Seasonal streamflow variability and surface water assessment in the Ia L o ^ ́ p and Ia H'leo basins of the Ea Súp semi-plain, Central Highlands of Vietnam

1 Introduction

In the context of climate change and increasing water demand, basin-scale water-resource management requires reliable estimates of seasonal water availability. Hydrological models are indispensable for simulating rainfall-runoff processes, quantifying water availability, and supporting decisions on water allocation, reservoir operation, and disaster risk reduction. Rainfall-runoff models are typically categorized as black-box (empirical), conceptual, or physically based (Beven, 2012; Sitterson et al., 2018). Black-box models emphasize empirical input-output relationships, while physically based models explicitly simulate hydrological processes, and yet require extensive datasets that are often lacking in developing regions. Conceptual models have therefore gained wide application due to their balance of physical realism and computational efficiency. They represent processes such as precipitation, infiltration, evapotranspiration, soil-water storage, and baseflow generation in a simplified yet practical manner (Jajarmizadeh et al., 2012). More recently, Jehanzaib et al. reaffirmed this tripartite classification and emphasized the emerging role of hybrid approaches (Jehanzaib et al., 2022).

Moreover, methodological advances, such as the multi-objective automatic calibration strategy proposed by Madsen (2000) using the SCE-UA algorithm, have further enhanced the model's robustness and adaptability across diverse hydrological conditions (Madsen, 2000).

MIKE-NAM has been widely applied as a lumped conceptual rainfall-runoff model in data-scarce basins because it balances modest input requirements with physically interpretable storage and routing parameters. Previous applications in Asia, Europe, Africa and Vietnam demonstrate its utility for reservoir inflow estimation, seasonal water assessment and climate-impact studies. International applications have reported the usefulness of MIKE-NAM and related MIKE 11/NAM configurations for flood forecasting, reservoir inflow simulation, flow-duration assessment and climate-impact analysis in both gauged and data-scarce basins (Larsen et al., 2010; Doulgeris et al., 2011; El-Nasr et al., 2013; Singh et al., 2014; Abdul Razad et al., 2018; Ghebrehiwot and Kozlov, 2020; Sajadi Bami et al., 2020; Teshome et al., 2020; Ghosh et al., 2022; Nannawo et al., 2022; Slieman and Kozlov, 2023; Dhruw et al., 2024; Werede et al., 2024; Yudianto et al., 2025). At the same time, this literature shows that precipitation inputs, parameter transferability and low-flow simulation remain major sources of uncertainty, particularly in monsoon-influenced basins with sparse hydrometric observations. Accordingly, this review emphasizes data-scarce basins, uncertainty sources, and low-flow simulation challenges that are directly relevant to the present study.

In Vietnam, MIKE-NAM has been used in several river basins for streamflow simulation, flood forecasting and climate-impact assessment. Vietnamese and Lower Mekong applications have also used MIKE-NAM, MIKE 11/HD or related MIKE frameworks for flood simulation, surface-water calculation, streamflow forecasting and climate-change impact assessment in several basins, including the Sê San, Ba, Mekong, Trà Khúc, Nhuệ-Đáy and Nam Ou systems (Tran and Tran, 2011; Dang et al., 2018; Hoang and Vo, 2018; Pham and Nguyen, 2019; Luong et al., 2020; Nguyen et al., 2021, 2023a,b). However, existing work has mainly emphasized large basins or water-allocation models, while sub-basin rainfall-runoff reconstruction in the lower Srêpôk/Ea Súp semi-plain remains limited.

Beyond conventional hydrological assessments, water resources in the Central Highlands are increasingly interpreted through the Water-Energy-Food (WEF) nexus, which emphasizes trade-offs among agricultural irrigation, urban water supply, and the energy required for water abstraction and treatment in rapidly urbanizing regions. Recent groundwater modeling studies in Đắk Lắk province using MODFLOW indicate growing pressure on aquifers from both irrigated agriculture and urban demand, highlighting the need to regulate total abstraction and its spatial distribution to ensure long-term water security (Nguyen et al., 2022). In this context, upstream sub-basins such as Ia H'leo and Ia Lốp contribute to streamflow regulation and groundwater recharge, which in turn support agricultural production and help sustain raw-water availability for nearby urban centers, particularly Buôn Ma Thuột (Đắk Lắk province) and, indirectly, Pleiku (Gia Lai province). Complementary evidence from managed aquifer recharge (MAR) studies suggests that reservoirs, temporary storage zones, and land-use management can enhance dry-season reliability and reduce energy-intensive pumping, reinforcing the importance of integrating upstream flow regulation and groundwater recharge in WEF-oriented water planning (Pavelic et al., 2022).

