# Regional Assessment of Groundwater Recharge in the Lower Mekong Basin

^{1}

^{2}

^{3}

^{*}

## Abstract

**:**

## 1. Introduction

^{2}in humid tropical settings, groundwater usually discharges to rivers all year round, resulting in gaining streams, compatible with groundwater recharge estimation by stream hydrograph separation [11,14]. However, baseflow separation methods inevitably involve subjective choices on the mathematical algorithms that cannot fully capture the nonlinearity of processes controlling surface-groundwater exchanges: riverbank storage, spatial variability in evaporation, recharge and storage capacity of the aquifer [15]. Consequently, values of baseflow estimated with different algorithms can vary by a factor of two [5]. There is no consensus on which geomorphic and land-cover characteristics are most closely linked to subsurface storage and baseflow [16]. Factors that promote infiltration and recharge to subsurface storages will increase baseflow, while factors associated with higher evapotranspiration will reduce baseflow.

## 2. Study Site

^{2}) and mean annual discharge (475 km

^{3}) [13]. Originating from the Tibetan Plateau in China, this river crosses Myanmar, Laos, Thailand, Cambodia and finally Vietnam where it flows into the South China Sea. The LMB includes the portion of the basin located downstream of China. It accounts for 79% of the total Mekong basin area and 84% of the flow volume [13].

#### 2.1. Climate and Hydrology

#### 2.2. Topography

^{2}of mangrove, swamps, sand dunes, spits, tidal flats, and irrigated rice paddy fields.

#### 2.3. Hydrogeology

## 3. Materials and Methods

#### 3.1. Catchment Selection and Baseflow Computation

^{0.2}where A is the catchment area in square miles [28]. The time interval includes (2N*−1)/2 days where 2N* is the odd integer between 3 and 11 nearest to 2N. In each of the 65 catchments, specific annual baseflow (mm/year) was computed for each hydrological year (1st April–31st March) available in the records and the median annual value Q

_{B}

_{estim}was selected as the independent variable to perform the multiple linear regression analysis (cf. Section 3.2).

#### 3.2. Prediction of Baseflow Across the Lower Mekong Basin

_{B}

_{predict}that predicts Q

_{B}

_{estim}from m catchment characteristics X

_{i}. Their logarithmic transformation produces a linear model (Equation (2)) whose m + 1 coefficients β

_{i}can be determined with multiple linear regressions.

_{0}is the intercept term of the model. ν and ε are the log-normally and normally distributed errors of the models, respectively. Normality in ε distribution is usually easier to obtain than in not log-transformed linear model, hence the advantage of power-law equations. The logarithm function being defined for strictly positive values only, adding one to catchment characteristics X

_{i}including zero values allows a correct mapping between the value of ln(X

_{i}

_{+1}) and X

_{i}[32]. The selection of the catchment characteristics X

_{i}that best predict Q

_{B}

_{estim}, and the calculation of their respective coefficients β

_{i}are performed by weighted least squares regressions applied to the 65 values of Q

_{B}

_{estim j}calculated in the 65 catchments (j = 1, …, 65), and their respective catchment characteristics X

_{ij}.

_{B}

_{estim j}equally, weighted least square regression enables the varying number k

_{j}of hydrological years used to calculate each value of Q

_{B}

_{estim j}to be accounted [33]. Values of Q

_{B}

_{estim j}derived from a greater number of hydrological years are more precise (have lower variance) and thus should have a greater weight in the regression. However, this reliability decreases as the variance of Q

_{B}

_{estim j}increases. To account for these two counteracting factors, weights (w

_{j}) were calculated as follows:

_{B}

_{estim j}) is the standard deviation of Q

_{B}

_{estim j}.

_{i}(i.e., catchment characteristics) that best predict baseflow were identified among a set of 15 candidate variables (described in Section 3.3 and listed in Table 1) using the two selection algorithms “best subsets regression” and “step-wise regression” available in MiniTab 16. This selection was intended to maximize the prediction R-squared (R

^{2}

_{pred}) calculated by leave-one-out cross-validations. Unlike the classical R-squared the maximization of which can lead to model over-fitting and loss of robustness, R

^{2}

_{pred}reflects the ability of the model to predict observations which were not used in the model calibration. Maximizing R

^{2}

_{pred}generally leads to greater parsimony in the number of explanatory variables. An explanatory variable was considered to be statistically significantly different from zero if its p-value, derived from the Student’s t-test, was lower than 0.05. The required homoscedasticity (homogeneity of variance) of the model residuals ε was verified by visual inspection of the residual plots. Possible multi-collinearity among the explanatory variables was controlled with the variance inflation factor (VIF). The influence statistic Cooks D was used to identify and remove outlier catchments exhibiting high influence on the estimation of the model coefficients [34].

