Impacts of climate change and reservoir operation on droughts: a case study in the Upper Part of Dong Nai River Basin, Vietnam

World Environment Day 2024 with the theme ‘Land restoration, desertification, and drought resilience’ implicates that drought is one of the most globally concerning issues. Drought, a complex natural phenomenon, results from climate variability, characterized by changes in the duration, intensity, and distribution of precipitation and other climatic factors. The key characteristics of drought include its frequency, duration, intensity, and severity, which collectively influence its impacts on ecosystems, agriculture, and water resources (Khan et al. 2021). Droughts are commonly classified into meteorological, hydrological, agricultural, and socioeconomic droughts, each representing different phases and impacts of water scarcity (Wilhite & Glantz 1985).

The meteorological drought is caused by deficiencies in precipitation, while the hydrological drought appears as the prolonged result in the meteorological drought leading to the depletion of surface water and subsequent drying of reservoirs and lakes. Furthermore, various indices to assess as well as predict meteorological and hydrological drought status have been developed, such as the standardized precipitation index (SPI), drought index (DI), Palmer drought severity index (PDSI), rainfall deficit index (RDI), effective drought index (EDI), reconnaissance drought index (RDI), rainfall index (RI), standardized precipitation evapotranspiration index (SPEI), surface water supply index (SWSI), and streamflow drought index (SDI) (Eslamian et al. 2017; Benyoussef et al. 2024). In general, each drought index has its own advantages and limitations and is subject to the user's purpose. Practically, the SPI and SDI were selected for this study due to their wide application, simplicity, and low data requirements in monitoring meteorological and hydrological droughts under changing climate conditions (Khoi et al. 2021; Mahdavi & Kharazi 2023; Tran et al. 2024).

Climate change is widely acknowledged as a driver of increasing drought risks by altering temperature, precipitation patterns, and evapotranspiration rates, leading to heightened meteorological and hydrological droughts (IPCC 2022). Studies have demonstrated significant regional variations in drought trends, with some areas experiencing increased precipitation and others facing prolonged drought periods (Li et al. 2024). For example, studies in the Hexi Corridor, China (Liming & Yaonan 2016), in the Ghis-Nekor Watershed, Morocco (Benyoussef et al. 2024), and in the Central Highlands, Vietnam (Sam et al. 2019; Tri et al. 2019) have reported variations in drought frequency and severity due to climate change.

In addition, anthropogenic activities such as reservoir construction and operation have significantly altered streamflow and influenced drought conditions (Ngan & Khoi 2019). Reservoirs can mitigate climate change impacts on water availability by regulating flows, as shown in the Srepok River Basin, Vietnam (Tran et al. 2023). They may also modify the relationship between meteorological and hydrological droughts, delaying and reducing drought severity, but potentially prolonging droughts when inflows are insufficient (Wu et al. 2016). Moreover, climate change can decrease and increase the amount of water reaching reservoirs, challenging hydropower supply, as seen in the Marun Reservoir in Iran (Golfam & Ashofteh 2022). While previous studies have explored the individual impacts of climate change and reservoirs on droughts, there is still limited knowledge regarding the simultaneous long-term impacts of climate change and reservoirs on droughts.

Conversely, the drought project in the future is pivotally based on the global climate change scenarios released by the Intergovernmental Panel on Climate Change (IPCC) (Xu et al. 2024). The latest Sixth Assessment Report (AR6), based on the Coupled Model Intercomparison Project Phase 6 (CMIP6) models, uses the Shared Socioeconomic Pathways (SSP) approach with five main scenarios, including: (i) ‘Green growth’ SSP1, (ii) ‘Middle-of-the-road’ SSP2, (iii) ‘Regional rivalry’ SSP3, (iv) ‘Inequality’ or ‘A road divided’ SSP4, and (v) ‘Fossil fuel-based development’ SSP5 (Riahi et al. 2017). CMIP6 is more updating than the previous CMIP5's Representative Concentration Pathways (RCPs), with better accuracy in rainfall and temperature projections (Eyring et al. 2016; Tran-Anh et al. 2023; Yu et al. 2024). The SSP2-4.5 and SSP5-8.5 scenarios from CMIP6 represent ‘moderate’ and ‘extreme’ emissions recommended by the IPCC to quantify varying levels of greenhouse gas emissions and associated socioeconomic factors (Kim et al. 2022; Tran et al. 2024). Thus, the SSP2-4.5 and SSP5-8.5 scenarios were chosen to project future climate conditions for this study.