Climate projections indicate that temperatures in the Central Highlands may increase by approximately 1.0–1.6 °C during the 2040–2059 period, accompanied by a continued decline in dry-season rainfall and an increase in summer-autumn precipitation, resulting in increasingly uneven seasonal rainfall distribution (Tran et al., 2024). At the scale of the Srêpôk River Basin, dry-season streamflow is projected to further decrease due to enhanced evapotranspiration and the regulating effects of existing reservoir systems, thereby exacerbating the risk of water shortages during the dry season (Tran et al., 2023). Recent studies on drought stress, climate-change adaptation and water management in Vietnam further emphasize the need for climate-sensitive streamflow assessment and adaptive reservoir and irrigation planning (Nguyen et al., 2023c, 2024, 2025). In parallel with these scientific assessments, water-resources management in the Srêpôk Basin is currently guided by the Integrated River Basin Planning for the period 2021–2030, with a vision to 2050, which adopts a basin and inter-basin management approach consistent with the Water-Energy-Food nexus. This policy framework explicitly recognizes the multifunctional roles of key upstream water sources, including the Ia H'leo and Ia Lốp rivers, in supporting domestic and urban water supply, agricultural production, industrial development, hydropower generation, and environmental protection, underscoring the strategic importance of regulating upstream flows and groundwater recharge for long-term urban water security under climate change (Government of Vietnam, 2021).

For the Srêpôk basin, recent studies have adopted multiple approaches to assess water resources. The MIKE HYDRO BASIN model was used to estimate regional water balance in the Central Highlands (Dao et al., 2020). MIKE BASIN was applied specifically to the Srêpôk basin to evaluate present and future water-balance scenarios (Nguyen, 2022). An integrated MIKE HYDRO BASIN-Water Scarcity Index framework identified spatial patterns of water stress (Le et al., 2024). CMIP6-based analyses characterized rainfall distributions to provide climatic inputs for hydrological modeling (Phan et al., 2025). However, applications in the lower Srêpôk basin have largely focused on basin-scale water balance and allocation models, with limited attention to sub-basin rainfall-runoff reconstruction.

The Ea Súp semi-plain, located in the lower Srêpôk basin, represents a critical area for water-resources assessment due to its strong climatic seasonality, prolonged dry season, and rapidly increasing agricultural and domestic water demand (Nguyen, 1986, 1985). Despite a relatively dense river network, water availability in the area is increasingly constrained by climatic variability and upstream regulation, heightening the risk of dry season shortages and competition among water users.

MIKE-NAM was selected because it is suitable for data-scarce, monsoon-influenced basins and can represent the dominant seasonal storage and release processes using a parsimonious parameter set. In the Ea Súp semi-plain, where direct streamflow observations are limited, this structure provides a practical basis for reconstructing monthly streamflow and for linking hydrological seasonality with WEF- and NbS-oriented water-resource planning.

1.1 Study area characterization

The Ia H'leo and Ia Lốp river basins are located within the Srêpôk River system, a major tributary of the Mekong. The Ia Lốp basin covers approximately 1,660 km² and originates on the southern slopes of Hàm Rồng Mountain (approximately 800 m) in Chu Sê commune (Gia Lai province); the river flows southeastward before turning southwest to converge with the Ia H'leo River.

The Ia H'leo basin drains ≈2.95 × 10³ km², with headwaters at elevations >900 m around Ia Drăng commune (Đắk Lắk province); its main channel flows north through Ea H'leo commune and then westwards to join the Ia Lốp basin. Downstream, the combined flow continues across the Vietnam-Cambodia border and joins the Srêpôk, which is formally recognized in the national planning framework for the Srêpôk basin and in basin authority summaries as a major Mekong tributary. The study area, river network, and hydrometeorological station locations are shown in Figure 1.

Despite a relatively dense river network, the hydrological regime is highly seasonal, with a pronounced dry period during which irrigation demand reaches its annual peak. This seasonal imbalance between water availability and demand makes local water security particularly sensitive to inter-annual hydro-climatic variability and upstream regulation.

In recent decades, increasing pressure from agricultural expansion, domestic water use, and reservoir operation has intensified dry-season water shortages in the Ea Súp semi-plain. While previous studies in the Srêpôk basin have provided valuable basin-scale assessments of water balance and allocation, rainfall-runoff processes and flow variability at the sub-basin scale-especially within key contributing basins such as Ia H'leo and Ia Lốp-remain insufficiently quantified. As a result, the hydrological basis for dry-season management and irrigation planning is still subject to considerable uncertainty.

This study aims to calibrate and validate MIKE-NAM using reliable reservoir-inflow observations, reconstruct monthly streamflow for 1980–2022, assess seasonal and long-term water-resource variability, and discuss implications for WEF- and NbS-oriented water planning.