^{2}

_{adj}), and R

^{2}

_{pred}. While R

^{2}

_{pred}assesses how well Equation (2) predicts responses to new observations, R

^{2}

_{adj}allows comparing the performance of linear regressions including different numbers of explanatory variables. While the value of the classical R

^{2}systematically increases when a new explanatory variable is added in Equation (2), R

^{2}

_{adj}will increase only if the new term improves the performance of the linear regression more than what would be expected by chance alone [35]. While R

^{2}

_{adj}and R

^{2}

_{pred}estimate the strength of the linear association between the estimations and predictions, NSEC measures the goodness of fit of linear and non-linear models (including power law models, i.e., Equation (1)), thus allowing performance comparison with any hydrological model. NSEC is computed as follows:

_{B}

_{estim j}is the median value of specific annual baseflow computed in catchment j with the local minimum method and Q

_{B}

_{pred j}is the corresponding value predicted with the power-law model (i.e., Equation (1)) in the same catchment. $\overline{{Q}_{B\text{}estim}}$ is the spatial mean of the estimated baseflow Q

_{B}

_{estim j}across all catchments.

#### 3.3. Catchment Characteristics

#### 3.3.1. Climate

_{B}

_{estim}: annual and monthly rainfall, rainfall cumulated over the n-day (n = 5, 10, and 15) rainiest periods of the hydrological year. Annual median rainfall, exhibiting the greatest correlation coefficient with Q

_{B}

_{estim}, was included as the only candidate rainfall variable in the regressions (Table 1). Median annual values of catchment areal temperature and standard evapotranspiration (ET

_{0}) were computed using 0.5° × 0.5° gridded monthly values from the Climate Research Unit [38]. ET

_{0}was calculated with the FAO grass reference evapotranspiration Equation applied to climate variables from the same data source covering the period 1901–2009 [39]. These three climate median values and Q

_{B}

_{estim j}were derived from the same k

_{j}hydrological years in each selected gauged catchment. In addition to ET

_{0}, median annual values of actual catchment areal evapotranspiration were computed using the land surface evapotranspiration product MODIS 16 available at daily time step with 1 km

^{2}resolution for the period 2000–2012 [40].

#### 3.3.2. Geomorphology and Geographic Coordinates

^{2}. This threshold value was selected to best capture the variability of drainage densities among the studied catchments. The geographic coordinates of the catchment centroid (latitude and longitude) were selected as two additional candidate explanatory variables to capture any longitudinal and latitudinal gradients in incomputable environmental variables possibly influencing groundwater recharge across the LMB (e.g., aquifer properties).

#### 3.3.3. Soil

#### 3.3.4. Land Cover

## 4. Results

#### 4.1. Baseflow Estimations

_{B estim}) varies between 53 mm/year (catchment of the Nam Mun River at Rasi Salai station, Thailand) and about 1000 mm/year (catchment of the Nam Sane River at Muong Borikhan station, Laos) with a median value of 439 mm/year. Circles depicted in Figure 4 show the spatial distribution of these estimated recharge rates. Strikingly, extreme values are grouped regionally. Greatest values are observed in Central Laos, ranging from 625 to 1000 mm/year, and Southern Laos, ranging from 600 to 1000 mm/year. Lowest values, between 83 and 190 mm/year, are grouped in Northeast Thailand.

#### 4.2. Multiple Regressions Analysis

#### 4.2.1. Prediction of Groundwater Recharge

^{2}

_{pred}percentage value by 27 points. A combination of four explanatory variables selected among the 15 variables listed in Table 1 is sufficient to predict Q

_{B estim}with the following performances: R

^{2}

_{pred}= 66.36%, R

^{2}

_{adj}= 70.30%, and NSEC = 63.70% (Equation (5)).

_{B pred}is the independent variable predicting Q

_{B estim}. The dependent variables Rain (medium annual rainfall), ET

_{0}, Lat (latitude) and Long (longitude) are listed in Equation (5) according to their decreasing explanatory power (T-ratios = 7.51; −4.35; −2.96, and −2.48, respectively). The coefficient of Rain is much greater than unity, indicating that an increase of x% in annual rainfall would induce an >x% increase in baseflow (i.e., groundwater recharge). Consistently, the coefficient of ET

_{0}is negative, reflecting the moderating effect of evapotranspiration on groundwater recharge. Unlike ET

_{0}, actual land surface evapotranspiration had no explanatory power. Lat and Long are negatively correlated to baseflow. The exclusion of actual ET and inclusion of Lat and Long in Equation (5) are discussed in Section 5. Predicted values of groundwater recharge Q

