A first investigation of hydrogeology and hydrogeophysics of the Maqu catchment in the Yellow River source region.

. The Tibetan Plateau is the source of most of Asia's major rivers and has been called the Asian Water Tower. Detailed 15 knowledge of its hydrogeology is paramount to enable the understanding of groundwater dynamics, which plays a vital role in headwater areas like the Tibetan Plateau. Nevertheless, due to its remoteness and the harsh environment, there is a lack of field survey data to investigate its hydrogeology. In this study, borehole core lithology analysis, soil thickness measurement, altitude survey, hydrogeological surveys, and hydrogeophysical surveys (e.g., Magnetic Resonance Sounding – MRS, Electrical Resistivity Tomography – ERT, and Transient Electromagnetic – TEM) were conducted in the Maqu catchment within the 20 Yellow River Source Region (YRSR). The hydrogeological surveys reveal that groundwater flows from the west to the east, recharging the Yellow River. The hydraulic conductivity ranges from 0.2 m d -1 to 12.4 m d -1 . The MRS soundings results, i.e., water content and hydraulic conductivity, confirmed the presence of an unconfined aquifer in the flat eastern area. Based on TEM results, the depth of the Yellow River deposits was derived at several places in the flat eastern area, ranging from 50 m to 208 m. The soil thickness measurements were done in the western mountainous area of the catchment, where 25 hydrogeophysical and hydrogeological surveys were difficult to be carried out. The results indicate most soil thicknesses, except on the valley floor, are within 1.2 m in the western mountainous area of the catchment, and the soil thickness decreases as the slope increases. These survey data and results can contribute to integrated hydrological modeling and water cycle analysis to improve a full–picture understanding of the water cycle at the Maqu catchment in the YRSR. The raw dataset is freely available at https://doi.org/10.17026/dans-z6t-zpn7 (Li et al., 2020a), and the dataset containing the processed ERT, 30 MRS, and TEM data is also available at the National Data Center


Introduction
The conventional way to acquire hydrogeological information in an unknown area is by augering or drilling boreholes and 65 carrying out hydraulic tests, for example, pumping tests (Vouillamoz et al., 2012). However, due to the harsh environment of the TP, and the high costs and time-consuming of the traditional hydrogeological survey methods, little hydrogeological work has been done on the TP.
The hydrogeophysical methods are up-and-coming in hydrogeological studies (Chirindja et al., 2016). They have been applied in various conditions, for example in: wetlands (Chambers et al., 2014), rivers (Steelman et al., 2015), proglacial moraine 70 (McClymont et al., 2011), karst regions (McCormack et al., 2017, and volcanic systems (Di Napoli et al., 2016;Fikos et al., 2012). Compared to other hydrogeophysical methods, such as seismics, gravity, and resistivity method, Magnetic Resonance Sounding (MRS) is the only method that is able to detect the free water in the subsurface directly (Lubczynski and Roy, 2003;Lubczynski and Roy, 2004), and quantify hydrogeological parameters and water storage (Lachassagne et al., 2005;Legchenko et al., 2002;Legchenko et al., 2018;Lubczynski and Roy, 2007). 75 The MRS method is based on the Nuclear Magnetic Resonance (NMR) principle, widely implemented in medical Magnetic Resonance Imaging (MRI). In contrast to MRI, the MRS excitation is done at the Larmor frequency corresponding to the local Earth's magnetic field which propagates and get attenuated through conductive subsurface media. Therefore, at a given position of the Earth, MRS signal response depends on the strength of the Earth's magnetic field, and on the strength of the natural and artificial noise (Lubczynski and Roy, 2004), but also on the subsurface resistivity. The electrical resistivity measurement is 80 suggested to be jointly used with MRS (Braun and Yaramanci, 2008;Descloitres et al., 2007;Vouillamoz et al., 2002).
Electrical Resistivity Tomography (ERT) is one of the predominantly employed hydrogeophysical methods to estimate the subsurface electrical resistivity (Herckenrath et al., 2012;Jiang et al., 2018). It has been widely applied together with MRS to explore regional hydrogeology (Vouillamoz et al. (2003), Descloitres et al. (2008), Pé rez-Bielsa et al. (2012)). Furthermore, the Transient Electro-Magnetic survey (TEM), referred to as the Time-Domain Electromagnetic Method (TDEM), provides 85 also subsurface resistivity, but is able to achieve deeper penetration than ERT. The ERT has already been used on the TP by Gao et al. (2019) and You et al. (2013) to investigate permafrost. Nevertheless, we haven't seen the combined use of MRS, ERT, and TEM for hydrogeophysical surveies over TP.
Investigations on various fields, such as geomorphology, climate change, glacier, and permafrost have been done on the TP based on different DEMs. Zhang et al. (2006) analyzed the geomorphic characteristics of the Minjiang drainage basin with 90 SRTM (Shuttle Radar Topography Mission) data. Wei and Fang (2013) assessed the trends of climate change and the spatiotemporal differences over the TP from 1961-2010, with a generalized temperature zone-elevation model and SRTM. Ye et al. (2015) calculated the glacier elevation change in the Rongbuk catchment from 1974 to 2006 based on topographic maps and ALOS (Advanced Land Observing Satellite). Niu et al. (2018) mapped permafrost distribution throughout the Qinghai-Tibet Engineering Corridor based on ASTER (The Terra Advanced Spaceborne Thermal Emission and Reflection 95 Radiometer) Global DEM. However, different DEMs used in different studies may lead to potential inconsistencies for understanding relevant physical processes (Zeng et al., 2019;Zeng et al., 2015). For Maqu catchment located within TP and being the focus of this study, it is crucial to evaluate the accuracy of different DEMs to select the best one and eventually calibrate it using ground measurements, since altitude is used to calculate hydraulic heads, which determine and control groundwater flow. Therefore, it is important to evaluate the accuracy of satellite DEM products with a ground-based Real-100 time Kinematic-Global Positioning System (GPS-RTK); such works have not been common over the TP.
This study jointly uses hydrogeological and hydrogeophysical methods, including aquifer tests, MRS, ERT, TEM, and other necessary approaches at Maqu catchment in the Yellow River Source Region (YRSR) on TP. The paper focuses on field hydrogeological and hydrogeophysical surveys, and corresponding datasets, aiming to fill the scientific data gap on TP. In what follows, the study area is introduced in Sect. 2. Borehole core lithology analysis, soil thickness measurement, altitude 105 survey, hydrogeological survey, and hydrogeophysical survey are presented in Sect. 3. The results are documented and uncertainties are discussed in Sect. 4. Data availability is given in Sect. 5. Conclusions are made in Sect. 6.