2 Materials and methods

2.1 Materials

Based on the input requirements for the MIKE-NAM model and the availability of gauging infrastructure in the study area, the following datasets were utilized for the 1980–2022 period:

• - Digital Elevation Model (DEM) data with a spatial resolution of 30 × 30 m, obtained from the U.S. Geological Survey (USGS), together with spatial data layers including basin boundaries and river network maps¹;
• - Meteorological Data: Daily rainfall records from six stations (Buôn Hồ, Chư Prông, Chư Sê, Ea H'leo, Ea Súp, and Bản Đôn) and daily evaporation records from the Buôn Hồ station. These were obtained from the National Hydro-Meteorological Monitoring Network²;
• - Hydrological Data: Daily average inflow records from the Ea Súp Thượng, Ia Mo, and Plei Pai reservoirs, extracted from the water resources monitoring system of the Ministry of Agriculture and Environment (formerly the Ministry of Natural Resources and Environment)³;

The meteorological datasets provide a 43-year input basis for long-term rainfall-runoff reconstruction. In contrast, reliable and quality-controlled reservoir inflow observations are available only for 2019–2022 and were used exclusively for model calibration and validation. Therefore, the reconstructed 1980–2022 monthly streamflow series should be interpreted as a model-based hydrological estimate rather than a direct observational record.

The spatial representativeness of the hydro-meteorological network is a primary source of uncertainty in this study. Although six stations provide relatively dense coverage for the basin, the complex topography of the Central Highlands can lead to localized rainfall variations not fully captured by point measurements. The use of the Thiessen Polygon method assumes uniform distribution within each polygon, which may introduce biases during extreme convective events. Furthermore, potential evapotranspiration (PET) data were derived solely from the Buôn Hồ station. While this is the most reliable long-term record available, its higher elevation compared to the semi-plain Ea Súp region may lead to a slight underestimation of PET, particularly during the peak dry season. These data constraints were addressed during the calibration process by tuning the evaporation focal factor (fEDF) and storage parameters to align simulated discharges with observed reservoir inflows.

2.2 Methods

Among conceptual models, MIKE-NAM (Nedbor-Afstrømnings Model), developed by the Danish Hydraulic Institute (DHI), has become one of the most widely applied tools for water resources management over the past decades. As a lumped conceptual rainfall-runoff model, it continuously simulates catchment water balance by representing storage and transfers among surface, root zone, groundwater, and snow reservoirs, each corresponding to distinct hydrological processes. Its relatively modest data requirements, combined with the ability to reproduce key runoff dynamics, make MIKE-NAM particularly valuable in regions with limited monitoring capacity. Over time, the model has demonstrated its effectiveness in simulating rainfall-runoff processes and supporting decision-making in water management, from small catchments to regional and national scales.

The MIKE-NAM rainfall-runoff model was employed to simulate surface-water resources in the Ia H'leo and Ia Lốp river basins. Originally developed in the 1970s at the Technical University of Denmark, NAM (Nedbør-Afstrømnings-Model, or “rainfall-runoff model” in Danish) is a lumped conceptual model widely applied in flood forecasting, reservoir inflow simulation, and basin-scale water management (DHI, 2017). In NAM, each basin is represented as a single computational unit, with model parameters describing the average hydrological behavior of the catchment.

The model continuously simulates rainfall-runoff processes through interconnected reservoirs representing surface storage, soil moisture in the root zone, and two groundwater storages (upper and lower). Fluxes between these storages are governed by parameters calibrated against observed discharge. This structure allows NAM to capture the seasonal water balance of catchments with pronounced wet-dry contrasts, making it well suited for the Central Highlands.

In this study, the methodological workflow (Figure 2) comprises four main steps: (i) input data screening and preparation (DEM, hydro-meteorological records, and reservoir inflow data), (ii) basin delineation and calculation of rainfall weight coefficients using ArcGIS, (iii) manual calibration and independent validation of MIKE-NAM against observed reservoir inflows, and (iv) synthesis of monthly outputs, rainfall contribution, surface water resources, and long-term fluctuation trends.

P=Qs+Qi+Qb+ETa+ΔS(1)

Where: P is precipitation, Q_s surface runoff, Q_i interflow, Q_b baseflow, ET_a actual evapotranspiration, and ΔS the change in catchment storage.

Key model parameters include the maximum root-zone storage capacity (Lmax), overland flow runoff coefficient (CQOF), threshold parameters controlling runoff generation, and time constants governing routing of surface and subsurface flows (CK1, CK2, and CKBF). These parameters are assumed to remain stationary over the simulation period, an assumption commonly adopted in long-term hydrological modeling where detailed land-use change data are unavailable.

Overall, the adopted framework combines data screening, physically interpretable NAM parameterization, manual calibration supported by multi-metric performance evaluation, and explicit discussion of data and parameter uncertainty. Because calibration data are limited, the long-term reconstructed series is used as a practical model-based hydrological baseline rather than as a substitute for continuous observed discharge.

3 Results

3.1 Streamflow simulation setup for the Ia H'leo and Ia Lo^́p river basins

3.1.1 Sub-basin delineation and determination of rain gauge influence weights

3.1.1.1 Sub-basin delineation

The delineation of sub-basins within the Ia H'leo and Ia Lốp river basins was based on watershed divides derived from topographic and hydrographic features, primarily reflecting the hydrological connectivity of main rivers and their tributaries. This approach yields hydrologically coherent sub-units that allow more accurate assessment of water resource potential and related factors. To ensure precision in boundary delineation, Geographic Information System (GIS) techniques were employed in conjunction with digital elevation data and hydrographic maps, facilitating spatial analysis and visualization of topographic and hydrological patterns within the study area. The resulting sub-basin division is shown in Figure 3.

