Prolonged and Severe Drought in the Most Dammed Tributaries of the Lower Mekong Basin

Abstract

Drought is a natural hazard that stresses ecosystems, agricultural production, food security, and local economies. Given ongoing hydropower dam development in the Sesan and Srepok Basins, the two most dammed tributaries in the Lower Mekong Basin, characterizing baseline drought events and understanding how dams modify downstream flow is needed to manage water resources and mitigate drought effects. We used the Soil & Water Assessment Tool (SWAT) to estimate streamflow data from 2001 to 2019. For both rivers, we found that runoff varied, but contributed about 75% of streamflow, followed by shallow and deep groundwater, which contributed up to 25%. We used the Standardized Runoff Index to characterize drought and detected frequent, severe, and prolonged drought events in the two basins. Severe and prolonged droughts in the 2009–2011 and 2015–2016 periods corresponded to the occurrence of Typhoon Ketsana and the El Niño-Southern Oscillation. Streamflow alteration can be caused by climatic conditions and anthropogenic activities such as hydropower dam development and operations (e.g., the timing and magnitude of water releases). Results from this study can be used as a baseline to gauge potential future droughts and design appropriate drought management plans to preserve ecosystems and food security in the Lower Mekong Basin and its tributaries.

1. Introduction

Droughts are natural events that can be devastating for ecosystems, agriculture, food security, and economic production [1,2,3,4]. Droughts deplete reservoir storage and increase evapotranspiration so that water must be managed and allocated more tightly among competing uses [5,6]. In regulated rivers, dam operations may mediate or exacerbate drought, yet these relationships are not well understood in the Lower Mekong Basin and its major tributaries, which are undergoing rapid hydropower dam development.

The Mekong River flows through six countries in Asia. The Lower Mekong Basin (LMB), in particular, is characterized by rich biodiversity due to its large floodplain, and long and broad tributaries [7,8,9,10]. A major tributary system of the LMB is formed by the Sekong, Sesan, and Srepok Rivers (the 3S Basin), which contribute up to 25% of streamflow and sediment to the mainstem Mekong River [11]. This basin is currently experiencing hydrologic alteration due to rapid hydropower development [12]. The 3S Basin has 51 operational dams as of 2021. Of the rivers in the 3S Basin, the Sesan River is the most dammed, with 28 dams, followed by the Srepok River with 17 dams. These two rivers are the focus of our study (Figure 1). The Sekong River is also part of the 3S Basin and has the fewest dams [12]. The Sesan and Srepok Rivers originate from the central highland of Vietnam, with peak elevations ranging from 2000 to 2400 m [13], and flow through the northeastern mountain area of Cambodia to join and enter the Mekong River in Stung Treng Province. The Sesan Basin encompasses an area of 18,800 km², which contains roughly 1 million people, and the Srepok River Basin has an area of 30,094 km², which contains roughly 2.8 million people [14]. Most of the residents in the two basins are indigenous, live in rural areas, and produce crops such as paddy rice, coffee, and other agro-industrial crops [13,15]. The basins are influenced by seasonal monsoons, with distinct dry (November–April) and wet (May–October) seasons. The annual rainfall of the two rivers ranges from 1900 mm to 2000 mm [14].

Research has shown that dams have significantly altered streamflow in the Mekong River by decreasing flow in the wet season and increasing flow in the dry season [17]. These altered flows have, consequently, led to low flow events and water quality changes in the Mekong Basin [18,19,20,21]. Assessing the effects of drought on streamflows versus changes from dam operations is needed to better manage reservoirs during drought, maintain adequate streamflows to preserve the unique biodiversity of the LMB, and understand the duration and severity of historical droughts.

To mitigate the impacts of drought, researchers have developed methods to detect drought using indices and models [22]. The Soil & Water Assessment Tool (SWAT), a rainfall-runoff model, has been widely implemented globally and in the LMB [23,24,25,26]. Similarly, the Standardized Runoff Index (SRI) is commonly used to characterize drought intensity, duration, and severity. Sam et al. [26] used SWAT to investigate climate change impacts from meteorological, hydrological, and agricultural drought in the Srepok River, using SRI, the standardized precipitation index, and the standardized soil moisture index to characterize drought. Akoko et al. [23] also used SWAT with SRI, the standardized precipitation index, and the standardized soil moisture index to assess drought impacts on hydrology, sediment, and nutrient loss with climate variability. Brouziyne et al. [24] applied SWAT to conduct drought analysis and project climatic change effects for a Mediterranean watershed.