_{B pred}derived from Equation (5) are represented with colored grid cells in Figure 4. To prepare this map at 0.25° × 0.25° spatial resolution, each value of ET

_{0}available at 0.5° × 0.5° spatial resolution was first replicated in the four corresponding 0.25° × 0.25° grid cells. Figure 5 displays regional variations in Rain and ET

_{0}. Visual comparison with Figure 4 confirms that Rain is the main driver of Q

_{B pred}, with maximum and minimum Q

_{B pred}values observed in the rainiest and driest locations, respectively. Anti-correlation between Q

_{B pred}and ET

_{0}is less obvious, though highly significant (F test p-value = 0.00), as confirmed by the correlation coefficients R

_{Ln}computed on the logarithms of the variables: R

_{Ln}(Q

_{B pred}, Rain) = 79.13%; R

_{Ln}(Q

_{B pred}, ET

_{0}) = −32.10%. Figure 6 maps spatial variations in the ratio between Q

_{B pred}and Rain. These variations follow the regional variations of Rain Figure 5a, R

_{Ln}(Rain, Q

_{B pred}/Rain) = 52.06%, because the coefficient of the dependent variable Rain in Equation (5) is greater than unity, reflecting the non-linear relationship between rainfall depth and groundwater recharge, and highlighting the dominant role of annual rainfall in the control of groundwater recharge rates. This ratio varies between less than 15% in Northeast Thailand, up to more than 50% in Central and Southern Laos.

_{0}, Lat and Long.

#### 4.2.2. Model Performance

_{B pred}values depicted by square grid cells in Figure 4 consistently exhibit local maximums and minimums in areas where Q

_{B estim}show similar extremes (e.g., Central and Southern Laos, and Northeast Thailand, respectively). Figure 7 compares Q

_{B estim}and Q

_{B pred}to assess the performance of the power-law model. The scatter plots align well along the first bisector with more than half of the catchments having an absolute normalized error (ANE = |Q

_{B estim}−Q

_{B pred}|/Q

_{B estim}) lower than 30%. These errors result from the assumptions of the local minimum and multiple linear regression methods, and from possible inaccuracies in the original flow values. Even though cross-validation has been performed, extrapolation to un-gauged catchments still adds non-measurable uncertainty. Figure 8 illustrates how ANE varies according to the aridity index (i.e., ET

_{0}/Rain) and the drainage area of the studied catchments. ANE ranges from 0.26 to 0.64 across the studied catchments, with a median of 0.45, typical of humid tropical areas where regression models predicting flow are known to perform best [45]. Although flow prediction is usually hampered by greater hydrological variability and higher presence of ephemeral rivers in drier areas, the power-law model predicting baseflow is not influenced by the aridity index in the LMB (Figure 8a). In contrast, ANE varies according to the drainage area of the catchments and exhibits a maximum in medium-size catchments (5000–10,000 km

^{2}) (Figure 8b).

^{2}

_{pred}= 66.36%; R

^{2}

_{adj}= 70.30%; NSEC = 63.70%) assess how well the power-law model described in Equation (5) performs, allowing comparisons with regional regression models developed in other parts of the world. Our NSEC value, equivalent to the classical R

^{2}reported in [45], falls within the range of values typically observed for regression models predicting low flows in other humid regions of the world [45]. Finally, it should be noted that we re-applied the regression analysis using the original values of the variables (not log-transformed) and verified that the power-law structure outperformed the linear one, likely because of the typical non-linear relationship between rainfall and flow.

## 5. Discussion

#### 5.1. Factors Determining Groundwater Recharge

_{0}. While ET

_{0}is the second most powerful predictive variable of Equation (5), actual land surface evapotranspiration had no predictive power. Likely explanations include: (i) the mismatch between periods used to compute Q

_{B}

_{estim}and actual evapotranspiration derived from MODIS 16, and available since 2000 only, considering that recent land-cover changes have occurred across the region; (ii) the incomplete validation of MODIS 16 product for tropical Southeast Asia [49]. Latitude and longitude are the third and fourth most powerful predictive variables of Table 1. They most likely act as surrogates for environmental processes controlling baseflow, exhibiting latitudinal and longitudinal gradients, and not listed in Table 1 since independency between the explanatory variables is a prerequisite for inclusion in the regression equation. Regional characteristics of the aquifers (Figure 2) exhibit such gradients: the southwest part of the LMB corresponds to sedimentary sandy deposits of the Khorat Plateau in Northeast Thailand and the alluvial plain of the Mekong River in Cambodia and Southern Vietnam. These sedimentary deposits are usually characterized by greater permeability than the more compact metamorphic rocks prevalent along the Annamite Range from the North to the Southeast of the LMB [3,18,19]. Although this contrast is reversed in a few and confined locations (e.g., the basalt flows in Southern Laos with greater permeability; the early Paleozoic rocks in Northeast Thailand with reduced permeability), these local contrasts have likely minor influence on groundwater recharge at the scale of the studied catchments. Based on these observations, and accounting for the limited quantitative hydrogeological data explained in Section 2, we hypothesize that Lat and Long in Equation (5) are surrogate variables for the permeability and transmissivity of the aquifers that likely increase from North to South and from East to West across the LMB, in accordance with the signs of the coefficients of Lat and Long in Equation (5). It should be noted that a power-law Equation with only Rain and ET