3.1.1.2 Calculation of rain gauge weights

Rainfall is a critical input for rainfall-runoff modeling and strongly influences the identification of appropriate model parameters at the sub-basin scale. Therefore, the first step was to delineate rainfall influence zones and determine the relative contribution (weights) of individual rain gauges in computing mean areal rainfall for each sub-basin. Based on observed rainfall data and the spatial distribution of meteorological and rainfall stations within and surrounding the Ia H'leo and Ia Lốp basins, spatial analysis tools in a GIS environment were used to estimate sub-basin rainfall. The Thiessen polygon method (Create Thiessen polygons) was applied to establish the area of influence of each rain gauge and to derive station- specific weighting factors corresponding to each sub-basin.

The weight is defined as the proportional coefficient representing the ratio between the portion of the basin area represented by a given rain gauge- located within or adjacent to the basin and the total basin area. The area of influence controlled by each rain gauge is determined as follows: rain gauge stations are connected by straight lines to divide the basin into a series of triangles, and the perpendicular bisectors of the triangle sides are then constructed. These bisectors define the boundaries of the area of influence associated with each rain gauge.

The mean areal rainfall over the basin is calculated using the following Equation 2:

X¯=∑i=1nXifi∑i=1nfi(2)

Where:

• X_*i* : the rainfall amount recorded at the i-th rain gauge;
• f_*i*: the area of influence associated with the i-th rain gauge within the basin;
• n : the number of rain gauge stations used in the calculation.

3.1.2 Calibration and validation of the MIKE-NAM model

Model parameters were calibrated manually through an iterative adjustment process guided by hydrological plausibility and multi-metric performance evaluation. Daily reservoir-inflow records from Ea Súp Thượng, Ia Mo and Plei Pai were used to run the model, whereas the main performance statistics were evaluated at the monthly scale to match the objective of reconstructing monthly streamflow for 1980–2022. Because no natural hydrological gauging stations are available in the study basins, reservoir-inflow records constitute the only available observations for calibration and validation; this limitation is explicitly considered in the uncertainty assessment. The locations of the model calibration and validation points are summarized in Table 1.

NSE=1-∑i=1n(Qsim-Qobs)2∑i=1n(Qobs-Q¯obs)2(3)

Table 1. Locations of model calibration and validation points.

Where: Q_*sim*: Simulated discharge; Q_*obs*: Observed discharge; Q¯obs: Mean observed discharge (Moriasi et al., 2007).

To provide a more comprehensive evaluation, especially regarding water balance and overall model fitness, Percent Bias (PBIAS) and Kling—Gupta Efficiency (KGE) were also calculated using Equations 4, 5, respectively.

• - Percent Bias (PBIAS):

PBIAS=∑i=1n(Qsim-Qobs)∑i=1nQobs100(4)

Where Q_obs and Q_sim are observed and simulated discharge at time. PBIAS measures the average tendency of the simulated data to be larger or smaller than their observed counterparts. According to Moriasi et al. (2007), for monthly streamflow simulation, |PBIAS| < 10% indicates very good performance in terms of bias, 10–15% indicates good performance, and 15–25% indicates satisfactory performance (Moriasi et al., 2007).

• - Kling-Gupta Efficiency (KGE):

KGE=1- (r-1)2+(α-1)2+(β-1)2(5)

Where r is the linear correlation coefficient, α is the variability ratio and β is the bias ratio (Gupta et al., 2009).

KGE provides a balanced assessment by accounting for correlation, bias, and variability simultaneously.

Table 2 summarizes the calibrated parameter values of the MIKE-NAM model for the Ea Súp Thượng, Ia Mo, and Plei Pai reservoirs; these values would then serve as the basis for streamflow simulation in the Ia H'leo and Ia Lốp basins.

Table 2. Model parameters determined after calibration and validation.

The variability of calibrated parameters across the three reservoirs highlights the influence of local physiographic and land-cover conditions on catchment hydrological response. The higher storage capacities (Umax, Lmax) observed at Ia Mo suggest the enhanced infiltration and soil-moisture retention, which is likely to be associated with deeper basaltic soils and greater forest cover. In contrast, the large baseflow recession constant (CKBF) at Plei Pai indicates a slower and more sustained groundwater contribution, consistent with gentler topography and more permeable geological formations. These differences underscore the need to account for catchment heterogeneity when applying conceptual models in the Central Highlands. Overall, the calibrated parameter sets (Table 2) demonstrate that the MIKE-NAM model reproduces the hydrological behavior of the Ia H'leo and Ia Lốp sub-basins with consistency. The parameter values for CQOF (Overland flow runoff coefficient), TOF (Root zone threshold value for overland flow), TIF (Root zone threshold value for interflow), and TG (Root zone threshold for groundwater recharge) indicate a realistic partitioning of overland flow, interflow, and groundwater recharge, while the ranges of other parameters remain physically plausible and hydrologically meaningful. These results provide a robust basis for subsequent streamflow simulations and water resource assessments in the study basins.