Given ongoing hydropower dam development in the 3S Basin, the simulation of historical streamflow and characterizing baseline drought events in the Sesan and Srepok Rivers is needed to understand drought effects on river management and the sustainable development of hydropower. In this study, we use (1) the SWAT model to simulate 2001–2019 streamflow and quantify streamflow contributions from surface runoff, and shallow and deep groundwater, and (2) modeled data to characterize baseline drought events using SRI to show spatial and temporal drought trends.

2. Materials and Methods

2.1. River Flow Data

Flow data from the Veurn Sai gauge station on the Sesan River and the Lumphat gauge station on the Sekong River were provided by the Mekong River Commission (MRC). Streamflow from these gauge stations are daily time-series data from 2001 to 2019.

2.2. Soil & Water Assessment Tool (SWAT) Hydrologic Modeling

Monthly streamflow in the Sesan and Srepok Basins was estimated with SWAT for 2001–2019. SWAT was developed by the United States Department of Agriculture and is now widely used [27]. The model simulates hydrologic pathways of surface and groundwater [23,28,29,30]. SWAT is capable of estimating streamflow, sediment, nutrients, and pesticide yields in ungauged catchments [27,31]. Catchments are generally divided into sub-catchments, then further subdivided into hydrological response units (HRUs) with homogenous land use, topography, and soil type characterization [32]. HRUs are represented as a percentage of sub-catchment areas. Canopy interception may prevent rainfall from reaching the soil surface. However, rainwater that reaches the soil surface will either evapotranspirate, infiltrate into the soil layer and contribute to groundwater, or flow as runoff to streams [33].

SWAT is based on a water balance as depicted below:

$$S{W}_t=SW{L}_0+{\displaystyle \sum _{i=1}^t(R_{day}-Q_{surf}-E_a-W_{seep}-Q_{gw})}$$

(1)

where $S{W}_t$ is soil water content (mm), t is time (days), $R_{day}$ is precipitation on day i (mm), $Q_{surf}$ is the surface runoff on day i (mm), $E_a$ is evapotranspiration on day i (mm), $W_{seep}$ is water entering the vadose zone on day i (mm), and $Q_{gw}$ is groundwater flow to streams on day i (mm).

To simulate streamflow, SWAT requires geospatial input data, including a digital elevation model (DEM) of topography, soil types, and land use/land cover data. We used a 90 m resolution DEM (

Table 1. Data inputs and sources for the Sesan and Srepok Basin SWAT models.

Input of Data	Resolution	Period	Sources
Digital elevation model (DEM)	90 m × 90 m	-	MERIT DEM [34]
Soil type	250 m × 250 m	2002	FAO Soil [35]
Land use/land cover	250 m × 250 m	2002	GLCC [36]
Precipitation	0.25°	2001–2019	TRMM [37]
Observed flow	2 Gauges	2001–2019	MRC

) with elevation ranging from 43 m to 2403 m above sea level, a gridded 250 m resolution soil map, and a gridded 250 m resolution land use/land cover map. We delineated 25 sub-catchments with an average area of 198,000 ha and 467 HRUs. Our model had 7 land uses, 5 soil types, and 5 slope categories (0–1%, 1–2%, 2–5%, 5–20%, and >20%) [10]. Meteorological input included daily precipitation (Table 1). We calibrated and validated the model with monthly observed streamflow from the Veurn Sai and Lumphat gauges of the Sesan and Srepok Rivers, respectively.

2.3. SWAT Calibration, Validation and Perfomance Evaluation

We calibrated and validated SWAT using observed streamflow data from 2001 to 2010, as well as observed streamflow data from 2011 to 2019, respectively. We manually calibrated the model by altering sensitive model parameters, a process that is commonly used (see Arnold et al. [32], Ich et al. [38], Sok et al. [39], and Nair et al. [40]). For the Sesan River, we calibrated one parameter, “effective hydraulic conductivity in main channel alluvium,” while we adjusted eight sensitive parameters for the Srepok River model (see Appendix A,

Table A1. SWAT calibration parameters, their ranges, and fitted values.

Parameter	Name	Input	Range	Fitted Value
V_CH_K2	Effective hydraulic conductivity in main channel alluvium	.rte	0–500	100
V_RCHRG_DP	Deep aquifer percolation fraction	.gw	0–1	0
V_REVAPMN	shallow aquifer for “revap” to occur (mm)	.gw	0–1000	0
V_GW DELAY	Groundwater delay (days)	.gw	0–500	150
R_SOL_AWC	Available water capacity of the soil layer (mm/mm soil)	.gw	0–1	0.25
R_SOL-Z	Saturated hydraulic conductivity (mm/hour)	.sol	−0.25–0.25	0.2
V_CANMIX	Maximum canopy storage	.hru	0–100	100
V_ESPO	Plant uptake compensation factor	.bsn	0–1	0.3
V_ESCO	Soil evaporation compensation factor	.bsn	0–1	0.2

).