_{0}as explanatory variables yields an R

^{2}

_{pred}value of 63.69%, about 3 points lower than the R

^{2}

_{pred}value of Equation (5). This comparison indicates that: (i) climate explains more than half of Q

_{B}

_{estim}variability across the LMB; and (ii) assuming that Lat and Long are surrogates for aquifer properties, the regional geology explains at least 3% of Q

_{B}

_{pred}variability.

_{0}, Long and Lat. The lower hydrological influence associated with land covers is consistent with the usually moderate hydrological effect of land-cover changes in catchments with mixed land covers and an area greater than 1000 km

^{2}. Over such large areas, the combinations of various land covers, with counteracting changes, generally render their individual hydrological effects difficult to detect [44]. The exclusion of soil characteristics from Equation (5) may be related to the poor accuracy of the soil maps used in this assessment, and/or to the surrogating effect of the latitude and longitude.

#### 5.2. Comparison with Previous Studies

^{2}catchment with the WetSpass model [50]. Resulting annual recharge rates averaged 360 mm/year, mostly influenced by rainfall and evapotranspiration. In Southeast Vietnam, measurements from 10 monitored wells were used to infer groundwater recharge using finite difference methods [10]. Estimated recharge rates varied between 307 and 325 mm/year, slightly below our estimations in this part of Vietnam (500 mm/year). Similar consistency was observed in Northern Vietnam where our estimates (315 mm/year) were moderately exceeded by a mean rate of 477 mm/year derived from the rainfall infiltration breakthrough model calibrated with measurements of rainfall and groundwater levels [9]. In the Day river sub-basin of the Red River in Northern Vietnam, Van der Wolf [51] calibrated the SWAT model using hydro-meteorological observations and detailed maps of land uses, topography and soils. The SCS curve number method was used to model surface runoff and to infer infiltration rates. The resulting recharge rates averaged 248 mm/year, ranging between 37 and 601 mm/year, in agreement with our results (213–318 mm/year) for this area. Broader-scale groundwater recharge assessments was performed over 15,000 km

^{2}in central Cambodia [52], yielding 448 mm/year from the SCS runoff curve method, aligned with Q

_{B}

_{estim}(464 mm/year) and slightly greater than Q

_{B}

_{pred}(278–357 mm/year) in this area. In Northwest Cambodia, groundwater recharge rates were estimated over 3375 km

^{2}of sandstones from Upper Triassic to Lower Cretaceous, using the water-table fluctuation method and the stable isotopes analysis from 12 piezometers [53]. Recharge rates of 10 to 70 mm/year, lower than our estimations in this area (200 mm/year), were explained by clayey soils overlying sandstone, whose presence can highly vary at the scale of few tens of meters. Similar low estimates (20 mm/year) were derived from the groundwater flow model MODFLOW in Southeast Cambodia, much lower than our estimations in this location (230 mm/year), and attributed to the presence of a surficial clay aquitard [54].

#### 5.3. Limitations of the Study

#### 5.4. Groundwater Potential for Irrigation

_{%}can be estimated as follows: I

_{%}= 0.5 × Q

_{B}/1000 [57]. In Northeast Thailand, I

_{%}= 7.5%. This percentage area is equivalent to the fraction of agricultural land actually irrigated with surface water in this region [59], suggesting that irrigation could potentially be doubled by improving groundwater access. However, any change in the local groundwater balance can have detrimental effects locally (e.g., increased groundwater salinity), and downstream (e.g., altered water uses and ecosystems). In the Vientiane Plain, I

_{%}= 17.5% while the current percentage area of irrigated land in the Vientiane Prefecture, which largely covers the southern boundary of the Vientiane Plain, is around 10% (unpublished sources from Department of Irrigation, Ministry of Agriculture and Forestry, Laos). This figure, although conservative, demonstrates the considerable groundwater potential for developing irrigation, albeit with numerous technical and non-technical issues which severely constrain development [60].