The streamflow simulation using the NAM model shows that simulated and observed reservoir inflows are generally aligned in terms of seasonal patterns. Although discrepancies remain in some months and basins, Table 3 reports acceptable to good monthly-scale model performance. NSE values range from 0.73 to 0.95 during calibration and from 0.84 to 0.91 during validation. For Ia Mo, the dry-season volume difference was +1% during calibration and −4% during validation. These results indicate acceptable to good model performance, but dry-season and low-flow estimates should be interpreted with caution.

Table 3. Monthly-scale model performance during calibration and validation.

NSE—Nash–Sutcliffe Efficiency coefficient for streamflow simulation (Nash and Sutcliffe, 1970).

PBIAS—Percent Bias (Moriasi et al., 2007) |PBIAS| < 10% very good; 10–15% good; 15–25% satisfactory.

KGE—Kling–Gupta Efficiency (Gupta et al., 2009); higher values indicate better model performance.

ΔQ0—Mean annual streamflow volume error (%).

ΔQdry—Mean streamflow volume error during the dry season (November–May) (%).

Performance classification is based on the combined interpretation of NSE, KGE, PBIAS, annual volume error and dry-season volume error, rather than on NSE alone.

PBIAS values for the three basins range from −8.4% to +22.5%. Following Moriasi et al. (2007), these correspond to “very good” performance (|PBIAS| < 10%) for Ea Súp Thượng (−8.4% to −5.2%) and Plei Pai (+0.7% to −5.4%), while Ia Mo falls within the “good to satisfactory” range (+16.8% to +22.5%). The higher positive bias at Ia Mo can be attributed to uncertainty in precipitation spatial distribution over the upper catchment, where a single rain gauge represents a relatively large contributing area.

In addition to NSE, the Kling-Gupta Efficiency (KGE) was calculated to provide a more balanced assessment of the model performance. The KGE values ranged from 0.71 to 0.82 during calibration and from 0.68 to 0.78 during validation, with most values exceeding the 0.75 threshold for ‘good' performance (Gupta et al., 2009).

To further illustrate the performance of the calibrated MIKE-NAM model, Figure 4 presents a comparison between observed and simulated inflows at the Ea Súp Tho^́ng, Ia Mo, and Plei Pai reservoirs during both the calibration and validation periods. The corresponding monthly calibration and validation comparisons are shown in Figure 5.

A fundamental constraint in the Ia H'leo and Ia Lốp basins is the absence of natural hydrological gauging stations. No continuous streamflow observations are available along free-flowing river reaches within the study area. Instead, the only discharge data suitable for model calibration and validation are reservoir inflow records from the Ea Súp Thượng, Ia Mo, and Plei Pai reservoirs.

Importantly, these reservoir inflow datasets are reliable, continuous, and quality controlled only for recent years, beginning approximately in 2019. Earlier records are either incomplete, discontinuous, or lack sufficient quality assurance for quantitative model calibration. As a result, it is not methodologically feasible to calibrate the MIKE-NAM model over the entire 1980–2022 period, despite the availability of long-term rainfall and evaporation data.

Under such conditions, the selection of the 2019–2021 period for model calibration is not a subjective or arbitrary choice, but a data-driven necessity. Attempting to extend calibration to earlier decades without reliable discharge observations would undermine parameter robustness and compromise the scientific validity of the model.

The calibration period (2019–2021) was selected because it represents the most reliable reservoir inflow data available, not because it fully captures the long-term hydroclimatic variability of the basin. Although this interval includes contrasting wet- and dry-season conditions, it remains short relative to the 1980–2022 reconstruction period. Therefore, the reconstructed series should be interpreted as a model-based hydrological estimate subject to data and parameter uncertainty.

Validation with the independent 2022 record provides useful evidence that the calibrated parameter set can reproduce the main seasonal runoff dynamics, but it does not eliminate uncertainty associated with short observed records, reservoir-inflow estimation, parameter stationarity, land-use change and ungauged dry-season abstractions.

The year 2022 was reserved as an independent validation period. The validation results support the use of the model for monthly seasonal assessment, but the short validation period means that conclusions regarding extreme years and long-term variability should remain cautious.

On this basis, the calibrated MIKE-NAM model was applied to reconstruct continuous monthly streamflow for 1980–2022. This strategy is useful in ungauged or poorly gauged basins, but the resulting series should be treated as a model-based hydrological estimate rather than a direct observational record.

The model's performance was further evaluated using Flow Duration Curves (FDCs) on a semi-logarithmic scale (Figures 6a–c). The simulated and observed curves show reasonable agreement across the three sub-basins, particularly for medium- and high-flow conditions. However, deviations in the lower tail (high exceedance probabilities) indicate that low-flow and dry-season simulations remain more uncertain than wet-season runoff estimates. The FDC analysis therefore supports the practical adequacy of the model for seasonal water-resource assessment, while also highlighting the need for caution when interpreting extreme low-flow conditions.

