Calibration was performed automatically using the Calibration module in the SWAT+ Toolbox based on the CALSI algorithm, following the sensitivity analysis stage. The objective of this process was to optimize model parameters so that the simulated streamflow closely matched the observed streamflow (Figure 7). Although streamflow was analyzed at multiple downstream stations, calibration was performed at a single basin outlet. The calibrated parameter dataset obtained from SWAT+ Toolbox was subsequently exported and applied to the entire model domain, and the simulation was rerun in SWAT+. This approach allowed model performance to be evaluated at multiple stations.

Due to the inconsistent availability of observed discharge data, which varied significantly across stations and periods, calibration and verification were conducted using stations and periods with sufficiently available observation. Among the three gauging stations in the lower basin in this research, Muara Kilis was selected for calibration and verification because it provided relatively longer periods of observed discharge records for model evaluation. In contrast, observed discharge data were unavailable at Simpang Berbak, while the Sungai Duren records exhibited periods of discontinuous availability.

For the first simulation (1993-2000), observed data from Muara Kilis station were used, with calibration conducted for 1994-1996 and verification for 1997-1998. For the second simulation period (2000-2023), the Muara Kilis station was also used, with a one-year warm up period. Calibration was conducted for 2008-2009, followed by verification for 2012-2013, corresponding to the periods with available observed discharge records. For the first simulation period, calibration during 1994–1996 resulted in NSE = 0.473, R² = 0.486, KGE = 0.498 and PBIAS = −3.85%, indicating satisfactory model performance. The R² value was close to 0.50, indicating moderate agreement between simulated and observed discharge, while the low absolute PBIAS indicated good agreement in terms of water-balance bias. During verification for 1997–1998, the model achieved NSE = 0.521, R² = 0.565, KGE = 0.726 and PBIAS = 6.76%, indicating satisfactory to good performance across the evaluation metrics. The model response was also assessed during selected ENSO–IOD periods. These results indicate that the model generally captured the observed streamflow response under ENSO–IOD conditions, including reduced streamflow during El Niño–IOD+ periods and increased streamflow during La Niña–IOD− periods, although some differences in timing and magnitude remained.

For the second simulation period, calibration during 2008–2009 yielded NSE = 0.444, R² = 0.582, KGE = 0.718 and PBIAS = 15.12%, indicating satisfactory to good model performance. During verification for 2012–2013, the model achieved NSE = 0.515, R² = 0.530, KGE = 0.519 and PBIAS = −0.30%, demonstrating satisfactory performance with very low bias. The model response was further examined during periods of ENSO and IOD. The results indicate that the model generally captured the observed streamflow response under varying ENSO –IOD conditions, although some differences in timing and magnitude remained. Overall, the model demonstrated consistent performance across the calibration and verification periods. Given that the model achieved satisfactory performance during verification, it was considered suitable for subsequent simulations under different land-use representations (Marek et al., 2016). The detailed interpretation of model performance refers to the criteria developed by (Almeida et al., 2018), as summarized in Table 3. In addition, the Kling-Gupta Efficiency (KGE) was calculated as a complementary performance metric. Following (Ahmed et al., 2023), KGE values were classified as unacceptable/poor (≤0.30), acceptable/satisfactory (>0.30–0.53), good (>0.53–<0.77), and excellent (≥0.77).

Given the discontinuous availability of observed streamflow records throughout the 1993-2023 study period, SWAT+ simulated streamflow was used as the primary streamflow dataset for the long-term climate-streamflow analysis. The available observed records, although limited to specific periods, were used for model calibration and verification. The calibrated and validated SWAT+ model was then used to generate a continuous simulated streamflow series throughout the study period, enabling consistent assessment of climate–streamflow relationships across the basin and different land-use periods. The calibrated and validated SWAT+ simulations provided a consistent basis for long-term analysis despite limited observed discharge records. Nevertheless, model-generated streamflow remains an important methodological limitation.