Study area
The study area is a catchment (33°43′ N -33°58′ N, 101°51′ E -102°16′ E) in Maqu county, China (Fig. 1). It is located at the northeastern edge of the TP, the first major bend of the Yellow River. Maqu county is regarded as the "reservoir" of the 110 YRSR. The Yellow river in contact with the Maqu catchment is about 36 km long, while the whole length passing the Maqu county is 433.3 km. When the Yellow River flows through Maqu county, the annual runoff increases by 10.8 billion m 3 , accounting for 58.7% of the total runoff of 18.4 billion m 3 of the Yellow River in the YRSR (Wang, 2008). The Maqu catchment is characterized by a cold climate with dry winter and warm summer, assigned as Dwb in the updated Köppen-Geiger climate classification (Peel et al., 2007). The annual mean temperature is about 1.8 °C , and the precipitation is around 115 620 mm annually. The catchment is covered by short grasses used for grazing by yaks and sheeps. The elevation ranges between 3367 to 4017 m a.s.l according to ALOS PALSAR (Phase Array type L band Synthetic Aperture Radar) RT1 (high terrain correction resolution). There is a reservoir in the catchment (Fig. 1c), with functions of grassland irrigation and flood control.
Based on the field survey of geomorphology and geology (Compton, 1962;Dackombe and Gardiner, 2020), the catchment can 120 be divided into two parts, the flat eastern area, and the western mountainous area. The western mountains are feldspathic quartzose sandstone and sandy slate with soil covered at the top, while the eastern, relatively flat part, is composed mainly of sediments, such as alluvial deposits with intercalated eolian units. The eastern part, together with its extension outside of the study area, is called the Ruoergai Basin (Fig. 1b). Surface processes cause erosion, mixing, unmixing, and redistribution of alluvial materials within the thick alluvia accumulation at the eastern part. Geomorphological characterization was carried out 125 in the Maqu catchment in 2018, and the three terraces were identified (Fig. 1c).

130
Some previous works have been done in or around the catchment. Su et al. (2011) monitored the soil moisture and soil temperature from 5 to 80 cm below the ground surface. Dente et al. (2012) assessed the reliability of AMSR-E (Advanced Microwave Scanning Radiometer for Earth Observing System) and ASCAT (Advanced Scatterometer) soil moisture products. Zheng et al. (2016) investigated the impacts of Noah model physics on catchment-scale runoff simulations. Zeng et al. (2016) combined the in situ soil moisture networks with the classification of climate zones to produce the in situ measured soil 135 moisture climatology at the plateau scale. Zhao et al. (2018) studied the soil hydraulic and thermal properties of the 0.8 m top soil column. Zhuang et al. (2020) blended the surface soil moisture data from satellites, land data assimilation, and in-situ measurements with the constraint of in-situ data climatology, and estimated the root zone soil moisture by scaling the blended surface soil moisture product. The present research focuses on the hydrogeological and hydrogeophysical aspects, complementing previous studies. 140 Figure 2 shows the fieldwork workflow towards establishing a hydrogeological conceptual model, which is beyond the scope of this study and will be presented in a future study. The workflow includes the borehole core lithology analysis, soil thickness measurement, altitude survey, hydrogeological survey, and hydrogeophysical survey (Table 1 and Fig. 2). Yellow boxes in Fig. 2 represent the fieldwork, green boxes represent the results of fieldwork, which will finally contribute to the setup of 145 hydrogeological conceptual model. The obtained information on lithology, soil thickness, and ground surface elevation provides basic knowledge in the study area. Hydrogeological measurements of water table depth and hydraulic conductivity provide important input that can be used to deduce the direction and rate of regional groundwater flow, which is beyond the scope of this study. For hydrogeophysical results, magnetic susceptibility measurements ensure the suitability of applying MRS, while underground resistivity measurements from ERT are used in the inversion of MRS for deriving water content and 150 transmissivity. TEM provides deeper resistivity information than other methods and estimates the depth of the hydrogeological boundary. The locations of the surveys and measurements are shown in Fig. 3, Fig. 4

Borehole core lithology
The borehole core lithology is helpful in terms of understanding the formation of an area, assessing geometry of aquifers and 170 estimating hydrogeological parameters. Some boreholes were available for water table depth measurement in the study area, but information of borehole core lithology was only available in one borehole ITC_Maqu_1 (shown in Fig. 3 as Core lithology analysis) drilled in 2017 down to the depth of 32 m from the ground surface. The lithology of the cores was determined based on particle size analysis using the sieving method with mesh sizes of 60, 40, 20, 10, 5, 2, 1, 0.5, 0.25, and 0.075 mm.

Soil thickness measurement 175
In the mountainous west where transportation and setup of hydrogeophysical surveys are difficult, we auger-sampled the thickness of the overlying soils in the west (Fig. 3) to build the hydrogeological conceptual model and to validate simulations of spatially distributed soil thickness by the landscape evolution model LAPSUS (LandscApe ProcesS modeling at mUlti dimensions and scaleS) (Schoorl et al., 2006;Schoorl et al., 2002) (relevant details will be presented in another study). The fieldwork was carried out at six sites ( Fig. 4a-4f). Measurements at sites 1 and 2 were conducted in 2018, while the rest in 180 2019. Soil thickness and slope of the ground surface were measured using an auger and a clinometer from Eijkelkamp Soil & Water Company (https://en.eijkelkamp.com). The exact measurement positions at each site ( Fig. 4) were decided based on slope forms and surface pathways.

