Indonesia. Water
**2021**, 13, 1375. [Google Scholar] [CrossRef] - Searcy, J.K. Flow-Duration Curves; United States Government Printing Office: Washington, DC, USA, 1959. [Google Scholar] [CrossRef]
- Goetz, J.; Schwarz, C.J. Fasstr: Analyze, Summarize, and Visualize Daily Streamflow Data, R package version 0.3.3; CRAN, 2021; Available online: https://cran.r-project.org/package=fasstr (accessed on 29 November 2021).
- Ropelewski, C.F.; Halpert, M.S. Precipitation Patterns Associated with the High Index Phase of the Southern Oscillation. J. Clim.
**1989**, 2, 268–284. [Google Scholar] [CrossRef][Green Version] - Kurniadi, A.; Weller, E.; Min, S.K.; Seong, M.G. Independent ENSO and IOD impacts on rainfall extremes over Indonesia. Int. J. Climatol.
**2021**, 41, 3640–3656. [Google Scholar] [CrossRef] - Bamston, A.G.; Chelliah, M.; Goldenberg, S.B. Documentation of a Highly ENSO-Related SST Region in the Equatorial Pacific: Research Note. Atmos.-Ocean
**1997**, 35, 367–383. [Google Scholar] [CrossRef] - Hendrawan, I.G.; Asai, K.; Triwahyuni, A.; Lestari, D.V. The interanual rainfall variability in Indonesia corresponding to El Niño Southern Oscillation and Indian Ocean Dipole. Acta Oceanol. Sin.
**2019**, 38, 57–66. [Google Scholar] [CrossRef] - Rao, J.; Ren, R. Parallel comparison of the 1982/83, 1997/98 and 2015/16 super El Niños and their effects on the extratropical stratosphere. Adv. Atmos. Sci.
**2017**, 34, 1121–1133. [Google Scholar] [CrossRef] - Qian, H.; Xu, S.-B. Prediction of Autumn Precipitation over the Middle and Lower Reaches of the Yangtze River Basin Based on Climate Indices. Climate
**2020**, 8, 53. [Google Scholar] [CrossRef][Green Version] - Hidayat, R.; Kizu, S. Influence of the Madden-Julian Oscillation on Indonesian rainfall variability in austral summer. Int. J. Climatol.
**2010**, 30, 1816–1825. [Google Scholar] [CrossRef] - R Core Team. R: A Language and Environment for Statistical Computing, R version 4.1.2; 2021; Available online: http://www.r-project.org/ (accessed on 29 November 2021).
- Wickham, H. Modelr: Modelling Functions that Work with the Pipe, R package version 0.1.8; CRAN, 2020; Available online: https://cran.r-project.org/package=modelr (accessed on 29 November 2021).
- Zambrano-Bigiarini, M. Hydrogof: Goodness-of-Fit Functions for Comparison of Simulated and Observed Hydrological Time Series, R package version 0.4-0; CRAN, 2020; Available online: https://cran.r-project.org/package=hydroGOF (accessed on 29 November 2021).
- Nash, J.E.; Sutcliffe, J.V. River flow forecasting through conceptual models part I—A discussion of principles. J. Hydrol.