Our aim was not to conduct a comprehensive lagged analysis across all possible monthly or seasonal combinations. Instead, the focus is on two targeted relationships: antecedent JJA climate conditions and concurrent SON climate conditions in relation to SON streamflow. The SON period coincides with the peak development of the IOD and the strengthening of ENSO-related climate signals, making large-scale climate signals more pronounced. Moreover, streamflow during this season begins to respond to increasing rainfall, allowing climate-related variability to be more clearly detected in hydrological responses (Sahu et al., 2012). The selected seasons represent key stages of climate variability over Indonesia. Boreal summer (JJA) corresponds to the development phase of ENSO and IOD, whereas boreal fall (SON) represents their mature phase and the transition from the dry to the wet season, during which streamflow begins to increase and climate signals become more evident in hydrological responses (Sahu et al., 2012). To account for this, the relationship between climate indices and streamflow was evaluated not only for the concurrent period (climate indices from September to November and streamflow from September to November), but also by considering antecedent climate conditions from the previous season (climate indices from June to August and streamflow from September to November). This approach allowed for the identification of time-lagged statistical relationships between ocean–atmosphere variability and streamflow response, providing insight into how large-scale climate drivers are associated with hydrological responses and extreme flow events in the Batanghari River Basin.

To assess the relationship between climate variability and river streamflow, Pearson and partial correlation analyses were conducted between the climate indices and SWAT+ simulated streamflow. Pearson correlation was used to evaluated the strength and direction of the linear relationship, whereas partial correlation was applied to assess the relationship between each climate index and streamflow, while controlling for the association with the other climate index. These correlation approaches are commonly used to examine relationships and variability patterns between climate indices and hydroclimatic variables (Caroletti et al., 2021; dos Santos et al., 2023). Following the framework of (Sahu et al., 2012), streamflow and rainfall anomalies were calculated based on long-term climatological means to identify deviations from normal conditions. In the original framework, extreme high-flow and low-flow conditions were identified using thresholds of +1.5σ and −1σ, respectively. In this study, the threshold was modified to a symmetric criterion, with values greater than +1.5σ classified as extreme high-flow conditions and values lower than −1.5σ classified as extreme low-flow conditions. This modification was made to ensure that extreme high and low flow conditions were defined using a consistent standardized deviation from climatological mean. For the correlation analysis, although partial correlation accounted for the contribution of the other climate index, the analysis did not quantitatively assess interactions or coupling between ENSO and IOD. The discussion on coupling is presented in a conceptual context rather than a quantitatively assessed interaction.

The Pearson (r) and partial (rXY.Z) correlation coefficients were calculated using equations (1) and (2):

where Xi and Yi represent the climate indices and streamflow variables; X¯ and Y¯ are their respective means, and n is the number of observations. The partial correlation is expressed as shown in Equation (2), where rXY_, rXZand rYZrepresent the Pearson correlation between the respective pairs of variables. In this study, X and Y represent the climate index and streamflow respectively, while Z represents the other climate index. The strength of the correlation coefficient was interpreted based on the classification (Handoko et al., 2023), shown in Table 4.

ENSO and IOD events exhibited significant fluctuation from 1993 to 2023. The temporal variation is shown in Figure 8. During the period 1993–2000, the strongest positive IOD (pIOD) events occurred from mid- to late 1994 and in 1997, while a strong El Niño event developed in the period 1997–1998. These climate anomalies were associated with a weakened westerly monsoon flow; enhanced lower-tropospheric stability, and reinforced dry, stable atmospheric conditions, thereby contributing to reduced precipitation across many regions through large-scale atmospheric teleconnections (Kasera & Minocha, 2025). Such teleconnections refer to physically meaningful links between climate anomalies over large distances, driven by energy transport and wave propagation within the atmosphere-ocean system (Alizadeh, 2024). From a hydrological perspective, these are often associated with lower soil moisture and reduced river streamflow, particularly within catchment areas such as river basins.

The period 2000–2023 is characterized by highly active ENSO and IOD dynamics, with several notable events over the past two decades. The period exhibits pronounced climate variability, in which ENSO and IOD are closely associated with local and regional climate patterns (Ummenhofer et al., 2013). In the early 2000s, several moderate El Niño events occurred (2002–2003 and 2006–2007), followed by a transition to La Niña in 2007–2008. Another relatively weak El Niño episode was observed from 2009 to mid‑2010.

During this period, a strong El Niño developed in 2015–2016, comparable in intensity to the 1997–1998 event. Extensive sea surface warming in the Pacific induced large-scale circulation anomalies, including a weakening of the Walker circulation, anomalous ascent over the central–eastern Pacific, and subsidence over the western Pacific, thereby altering wind and rainfall patterns worldwide (C. Wang, 2002; Zhang et al., 2017). These years represent key milestones in modern climatology. Subsequently, La Niña conditions recurred in 2010–2011, 2016–2018 and 2020–2022, indicating sustained cooling in the central and eastern Pacific for more than two consecutive years. Such cooling is linked to complex interactions between subsurface temperature anomalies, ocean currents and strengthened trade winds, which intensify the Walker circulation and promote further cooling (C. Gao et al., 2022; Song et al., 2022).