Moreover, to estimate the impacts of climate change on streamflow, practical hydrological models have been developed and applied such as the European Hydrological System Model (MIKE SHE) (Suman & Akther 2014; Liuxin et al. 2015; Zhang et al. 2021), Hydrologic Engineering Center Hydrologic Modeling System (HEC-HMS) (Kabiri et al. 2013; Hamdan et al. 2021), Identification of Unit Hydrographs and Component Flows from Rainfall, Evaporation, and Streamflow data (IHACRES) (Moghadam et al. 2022; Kalhori et al. 2023; Ashofteh et al. 2024), and Soil and Water Assessment Tool (IHACRES) (Tri et al. 2019; Hung et al. 2021; Khoi et al. 2021). Among hydrological models, the SWAT model was employed in this research because of its effectiveness and for being widely used for assessing climate change impacts on hydrology and droughts (Gassman et al. 2007; Janjic & Tadic 2023; Tran et al. 2024).

Vietnam, a tropical climate nation with a long coastline, many river delta regions, and low-lying cities, is one of the countries most vulnerable to climate change (Vietnam MONRE 2016). In 2020, Vietnam lost approximately US$10 billion, accounting for nearly 3.2% of GDP, due to climate change impacts (World Bank 2022). Over the past two decades (2000–2019), Vietnam experienced more than 13,000 deaths and economic losses amounting to about US$6.4 billion due to disasters. For 18 provinces affected by the 2015–2016 drought, direct economic losses were VND 15,032 billion (about US$0.674 billion), which was 0.35% of GDP at that time (World Bank & GFDRR 2017). Under climate change, the intensity and frequency of droughts are expected to increase in some parts of the Central Highlands of Vietnam (Sam et al. 2019; Tri 2020). The Upper Part of Dong Nai (UPDN) River Basin, located in the Central Highlands, has complex hydrological systems and operates more than 15 cascade hydropower plants with a total installed capacity of 2,380 MW (Truong et al. 2018). The long-term impacts of climate change and reservoirs on meteorological and hydrological droughts in the UPDN basin have not been determined yet. Therefore, this study has selected the UPDN River Basin in Vietnam as a case study.

The paper aims to access the long-term effects of climate change and reservoirs on meteorological and hydrological droughts in the UPDN River Basin by using two indices of SPI and SDI to assess the historical drought and project into the future with two climate change scenarios of SSP2-4.5 and SSP5-8.5. The obtained results will provide managers and policy decision-makers useful information for sustainable water resources management and development adapting to climate change.

The UPDN River Basin is divided into 18 sub-basins under two tributaries including the main UPDN River Basin and another sub-basin named La Nga River Basin. The main UPDN River Basin covers 17 sub-basins from DN1 to DN17, ending at the Tri An hydropower reservoir. Another tributary has only one sub-basin of La Nga (LN1) stopped at the Dami hydropower reservoir. The terrain of the UPDN River Basin has a clear gradation from north to south. Its north is the LangBiang Plateau ranging from 1,300 m to more than 2,000 m above sea level. Its core is the Di Linh Plateau, which rises between 700 and 1,000 m above sea level. Its south is the transition zone between the plateau and the plain with elevations varying from 200 to 500 m. Location of the tropical monsoon climate zone, this region has two distinct seasons with the rainy season from May to November and the dry season from December to April of the following year.