Altitude survey
The accuracy of ground surface elevation is crucial for the assessment of hydraulic heads, hydraulic gradient, and also 185 groundwater flow and its direction, therefore it is also important for groundwater modeling. The dynamic type of GPS positioning technique, GPS-RTK, is able to achieve centimeter-level accuracy in real-time, in terms of geolocation and elevation. GPS-RTK instrument CHCNAV T4 from Shanghai Huace Navigation Technology Limited (https://www.chcnav.com), with a vertical accuracy of 3 cm and a horizontal accuracy of 2 cm, was employed to measure elevations in 2019. Among the 46 ground surface elevation measurements made in total, 33 were located in the flat eastern 190 area, and 13 in the mountainous area (Fig. 3). The ground data were intended to evaluate seven DEM datasets (Table 2)   surface elevation, a hydraulic head is obtained. A set of hydraulic heads distributed over the study area can be used to determine 200 the regional groundwater piezometric map to enable a general understanding of the groundwater flow system in the study area.
We measured 40 instantaneous water table depths in 34 boreholes during 05-08 August 2018 and 20 August -05 September 2019 using a dipper (Fig. 3). For each measurement, the depth was measured 3-4 times to ensure data quality. Eight levelloggers were installed to monitor the long-term groundwater level fluctuation (see Data.xlsx in repositories), but the data are not available yet. 205

Aquifer tests
Aquifer tests, including pumping tests and slug tests, were conducted to obtain aquifer hydraulic conductivity (Fig. 3). The first pumping test was done in 2017, in the borehole ITC_Maqu_1, where core lithology information was available. The pumping rate was constant 55.6 m 3 d -1 measured with a flowmeter, and the pumping duration was about 30 minutes. The pumping rate was limited because the borehole ITC_Maqu_1 could easily collapse if the pumping rate were too high. The 210 water level became stable soon after the start of pumping and was recorded every minute using a data logger (TD-Diver manufactured by Van Essen Instruments, with a measuring range of 10 m). Other tests were carried out in 2019, including two pumping tests and eight slug tests (Fig. 3). For the two pumping tests with the constant rate of 31.6 m 3 d -1 and 101.52 m 3 d -1 , only water level recovery data were analyzed, due to practical reasons that the water pump run out of energy before reaching a stable water level. Also recovery tests were done in the eight slug tests, where in all the tests, groundwater levels were 215 abruptly lowered by extracting 11.75 litter of water. The recovering water levels were recorded every second or two seconds in the six slug tests and every five seconds or 20 seconds in pumping-recovery tests, all using a data logger (3001 Levelogger Edge manufactured by Solinst, with a range of 10 m).
The pumping test data acquired from the borehole ITC_Maqu_1 was analyzed using the Boulton (1963) method as follows: where is drawdown (m), is pumping rate (m 3 d -1 ), is transmissivity (m 2 d -1 ), is a dimensionless variable obtained by Theis curve matching. The method can be used for pumping tests performed in unconfined aquifer (isotropic or anisotropic), and for both fully or partially penetrating wells.
Slug tests data were analyzed using the Bouwer and Rice (1976) method for hydraulic conductivity in unconfined aquifer as follows: 225 In case of partially penetrating wells, where is hydraulic conductivity (m d -1 ), is the radius of the borehole casing (m), is the effective radial distance over which the head difference is dissipated (m), is radius measured from borehole center to undisturbed aquifer (m), is the 230 length of the screen (m), is time (d), ℎ 0 is the water level at time 0 (m), and ℎ is the water level at time t (m), is the length from bottom of well screen to water table (m), is the thickness of aquifer (m), and are dimensionless coefficients that can be obtained from empirical curves developed by Bouwer and Rice (1976).
Another two pumping test data were analyzed using the Boulton and Agarwal (1980) method. The Agarwal method employs the following simple data transformations for recovery test data: 235 where is the recovery drawdown (m); ℎ is the head during the recovery period (m); ℎ is the head at the end of the pumping period (m); is the recovery time (d); t is the time since pumping started (d); is the duration of pumping (d). Applying the data transformations in the Boulton method accounts for the drawdown period and allows recovery plots to be analyzed with 240 curve matching (Thorne and Newcomer, 2002). Data were processed automatically in AquiferTest software (https://www.waterloohydrogeologic.com/products/aquifertest/)

MRS
MRS was conducted to define aquifer geometry, estimate transmissivity, hydraulic conductivity and water content with depth 245 that can be used to derive storage parameters. In total, 18 soundings ( To estimate hydraulic conductivity, the decay time constant was used. There are three kinds of : longitudinal decay time constant 1 , transverse decay time constant 2 , and free induction decay time constant 2 * . With the current instrument, only 1 (actually an approximate value 1 * ) and 2 * were available. The Seevers equation (Seevers, 1966) (Eq. 6) and the Kenyon equation (Kenyon et al., 1989) (Eq. 7) were used for estimating hydraulic conductivity (m d -1 ): 255 where is the calibration coefficient, which is a lithology dependent factor that needs to be calibrated from the pumping test (dimensionless). is the MRS estimated water content (%). Compared to the Kenyon equation, Seevers equation is more accurate (Plata and Rubio, 2008) and has been widely used (e.g., Legchenko et al. (2002), Vouillamoz et al. (2007), Nielsen 260 et al. (2011)) and is used in this study. Once is estimated, the transmissivity (m 2 d -1 ) can be calculated using the equation: where ∆ is the layer thickness (m) derived from MRS inversion.
MRS data were interpreted with an open-access software Samovar V6.6 from the IRIS Instrument company (http://www.irisinstruments.com), which is based on the Tikhonov regularization method (Legchenko and Shushakov, 1998). Samovar 265 assumes the default calibration coefficient of 7E-09 for sandy aquifers and aquifers composed of weathered and highly fractured rock based on MRS calibration experience in France (Legchenko et al., 2004). In this study, was estimated using only pumping test data.