A significant positive IOD (pIOD) event occurred in 2019, distinguished as one of the most notable events since the 1960s. This pIOD evolved primarily independently of El Niño and was initiated by Rossby waves in the southern Indian Ocean, which caused a sea-level rise, with warming in the west and cooling in the east, further intensified by ocean-atmosphere feedback mechanisms (Du et al., 2020; Lu & Ren, 2020). In addition, strong negative IOD (nIOD) events occurred, particularly in 2010 and 2016. These events were marked by extreme warming in the eastern Indian Ocean and very high sea surface temperature anomalies, while anomalies in the western basin remained relatively weak, producing a monopolar pattern distinct from the classical dipole structure. The 2016 event was further driven by both surface and subsurface conditions in the Indian Ocean and reinforced by long‑term warming in the region.

Monthly anomalies of streamflow and precipitation at the three gauging stations in the lower Batanghari River Basin are presented in Figure 9, namely those of (a) Muara Kilis, (b) Sungai Duren, and (c) Simpang Berbak. The orange line represents streamflow anomalies, while the blue indicates precipitation anomalies. Red and green markers denote extreme high-flow and low-flow anomalies respectively.

The three downstream stations exhibit broadly balanced distributions of positive and negative streamflow anomalies, but differ in the occurrence of extreme-flow events. Muara Kilis, located in the upper part of the lower basin and adjacent to the middle sub-basin, shows a relatively balanced distribution of positive and negative anomalies, with extreme low-flow events more frequent than extreme high-flow ones. In contrast, Sungai Duren is characterized by a similar overall balance between positive and negative anomalies, while extreme high-flow events are somewhat more frequent than extreme low-flow one. Finally, Simpang Berbak, situated near the river mouth, also shows a closely balanced distribution of positive and negative anomalies, with a slightly higher occurrence of extreme high-flow events than extreme low-flow ones. Across all the stations, extreme high-flow events occur at relatively similar frequencies, whereas the occurrence of extreme low-flow ones is comparatively higher at Muara Kilis. Periods of relatively high and low streamflow generally coincide with positive and negative precipitation anomalies respectively, suggesting a possible association between rainfall variability and streamflow dynamics in the lower Batanghari River Basin. This pattern indicates a link between atmospheric forcing and hydrological response, whereby variations in precipitation are associated with changes in river flow conditions in this humid tropical system.

Further analysis during the September-November season, as shown in Figure 10, reveals a generally consistent pattern of extreme flow conditions among the three stations. At Muara Kilis, extreme high-flow conditions are relatively infrequent and occurred only in specific years, notably in 2000 and 2022. In contrast, extreme low flow conditions occur more frequently, particularly during several episodes in the analysis period. This indicates that low flow extremes were more prevalent than high-flow ones at this station during SON.

At the Sungai Duren station, both extreme high and low flow conditions are observed, with low flow extremes occurring more frequently than high-flow ones. Extreme high-flow anomalies are observed in 2010, 2011, 2018 and 2022, while extreme low-flow events occur across several years, particularly during the earlier and later parts of the analysis period. At the Simpang Berbak station, extreme high-flow events are relatively infrequent, whereas extreme low flow ones occur more frequently. The timing of several extreme events is broadly consistent with those identified at Sungai Duren, particularly during the 2010-2011 and 2015-2019 periods. The occurrence of extreme events during similar periods at the two stations suggests a degree of temporal coherence in downstream flow variability.

Overall, the three stations exhibit a relatively similar frequency of extreme events during SON, with the total number of events ranging from 11 to 13. Extreme low-flow events consistently outnumber extreme high-flow ones at all three stations. Rather than indicating a systematic increase in the frequency of streamflow intensity anomalies from upper to lower, the results suggest a relatively consistent seasonal pattern of extreme streamflow conditions across the lower Batanghari River Basin.

The correlation analysis between two large-scale climate indices and seasonal streamflow anomalies in the Batanghari River at the three downstream stations (Table 5) reveals a consistent pattern of association between climate variability and hydrological response. It should be noted that this analysis is based on overall streamflow anomalies across the full dataset, rather than being limited to extreme high or low events. These events were examined separately in the previous section to illustrate the range and variability of streamflow responses.