**1970**, 10, 282–290. [Google Scholar] [CrossRef] - Gupta, H.V.; Kling, H.; Yilmaz, K.K.; Martinez, G.F. Decomposition of the mean squared error and NSE performance criteria: Implications for improving hydrological modelling. J. Hydrol.
**2009**, 377, 80–91. [Google Scholar] [CrossRef][Green Version] - Knoben, W.J.M.; Freer, J.E.; Woods, R.A. Technical note: Inherent benchmark or not? Comparing Nash-Sutcliffe and Kling-Gupta efficiency scores. Hydrol. Earth Syst. Sci.
**2019**, 23, 4323–4331. [Google Scholar] [CrossRef][Green Version] - Kling, H.; Fuchs, M.; Paulin, M. Runoff conditions in the upper Danube basin under an ensemble of climate change scenarios. J. Hydrol.
**2012**, 424–425, 264–277. [Google Scholar] [CrossRef] - Hamada, J.I.; Yamanaka, M.D.; Matsumoto, J.; Fukao, S.; Winarso, P.A.; Sribimawati, T. Spatial and temporal variations of the rainy season over Indonesia and their link to ENSO. J. Meteorol. Soc. Japan
**2002**, 80, 285–310. [Google Scholar] [CrossRef][Green Version] - Jun-Ichi, H.; Mori, S.; Kubota, H.; Yamanaka, M.D.; Haryoko, U.; Lestari, S.; Sulistyowati, R.; Syamsudin, F. Interannual rainfall variability over northwestern Jawa and its relation to the Indian Ocean Dipole and El Niño-Southern Oscillation events. Sci. Online Lett. Atmos.
**2012**, 8, 69–72. [Google Scholar] [CrossRef][Green Version] - Hidayat, R.; Ando, K.; Masumoto, Y.; Luo, J.J. Interannual Variability of Rainfall over Indonesia: Impacts of ENSO and IOD and Their Predictability. IOP Conf. Ser. Earth Environ. Sci.
**2016**, 31, 012043. [Google Scholar] [CrossRef] - Maryono, A. River Restoration, 4th ed.; Gadjah Mada University Press: Yogyakarta, Indonesia, 2020; ISBN 978-979-420-667-6. [Google Scholar]
- Kali Cisadane. Catalogue of Rivers for Southeast Asia and the Pacific-Volume V; Tachikawa, Y., James, R., Abdullah, K., Desa, M.N.b.M., Eds.; The UNESCO-IHP Regional Steering Committee for Southeast Asia and the Pacific, 2004; pp. 45–56. ISBN 4-902712-00-8. Available online: https://hywr.kuciv.kyoto-u.ac.jp/ihp/riverCatalogue/Vol_05/index.html (accessed on 29 November 2021).
- Bengawan Solo. In Catalogue of Rivers for Southeast Asia and the Pacific-Volume I; Takeuchi, K.; Jayawardena, A.W.; Takahasi, Y. (Eds.) The UNESCO-IHP Regional Steering Committee for Southeast Asia and the Pacific, 1995; pp. 91–102. ISBN 962-8014-07-2. Available online: https://hywr.kuciv.kyoto-u.ac.jp/ihp/riverCatalogue/Vol_01/index.html (accessed on 29 November 2021).
- Gornitz, V.; Rosenzweig, C.; Hillel, D. Effects of anthropogenic intervention in the land hydrologic cycle on global sea level rise. Glob. Planet. Change
**1997**, 14, 147–161. [Google Scholar] [CrossRef] - Vorosmarty, C.J.; Sahagian, D. Anthropogenic disturbance of the terrestrial water cycle. Bioscience
**2000**, 50, 753–765. [Google Scholar] [CrossRef][Green Version] - Tang, Y.; Zhang, R.H.; Liu, T.; Duan, W.; Yang, D.; Zheng, F.; Ren, H.; Lian, T.; Gao, C.; Chen, D.; et al. Progress in ENSO prediction and predictability study. Natl. Sci. Rev.
**2018**, 5, 826–839. [Google Scholar] [CrossRef] - Feba, F.; Ashok, K.; Collins, M.; Shetye, S.R. Emerging Skill in Multi-Year Prediction of the Indian Ocean Dipole. Front. Clim.
**2021**, 3, 1–8. [Google Scholar] [CrossRef]