The essential datasets required by the SWAT model are a digital elevation model (DEM) map, soil map, land use and land cover (LULC) map, weather data, point sources, reservoirs data, among others (Neitsch et al. 2011). The DEM map with a 12.5-m resolution was downloaded from the NASA website: https://urs.earthdata.nasa.gov/users/new. Climate data applied in the SWAT model consist of daily precipitation, maximum, and minimum temperatures in the historical period of 1990–2020 covering a baseline 30-year period obtained from five weather stations (Da Lat, Lien Khuong, Bao Loc, Cat Tien, and Dak Nong). The missing data in daily solar radiation, wind speed, and relative humidity were generated automatically by SWAT (Mango et al. 2011). The spatial location of these stations can be relatively representative for the rainfall and temperature across the study area, see Figure 1. The SWAT model needs soil property data such as the texture, chemical composition, physical properties, available moisture content, hydraulic conductivity, bulk density, and organic carbon content for the different layers of each soil type (Neitsch et al. 2011). Soil information such as the texture, available moisture content, hydraulic conductivity, chemical composition, physical properties, bulk density, and organic carbon content was obtained from the 1:100,000 soil maps of the local Department of Natural Resources and Environment (DONRE). The LULC map applying for SWAT was derived from LULC classified result of the Landsat image in 2020 with seven categories: (1) WATR, (2) FRSE, (3) FRST, (4) PINE, (5) URML, (6) AGRR, and (7) AGRC (Hung et al. 2021, 2023). In addition, other socioeconomic datasets were based on the Statistical Yearbooks of Lam Dong and Dak Nong Provinces. The description of the main datasets is presented in Table 2.

To optimize the allocation of water resources and evaluate the performance of water supply systems, including reservoirs, irrigation networks, and hydropower plants under various scenarios of climate change, the previous studies used the Modeling and Simulation (MODSIM) model as a tool that supports decision-making for efficient reservoir operation and river basin management (Mortezaeipooya et al. 2022; Azadi et al. 2024). The UPDN River has more than 15 reservoirs with capacities greater than 10⁶ m³ (such as Da Nhim, Tuyen Lam, Dai Ninh, Dankia, Da Dang, Dong Nai 2, Dong Nai 3, Dong Nai 4, Dak R'tih, Dong Nai 5, Dak Sin, Da Teh, Dam B'ri, Dai Nga, Da Khai, and Dai Binh), most of them are cascade hydroelectric reservoirs located in the Ta Lai watershed, the main sub-basin of the UPDN River, as depicted in Figure 1 and Table 1. Thus, to assess the impact of these reservoirs and their regulation on droughts based on the characteristics of the UPDN River Basin, key reservoir features such as hydropower capacity, year of operation, area, and volume required to fill the reservoir to the emergency and principal spillways were collected for input into the SWAT model. These data were obtained from the local Department of Industry and Trade (DIT) and the Department of Agriculture and Rural Development (DARP). In addition, the current and future reservoirs' operation is based on regulations of inter-reservoir operation procedures in the Dong Nai River Basin according to Decision No. 1895/QD-TTG (Vietnam Prime Minister 2019). Maintaining minimum environmental flow in the rivers, streams, and downstream of reservoirs, dams are implemented according to Circular 64/2017/TT-BTNMT (Vietnam Ministry of Natural Resources & Environment 2017). The operation of Dai Nga hydropower plant for the future is based on the monthly operating mean in the past period.

The suitability evaluation of the SWAT model was calibrated and validated by using the coefficient of determination (R²) and the Nash–Sutcliffe efficiency (NSE) index. The range of R² is from 0 (implying no agreement between simulated and observed values) to 1 (the result of the simulation is perfect). Meanwhile, the value of NSE ranges from −∞ to 1. The ENS values greater than 0.5 are satisfactory. Furthermore, the NSE values from 0.5 to 0.65, 0.65 to 0.75, and 0.75 to 1.0 are acceptable, good, and very good, respectively (Moriasi et al. 2007). Monthly streamflow data from 1990 to 2020 in the upstream and downstream of the basin were collected from the Thanh Binh, Dai Nga, and Ta Lai gauges. Of these gauges, data values from the periods 1990–2000 and 2001–2010 were used for model calibration and validation of simulated runoff, respectively. In the study, these steps were performed by using the SUFI-2 algorithm, a suitable method for model calibration and uncertainty analysis (Khoi & Thom 2015), which was implemented in the SWAT-CUP 2012 tool for extension of ArcGIS 10.2 software.

As mentioned previously, there are many indicators for assessing meteorological and hydrological drought conditions, including the standard precipitation index (SPI), SPEI, RDI, drought index (DI), reconnaissance drought index (RDI), standardized runoff index (SRI), standardized terrestrial water storage index (STI), and SDI. Due to the low availability and data requirements of the SPI and SDI, this study uses these two indices to assess hydrological and meteorological drought characteristics (McKee et al. 1993; Nalbantis & Tsakiris 2009). These methods have been widely used in previous studies (Khoi et al. 2021; Mahdavi & Kharazi 2023; Tran et al. 2024).