Magnetic susceptibility
The magnetic susceptibility of rocks changes the local geomagnetic field. The magnetic rocks, which lead to different gradients 270 and intensities of the geomagnetic field, result in different Larmor frequencies and further can make the MRS signal undetectable (Lubczynski and Roy, 2007;Plata and Rubio, 2007). The MRS sounding is usually not possible when the magnetic susceptibility is larger than 10 -2 SI units, but possible when it is lower than 10 -3 SI units. The sounding may be or may not be possible within the interval between 10 -2 SI and 10 -3 SI units, probably depending on the remanent magnetization of the material (Bernard, 2007). Therefore, it is always recommended to measure the magnetic susceptibility before embarking 275 on MRS survey (Roy et al., 2008). In this study, portable magnetic susceptibility meter SM-20 was used to measure the magnetic susceptibility at 11 sites in the field (Fig. 5). At each site, an average magnetic susceptibility was obtained from 3-5 repeated measurements.

ERT
Subsurface resistivity depends on many different parameters, e.g., lithology, water content, and water conductivity. Its 280 distribution in the subsurface can be visualized by 2D ERT. ERT was employed in this study because it not only provides general understanding of the aquifer but also supports the analysis of MRS measurements.
We performed seven ERT surveys with ERT instrument WGMD-9 manufactured by Chongqing Benteng Digital Control Technical Institute (http://www.cqbtsk.com.cn) in China using two standard configurations, Wenner and dipole-dipole.
Wenner usually has a good signal-to-noise ratio (S/N) and is good at detecting vertical changes in resistivity, i.e., suitable to 285 image horizontal structures. Dipole-dipole is sensitive to horizontal changes in resistivity, so it is ideal for vertical structure delineation. Multicore cables with a fixed electrode spacing of 10 m were used in the field. The length of cable was 890 m for ERT1 -ERT4, and 810 m for ERT5 -ERT7 ( cm. The industry-standard RES2DINV (Loke, 1999) was employed for ERT inversion. 290

TEM
Compared to ERT, TEM also provides subsurface resistivity but with deeper penetrations, a relatively lower resolution, and a shorter time of data acquisition. TEM instrument is usually operated in a 1D sounding mode as compared to the ERT 2D profiling mode. Since magnetic fields propagate faster in resistive media than in conductive ones, TEM is advantageous in low resistivity media and in mapping deep conductive targets. Similar to MRS but with different constraints, there is a dead time 295 between the excitation or transmitter function and the detection or receiver function, which are time-shared. Such TEM deadtime is much shorter than in the case of MRS. TEM commonly involves placing a square loop on the targeted place and performing soundings. It generates a primary magnetic field that is abruptly interrupted to produce induced eddy currents in the subsurface. The eddy currents will lead to a secondary magnetic field, which can be detected by the loop on the ground surface. The received signals can be used to estimate subsurface resistivities by using appropriate inversion techniques 300 (Nabighian and Macnae, 1991).
The TEM soundings were performed at ten locations ( (3 µs -9 µs ), stacks (5 -10), and currents (0.7 A -1.1 A) were applied to select the optimal dataset to have the maximum investigation depth. If abrupt changes occurred in the obtained curve presenting the relation between apparent specific resistivity and time, the measurement was repeated to ensure data quality. After field collection, data were processed using The core lithology of the borehole ITC_Maqu_1 is shown in Fig. 6a, 6b and 6c. The top layer is eolian sand and loam. There are dunes that have been blown out of the river bed on top of the terraces. The deep layer is fluvial sediment. Based on the lithology information, the range of lithology related parameters can be estimated. According to Chen et al. (1999), the Ruoergai 315 Basin was occupied by a large inland lake during the Quaternary before around 40 ka BP, while currently, it is a dry lake basin, with lake deposits exceeding 300 m in thickness. The extent of the ancient lake and Quaternary lake deposits is shown in Fig.   6c. Based on Fig. 6c and the log of the ITC_Maqu_1 borehole shown in Fig. 6a, it is assumed that deep sections of the Quartenary lacustrine deposits at the eastern part of the Maqu catchment are covered with the alluvial Yellow River deposits with a thickness of more than 32 m. This conclusion is generally consistent with the log of two other boreholes located to the 320 east of the study area in Ruoergai Basin, RM (33°57′, 102°21′) and RH (33°54′, 102°33′) (Fig. 6). RH is about 40 km east of the study area, with a depth of 120 m, not reaching bedrock. The top 12.4 m of coarse alluvial sediments, i.e., sands were deposited by rivers, while the deeper deposits are lake sediments, mainly composed of silt clay, clay silt, and clay (Wang et al., 1995). RM is about 20 km east of the study area, with a depth of 310 m also not reaching bedrock. Like RH, RM core also reveals thick lake sediments, with 5.5 m river deposits on the top (Xue et al., 1998). Besides, the river deposits and alluvial 325 sediments in RM and RH, located near a tributary of the Yellow River, are thinner than the fluvial sediment in borehole ITC_Maqu_1 which is located near the mainstream of the Yellow River.

Soil thickness measurement
Results of soil thickness measurements indicate that in the mountainous west, feldspathic quartzose sandstone and sandy slate parent materials show variable soil depths related to landscape position, i.e., the soil thicknesses decrease as the slope increases, and are within 1.2 m in most cases (Fig. 7a). Based on the measurements, the relationship between the soil thickness and slope 335 can be expressed using the equation: Where is the slope (°), is soil thickness (cm). Equation 9 is a regression line from data obtained over residual soils in the west. The measured thickness is a result of in-situ soil forming processes. While in the east, a transported soil is observed, the thickness of which is controlled by different processes from those acting on residual soils. In general, assuming similar geology 340 and except for the valley bottom, equation 9 would apply to the western study area (Fig. 7b).
In the west, under the soil layer, a less weathered layer exists where water can also flow and needs to be taken into account in the conceptual model. In the field, the difference between the less weathered layer and the soil layer is that the less weathered layer contains partially weathered stones. According to the owners of three boreholes located in or near the valley (numbered 32-34 in Fig. 8), their depths are larger than 10 m and do not reach bedrock. By subtracting the estimated soil thickness (Fig.  345 7b) from available depth to bedrock estimates, for example from Yan et al. (2020) and Shangguan et al. (2017), the thickness of the less weathered layer can be estimated (Fig. 8). In the mountainous west, because the estimated depth to bedrock is often at least an order of magnitude larger than the soil thickness, the uncertainty of the less weathered layer thickness mainly depends on the uncertainty of the estimated depth to bedrock, which is high due to the lack of boreholes for appropriate training (Shangguan et al., 2017;Yan et al., 2020). 350 According to the study from Yan et al. (2020), the ensemble model estimated depth to bedrock in the mountainous west ranges from 6.2 m to 70.0 m, and in the east generally within the range of 70 m to 95 m. While according to Shangguan et al. (2017), the depth to bedrock in the mountainous west ranges from 6.4 m to 35.0 m, and in the east generally within the range of 35 m to 74 m. Both studies appear to underestimate depth to bedrock in the east which is also due to the lack of boreholes for proper constraining. Though with high uncertainty, the calculated less weathered layer thickness in the west can be applied for 355 characterizing the hydrogeological conceptual model, since the west mainly plays a role in collecting water and there is no other data available to determine the thickness of the less weathered layer.