## 6. Conclusions

## Acknowledgments

## Author Contributions

## Conflicts of Interest

## References

- Gleeson, T.; Wada, Y.; Bierkens, M.F.P.; van Beek, L.P.H. Water balance of global aquifers revealed by groundwater footprint. Nature
**2012**, 488, 197–200. [Google Scholar] [CrossRef] [PubMed] - Siebert, S.; Burke, J.; Faures, J.M.; Frenken, K.; Hoogeveen, J.; Döll, P.; Portmann, F.T. Groundwater use for irrigation—A global inventory. Hydrol. Earth Syst. Sci.
**2010**, 14, 1863–1880. [Google Scholar] [CrossRef][Green Version] - Landon, M. Preliminary Compilation and Review of Current Information on Groundwater Monitoring and Resources in the Lower Mekong River Basin; U.S. Geological Survey Report to the Mekong River Commission: Vientiane, Laos, 2001.
- Vote, C.; Newby, J.; Phouyyavong, K.; Inthavong, T.; Eberbach, P. Trends and perceptions of rural household groundwater use and the implications for smallholder agriculture in rain-fed Southern Laos. Int. J. Water Resour Dev.
**2015**, 31, 558–574. [Google Scholar] [CrossRef] - Döll, P.; Fiedler, K. Global-scale modelling of groundwater recharge. Hydrol. Earth Syst. Sci.
**2008**, 12, 863–885. [Google Scholar] [CrossRef] - Gleeson, T.; Alley, W.M.; Allen, D.M.; Sophocleous, M.A.; Zhou, Y.X.; Taniguchi, M.; VanderSteen, J. Towards sustainable groundwater use: Setting long-term goals, backcasting, and managing adaptively. Groundwater
**2012**, 50, 19–26. [Google Scholar] [CrossRef] [PubMed] - Wada, Y.; van Beek, L.P.H.; van Kempen, C.M.; Reckman, J.W.T.M.; Vasak, S.; Bierkens, M.F.P. Global depletion of groundwater resources. Geophys. Res. Lett.
**2010**, 37, L20402. [Google Scholar] [CrossRef] - Saraphirom, P.; Wirojanagud, W.; Srisuk, K. Impact of climate change on waterlogging and salinity distributions in Huai Khamrian subwatershed, NE Thailand. Environ. Earth Sci.
**2013**, 70, 887–900. [Google Scholar] [CrossRef] - Rõi, N.D. Estimation of groundwater recharge of the Holocene aquifer from rainfall by RIB method for Hưng Yên Province. VNU J. Sci. Earth Environ. Sci.
**2014**, 30, 49–63. [Google Scholar] - Vinh, C.T.; Khuyen, N.M.; Hieu, N.H.; Long, D.V.; Bach, N.T.; Van, T.T.T. Assessment of Groundwater Recharge from Rainfall in the Plain of Cai Phan Rang, Vietnam. J. Environ. Sci. Eng.
**2014**, 3, 141–149. [Google Scholar] - Scanlon, B.R.; Healy, R.W.; Cook, P.G. Choosing appropriate techniques for quantifying groundwater recharge. Hydrogeol. J.
**2002**, 10, 18–39. [Google Scholar] [CrossRef] - Von Freyberg, J.; Moeck, C.; Schirmer, M. Estimation of groundwater recharge and drought severity with varying model complexity. J. Hydrol.
**2015**, 527, 844–857. [Google Scholar] [CrossRef] - Mekong River Commission (MRC). Overview of the Hydrology of the Mekong River Basin; Mekong River Commission (MRC): Vientiane, Laos, 2005. [Google Scholar]
- Halford, K.J.; Mayer, G.C. Problems associated with estimating ground water discharge and recharge from stream discharge records. Groundwater
**2000**, 38, 331–342. [Google Scholar] [CrossRef] - Nathan, R.J.; McMahon, T.A. Evaluation of automated techniques for base flow and recession analyses. Water Resour. Res.
**1990**, 26, 1465–1473. [Google Scholar] [CrossRef] - Price, K. Effects of watershed topography, soils, land use, and climate on baseflow hydrology in humid regions: A review. Prog. Phys. Geol.
**2011**, 35, 465–492. [Google Scholar] [CrossRef] - Mekong River Commission (MRC). State of the Basin Report; Mekong River Commission (MRC): Vientiane, Laos, 2010. [Google Scholar]
- Johnson, J.H. Preliminary Appraisal of the Hydrogeology of the Lower Mekong Basin; Interim Mekong Committee: Vientiane, Laos, 1986; p. 147. [Google Scholar]
- Charuratna, A.; Phu, T.H. Hydrogeological Map of Lower Mekong Basin. 1:1,000,000; Mekong Secretariat, Interim Committee for Coordination of Investigations of the lower Mekong Basin: Bangkok, Thailand, 1992. [Google Scholar]
- Japan International Cooperation Agency (JICA). Basic Design Study Report on the Project for Groundwater Development in the Vientiane Province in Lao PDR; Japan International Cooperation Agency (JICA): Tokyo, Japan, 1993; p. 199.