Model performance was evaluated using the coefficient of determination (R²), Nash-Sutcliffe efficiency (NSE), root mean standard deviation ratio (RSR), and percentage bias (PBIAS) [41]. R² was used to evaluate the correlation between the simulated and observed flow data, NSE and RSR were used to evaluate extreme and peak values, and PBIAS was used to evaluate whether models over- or under-estimated streamflow [32,42].

The watershed conditions generated from the SWAT model were used to analyze hydrological pathways of water for each river. These pathways included surface runoff, shallow and deep groundwater, and total water yield.

2.4. Standardized Runoff Index and Drought Characterization

We used the Standardized Runoff Index (SRI) as an indicator of hydrologic drought because it is widely implemented [43]. We calculated SRI over a 12-month timescale to reflect a true drought, versus a few dry months, and to represent a time frame useful for water and land planning. The SRI values were computed from SWAT output based on the following equation:

$$SR{I}_i=\frac{V_i-\overline{V}}{SD}$$

(2)

where V_i is the cumulative runoff for each year i, and $\overline{V}$ and SD are the mean and standard deviation of the monthly runoff, respectively. The index values range from −3.0, indicating the driest years, to 3.0, indicating the wettest years. For instance, an SRI value of −2 means that surface runoff is less than the mean with two times the variance (

Table 2. Drought classification based on the Standardized Runoff Index (SRI).

Standardized Index	Description
−0.50 to −0.79	Abnormally dry
−0.80 to −1.29	Moderate drought
−1.30 to −1.59	Severe drought
−1.6 to −1.99	Extreme drought
−2.0 or less	Exceptional drought

) [44]. Heatmaps and maps of SRI drought classification were used to visualize results. They were produced using Python JupyterLab with version 3.1.18.

3. Results

3.1. Streamflow Simulation and Hydrologic Pathways

The simulated SWAT streamflow matched observed flows well, with NSE ranging from 0.72 to 0.79 and R² ranging from 0.75 to 0.81 (

Table 3. SWAT performance for monthly Sesan River streamflow (Veurn Sai station) and Srepok River streamflow (Lumphat station). Dark green shading indicates very good model performance, light green shading indicates good model performance, and no shading indicates performance categorization was unavailable (Moriasi, et al. [45]).

Station	Dataset	Period	R2	NSE	PBIAS	RSR
Veurn Sai	Calibration	2001–2010	0.81	0.79	10	0.45
	Validation	2011–2019	0.80	0.78	3	0.46
Lumphat	Calibration	2001–2010	0.75	0.72	−16.2	0.52
	Validation	2011–2019	0.79	0.75	−13.3	0.49

). In the Sesan River, surface runoff averaged 79% of streamflow, deep groundwater averaged 17%, and shallow groundwater (lateral flow) averaged 4%. Similarly, for the Srepok River, the largest average contributor to streamflow was surface runoff (71%), followed by deep groundwater (25%), and shallow groundwater (lateral flow) (4%) (

Table 4. Average annual surface runoff, and shallow and deep groundwater contributions to Sesan and Srepok River streamflow at the two gauging stations, with average annual 2001–2019 discharge at the river outlets.

Basin	Gauge Name	Surface Runoff		Shallow Groundwater		Deep Groundwater		Total Water Yield
		(mm/year)	%	(mm/year)	%	(mm/year)	%	(mm/year)
Sesan	Veurn Sai	1133	79%	57	4%	240	17%	1430
Srepok	Lumphat	663	71%	35	4%	231	25%	929
Sesan + Srepok	Outlet	853	75%	42	4%	233	21%	1128

, Figure 2 and Figure 3). The average annual streamflow at the outlet of the Sesan River was 1430 mm/year and was 929 mm/year at the outlet of the Srepok River.

Surface runoff in the Sesan and Srepok Basins was highly variable (Figure 2 and Figure 3). In the Sesan Basin, the average surface runoff across the study period was 1133 ± 313 mm. The wettest year occurred in 2009, when surface runoff was 1539 mm, followed by 2018 with 1319 mm, and 2014 with 1236 mm. The driest years occurred in 2005, 2010, and 2015. The year 2010 had the lowest annual surface runoff, with 695 mm (Figure 2). In the Srepok Basin, the average surface runoff across the study period was 663 ± 198 mm. The wettest year occurred in 2009, with annual surface runoff of 856 mm, followed by 2018 with 716 mm, and 2014 with 707 mm. Again, the driest years were 2005, 2010, and 2015, where 2015 had the lowest annual surface runoff at 385 mm (Figure 3).