Water table depth measurement
For the altitude survey, ALOS RT1, with a spatial resolution of 12.5 m, performed better than other DEM products across the 365 whole study area and had a higher resolution than the others. It was the most suitable DEM to be used in this study area for determining water table (WT) depths. For details, see Appendix A1.
There were 22 WT depths measured in 2018, and 18 in 2019 (Fig. 3). In the flat eastern area, the WT depths were interpolated with the software Surfer (https://www.goldensoftware.com/products/surfer) using the default Ordinary Kriging method with the linear variogram model (slope=1, anisotropy ratio=1, anisotropy angle=0) (Cressie, 1990(Cressie, , 1991, which provides reasonable 370 grids in most circumstances. Owing to the fact that most people living in the mountainous west use water from streams (via field survey), the need for groundwater is low, and only few boreholes exist. As such, only three boreholes numbered 32-34 were found in that western area (Fig. 9) and WT depths were measured. Normally, a good interpolation of WT depths or piezometric head over a large area needs and uses every measurement. But in this case, a reasonable WT depth map or piezometric head map in the mountainous west will need more than 100 borehole measurements (Hopkins and Anderson, 375 2016), because the ground surface elevation changes dramatically in the west and so does the groundwater level. The three boreholes are far from enough to provide a reasonable WT depth map or piezometric head map, and, therefore, were excluded from the interpolation. In contrast, the measured groundwater depths (and the interpolation) in the eastern study area can give a reasonable WT depth map or piezometric head map ( Fig. 9a and Fig. 9b). Among the 34 boreholes with WT depth measurements, ground surface elevations measured by GPS-RTK were only available for 13 boreholes. For the rest 21 380 boreholes, ground surface elevations were extracted from ALOS RT1. These elevations were used to derive hydraulic heads by subtracting the WT depths from the ground surface elevation. Similarly, using the Kriging method, hydraulic heads were interpolated to obtain piezometric maps in the flat east ( Fig. 9c and Fig. 9d).
In general, the range of WT depth was between 0.  Fig. 9c and Fig. 9d) decrease from the middle of the study area to the eastern boundary, meaning that the groundwater flow is from the west to the east with the hydraulic gradient of about 0.002 (dimensionless), 390 recharging the Yellow River (Fig. 9f). This is consistent with the conclusion from Chang (2009). Ground surface elevations in Fig. 9f were extracted from ALOS RT1, and hydraulic heads were extracted from Fig. 9c and Fig. 9d. Some hydraulic heads are higher than the ground surface elevations as shown in Fig. 9f, which is due to: 1) the accuracy of ALOS RT1; 2) the lack of control points of hydraulic heads.  Fig.9c and fig. 9d. Numbers from 1 to 34 are identification numbers of boreholes.

Aquifer tests
There were 11 aquifer tests conducted (Fig. 10, Fig. 3) in unconfined aquifers, in partially penetrating boreholes. Eight slug tests were done in boreholes numbered 16,21,24,26,27,32,33,34 (Fig. 9), two pumping tests were carried out in 2019 at boreholes 6 and 23 (Fig. 9), and one pumping test was carried out in 2017 at the borehole ITC_Maqu_1. Data were processed automatically in AquiferTest software (https://www.waterloohydrogeologic.com/products/aquifertest/) with assumptions 410 made considering the average conditions in the study area: aquifer is unconfined and 35 m thick; borehole is partially penetrating; screen radius is 0.27 m; screen length is 15 m; the distance from aquifer top to screen bottom is 15 m; casing radius is 0.25 m; borehole radius is 0.3 m. Considering the spatial heterogeneity of hydraulic conductivities, they were not interpolated in the study area. As a result, the hydraulic conductivities ranged from 0.1 m d -1 to 15.6 m d -1 (Fig. 10). According to Healy et al. (2007), this range of hydraulic conductivity can be classified as of fine silty sand to coarse clean sand. The slug 415 test in the borehole 24 provided lower estimates of hydraulic conductivity than the nearby pumping test in borehole ITC_Maqu_1, which likely underestimated the hydraulic conductivity as borehole 24 was not used for a period of time.
Therefore, compared to the slug test, the hydraulic conductivity obtained from the pumping test is more accurate and is a volumetric average, which makes it more suitable to calibrate , because MRS results are also volumetric averages.