- Perttu, N.; Wattanasen, K.; Phommasone, K.; Elming, S.A. Characterization of aquifers in the Vientiane Basin, Laos, using magnetic resonance sounding and vertical electrical sounding. J. Appl. Geophys.
**2011**, 73, 207–220. [Google Scholar] [CrossRef] - Lacombe, G.; McCartney, M. Evaluating the Flow Regulating Effects of Ecosystems in the Mekong and Volta River Basins; IWMI Research Report 166; International Water Management Institute: Colombo, Sri Lanka, 2016; p. 40. [Google Scholar]
- Yermoli, C. Draft User Manual. Hydropower Project Database; Basin Development Plan Phase 2; Mekong River Commission: Vientiane, Laos, 2008; p. 27. [Google Scholar]
- WLE. Dams in the Mekong River Basin: Commissioned, under Construction and Planned Dams in April 2016; CGIAR Research Program on Water, Land and Ecosystems-Greater Mekong (WLE): Vientiane, Laos, 2016. [Google Scholar]
- Pettyjohn, W.A.; Henning, R. Preliminary Estimate of Groundwater Recharge Rates, Related Streamflow and Water Quality in Ohio; Completion Report Number 552; University Water Resources Center: Columbus, OH, USA, 1979. [Google Scholar]
- HydroOffice. BFI+ 3.0. User’s Manual by Milos Gregor; HydroOffice: Bratislava, Slovakia, 2010. [Google Scholar]
- Milnes, E.; Negro, F.; Perrochet, P. Vietnam to Produce More Coffee with Less Water—Towards a Reduction of the Blue Water Footprint in Coffee Production. Hydrogeological Study of the Basaltic Plateau in Dak Lak Province, Vietnam; Université de Neuchâtel: Neuchâtel, Switzerland, 2015. [Google Scholar]
- Linsley, R.K.; Kohler, M.A.; Paulhus, J.L.H. Hydrology for Engineers, 2nd ed.; McGraw-Hill Book Company: New York, NY, USA, 1975. [Google Scholar]
- Eckhardt, K. A comparison of baseflow indices, which were calculated with seven different baseflow separation methods. J. Hydrol.
**2008**, 352, 168–173. [Google Scholar] [CrossRef] - Risser, D.W.; Gburek, W.J.; Folmar, G.J. Comparison of Methods for Estimating Ground-Water Recharge and Base Flow at a Small Watershed Underlain by Fractured Bedrock in the Eastern United States; US Geological Survey Scientific Investigations Report 2005–5038; U.S. Geological Survey: Reston, VA, USA, 2005.
- Vogel, R.M.; Wilson, I.; Daly, C. Regional regression models of annual streamflow for the United States. J. Irrig. Drain. Eng.
**1999**, 125, 148–157. [Google Scholar] [CrossRef] - Homa, E.S.; Brown, C.; McGarigal, K.; Compton, B.W.; Jackson, S.D. Estimating hydrologic alteration from basin characteristics in Massachusetts. J. Hydrol.
**2013**, 503, 196–208. [Google Scholar] [CrossRef] - Tasker, G.D. Hydrologic regression with weighted least squares. Water Resour. Res.
**1980**, 16, 1107–1113. [Google Scholar] [CrossRef] - Cook, R.D.; Weisberg, S. Residuals and Influence in Regression; Chapman and Hall: New York, NY, USA, 1982. [Google Scholar]
- Helsel, D.R.; Hirsch, R.M. Statistical Methods in Water Resources; US Geological Survey: Reston, VA, USA, 2002.
- Yatagai, A.; Kamiguchi, K.; Arakawa, O.; Hamada, A.; Yasutomi, N.; Kitoh, A. APHRODITE: Constructing a long-term daily gridded precipitation dataset for Asia based on a dense network of rain gauges. Bull. Am. Meteorol. Soc.
**2012**, 93, 1401–1415. [Google Scholar] [CrossRef] - Lauri, H.; Räsänen, T.A.; Kummu, M. Using reanalysis and remotely sensed temperature and precipitation data for hydrological modeling in monsoon climate: Mekong river case study. J. Hydrometeorol.
**2014**, 15, 1532–1545. [Google Scholar] [CrossRef] - Harris, I.; Jones, P.D.; Osborn, T.J.; Lister, D.H. Updated high-resolution grids of monthly climatic observations—The CRUTS3.10 Dataset. Int. J. Climatol.
**2014**, 34, 623–642. [Google Scholar] [CrossRef][Green Version] - Ekström, M.; Jones, P.D.; Fowler, H.J.; Lenderink, G.; Buishand, T.A.; Conway, D. Regional climate model data used within the SWURVE project 1: Projected changes in seasonal patterns and estimation of PET. Hydrol. Earth Syst. Sci.
**2007**, 11, 1069–1083. [Google Scholar] [CrossRef] - Mu, Q.; Zhao, M.; Running, S.W. Improvements to a MODIS global terrestrial evapotranspiration algorithm. Remote Sens. Environ.
**2011**, 115, 1781–1800. [Google Scholar] [CrossRef] - Lehner, B.; Verdin, K.; Jarvis, A. HydroSHEDS Technical Documentation; World Wildlife Fund US: Washington, DC, USA, 2006. [Google Scholar]
- MRC. Planning Atlas of the Lower Mekong River Basin. Basin Development Plan Programme; Mekong River Commission (MRC): Phnom-Penh, Cambodia, 2011. [Google Scholar]