**Figure 3.**Boxplot series showing model’s evaluation metrics of (

**a**) adjusted R

^{2}and (

**b**) KGE coefficient with two variants of predictor combinations: NINO 3.4-DMI and SOI-DMI. The x-axis shows the temporal variation sets (MON: monthly flow, MA6: six-month moving average flow, MAM-DJF: monthly flow in the window of each respective season, MAMa-DJFa: three-month average flow in the window of each respective season). C = Code River. The lines in the middle of the box are median. The top of the box is the third quartile (Q3), and the bottom is the first quartile (Q1). The vertical lines above and below the box show the data range, excluding outliers. The dots represent outliers, values that are farther than 1.5 times of their interquartile ranges (IQR = Q3 − Q1).

**Figure 4.**Scatterplots between the predicted (using second- and third-order MR models with Niño 3.4 and DMI as predictors) and the observed Q

_{50}for Madiun River (Ma) in all temporal variants: monthly (MON), six-month moving average (MA6), monthly seasonal (MAM-DJF), and averaged monthly seasonal (MAM

_{a}-DJF

_{a}).

**Figure 5.**Plot between the number of daily observations (x-axis) for Q

_{50}models (second-order) in the SON

_{a}set and their model skills (y-axis).

**Figure 6.**(

**a**) Evaluation metrics boxplots and (

**b**) predicted–observed scatterplots of MR models (2nd-order, Niño 3.4–DMI predictors, SON

_{a}set) with three different window periods: 1970–2018, 1970–1989, and 1990–2018. The boxplots show the metrics for all rivers, while the scatterplots only show the plots for Cs and Cu Rivers.

**Figure 7.**The adjusted R

^{2}and KGE of Code River’s Q

_{50}second-order MR models in different temporal data sets.

Code-Name | River Name | Flow Records Used (year) | Gauge’s Catchment Area (km ^{2}) | Region in Java | Gauge’s Location Coordinates |
---|---|---|---|---|---|

Cu | Ciujung | 42 | 1562.7 | West | −6.133, 106.300 |

Cs | Cisadane | 44 | 850.2 | West | −6.514, 106.678 |

Bo | Bodri | 31 | 522.3 | Central | −7.011, 110.137 |

Pr | Progo | 38 | 423.4 | Central | −7.340, 110.209 |

BS | Bengawan Solo | 42 | 11,125.0 | East | −7.150, 111.599 |

Ma | Madiun | 30 | 2126.0 | East | −7.645, 111.513 |

Ba | Baru | 33 | 454.0 | East | −8.413, 114.094 |

Co | Code | 21 | 40.0 | Central | −7.849, 110.375 |

Set Name | Code-Name | Time Scale | Calculation Method |
---|---|---|---|

Ordinary monthly | MON | monthly flow | Q_{50}: median of daily flow distribution in a month.Q _{90}: 10% percentile of daily flow distribution in a month.Q _{10}: 90% percentile of daily flow distribution in a month.Thus, there are 12 data sets for each year. |

Six-month moving average | MA6 | monthly flow with six-month information | Mean of each flow regime monthly series for six months. Thus, there are 12 data sets for each year. |

Seasonal monthly (SM) | MAM/ JJA/ SON/ DJF | monthly flow | The same as with the ordinary monthly flow series, except the series is limited only to three months according to each season (e.g., Q_{50} of the MAM series includes the monthly Q_{50} of only March, April, and May, from 1970 to 2018). Thus, there are three data sets for each year. |

Seasonal average (SM_{a}) | MAM_{a}/JJA _{a}/SON _{a}/DJF _{a} | three-month flow | Mean of the seasonal monthly series (e.g., the first datum in the MAM_{a} series is the mean of three months: March, April, and May, in 1970). Thus, there is only one datum set for each year. |

**Table 3.**Adjusted R

^{2}for the models using Niño 3.4 and DMI as predictors in SON

_{a}set. The values higher than 0.5 are bolded, indicating good model skill.