MRS
ERT results were used to establish geoelectrical models for MRS inversion (see Appendix A3). The magnetic susceptibility measurements reveal very low susceptibility in the catchment, ensuring the suitability of applying MRS in the study area (see 425 Appendix A2).
We didn't use default inversion parameters because they sometimes result in abrupt changes or discontinuities of water content at two adjacent MRS sounding sites, and lead to unrealistic results (Shah et al., 2008). Some excitations were excluded during inversion based on S/N and the mismatch in terms of amplitude, Larmor frequency, and phase. The inversion parameters are listed in Table A7. The temperature of the water leads to different water densities and viscosities, and influences therefore also 430 hydraulic parameters. In Samovar V6.6, a default temperature of 20 ℃ is used. But in the study area, the average groundwater temperature is 6.2 ℃. Therefore, it was necessary to take the true groundwater temperature into account when estimating hydraulic parameters. Thus, by correcting gravitational acceleration, water densities, and water viscosities using eq. 10, a correction factor of 0.69 was used during the inversion process to improve accuracy.
Where is hydraulic conductivity (m d -1 ), is the permeability of porous media (m 2 ), is water density (kg m -3 ), is the gravitational acceleration (m s -2 ), and is water viscosity (Pa· s).
MRS3-1 sounding (Fig. 5) was used to calculate the calibration coefficient , because it is the nearest MRS sounding to the borehole ITC_Maqu_1 for which pumping test data is available. Using a single point of calibration, the calibration coefficient can be estimated with the uncertainty ≤ 150% (Boucher et al., 2009). The calibrated is 8.78E-09 for T1 and 8.13E-9 for 440 T2 * . Figure 11 shows inversion results of water content and T1 of MRS2-1, MRS3-1, and MRS3-2, and complete results are shown in Fig. A3 in the Appendix. Details are available in MRS inversion results.xlsx at the National Tibetan Plateau Data Center, including T1, T2 * , water content, T1 and T2 * derived hydraulic conductivities KT1, KT2* and transmissivities TT1, TT2*.
The water content distribution of MRS9-2, MRS7-2, MRS7-1, and MRS4-1 (Fig. A3) extends down to 150 m deep. Except for MRS4-1, soundings MRS9-2, MRS7-2, and MRS7-1 are adjacent, indicating that in the southeast, near the Yellow River, 445 the groundwater extends to more than 150 m depth. So it is concluded that the flat east plays the main role in storing groundwater and the groundwater can extend to more than 150 m depth.
limiting values of 0.00 ms and 1000.00 ms for T2 * and 0.00 ms and 3000.00 ms for T1 are indicators that a valid numerical solution to the measured records (i.e., the inversion) was not reached and no valid outcome is available. Except for invalid values, T1 derived hydraulic conductivity (KT1) ranges from 0.00 m d -1 to 210.98 m d -1 , T2 * derived hydraulic conductivity 450 (KT2*) ranges from 0.00 m d -1 to 19.64 m d -1 . The value of 0.00 m d -1 comes from the estimation of very low water content.
Here, an order of magnitude difference is observed between the range of KT1 and KT2*, which is due to the big difference between T1 and T2 * . In theory, T1 is less affected by magnetic heterogeneities, thus permits a better estimation of the hydraulic conductivity compared to T2 * . However, it is to note that no magnetic disturbance is expected in Maqu catchment (Fig. A1).
Furthermore in the case of T1, because two timed delayed responses are compounded, any model mismatch, e.g. the MRS loop 455 sampled volume being significantly different from a layered model parallel to the loop due to near-source river deposit media heterogeneity, can make the measured responses 'doubly' distorted and may not fit a T1 expected response. In both cases, T1 and T2 * , a distortion is occurring. Nevertheless, according to specific circumstances, T2 * , which is evaluated from rest with a single pulse, may undergo less severe overall distortion. So KT2* and TT2* tend to be more reliable than KT1 and TT1, and should be used for future study. By checking the values of KT2*, it is concluded that there is an unconfined aquifer in the eastern study 460 area. Based on KT1 (and water content results), with a proper threshold to define aquifer and non-aquifer, the aquifer geometry can be defined.
MRS has its own limitations in that some of the in-situ water information is missing, and that the current 'window of the technique' is only sensitive to the larger pore fraction of water content. Near-source river environment leads to the unknown mixture of varied lithology. Missing water is unknown, but accounting for a variety of lithology, including fine pore ones, 465 from water table depth (Fig. 9) to the base of the aquifer (50 to 208 m range, see the following TEM section) may lead to well over 50% missed water (Boucher et al., 2011). Besides, the invalid values for T2* and T1 may be attributed to the hydrogeological conditions, such as highly heterogeneous lithology or too low signal/noise ratio, and may be eased using an updated version of Samovar V6.6, such as Samovar V11.4, which not only improves the capability of signal analysis, for example, allows optimizing the number of inverted layers, but also adds uncertainty estimation function by incorporating 470 singular value decomposition. Nevertheless, in highly heterogeneous environments, the indetermination of some parameters may remain with current technology. In terms of using default inversion parameters, part of the difficulty is in fitting the observed data to a too large number of layers: i.e. partly fitting to the noise component of the records. The heterogeneity of the near-source river environment is also contributing to this difficulty. With more recent tools, like Samovar V11.4, the difficulty can be better handled (Legchenko et al., 2017). 475 In general, MRS provides preliminary and valuable information on water content, hydraulic conductivity, and transmissivity.
Once the geometrical mapping and its fabric have been mapped, groundwater flow parameters, and groundwater storage or volume can be better determined.