In addition, temporal variation in drought may differ from time scales in accumulation periods of short-term (1 and 3 months), medium-term (6 months), upper medium-term (9 months), and long-term (12, 24, and 48 months). The medium-term time scale of 6 months (SPI6 and SDI6) was found to be appropriate to portray the hydrometeorological regimes and suitable for hydrometeorological drought monitoring (Padhiary et al. 2022), and it has also been used in some studies related to climate change and droughts (Pandhumas et al. 2020; Khoi et al. 2021). Therefore, the medium-term of the SPI6 and the SDI6 were applied in this study using monthly precipitation and streamflow as input data, respectively. SPI6 and SDI6 calculations were performed by using the design module integrated in the Drought Index Calculator (DrinC) software (Tigkas et al. 2015). This software is widely used in more than 145 countries for drought-related studies (Tigkas et al. 2022) and is freely available online (https://drought-software.com). The value of SPI or SDI less than −1 is considered a drought event, and the drought classification is presented in Table 3 (McKee et al. 1993; Khoi et al. 2021). Furthermore, this study chooses drought frequency and drought severity as the characteristics of drought events.

Based on the previous studies in performance evaluation and ranking of some CMIP6 general circulation models (GCMs) in projecting rainfall and temperature for Vietnam as well as the UPDN River Basin (Nguyen-Duy et al. 2023; Hung et al. 2024), this study selected five of the most suitable, as listed in Table 4, namely, EC-Earth3-Veg, CanESM5, EC-Earth3, HadGEM3-GC31-LL, and CNRM-CM6-1-HR, for projecting rainfall and temperature factors in the UPDN Basin. These GCMs have been downscaled for Vietnam (CMIP6_VN) with a spatial resolution of 0.1° (corresponding to 10 × 10 km) and temporal resolution of daily time by the statistical downscaling method bias-corrected spatial disaggregation (BCSD) (Department of Climate Change 2022; Tran-Anh et al. 2023). These climate data can be accessed free of charge at: http://remosat.usth.edu.vn/∼thanhnd/Download/dat_GEMMES_WP1. The utilization of 10-km downscaled resolution data in this study enables the provision of sufficiently detailed inputs for hydrological processes, identification of drought events, along with their duration and intensity at the basin scale, exemplified by the UPDN Basin (36,530 km²). This study used the ‘Multidimension Toolbox’ tool in ArcGIS 10.2 software to convert nationally multidimensional climate data ‘netCDF.NC’ to point data (‘Excel.xls’) at the coordinate location of the monitoring stations. Previous studies have shown significant uncertainty in the individual use of GCMs, emphasizing the need for an ensemble approach using multiple models (Ding et al. 2023; Tran et al. 2024). Thus, an ensemble resulting from the chosen five GCMs has been utilized for assessments.

As mentioned earlier, the latest climate change scenarios of IPCC AR6 include five main scenarios (SSP1, SSP2, SSP3, SSP4, and SSP5) (IPCC 2022). In this study, the SSP2-4.5 and SSP5-8.5 scenarios known as ‘moderate emissions scenarios’ and ‘high emissions scenarios’ were chosen to project climate aspects in the future for the UPDN River Basin. The daily ensemble mean of rainfall and near-surface temperature (T_min and T_max) are obtained by averaging the outputs of five selected models for the SSP2-4.5 scenario and the SSP5-8.5 scenario. Based on the 20-year period of the near future 2030s (2021–2040), the mid-21st century 2050s (2041–2060), the far future 2070s (2061–2080), and the late 21st century 2090s (2081–2099), four stages of future climate change are proposed to assess changes in SPI6 and SDI6 of the UPDN River Basin.

The Mann–Kendall (M-K) test accompanied with Sen's slope estimation (Kendall 1948; Sen 1945, 1968), a non-parametric trend test method, was used to access the trends of annual hydrometeorological series as well as analyze the trend of SDI and SPI series in this study by applying XLSTAT add-in for MS Excel. The statistical significance level α is set as 0.05 and 0.01, and their corresponding p-values (two-tailed) are 0.05 and 0.01. When the p-value is less than 0.05 or 0.01, it indicates that there is a significant trend of change at the 95 or 99% confidence level, respectively (Yue & Wang 2004). Magnitude and negative and positive trends can be detected by values of Sen's slope.