In the Ea Súp Thượng basin (Figure 6a), the model demonstrates exceptional fidelity, with the two curves nearly overlapping across the entire probability range, from peak flows to baseflow. For the Ia Mo (Figure 6b) and Plei Pai (Figure 6c) basins, the model exhibits high precision in simulating flood events at low exceedance probabilities (P < 10%), ensuring the reliability of the reconstructed series for flood risk assessment.

However, at high exceedance probabilities (P > 90%), representing extreme low-flow conditions, a slight overestimation by the model is observed, particularly in the Ea Súp Thượng basin. From a hydrological research perspective, this discrepancy at the “lower tail” serves as a quantitative indicator of data constraints and hydrological uncertainties arising from the spatial sparsity of the rain-gauge network and potential ungauged local irrigation abstractions during the dry season. Despite these minor deviations during the recession period, the consistent NSE, PBIAS, and KGE metrics confirm that the 43-year reconstructed flow series provides a scientifically sound foundation for water resource planning and Nature-based Solutions (NbS) development in the Ea Súp semi-plain region.

3.1.3 Parameter sensitivity and uncertainty analysis

To provide a quantitative screening-level assessment of parameter sensitivity and uncertainty, the SCE-UA calibration trials generated for the MIKE-NAM model were analyzed. For each reservoir catchment (Ea Súp Thượng, Ia Mo and Plei Pai), 2,001 parameter sets were retained together with the associated aggregated objective function, which combines water-balance deviation and root-mean-square difference between simulated and observed flows. Parameter sensitivity was quantified using Spearman rank correlation between each parameter and the objective function. Larger absolute correlation values indicate stronger parameter influence on model behavior, while the sign indicates the direction of the relationship within the explored parameter range. The parameter sensitivity results are summarized in Table 4.

Table 4. Parameter sensitivity (Spearman rank correlation between each MIKE-NAM parameter and the SCE-UA aggregated objective function; larger absolute values indicate higher sensitivity).

Values are Spearman rank correlations between each MIKE-NAM parameter and the SCE-UA aggregated objective function, based on 2,001 calibration trials for each reservoir catchment. Larger absolute values indicate higher parameter sensitivity. A positive value indicates that increasing the parameter tends to increase the objective function within the explored range, whereas a negative value indicates that increasing the parameter tends to reduce the objective function and improve model fit. Routing time constants (CKIF and CK1), TOF and TIF showed lower and less consistent sensitivity and are therefore omitted for brevity.

The results show that the maximum root-zone storage capacity (Lmax), the baseflow recession constant (CKBF) and the groundwater recharge threshold (TG) are the most influential parameters across the three catchments. These parameters control soil-water storage, groundwater contribution and the recession of dry-season flow. Their high sensitivity suggests that the reconstructed streamflow series, particularly during the dry season, is strongly affected by subsurface storage and baseflow processes. This finding is consistent with the observed low-flow deviations and supports cautious interpretation of dry-season estimates.

The spread of model performance among the best-performing parameter sets is relatively small, indicating that the calibrated parameter values are reasonably stable over the calibration period. However, this uncertainty assessment remains conditional on the available 2019–2022 reservoir-inflow records. It does not fully account for other sources of uncertainty, including rainfall spatial variability, model structural assumptions and the limited length of observed discharge data. Therefore, the 1980–2022 reconstruction should still be interpreted as a model-based hydrological estimate rather than a direct observational record. Future studies should apply more formal uncertainty frameworks, such as GLUE or Monte Carlo simulation, when longer and more complete observed streamflow records become available.

3.2 Water resource assessment for sub-basins

3.2.1 Rainwater resources

The long-term mean annual rainfall was derived from observational data from rainfall stations, supplemented by correlation analysis to ensure consistency over the period 1980–2022. The calculated values for meteorological stations located within and adjacent to the Ia H'leo and Ia Lốp basins are summarized in Table 5.

Table 5. Average rainfall at meteorological stations in and around the river basin.

Rainfall exhibits clear spatial variability across the region, ranging from 1,551 mm at Bản Đôn to 2,333 mm at Chu Prông. The higher rainfall values observed at stations located on elevated terrain (e.g., Chu Prông, Chu Sê) contrast with the lower values in lowland areas (e.g., Bản Đôn, Ea Súp), reflecting a strong influence of orographic effects and topographic setting on precipitation distribution within and around the Ia H'leo and Ia Lốp basins. The spatial distribution of mean annual rainfall is shown in Figure 7.

3.2.1.1 Seasonal rainfall distribution

According to the “exceedance loss” criterion, with an occurrence frequency of P > 75% at rainfall stations in the basin, the rainy season is determined to occur from May to October, while the dry season extends from November to April of the following year. This seasonal division is consistent with hydrological patterns observed in the Central Highlands and other regions of Vietnam. The dry season lasts for 6 months and contributes only 8–16% of the total annual rainfall, whereas the rainy season, also lasting 6 months, accounts for 84–92%. The seasonal classification of rainfall at the stations is summarized in Table 6. Such a distinct seasonal contrast has significant implications for water resource management and agricultural activities in the basin.