- Lacombe, G.; Douangsavanh, S.; Vogel, R.M.; McCartney, M.; Chemin, Y.; Rebelo, L.M.; Sotoukee, T. Multivariate power-law models for streamflow prediction in the Mekong Basin. J. Hydrol. Reg. Stud.
**2014**, 2, 35–48. [Google Scholar] [CrossRef] - Bruijnzeel, L.A. Hydrological functions of tropical forests: Not seeing the soil for the trees? Agric. Ecosyst. Environ.
**2004**, 104, 185–228. [Google Scholar] - Salinas, J.L.; Laaha, G.; Rogger, M.; Parajka, J.; Viglione, A.; Sivapalan, M.; Blöschl, G. Comparative assessment of predictions in ungauged basins—Part 2. Flood and low flow studies. Hydrol. Earth Syst. Sci.
**2013**, 17, 2637–2652. [Google Scholar] [CrossRef] - Nolan, B.T.; Healy, R.W.; Taber, P.E.; Perkins, K.; Hitt, K.J.; Wolock, D.M. Factors influencing ground-water recharge in the eastern United States. J. Hydrol.
**2007**, 332, 187–205. [Google Scholar] [CrossRef] - Kim, J.H.; Jackson, R.B. A global analysis of groundwater recharge for vegetation, climate, and soils. Vadose Zone J.
**2012**, 11. [Google Scholar] [CrossRef] - Zomlot, Z.; Verbeiren, B.; Huysmans, M.; Batelaan, O. Spatial distribution of groundwater recharge and base flow: Assessment of controlling factors. J. Hydrol. Reg. Stud.
**2015**, 4, 349–368. [Google Scholar] [CrossRef] - Kim, H.W.; Hwang, K.; Mu, Q.; Lee, S.O.; Choi, M. Validation of MODIS 16 Global Terrestrial Evapotranspiration Products in Various Climates and Land Cover Types in Asia. KSCE J. Civ. Eng.
**2012**, 16, 229–238. [Google Scholar] [CrossRef] - Kornkul, J.; Chotpantarat, S. Distributed Groundwater Recharge Estimation in Phrae Province Using WetSpass. In Proceedings of the Second Environment Asia International Conference on “Human Vulnerability and Global Environmental Change”, Chonburi, Thailand, 15–17 May 2013; pp. 553–565. [Google Scholar]
- Van der Wolf, M. Effects of Land Use Change on Groundwater Recharge. A case Study in the Day River Basin, Vietnam. Master’s Thesis, Utrecht University, Utrecht, The Netherlands, 2015. [Google Scholar]
- JICA. The Study on Groundwater Development in Central Cambodia; Interim Report of the Japan International Cooperation Agency (JICA); Kokusai Kogyo Co. Ltd.: Tokyo, Japan, 2001. [Google Scholar]
- Vouillamoz, J.M.; Valois, R.; Lun, S.; Caron, D.; Arnout, L. Can groundwater secure drinking-water supply and supplementary irrigation in new settlements of North-West Cambodia? Hydrogeol. J.
**2016**, 24, 195–209. [Google Scholar] [CrossRef] - Erban, L.E.; Gorelick, S.M. Closing the irrigation deficit in Cambodia: Implications for transboundary impacts on groundwater and Mekong River flow. J. Hydrol.
**2016**, 535, 85–92. [Google Scholar] [CrossRef] - Pilgrim, D.H.; Cordery, I.; Baron, B.C. Effects of catchment size on runoff relationships. J. Hydrol.
**1982**, 58, 205–221. [Google Scholar] [CrossRef] - Suhardiman, D.; Giordano, M.; Leebouapao, L.; Keovilignavong, O. Farmers’ strategies as building block for rethinking sustainable intensification. Agric. Hum. Values
**2016**, 33, 563–574. [Google Scholar] [CrossRef] - Pavelic, P.; Smakhtin, V.; Favreau, G.; Villholth, K. Water-balance approach for assessing potential for smallholder groundwater irrigation in Sub-Saharan Africa. Water SA
**2012**, 38, 399–406. [Google Scholar] [CrossRef] - Brouwer, C.; Heibloem, M. Irrigation Water Needs Irrigation Water Management Training Manual 3; Food and Agriculture Organization (FAO): Rome, Italy, 1986. [Google Scholar]
- Agricultural Statistics of Thailand 2011. Office of Agricultural Economics (OAE). Available online: http://www.oae.go.th (accessed on 9 July 2016).
- Pavelic, P.; Xayviliya, O.; Ongkeo, O. Pathways for effective groundwater governance in the least developed country context of the Lao PDR. Water Int.
**2014**, 39, 469–485. [Google Scholar] [CrossRef]