River | 2nd-Order MR | 3rd-Order MR | ||||
---|---|---|---|---|---|---|

Q_{50} | Q_{10} | Q_{90} | Q_{50} | Q_{10} | Q_{90} | |

Cu | 0.28 | 0.17 | 0.31 | 0.19 | 0.10 | 0.26 |

Cs | 0.14 | 0.09 | 0.08 | 0.09 | 0.03 | 0.00 |

Bo | 0.18 | 0.16 | 0.26 | 0.21 | 0.26 | 0.33 |

Pr | −0.10 | −0.12 | 0.02 | 0.07 | 0.09 | 0.19 |

BS | 0.43 | 0.25 | 0.41 | 0.53 | 0.38 | 0.43 |

Ma | 0.59 | 0.61 | 0.54 | 0.67 | 0.65 | 0.57 |

Ba | 0.50 | 0.37 | 0.58 | 0.53 | 0.40 | 0.60 |

Co | 0.48 | 0.08 | 0.57 | 0.74 | −0.06 | 0.84 |

**Table 4.**KGE for the models using Niño 3.4–DMI as predictors in SON

_{a}set. The values higher than 0.5 are bolded, indicating good model skill.

River | 2nd-Order MR | 3rd-Order MR | ||||
---|---|---|---|---|---|---|

Q_{50} | Q_{10} | Q_{90} | Q_{50} | Q_{10} | Q_{90} | |

Cu | 0.24 | −0.04 | 0.30 | 0.24 | 0.05 | 0.35 |

Cs | −0.15 | −0.40 | −0.44 | 0.00 | −0.19 | −0.31 |

Bo | 0.11 | 0.06 | 0.28 | 0.41 | 0.46 | 0.54 |

Pr | −2.65 | −3.48 | −0.64 | 0.04 | 0.09 | 0.28 |

BS | 0.49 | 0.17 | 0.47 | 0.67 | 0.52 | 0.58 |

Ma | 0.70 | 0.72 | 0.65 | 0.82 | 0.80 | 0.75 |

Ba | 0.61 | 0.44 | 0.68 | 0.70 | 0.59 | 0.76 |

Co | 0.64 | 0.08 | 0.73 | 0.89 | 0.35 | 0.94 |

**Table 5.**Correlation coefficient (R Pearson) between the observed and the predicted flow regimes by Niño 3.4 and DMI in SON

_{a}set. The values higher than 0.7 are bolded, indicating good model skill.

River | 2nd-Order MR | 3rd-Order MR | ||||
---|---|---|---|---|---|---|

Q_{50} | Q_{10} | Q_{90} | Q_{50} | Q_{10} | Q_{90} | |

Cu | 0.60 | 0.63 | 0.52 | 0.61 | 0.65 | 0.54 |

Cs | 0.49 | 0.43 | 0.44 | 0.53 | 0.46 | 0.48 |

Bo | 0.56 | 0.62 | 0.55 | 0.67 | 0.73 | 0.69 |

Pr | 0.22 | 0.40 | 0.19 | 0.54 | 0.62 | 0.56 |

BS | 0.71 | 0.70 | 0.58 | 0.79 | 0.75 | 0.72 |

Ma | 0.81 | 0.79 | 0.82 | 0.88 | 0.84 | 0.87 |

Ba | 0.76 | 0.80 | 0.69 | 0.81 | 0.84 | 0.75 |

Co | 0.78 | 0.83 | 0.56 | 0.93 | 0.96 | 0.65 |

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |

© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Nugroho, A.R.; Tamagawa, I.; Harada, M.
Spatiotemporal Analysis on the Teleconnection of ENSO and IOD to the Stream Flow Regimes in Java, Indonesia. *Water* **2022**, *14*, 168.
https://doi.org/10.3390/w14020168

**AMA Style**

Nugroho AR, Tamagawa I, Harada M.
Spatiotemporal Analysis on the Teleconnection of ENSO and IOD to the Stream Flow Regimes in Java, Indonesia. *Water*. 2022; 14(2):168.
https://doi.org/10.3390/w14020168

**Chicago/Turabian Style**

Nugroho, Adam Rus, Ichiro Tamagawa, and Morihiro Harada.
2022. "Spatiotemporal Analysis on the Teleconnection of ENSO and IOD to the Stream Flow Regimes in Java, Indonesia" *Water* 14, no. 2: 168.
https://doi.org/10.3390/w14020168