TEM 480
A variety of acquisition parameters were applied and the signals were analyzed to determine the optimal TEM measurements for each site. Detailed information of ten optimal TEM measurements and inversion parameters are listed in Table 3 and Table   A8, respectively. The industrial noise filter was set at 50 Hz, and the amplifier was off. In the study area, using the square loop with a side length of 48 m or 95 m, the maximum time of 1 ms or 4 ms, stack between 5 -10, and current of 0.8 A or 1.1 A, the TEM method can reach the maximum investigation depth ranging from 150 m to more than 1000 m. For data processing, 485 the invalid data points were first removed, and the field data were smoothed, then the voltage data were inverted to obtain an apparent resistivity, which was further inverted to get true resistivity (Abiye and Haile, 2008;Cosentino et al., 2007). Appling a variety of acquisition parameters ensures that bad data can be revealed (Auken et al., 2006). Induced polarization (IP) and superparamagnetic (SPM) effects were not considered in the inversion process (Macnae, 2016). Because of the dead time and the fact that at most sites, a relatively dry layer of sediments exists near the ground surface with a corresponding high resistivity 490 depth interval, the upper 15 m to 30 m of the sounding is lost, although subsequent layered Earth modeling attempts filling the gap.
The RMS error of the inversion results shown in Fig. 12 is below 2% in the flat area and below 10% in the mountainous area.
The results in the mountainous area, i.e., results of TEM6, TEM7, and TEM8, indicate that the resistivity becomes larger in the deep subsurface, and is consistent with our understanding that the bedrock is located at relatively shallow depth from the 495 ground surface. The maximum investigation depth of TEM6 is shallow, only ten time windows were available and resulted in about 150 m investigation depth from the ground surface. This may be due to the local unknown geological condition. In addition to consolidated rock resistivity of the order of 2 kΩ•m to 4 kΩ•m, TEM7 and TEM8 responses may show instances of fracturing, weathering or faulting. To determine exactly what structure it is, and the scope of the structure, further investigation is needed. For example, a systematic high spatial resolution geophysical survey with appropriate depth capability, such as the 500 airborne electromagnetic survey, followed by systematic borehole drilling. The rest of the TEM measurements are scattered in the east where it is likely that lake deposits are covered by river deposits on the top. Because the clay silt lithology has a lower resistivity than sand-rich lithology, and Chen et al. (1999) suggested that the ancient lake in Ruoergai Basin was a freshwater or slightly saline lake for most of its life, the decrease of resistivity may indicate the change from river deposits to lake deposits. Table 4 listed the TEM derived depth of river deposits bottom in the east. For TEM0, TEM1, TEM2, TEM3, 505 TEM4, TEM9, the bottom of river deposits are deeper than 100 m, with lake deposits underneath. But for TEM5, located near the mountainous area, the bottom of river deposits is at 50 m deep, followed by 64 m thick lake deposits, with the bedrock down most, and the nearest MRS sounding MRS6-1 indeed shows that there is no free water under 50 m depth.
The large contact area between river and lake deposits in the eastern study area and the Ruoergai basin allows large water exchange. According to Fig. 12, the thickness of the lake deposits can reach more than 150 m. The thick lake deposits constitute 510 a nearly infinite hydraulic buffer reservoir allowing multi-years storage equivalent in Maqu catchment. Thus it would be relevant to include the deep lake deposits in future hydrogeological models.

Uncertainties
As shown in Fig. 2, direct techniques, i.e. particle size analysis, altitude survey, soil thickness measurement, water table depth measurement, aquifer test, and magnetic susceptibility measurements have low uncertainties. There are random errors for 520 particle size analysis (Wang, 2011), but they are small and not expected to affect the final lithology result (ASTM, 2017), and thus can be neglected. For measured ground surface elevations, soil thicknesses, and water table depths, the uncertainty is supposed to be within a few centimeters considering the accuracy of equipment and errors during the measurement process (Burt, 2014;Cunningham and Schalk, 2011;Rydlund Jr and Densmore, 2012). In terms of hydraulic heads derived from ALOS RT1 in boreholes, the uncertainty not only comes from water table depths measurement, but also from ALOS RT1 which 525 contains the mean absolute error of 4.4 m in the study area based on our results (Table A1). For hydraulic conductivities obtained from aquifer tests, the uncertainty mainly comes from data collection and processing. Though the duration of pumping in the borehole ITC_Maqu_1 did not reach 48 hours, the water levels became steady very soon after the pumping started, so the uncertainty is estimated to be within 25% according to studies from Brown et al. (1995) and Delnaz et al. (2019). For magnetic susceptibility, although the resolution of SM-20 is 1e -6 SI units, the actual reading accuracy is dependant on 530 appropriate corrections, e.g. temperature, shape, volume, effective distance to sensor, etc. The corrections may reach a few orders of magnitudes for volume and up to ± ~50% for shape (Hoffman, 2006). In the case of Maqu catchment, these are so far from the levels at which MRS problem may occur that, corrected or not, the final results will still be below the threshold for concerns. In terms of indirect techniques, ERT, MRS, and TEM, performances of the raw data could be evaluated with parameters such 535 as S/N for MRS and amount of bad data for ERT and TEM. Knowledge of the subsurface geometry and fabric would lead to the resolution of the main uncertainty issues for inverted data. Because there are implicit modeling assumptions for each method. For example, the assumption for MRS is that the subsurface is made of 1D planar layers parallel to the MRS loop with depth-increasing thickness. We cannot quantify to what extent these assumptions are met, and therefore also to what extent the inversion data are accurate measurements of the site's hydrogeological parameters, thus appropriate uncertainty 540 figures cannot be reliably generated for inverted data. The inverted data, as an illustration of what can be extracted from the raw data, are preliminary results with only inversion RMS errors quantified (ERT and TEM). Lake deposits, being far from the source, should not suffer from the near-source river deposits heterogeneity, but its lithology makes it insensitive to MRS.
To further reduce uncertainties from indirect techniques, ERT, MRS, and TEM, it is important to determine subsurface geometry and its fabric. State-of-the-art airborne electromagnetic technology allows high spatial resolution mapping down to 545 500 m depth and is probably the most appropriate tool for now (Legault, 2015). After the site geometry is properly mapped and the subsurface fabric is properly understood, optimum borehole drilling locations can be selected. When the detailed geometrical mapping of the subsurface and systematic borehole information are available, the inversion process can be better constrained and improved (Galazoulas et al., 2015;Vouillamoz et al., 2005;Wang et al., 2021).

Data availability 550
The raw dataset is archived and freely available in the DANS repository under the link https://doi.org/10.17026/dans-z6t-zpn7 (Li et al., 2020a), and the dataset containing the processed ERT, MRS, and TEM data is available at the National Tibetan Plateau Data Center with the link https://doi.org/10.11888/Hydro.tpdc.271221 (Li et al., 2020b).