Table 6. Seasonal classification of rainfall at stations in and around the Ia Lo^́p and Ia H'leo river basins based on the exceedance loss criterion (P > 75%).

Seasonal rainfall distribution further amplifies this spatial contrast. The rainy season (May–October), driven by the Southwest monsoon, contributes on average 88.6% of annual rainfall, whereas the prolonged dry season (November–April) accounts for only about 11.4%. Such strong seasonality directly governs runoff generation, storage dynamics, and water availability, explaining the marked concentration of streamflow during the flood season and the severe reduction of discharge during dry months. These features highlight the vulnerability of the study area to both floods and droughts and underscore the importance of seasonal regulation in water-resources planning.

3.2.2 Surface water resources

The MIKE-NAM model simulations provide detailed estimates of annual runoff characteristics for the Ia Lốp and Ia H'leo river basins. At the outlet of the Ia Lốp basin, the total runoff is approximately 2.3 × 10⁹ m³/year, corresponding to a standard flow (Qo) of 71.8 m³/s, a runoff modulus (Mo) of 43.2 l/s/km², and an average runoff depth of 1,363 mm. In contrast, the Ia H'leo basin generates about 3.1 × 10⁹ m³/year, with Qo = 96.9 m³/s, Mo = 33.1 l/s/km², and a runoff depth of 1,043 mm. Compared to major river basins in Vietnam, these values indicate moderate to high runoff potential (33–43 l/s/km²), underscoring the relatively abundant surface water resources of the two basins (Table 7).

Table 7. Annual streamflow characteristics of sub-basins in the Ia Lo^́p and Ia H'leo river systems, derived from MIKE-NAM model simulations.

Bold values indicate aggregated values for the two main river basins, Ia Lo^́p and Ia H'leo, rather than individual sub-basins or river sections.

While these values are comparable to those reported for other river basins in the Central Highlands, their interpretation must consider the specific local context. The semi-plain morphology, characterized by gentler slopes and lower elevations than surrounding plateaus, promotes higher evapotranspiration losses and reduces dry-season runoff persistence, despite relatively abundant rainfall during the wet season.

The flood season, defined here as June–October, contributes between 71% and 93% of annual runoff, starting approximately 1 month after the onset of the rainy season. This lag reflects the combined effects of soil-moisture replenishment, groundwater recharge, and basin storage processes. Conversely, the dry season (November–May) contributes only 19–28% of annual flow, highlighting the limited buffering capacity of the basin during prolonged dry periods. Such seasonal asymmetry is more pronounced than in several upland catchments of Vietnam, where steeper slopes and thinner soils often result in faster runoff responses but less severe dry-season depletion. These seasonal streamflow patterns are illustrated by the monthly average flow distribution and total seasonal flow in Figures 8 and 9, and are summarized quantitatively for the dry and flood seasons in Table 8.

Table 8. Seasonal distribution of streamflow in the Ia Lo^́p and Ia H'leo river basins, showing dry (Nov–May) and flood (Jun–Oct) seasons.

The dry season is defined to last from November–May and the flood season as June–October, based on the exceedance probability criterion (P > 50%). The percentages represent the seasonal contribution to the annual mean discharge. Bold values indicate aggregated values for the two main river basins, Ia Lo^́p and Ia H'leo, rather than individual sub-basins or river sections.

3.2.3 Trend of water resource variability in the study basin

The analysis of the standardized cumulative departure curve reveals that the annual and seasonal water availability in the basin exhibits distinct multi-decadal fluctuations. The cumulative deviation of annual flow is shown in Figure 10. From 1980 to 2022, the basin underwent a prolonged dry phase (1990–2008), followed by a wetter phase (2009–2019). Since 2020, water availability has entered a near-average state, with indications of a possible transition toward a drier cycle. This pronounced intra-annual variability in streamflow poses significant challenges to the sustainable management and utilization of water resources, particularly under the conditions of declining availability.

4 Discussion

The hydrological response within the Ia H'leo and Ia Lốp basins is fundamentally dictated by the distinct physiography of the Ea Súp semi-plain. Unlike the elevated plateaus of the Central Highlands, this transitional lowland is characterized by gentler gradients and elevated potential evapotranspiration (PET), which directly modulate the rainfall-runoff transformation. Rainfall exhibits marked spatial variability, with orographic enhancement driving higher annual totals at elevated stations (Chu Prông, Chu Sê) compared to the lowland rain shadows (Ea Súp, Bản Đôn). The concentration of nearly 90% of annual precipitation during the monsoon season (May–October) amplifies hydrological seasonality, resulting in a flood season that contributes over two-thirds of annual runoff. Compared to steeper upland catchments in Vietnam, the semi-plain morphology is associated with limited subsurface storage and higher losses, leading to more severe dry-season depletion and heightened vulnerability to hydrological droughts.