Across all stations and index configurations, the Pearson correlation coefficients are uniformly negative. This suggests that El Niño conditions and positive IOD phases are generally associated with reduced streamflow during the September-November period, whereas La Niña conditions and negative IOD phases tend to correspond to increased streamflow. All correlations are statistically significant (p < 0.001). In terms of magnitude, the correlation coefficients range from weak–moderate to moderate–strong. For the June-August (JJA) indices correlated with September-November (SON) streamflow, IOD correlations range from -0.33 to -0.39, whereas ENSO correlations are stronger, ranging from -0.51 to -0.58 across the three stations. When SON indices are correlated with SON streamflow, IOD correlations increase to -0.56 to -0.64, while ENSO correlations range from -0.54 to -0.59. Among the three stations, Muara Kilis exhibits the strongest correlations, with the IOD–SON and ENSO– SON streamflow period correlations reaching −0.64 and −0.59, respectively. Sungai Duren and Simpang Berbak show relatively similar correlation magnitudes, with SON IOD-SON streamflow correlations of -0.56 and -0.58, and SON ENSO-SON streamflow correlations of -0.54 for both stations (Figure 11).

Nevertheless, differences in the relative association of ENSO and IOD are apparent depending on the seasonal timing of the climate indices relative to the streamflow period. For JJA indices and SON streamflow, ENSO exhibits stronger correlations than IOD at all three stations. In contrast, for SON indices and SON streamflow, IOD shows slightly stronger correlations than ENSO at the three stations. The stronger correlations obtained when climate indices and streamflow are considered within the same SON season also indicate a closer statistical association between climate variability and seasonal streamflow anomalies within the same season than between preceding JJA climate conditions and subsequent SON streamflow. Overall, the results suggest that the associations of ENSO and IOD with streamflow variability in the lower Batanghari River basin exhibit clear seasonal and spatial variability. The relationship between climate indices and streamflow is consistently stronger for the SON season, with IOD showing a slightly stronger association than ENSO across all three stations (Figure 11). Spatially, the strongest correlations were observed at Muara Kilis, whereas Sungai Duren and Simpang Berbak exhibited relatively similar correlation magnitudes. These findings highlight the importance of considering both the seasonal timing of climate indices and spatial variability among lower stations when characterizing climate–streamflow relationships in tropical river systems.

To further illustrate the linear relationships identified by the Pearson correlation analysis, scatter plots with fitted regression lines were used to visualize the relationship between the climate indices and SON streamflow. The regression lines provide a visual representation of the direction and relative strength of the linear associations identified by the correlations analysis across the three lower stations. At Muara Kilis, the scatter plots and fitted regression lines show a consistently negative linear relationship between the climate indices and SON streamflow (Figure 12). The relationship is relatively weak for IOD JJA, whereas ENSO JJA exhibits a stronger negative linear association with SON streamflow. For the SON indices, both IOD and ENSO demonstrate stronger negative relationships, with IOD SON exhibiting the strongest association among the four relationships. These patterns are consistent with the Pearson correlation results (Table 5), with the regression lines providing a visual representation of the direction and relative strength of the climate-streamflow relationships.

The scatter plots for Sungai Duren similarly show consistently negative linear relationships between the climate indices and SON streamflow (Figure 13). The relationship is weaker for IOD JJA, whereas ENSO JJA exhibits a stronger negative association with SON streamflow. For the SON indices, IOD shows a stronger negative relationship than ENSO, with the IOD SON relationship exhibiting the highest coefficient of determination among the four relationships. Overall, these regression patterns are consistent with the Pearson correlation result.

A similar pattern was observed for Simpang Berbak, where all four relationships of climate index-SON streamflow show negative regression slopes (Figure 14). ENSO JJA exhibits a stronger negative association with SON streamflow than IOD JJA, while for the SON indices, IOD shows a slightly stronger relationship than ENSO. The IOD SON relationship also shows a higher coefficient of determination than the ENSO SON relationship, consistent with the stronger Pearson correlation observed for IOD.