**Figure 2.**Geology of the Lower Mekong Basin, adapted from [19].

**Figure 4.**Median annual groundwater recharge in the Lower Mekong Basin. Circles are located in the centroid of the gauged catchments and their corresponding values are estimated with the local minimum method (Q

_{B estim}). Un-gauged values in square grid cells are predicted by multiple log-linear regressions (Q

_{B pred}). The graduated color-scale indicates the values for both Q

_{B estim}and Q

_{B pred}.

**Figure 6.**Ratio between estimated median annual recharge (Q

_{B}

_{estim}) and median annual rainfall (Rain) in the Lower Mekong Basin.

**Figure 7.**Comparison of observed (Q

_{B}

_{estim j}) and predicted (Q

_{B}

_{pred j}) specific median annual baseflow in each gauged catchment j.

**Figure 8.**Absolute normalized error (ANE) of predicting median baseflow as a function of aridity index (

**a**) and drainage area (

**b**). Boxes: 40–60% quantiles; whiskers: 20–80% quantiles; white circles: medians.

**Table 1.**Candidate explanatory variables considered in the multiple regression analyses: Variation ranges across the 65 catchments.

Variables | Unit | Minimum | Median | Maximum |
---|---|---|---|---|

Climate | ||||

Median annual rainfall | mm/year | 880 | 1416 | 2093 |

Median annual temperature | °C | 21.0 | 24.2 | 27.4 |

Median annual standard evapotranspiration | mm/year | 1017 | 1168 | 1338 |

Median annual actual evapotranspiration | mm/year | 818 | 1280 | 1374 |

Geomorphology | ||||

Drainage area | km^{2} | 207 | 3278 | 106,748 |

Drainage density | km^{−1} | 0.09 | 0.13 | 0.17 |

Mean elevation | m | 84 | 562 | 1168 |

Mean slope | % | 2 | 15 | 32 |

Perimeter | km | 76 | 401 | 2090 |

Geographic coordinates of catchment centroid | ||||

Latitude | decimal degree | 12.33 | 16.70 | 20.70 |

Longitude | 99.35 | 104.03 | 108.00 | |

Soil | ||||

Top-soil texture | 4-unit * scale | 0 | 2.08 | 2.91 |

Depth | 0 | 3.07 | 4.00 | |

Land cover | ||||

Forest | % area | 3 | 75 | 98 |

Rain-fed lowland paddy | 0 | 4 | 77 |

Soil Depth | Top Soil Texture | 4-Unit Scale |
---|---|---|

<30 cm | Coarse | 1 |

30–50 cm | Medium | 2 |

30–50 cm with gravel | Fine | 3 |

>50 cm | Peat | 4 |

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Lacombe, G.; Douangsavanh, S.; Vongphachanh, S.; Pavelic, P. Regional Assessment of Groundwater Recharge in the Lower Mekong Basin. *Hydrology* **2017**, *4*, 60.
https://doi.org/10.3390/hydrology4040060

**AMA Style**

Lacombe G, Douangsavanh S, Vongphachanh S, Pavelic P. Regional Assessment of Groundwater Recharge in the Lower Mekong Basin. *Hydrology*. 2017; 4(4):60.
https://doi.org/10.3390/hydrology4040060

**Chicago/Turabian Style**

Lacombe, Guillaume, Somphasith Douangsavanh, Sinxay Vongphachanh, and Paul Pavelic. 2017. "Regional Assessment of Groundwater Recharge in the Lower Mekong Basin" *Hydrology* 4, no. 4: 60.
https://doi.org/10.3390/hydrology4040060