The model's performance was further evaluated using Flow Duration Curves (FDCs) on a semi-logarithmic scale (Figure 6a–c). The simulated and observed curves show reasonable agreement across the three sub-basins, particularly for medium- and high-flow conditions. However, deviations in the lower tail (high exceedance probabilities) indicate that low-flow and dry-season simulations remain more uncertain than wet-season runoff estimates. The FDC analysis therefore supports the practical adequacy of the model for seasonal water-resource assessment, while also highlighting the need for caution when interpreting extreme low-flow conditions.

For regional water managers, the reconstructed series is most useful for identifying seasonal water availability, wet/dry multi-decadal variability and relative dry-season vulnerability. It should not be treated as a direct observational record or as a substitute for future hydrometric monitoring. Additional stream gauges, longer reservoir-inflow records and formal parameter-uncertainty analysis would strengthen future applications of the model.

In Vietnam, previous MIKE-NAM applications have often focused on large river systems such as the Sê San, Ba and Mekong basins, or on basin-scale water-balance and allocation models. Consequently, the internal variability of agriculturally important sub-basins in the lower Srêpôk basin remains insufficiently quantified. This study addresses this gap by providing a long-term model-based reconstruction of monthly streamflow for the Ia H'leo and Ia Lốp sub-basins. Unlike studies in mountainous or plateau-dominated basins, the results highlight the importance of the semi-plain physiographic setting, seasonal storage limitation and groundwater-recession processes in controlling dry-season flow. Therefore, cross-regional comparisons should be made cautiously, particularly for low-flow and dry-season estimates.

From a Water-Energy-Food (WEF) nexus perspective, the low-flow portion of the flow-duration curves indicates that dry-season irrigation security is closely linked to reservoir regulation and pumping-energy demand. A simple pumping-energy estimate can be expressed as:

E=ρgHVη × 3.6 × 106(6)

Where: E is energy demand (kWh), ρ is water density, g is gravitational acceleration, H is pumping head, V is pumped water volume and η is pump efficiency. For 1 million m³ of water, the estimated energy demand is approximately 19.5 MWh for H = 5 m and η = 0.70, 41.9 MWh for H = 10 m and η = 0.65, and 90.8 MWh for H = 20 m and η = 0.60. This calculation demonstrates that reduced dry-season flow can increase energy demand for irrigation and water regulation, thereby intensifying WEF trade-offs in the Ea Súp semi-plain.

The hydrological results also provide a clearer basis for Nature-based Solutions. Because runoff is strongly concentrated during June-October and dry-season flows are limited, NbS should focus on increasing catchment storage and delaying runoff release. Headwater forest protection, riparian buffer restoration, wetland conservation and managed aquifer recharge can complement reservoirs by enhancing infiltration, supporting baseflow and reducing dependence on energy-intensive pumping during the dry season.

5 Conclusion

This study applied the MIKE-NAM hydrological model to assess surface-water availability in the Ia H'leo and Ia Lốp sub-basins, providing a useful model-based hydrological baseline for water management in Vietnam's Central Highlands. The reconstructed monthly flow series for (1980–2022) indicates mean annual runoff of approximately 3.1 × 10⁹ m³ for the Ia H'leo basin and 2.3 × 10⁹ m³ for the Ia Lốp basin, yielding a combined total of about 5.37 × 10⁹ m³. The regional climate exhibits pronounced seasonality, with around 88.6% of annual rainfall occurring during the May–October rainy season (84–92% across stations), while the flood season (June–October) contributes approximately 71–93% of total annual streamflow.

The results confirm a strongly seasonal hydrological regime, with most runoff concentrated during the June—October flood season and limited water availability during the dry season. Long-term variability analysis identifies a prolonged dry phase from 1990 to 2008, followed by a wetter interval during 2009–2019, highlighting the sensitivity of the Ea Súp semi-plain to hydro-climatic variability.

The model performance is generally acceptable to good for monthly seasonal assessment, but dry-season and low-flow estimates should be interpreted with caution because calibration and validation rely on short reservoir-inflow records from 2019–2022 and no continuous natural streamflow gauging stations are available within the study basins.

Viewed through the Water-Energy-Food (WEF) nexus, the reconstructed streamflow series provides an initial quantitative basis for linking dry-season water availability with irrigation demand, reservoir operation, and pumping-energy requirements. Future research should incorporate longer observed discharge records, formal uncertainty analysis (e.g., GLUE, Monte Carlo or parameter sensitivity analysis), water-demand data, land-use change information and coupled water-allocation modeling to strengthen WEF-oriented planning in the Ea Súp semi-plain.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. Generative AI was used only for language polishing, editorial support, and improving the clarity and structure of selected sections of the manuscript. It was not used to generate original research data, conduct hydrological modeling, perform statistical analyses, or draw scientific conclusions. All AI-assisted outputs were critically reviewed, edited, and verified by the author(s), who take full responsibility for the accuracy and integrity of the manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Naser Valizadeh, Shiraz University, Iran

Reviewed by: Negin Fallah Haghighi, Iranian Research Organization for Science and Technology, Iran

Shobeir Karami, Persian Gulf University, Iran

Footnotes

https://earthexplorer.usgs.gov/

http://hymetnet.gov.vn/

https://iot.monre.gov.vn/tnn/