However, because ENSO and IOD may occur concurrently, Pearson correlations cannot distinguish the individual association of each climate index with streamflow while accounting for the other climate index. Therefore, a partial correlation analysis was conducted to evaluate the association of each climate index with SON streamflow, while statistically controlling for the other climate index. The partial correlation results show a clearer distinction between the associations of ENSO and IOD with SON streamflow (Table 6). For the relationship between JJA climate indices and SON streamflow, ENSO remains significantly and negatively correlated with SON streamflow at all three stations, with partial correlation coefficients ranging from -0.44 to -0.50 (p < 0.001). In contrast, the partial correlations between IOD and SON streamflow become much weaker, ranging from -0.17 to -0.22, and are not statistically significant at Sungai Duren or Berbak (p = 0.12 and 0.12, respectively). At Muara Kilis, the IOD association remains statistically significant but is weak (r = -0.22, p = 0.04). This indicates that the association between JJA climate indices and SON streamflow is more consistently represented by ENSO than IOD after accounting for the variability of the other climate index.

In contrast, regarding the relationship between the SON climate indices and SON streamflow, both indices remain significantly associated with streamflow after accounting for the other index. However, IOD consistently exhibits stronger negative partial correlations than ENSO at all three stations. The partial correlations for IOD range from -0.35 to -0.43, whereas those for ENSO range from -0.27 to -0.30. The strongest IOD association was observed at Muara Kilis (r = -0.43), followed by Simpang Berbak (r = -0.354) and Sungai Duren (r = - 0.349), while the corresponding ENSO correlations were -0.30, -0.27, and -0.27, respectively. All of these relationships are statistically significant (p ≤ 0.012). Overall, the partial correlation results indicate that the associations of ENSO and IOD with SON streamflow depend on the seasonal timing of the climate indices. ENSO JJA shows a more consistent association with SON streamflow, whereas for the concurrent SON period, IOD exhibits a stronger association with the streamflow than ENSO. These findings further demonstrate that the two climate indices show distinct patterns of association with seasonal streamflow variability. To quantify the relative strength of these associations, while accounting for the simultaneous presence of the other climate index, multivariate regression analysis was performed (Figure 15).

In relation to the resulting regression coefficients (β) and their 95% confidence intervals (Figure 15a), ENSO generally exhibits more negative regression coefficients than IOD across all three stations, indicating a stronger association with SON streamflow when the preceding JJA climate conditions are considered simultaneously. In contrast, for the same-season relationships between the SON climate indices and SON streamflow (Figure 15b), IOD consistently shows more negative coefficients than ENSO at all three stations, indicating a stronger association of IOD with SON streamflow for the same season. The error bars represent the 95% confidence intervals of the regression coefficients and indicate differences in the precision of the estimated associations among stations and climate indices. Overall, these results are consistent with the patterns identified by the Pearson and partial correlation analysis.

The analysis shows that streamflow variability in the lower Batanghari River Basin is characterized by significant temporal fluctuations, with both extreme high-flow and extreme low-flow occurring at different periods across the three stations. In general, the frequency of extreme events is relatively similar among the stations, although differences are observed in the timing and type of extremes. During SON, extreme low-flow extremes occurred more frequently than high-flow extremes at all three stations. This pattern indicates that the frequency of extreme events does not exhibit a systematic change from Muara Kilis to Simpang Berbak. Several extreme events also occurred during relatively similar periods across the stations, indicating a degree of temporal coherence in streamflow variability during SON. Similar spatiotemporal variability in extreme streamflow has been reported in other large river basins, where the timing and magnitude of extreme flows vary across seasons and locations and can be influenced by large-scale climate variability (Huang et al., 2024). At the global basin scale, ENSO has been linked to variations in flood frequency and duration, including simultaneous and lagged responses (Yan et al., 2020), while IOD has also been identified as an important climate mode influencing streamflow variability (Bates & Dowdy, 2024). This hydrological pattern provides a context for understanding the relationship between streamflow variability and large-scale climate signals, particularly ENSO and IOD.