3.2. Hydrologic Drought Characterisation

We identified two severe and extreme drought events from June 2010 to August 2011 and from June 2015 to April 2017, respectively (Figure 4). In 2010–2011, the drought lasted 13 months and, in 2015–2017, the drought lasted 22 months. Another moderate drought occurred from 2004 to 2005, with abnormally dry periods from 2008 to 2009 and from 2012 to 2014 (Figure 4 and Figure 5).

Transitions from wet and dry years can be abrupt (Figure 5). For example, 2009 started out dry, but transitioned to a very wet year, then 2010 transitioned from a wet year to severe drought. Dark blue shading indicates very wet conditions, while dark red shading indicates extreme drought.

Droughts were not uniform throughout the study basins (Figure 6). The 12-month accumulation periods of SRI ending in June 2010 (SRI12-June) showed wet-to-normal conditions. While the dry 12-month accumulation periods were just finishing in August 2010 (SRI12-August), only sub-catchments near the outlet of basins registered drought. However, the 12-month accumulating timescale ending in in November 2010 (SRI12-November) showed that severe drought was widespread throughout the Sesan and Srepok Basins. These space and time trends were even more pronounced during the extreme 2015–2016 drought (Figure 6).

4. Discussion

4.1. SWAT Performance, Streamflow, and Drought in the Sesan and Srepok Rivers

SWAT performance was good to very good for the Sesan and Srepok Rivers, which is consistent with the model performance from other studies [10,26,30]. Surface runoff was the largest contributor to streamflow in the two rivers, with more minor groundwater contributions. This finding has also been previously reported [10,47,48,49]. Precipitation was the source of surface runoff [10,47], and maximum precipitation in headwater streams of the central Vietnamese Highlands and northeastern Cambodia can exceed 2000 mm per year [10].

Severe and extreme droughts were evident in the Sesan and Srepok Rivers, with three major droughts, thus far, in the 21st century. Twelve-month-or-longer drought events occurred from 2010 to 2011 and from 2015 to 2016 in the Sesan and Srepok Basins. Reduced precipitation across the Mekong Basin reduced surface runoff. Interestingly, these drought events spanned the dry and wet seasons (Figure 5 and Figure 6). Our findings reiterate those of other studies that reported droughts in 2010, including in Russia [50], the Amazon River Basin [51], and the broader Mekong River Basin [3].

From 2015 to 2016, the severe and prolonged drought was likely caused or exacerbated by El Niño conditions, which lead to very dry conditions in Southeast Asia [52,53]. The 2015–2016 drought is considered to be the longest and most severe drought to date in the Lower Mekong Basin [26,54]. Severe and prolonged droughts that were caused by El Niño conditions were typically preceded by wet periods caused by La Niña conditions. La Niña years increase precipitation and streamflow [53]. From 2009 to 2010, Typhoon Ketsana and another tropical storm flooded the 3S Basin and the LMB [55]. That year, there was a tremendous release of water from Yali Falls Dam in the upper Sesan Basin due to the very wet conditions [56].

4.2. Drought in Regulated Watersheds

Aside from drought, streamflow alteration is caused by anthropogenic activities such as hydropower dam operations and water extraction for agricultural irrigation [57]. For instance, summer baseflow in dammed rivers is generally higher than free-flowing rivers [18,58], which is the case for the Sesan River [12]. Li et al. [18] and Wild et al. [58] reported that dams have significantly decreased streamflow in the wet season in the Mekong River and Tonle Sap Lake, and increased streamflow in the dry season. Although 45 dams have been built in the Sesan and Srepok Basins, we observed no evidence that they mediated the effects of drought. Droughts that began in the dry season were likely to decrease streamflow despite the potential of the upstream dams to re-regulate flows. This may portend that hydropower generation and ecosystems that depend on consistent streamflows could be susceptible to the anticipated increase in drought intensification from climate change [6].

5. Implications and Conclusions

Prolonged droughts impact water resources, river ecosystems, agricultural productivity, food security, and the socio-economic status of local residents [1,2]. In the 3S Basin, the 2015–2016 drought reduced the production of cashews, pepper, and coffee by approximately 15% [59], which consequently impacted global supply chains [60], and affected the livelihoods of about two million people [59]. Given the global impacts of droughts, characterizing hydrological drought in the Srepok and Sesan Basins can provide baseline metrics for future droughts. Our findings of prolonged and severe droughts from 2009 to 2011 and from 2015 to 2016 corresponded to Typhoon Ketsana and El Niño conditions that reduced surface runoff in the 3S Basin. These findings can provide insights into future drought impact mitigations. For example, early drought warnings could reduce water demands or improve streamflow management to mitigate drought effects. Drought metrics, such as those provided here, can be used to support appropriate drought management and decision-making for sustainable river ecosystems, water resources, agricultural productivity, and food security in the 3S Basin and the Lower Mekong River.