Pearson correlation coefficient (PCC) was used to quantitatively measure the correlation of SDI and SPI at the lag time variations of hydrological drought in response to meteorological drought coupled with reservoir patterns by using SPSS 22.0 software package for Windows. The higher the PCC at the lag time, the more sensitive the response time is, and vice versa in the relationship between meteorological drought and hydrological drought (Wu et al. 2016).

The results of the sensitivity analysis for the SWAT model showed that the five most sensitive parameters were CN2 (Initial SCS CN II value), GW_DELAY (threshold water depth in the shallow aquifer for flow), OV_N (Manning's ‘n’ value for overland flow), CH_K2 (channel effective hydraulic conductivity), and ALPHA_BNK (baseflow alpha factor for bank storage), respectively. Among these, the surface flow parameters (represented by CN2, CH_K2, and OV_N) are highly sensitive in the UPDN Basin in the context of multiple hydro-reservoir systems.

The impact of rainfall changes on runoff in the UPDN River Basin was analyzed using data from various meteorological stations. Specifically, rainfall data from the Bao Loc station was used for the Dai Nga sub-basin (39,266 ha), average values from Da Lat and Lien Khuong were used for the Thanh Binh sub-basin (32,493 ha), and data from five stations were used for the Ta Lai sub-basin (973,265 ha). The findings reveal that streamflow varies significantly between sub-basins, with annual streamflow increasing from upstream to downstream: 3,607.5 m³/s at Thanh Binh, 6,008.4 m³/s at Dai Nga, and 123,060.5 m³/s at Ta Lai. The M-K test indicates an increasing trend in annual rainfall across all three sub-basins, with significant increases in Thanh Binh and Ta Lai (p-values of 0.038 and 0.035; Sen's slope coefficients of 9.9 and 12.2, respectively), as presented in Table 5. These trends are similar to those reported in the Be River Basin, Vietnam (Khoi et al. 2021), in the Hexi Corridor, China (Liming & Yaonan 2016), and in the Ghis-Nekor Watershed (North of Morocco) (Benyoussef et al. 2024).

In general, reduction in streamflow was noted at the Ta Lai gauge, although it was not statistically significant, likely due to the relatively low volume of water diverted by the upstream Dai Ninh hydropower plant, which accounts for only about 7% of the total river flow. Streamflow at Thanh Binh has increased but decreased at Dai Nga, despite higher rainfall, indicating a complex relationship between rainfall and hydropower management. The results showed a statistically significant increasing trend of streamflow at Thanh Binh gauge (p = 0.01, Sen's slope = 41.2) but a significant decrease at the Dai Nga gauge (p = 0.001, Sen's slope = −163.8), attributed to the operational impacts of the Dai Nga hydropower plant established in 2015, which diverted water downstream, leading to a marked reduction in streamflow. This situation was similarly observed in a study focused on the impacts of reservoir operations in the Jinsha River Basin in Southwest of China (Zhang et al. 2023). These results align with findings in the Jinjiang River Basin in Quanzhou City, Fujian Province, China, where it was noted that increased rainfall led to varying impacts on streamflow due to reservoir operations (Wu et al. 2016).

The pattern of historical drought events is closely linked to the Thanh Binh, Dai Nga, and Ta Lai stations, although the timing and severity are generally not synchronized. For example, severe droughts occurred simultaneously in the Da Nga and Ta Lai watersheds in 2010–2011, while the drought in the Thanh Binh watershed during the same period was only moderate. Additionally, the drought duration and magnitude in these regions (Thanh Binh, Dai Nga, and Ta Lai) have varied considerably since 2008 and 2015, when the Dai Ninh and Dai Nga hydropower plants, respectively, began operation. For the Thanh Binh River Basin, based on long-term SPI6 and SDI6 data, a clear upward trend in SDI and SPI values was observed between 1990 and 2020. The probability of drought in this basin is decreasing, indicating that the region is becoming more humid. This finding aligns with results from the Hexi Corridor in China (Liming & Yaonan 2016). Conversely, the drought trend at the Dai Nga and Ta Lai stations has been increasing, even though rainfall in the area has been rising. The cause of the drought is attributed to anthropogenic reservoir operations in the observed period.