The observed streamflow variability shows a consistent relationship with ENSO and IOD. The negative correlations observed at all stations indicate that El Niño and positive IOD conditions tend to be associated with lower streamflow, whereas La Niña and negative IOD conditions tend to be linked to higher streamflow. Similar relationship between ENSO-IOD phases and streamflow variability have been reported in the Jinsha River Basin, China, where both climate modes influenced annual and seasonal streamflow, with their effects varying between developing and decaying phases (Jia et al., 2023). Similar climate-streamflow linkages have also been reported for the Lancang-Mekong River, where ENSO and IOD were identified as important drivers of interannual streamflow variability and extreme hydrological conditions (S. Wang et al., 2024). From a hydroclimatic perspective, this pattern may be related to changes in atmospheric circulation and rainfall conditions in Indonesia during different ENSO and IOD phases. (As-syakur et al., 2014), on the basis of TRMM rainfall data, showed that rainfall responses to ENSO and IOD were relatively strong during JJA and SON, with pronounced patterns over southeastern Sumatra and Java. These findings provide a relevant hydroclimatic context for the lower Batanghari River Basin, as changes in rainfall conditions associated with Indo-Pacific climate variability may be reflected in changes in streamflow across contributing catchments. The negative relationships identified in this study are also consistent with the findings of (Sahu et al., 2012) for the Citarum River, which showed associations between El Niño and positive IOD conditions with lower flow conditions, and between La Niña and negative IOD conditions with higher flows.

However, the ENSO and IOD associations exhibit different patterns depending on the timing of the climate signal relative to the streamflow period. When JJA indices were related to SON streamflow, ENSO showed stronger relationships than IOD at all stations. More importantly, after statistically controlling for IOD through partial correlation, the relationship between JJA ENSO and SON streamflow remained significant, with coefficients ranging from −0.44 to −0.50, whereas the relationship between JJA IOD and SON streamflow weakened from −0.17 to −0.22 and became non-significant at Sungai Duren and Simpang Berbak. In contrast, for the SON–SON relationship, IOD showed stronger partial associations than ENSO at all stations. This difference indicates that the relative associations between the two climate modes and streamflow vary according to the timing of the climate signal and the period of hydrological response. This seasonal and lagged dependence is consistent with recent studies showing that ENSO and IOD influence hydrological variability differently across seasons and climate mode combinations, while ENSO can also modify seasonal river discharge variability (Zhu et al., 2025).

The stronger relationship between JJA ENSO and SON streamflow may be linked to the development of ENSO-related hydroclimatic signals that are already evident during JJA, as well as the broader spatial extent of ENSO-related hydroclimatic variability across Indonesia. (Kurniadi et al., 2021), who also examined this relationship in the tropics, showed that ENSO was independently associated with extreme rainfall across a broader area of Indonesia. Similarly, (As-syakur et al., 2014) showed that rainfall responses to ENSO and IOD were evident during both JJA and SON. Therefore, the hydroclimatic signal associated with ENSO during JJA may provide a clearer antecedent condition when entering the SON period. This is consistent with our results, in which the relationship between JJA ENSO and SON streamflow remains significant after controlling for IOD, whereas the relationship with JJA IOD is weakened. This interpretation is also consistent with (Nugroho et al., 2021), who reported a significant relationship between ENSO and the streamflow regime of the Code River, Indonesia Conversely, the stronger association of IOD in the SON–SON relationship may be related to its development from JJA toward SON, bringing the climate signal closer in time to the streamflow response. The IOD begins to develop during boreal summer and strengthens toward boreal fall. (Zhang et al., 2017) showed that IOD development is associated with strengthening southeasterly wind anomalies around Sumatra and Java, which promote cooling in the eastern equatorial Indian Ocean through interactions between wind, evaporation, SST and Bjerknes feedback. These processes subsequently strengthen the SST gradient characteristic of IOD development, leading to its mature phase during SON. Therefore, IOD conditions that begin to develop during JJA may provide an early hydroclimatic signal, while the subsequent strengthening of IOD toward its mature phase in SON may enhance its association with hydroclimatic conditions during the same season.

The findings of this study are generally consistent with previous research on the relationship between IOD and hydrological variability in Indonesia. (Sahu et al., 2012), based on observed daily discharge from the Citarum River, identified a relationship between antecedent IOD conditions and SON discharge. This finding supports our result that JJA IOD remains associated with SON streamflow. (Weller & Cai, 2014) showed that some IOD events develop during JJA and reach their mature phase during SON, supporting the cross-seasonal IOD relationship observed in our research. (Zhang et al., 2017) also showed that IOD development can extend from JJA to SON, consistent with the temporal pattern of IOD considered in this study. At the regional scale, (As-syakur et al., 2014) found that during SON, the relationship between IOD and rainfall was stronger than that of ENSO over southeastern Sumatra and Java. This pattern is consistent with our findings, in which IOD showed a stronger association with SON streamflow than ENSO in the SON–SON relationship. Overall, the results of this study are consistent with these regional characteristics: ENSO is more clearly expressed as an antecedent signal during JJA, whereas IOD shows a stronger concurrent association when both the climate index and streamflow are considered during SON.