Table 6 presents PCC values of SPI6 and SDI6 with time series data from 1990 to 2020 for the Thanh Binh, Dai Nga, and Ta Lai watersheds at lag times of 0, 1, 2, 3, 4, and 5 months. Good agreement was observed between the SPI6 and SDI6 at lag times of 0 and 1 month (prior to the operation of the Dai Nga hydropower plant in 2015 and the Dai Ninh hydropower plant in 2008). These correlation coefficients range from 0.632 to 0.755, as shown in the 0-month and 1-month columns.

The highest PCC values for SPI6 and SDI6 were 0.755 and 0.633, respectively, at Thanh Binh and Dai Nga with a 0-month lag time. For the period that included the operation of the two hydropower plants (Dai Nga and Dai Ninh), the correlation values between SPI6 and SDI6 dropped compared with the earlier period. This change confirms that reservoir operations have a clear influence on drought conditions in the river basins. For example, the PCC values dropped from 0.633 to 0.268 at the 0-month lag time at Dai Nga and from 0.753 to 0.640 at the 1-month lag time at Ta Lai. The 1-month lag between meteorological drought and hydrological drought was also found in the Be River Basin, Vietnam (Khoi et al. 2021). Hydrological drought is typically observed 1 month after meteorological drought due to streamflow delays, which are controlled by basin characteristics such as reservoirs (Wu et al. 2016). The results show that reservoir regulation affects drought in Ta Lai by reducing the intensity of droughts during the dry season and the severity of floods during the rainy season. This conclusion aligns with findings from the Jinjiang River Basin in Quanzhou City, Fujian Province, China (Brunner 2021).

Projected changes in annual precipitation for the future time slices (2030s, 2050s, 2070s, and 2090s) under the SSP2-4.5 and SSP5-8.5 scenarios are detailed in Table 7, respectively. In the future, yearly rainfall in higher elevation areas such as Thanh Binh (1,100 m above sea level) is projected to fluctuate with slight increases and decreases throughout the century. By contrast, lower elevation areas such as Dai Nga (800 m above sea level) are expected to see consistent increases in rainfall, contributing to an overall rise in rainfall across the UPDN Basin before a slight decline by the end of the century. Generally, annual rainfall tends to increase slightly in the future in the UPDN Basin, as shown in Figure 5. This trend is also consistent with the study in the adjacent Srepok Basin (Tran et al. 2023).

Notably, seasonal rainfall patterns in the UPDN Basin show clear changes under climate change, with an increase in rainfall during the rainy season and a decrease during the dry season, despite relatively stable annual totals. Rainfall is projected to increase from June to December and decrease from February to May under both SSP2-4.5 and SSP5-8.5 scenarios (Table 7). Also, this aligns with findings from the Ca River Basin located in North Central Vietnam (Shin et al. 2024), where future projections show reduced spring precipitation (March to May) and increased autumn precipitation (September to November). The findings of this study are quite consistent with previous research results in the Central Highlands of Vietnam, where average monthly rainfall tends to increase during August and September in the rainy season (Tran et al. 2023), while the decrease in the dry season often leads to drought events, particularly from January to April (Tri 2020). The results of this study indicate that both floods and droughts are projected to become more extreme per the assertions by Doan et al. (2022) that ‘extremes get more extreme’ as a consequence of global climate change.

Overall, the UPDN River Basin is projected to experience a clear rise in temperature, while precipitation is expected to show a slight increasing trend in the period 2021–2099, as illustrated in Figure 5. These findings are consistent with previous studies (Department of Climate Change 2022; Tran-Anh et al. 2023), which also report rising temperatures and fluctuating precipitation under various climate change scenarios. Similar trends have been observed in studies from other regions, such as the Srepok River Basin, Vietnam (Tran et al. 2024) and the Lower Nam Phong River Basin, Thailand (Pandhumas et al. 2020), where climate change leads to increased rainfall during the wet season and reductions during the dry season, reinforcing the importance of understanding seasonal variability in climate projections.