The differences between the JJA–SON and SON–SON relationships also highlight the importance of hydrological response timescales. River discharge integrates rainfall variability from the contributing catchment, meaning that the statistical relationship between climate indices and streamflow may depend on the timing of the climate signal relative to the river response period (Canchala et al., 2024; Moges et al., 2022). Therefore, the relationship between JJA ENSO and SON streamflow should not necessarily be interpreted as a direct response occurring at the same time as the index, but rather as an association between antecedent climate conditions and streamflow conditions in the subsequent season. Conversely, the stronger IOD SON–SON relationship indicates closer temporal alignment between the climate signal and hydroclimatic

response during the same season. In addition to temporal alignment, spatial contributions should also be considered. (Rice & Emanuel, 2017) showed that watersheds can act as spatiotemporal filters that modify how hydroclimatic signals are translated into streamflow responses, together with other geohydrological factors. This concept may also help explain the spatial variability observed in the lower Batanghari River Basin.

Although all three stations are located in the lower part of the basin, Muara Kilis exhibits the strongest associations, whereas Sungai Duren and Simpang Berbak show relatively similar magnitudes. Previous studies have also shown that the relationship between large-scale climate variability and streamflow can vary spatially among river basins because of differences in catchment characteristics and hydrological response (Pérez-Ciria et al., 2022; Worako et al., 2021). Muara Kilis is located relatively farther upstream than the other two stations and is adjacent to the transition between the middle and lower basin. Its position means that streamflow at the station is augmented by upstream and middle-basin catchments before the flow continues toward the downstream stations. This transition is also accompanied by a change in river morphology and topographic setting, from relatively steeper upstream and middle-basin reaches, towards a gentler and more alluvial downstream environment. The convergence of flows from upstream and middle basin contributing areas at this transition may allow the regional hydroclimatic signal to remain relatively pronounced at Muara Kilis. Further downstream, the river channel becomes increasingly meandering, as observed along the reaches toward Sungai Duren and Simpang Berbak. The greater channel sinuosity may increase the flow path and contribute to the attenuation and temporal smoothing of upstream streamflow signals during downstream propagation (Barneveld et al., 2024; Paiva & Lima, 2024). In addition, increasing contributions from groundwater and baseflow in the alluvial lower reaches may further modify the streamflow signal. These processes may partly explain why the associations with ENSO and IOD are relatively stronger at Muara Kilis. However, this difference does not indicate a simple upstream-to-downstream gradient (Wohl et al., 2025), as Sungai Duren and Simpang Berbak exhibit relatively similar associations.

From an applied hydrological perspective, these findings may be useful for understanding seasonal streamflow conditions in the lower Batanghari River Basin by considering the phase of climate signals rather than simply the ENSO or IOD phase. The stronger JJA ENSO–SON streamflow association indicates that ENSO conditions during JJA are more strongly linked to subsequent SON streamflow conditions, whereas SON IOD shows a closer association with streamflow variability during the same season. This finding highlights the potential relevance of antecedent climate conditions for understanding seasonal streamflow variability, consistent with previous studies examining the role of climate information in hydrological assessment (Beckers et al., 2016; Petry et al., 2023; Slater & Villarini, 2018). However, the findings do not constitute an evaluation of operational forecasting skill; the development of a forecasting system would require explicit assessment of lead time, hindcasts, prediction uncertainty and predictive performance.

Nevertheless, a more comprehensive understanding of the basin’s hydrological system requires further investigation, particularly through spatial analysis of hydrological responses from up-stream to downstream. Such an approach is essential for capturing variations in hydrological pro-cesses along the river continuum. In addition, incorporating more dynamic land use/land cover (LULC) representations with multiple time periods is necessary to better capture surface chang-es within the basin. This would allow for a more robust assessment of the relationship between land use change and streamflow variability. Furthermore, recent studies based on observed discharge have also highlighted the importance of considering temporal scales and hydrological persistence in climate–streamflow relationships (Mendoza et al., 2026). A broader lag analysis across multiple monthly and seasonal windows is also needed in future studies to further evaluate the predictive value of ENSO and IOD for streamflow variability in the lower Batanghari River Basin.