Owing to the slight rainfall increase, the streamflow in the entire UPDN River Basin also insignificantly changes in the future periods of 2030s, 2050s, 2070s, and 2090s. However, in the Ta Lai watershed, the main part of the UPDN Basin, the results show an increase in total annual streamflow during 2021–2099 compared with the baseline period of 1990–2007 with no Dai Ninh hydropower plant. Projected total annual streamflow values are 190,841 m³/s for SSP2-4.5 and 167,222 m³/s for SSP5-8.5, compared with 130,826 m³/s for 1990–2007. This indicates that climate change has caused an increase in runoff due to higher rainfall across the entire basin, even though approximately 7% of the total annual flow is diverted by the Da Nhim and Dai Ninh hydropower plants.

However, streamflow volume is expected to vary between sub-basins, influenced by seasonal factors. In the dry season, the proportion of water flow compared with the rainy season in the three regions of Thanh Binh, Dai Nga, and Ta Lai is 15.0 vs. 85.0%, 11.1 vs. 88.9%, and 16.9 vs. 83.1%, respectively, under the SSP2-4.5 scenario. For the SSP5-8.5 scenario, these values are 15 vs. 85%, 16.1 vs. 83.9%, and 18 vs. 82%. The development of cascading hydropower plants in the Ta Lai basin, including Da Dang 2 (34 MW, 2009), Dong Nai 3 (180 MW, 2010), Dong Nai 4 (340 MW, 2012), Dong Nai 2 (70 MW, 2013), and Dong Nai 5 (150 MW, 2014), has helped regulate water flow, leading to an increase in dry-season flows. Indeed, compared with the previous period (1990–2007) with no reservoirs and their operation, the flow value in the dry season of Ta Lai is only 10.7% compared with 16.9% of SSP2-4.5 and 18% of SSP5-8.5. These findings are consistent with increases in dry-season flow due to reservoir operations in the Srepok River Basin in the Central Highlands of Vietnam (Nhi 2023).

Figure 7(a2), 7(b2), 7(c2) and Figure 8(a2), 8(b3), 8(c2) depict the projected hydrological drought frequencies (SDI6) for the periods 2030s, 2050s, 2070s, and 2090s under the SSP2-4.5 and SSP5-8.5 scenarios. The frequencies of D compared with WC in SDI6 for Thanh Binh, Dai Nga, and Ta Lai are 16 vs. 15%, 13 vs. 14, and 14 vs. 15%, respectively. Under SSP5-8.5, these values shift to 15 vs. 13%, 17 vs. 16, and 15 vs. 17%, respectively. The analysis indicates consistent fluctuations in SDI6 and SPI6 for the Thanh Binh and Dai Nga Rivers, while in Ta Lai, SDI6 decreases slightly compared with SPI6. This suggests that hydrological drought in Ta Lai is mitigated relative to meteorological drought, likely due to storage and streamflow regulation by reservoirs within the Ta Lai sub-basin. The analysis indicates consistent fluctuations in SDI6 and SPI6 for the Thanh Binh and Dai Nga Rivers, while in Ta Lai, SDI6 decreases slightly compared with SPI6. This suggests that hydrological drought in Ta Lai is alleviated compared with meteorological drought, likely due to storage and streamflow regulation within the Ta Lai sub-basin by reservoirs.

As to the four future periods of 2030s, 2050s, 2070s and 2090s, the SPI6 and SDI6 trends inconsistently and unclearly propagate between Thanh Binh, Dai Nga and Ta Lai compared with the historical period of 1990–2020. For the main part of the UPDN Basin (Ta Lai basin), observed results show that SPI6 tends to increase slightly in the near future (2030s), decrease in the mid-century period, increase in the late century period under SSP2-4.5, with corresponding values of 13.9, 16.2, 13.6, 18.3, and 17.0% for the periods 1990–2020, 2030s, 2050s, 2070s, and 2090s, respectively. These findings align with the results of Sam et al. (2019), which report an upward trend in drought frequency from 2016 to 2040 in the Central Highlands of Vietnam. Similarly, studies in the neighboring Be River Basin also predict an increase in meteorological drought frequency in the 2030s, and a decrease in the 2050s (Khoi et al. 2021). Specifically, drought frequencies in the Be River Basin are expected to rise to 34.5% during the 2030s and reduce by 13.8% during the 2050s for the RCP4.5. Under the SSP5-8.5, drought frequency in the UPDN Basin tends to increase by 17.0 and 17.0% in the 2030s and 2050s, respectively, and decrease by 14.5% in the 2070s, then increase approximately by 18.3% in the 2090s. These trends are also consistent with the report on the Be River Basin by Khoi et al. (2021).

In terms of the severity of drought, in general MD is more common than SD and ED for both meteorological and hydrological droughts. The SSP5-8.5 scenario shows higher levels of severe and extreme droughts compared with SSP2-4.5, particularly in areas such as Thanh Binh, Dai Nga, and Ta Lai. Notably, Ta Lai experiences lower levels of severe and extreme hydrological droughts compared with the other two areas under both climate scenarios. The presence and operation of reservoirs have significantly influenced the occurrence and severity of hydrological droughts in the river basin.

Based on these results, it is evident that meteorological drought is primarily influenced by rainfall conditions, whereas hydrological drought is also shaped by human activities, such as reservoir construction and operation. This underscores the need for appropriate embankment and irrigation measures to mitigate the impacts of droughts in the UPDN River Basin. Upgrading reservoirs and improving irrigation system efficiency will be critical in reducing the effects of droughts. Additionally, the establishment of a comprehensive network of meteorological and hydrological monitoring stations is essential for early drought warning, particularly in regions such as Dai Nga that currently lack sufficient hydrological monitoring infrastructure. Further, the implementation of effective drought management strategies, improvement in water-use efficiency, and promotion of smart agricultural practices will be vital to adapting to and mitigating the impacts of climate change.

Based on hydrometeorological monitoring data in the past period 1990–2020 and the ensemble mean values of five GCMs in the future period 2021–2099 provided by CMIP6 downscaled for Vietnam combined with two indices of SPI6 and SDI6, this study investigated and analyzed the responses of meteorological and hydrological droughts under the effects of climate change and reservoirs in the UPDN River Basin. The main conclusions of this study are underlined as follows:

• (1) In the historical period 1990–2020, hydrological drought was significantly affected by the formation and operation of reservoirs. The frequency and severity of droughts vary among sub-basins of the UPDN River Basin. For large basins with many cascade reservoirs, the lag time between hydrological drought and meteorological drought is 1 month, while for small basins without reservoirs, the lag time between hydrological drought and meteorological drought is 0 months.

• (2) Climate change has caused temperature increases clearly than precipitation increases in the UPDN River Basin. Rainfall projects an increasing trend from June to December and a decreasing trend from February to May. Therefore, policymakers need to prevent the occurrence of extreme drought events during the dry season in the future properly such as by developing early warning systems and improving drought preparedness plans, to mitigate the risk of extreme drought events during the dry season.

• (3) In the future, reservoirs will still be the dominant factor in streamflow in the main part of the UPDN River Basin. The reservoirs and their operations can alleviate hydrological droughts in the sub-basins of the UPDN River Basin during the dry season. Thus, appropriate dyke and irrigation measures can regulate and reduce damages caused by droughts in the UPDN River Basin. Moreover, upgrading reservoir infrastructure, implementing efficient irrigation practices, and establishing comprehensive water resource management frameworks should be prioritized to optimize water use and reduce drought-related losses.

This study proposes a practical method to assess the long-term impacts of climate change and reservoir operations on droughts in the UPDN Basin. However, it has limitations due to the lack of detailed reservoir operation data, which relied on historical monthly averages and current inter-reservoir regulations (Decision No. 1895/QD-TTG, Circular 64/2017/TT-BTNMT) rather than daily data. Additionally, the weather input for the SWAT model, derived from GCMs with a 10 × 10 km resolution, can cause uncertainty. Thus, it is necessary to implement these data to provide more precise results in drought projection for the UPDN River Basin.

The authors thank the Department of Science and Technology (DOST) and the Department of Natural Resources and Environment (DONRE) of Lam Dong provinces as well as the Research Institute for Innovation and Sustainable Development (RIFISD) for their support to this study by providing statistical data and implementing resources.

We also acknowledge the support of the time and facilities provided by the Ho Chi Minh City University of Technology (HCMUT), VNU-HCM, for this study.

P.H. and V.L.P. developed ideas, conceptualized the framework, and methodology; P.H. selected methods, techniques, and processed data; P.H. wrote the draft paper; V.L.P. edited and reviewed the paper.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.