Conclusion
We conducted borehole core lithology analysis, altitude survey, soil thickness measurement, hydrogeological survey, and 555 hydrogeophysical survey in the Maqu catchment of the Yellow River source region on the Tibetan Plateau, where little subsurface data are available. Seven DEMs were evaluated using GPS-RTK measured ground surface elevations, and ALOS RT1 is suggested to be used in this study because of its better accuracy as well as higher resolution than other DEMs. The medium-deep lithology of subsurface down to 32 m below the ground surface, mainly composed of sand, is available from one borehole ITC_Maqu_1. It helps provide useful information for the cross-validation with hydrogeophysical survey results. 560 According to the in-situ measurements, soil thicknesses, except on the valley floor, are within 1.2 m depth in most cases in the west, and the soil thickness decreases as the slope increases. The hydrogeological survey reveals that groundwater flows from the west to the east, recharging the Yellow River, and the hydraulic conductivity ranges from 0.2 m d -1 to 12.4 m d -1 . The hydrogeophysical survey demonstrates the presence of an unconfined aquifer in the east. The water content and hydraulic parameters of that aquifer were estimated at MRS sounding locations. The depth of the Yellow River deposits was derived at 565 TEM sounding positions in the flat eastern area.
Although water table depths were only measured once or twice, and hydrogeophysical methods, like ERT, TEM, and MRS, have inherent non-uniqueness problems during the inversion process, they all provide valuable information, especially in such data-scarce areas as TP. The data from direct techniques with low uncertainties, as shown in Fig. 2: lithology, ground surface elevation, soil thickness, water table depth measurement, hydraulic conductivity from aquifer test, magnetic susceptibility, can 570 contribute to related global or regional databases where the in-situ data over TP is scarce, or be regarded as verification and validation data for groundwater modeling over Maqu catchment. The data from indirect techniques, ERT, MRS, and TEM, is a rare unique and particularly rich training data source for geoscientists interested in the data processing and interpretation of the particular hydrogeological and hydrogeophysical techniques used here. It is a dynamic set where additional complementary data will gradually add constraints to the inversion processes. For example, a researcher developing new techniques for S/N 575 improvements of some of these techniques will get free and highly relevant data to work with.
In Maqu catchment, a near-source river environment, without adequate geometrical mapping, the representativity of the various sampled volumes is unknown as well as whether the sampled volume fits the models used for data inversion. This is much less the case further away from the source where homogeneity and fitting of the model to the actual hydrogeological setting is achieved. In such away-from-source case, pumping tests data may be assumed to be representative of the tested formation 580 while in techniques such as MRS, depth and thickness information may be extracted from the datasets as well as hydraulic estimates. This is an ongoing project and it may become available later if such above-mentioned mapping is completed. Any further similar surveys and borehole drillings would benefit from such geometrical mapping since their precise localization may then be optimized in view of proper data inversion and information gaps filling.
Generally speaking, all the methods work well in the study area, and have confirmed the presence of an unconfined fluvial 585 aquifer within the 250 m below surface and the presence of lake deposits with much finer pores lithology. By combining our dataset with available depth to bedrock datasets, a preliminary hydrogeological conceptual model can be established. If combining our dataset with detailed geometrical mapping of the subsurface and deep borehole information, a more complete and accurate conceptual model can be obtained. Furthermore, we will be monitoring the groundwater and surface water in the study area and aim for establishing a long-term monitoring network, which will eventually contribute to the verification and 590 validation of future studies on groundwater modeling over the Maqu catchment. To our knowledge, this is the first time such detailed surveys were conducted in a TP catchment in order to set up a hydrogeological conceptual and numerical groundwater model. This paper is expected to contribute not only to the hydrogeological conceptual and numerical model of the Maqu catchment at the TP, but also to provide data for hydrogeological and hydrogeophysical communities, and promote interdisciplinary research. 595 Appendix A A1 Altitude survey 46 ground surface elevations were measured (33 in the flat east, 13 in the mountainous west), and were used to evaluate the accuracies of seven DEM datasets (data available in DEM.xlsx in the National Tibetan Plateau Data Center) and the most accurate one was applied in this study. The statistical analysis results of the seven DEMs in the study area are shown in Table  600 A1. The root mean squared error (RMSE) of ALOS RT1 and ALOS RT2 are 5.695 m and 5.477 m, respectively, much smaller than the RMSE of the other five DEMs. The correlation coefficient, the mean error, and the mean absolute errors of ALOS RT1 and ALOS RT2 also show better performance than those of the other five DEMs. Comparing ALOS RT1 with ALOS RT2, ALOS RT1 slightly outperforms ALOS RT2 with regards to RMSE, correlation coefficient, and the mean error. Table   A2 and Table A3 list the statistical analysis results of seven DEMs, separately for the flat eastern area and the mountainous 605 western area. Seven DEMs, all behave better in the west than the east in terms of the correlation coefficient. In the west, the correlation coefficients of seven DEMs are all larger than 0.94, while in the east, the correlation coefficients are all lower than 0.24. This is because the range of ground surface elevation in the flat east is much smaller than the range of elevation in the mountainous west. With regard to the RMSE, mean error, and mean absolute error, all seven DEMs have better behavior in the east than in the west. In general, ALOS RT1 and ALOS RT2 also outperform the other five DEMs, according to Table A2  610   and Table A3.
Since ALOS RT1 performs slightly better than ALOS RT2 in the whole study area and has a higher resolution than ALOS  Fujita et al. (2008) found that the elevation differences between DEMs and ground survey data from differential GPS were 11.0 m for ASTER and 11.3 m for SRTM in the Lunana region, Bhutan Himalaya. The DEM evaluation results also indicated that in different places over the TP, the satellite DEM estimates are 625 acquired with varying accuracy. This may be due to different topographic complexity in different areas.
The DEMs' quality can be influenced by several factors, such as sensor type, algorithm, terrain type, and grid spacing. (Hebeler and Purves, 2009). In this study, grid spacings of DEMs are similar except for ALOS RT1, so the main factors that affect the accuracy of the DEMs should be sensor types and algorithms. For SRTM, the issue inherent to the production method is mast oscillations, while for ASTER and AW3D30, the issue is scene mismatch (Grohmann, 2018). As for radiometrically terrain 630 corrected (RTC) products ALOS RT1 and ALOS RT2, the quality is directly related to the quality of the source DEM SRTM which was used in the RTC process. This results in very similar correlation coefficients of SRTM, ALOS RT1, and ALOS RT2, and obvious improvements in RMSE, MAE, and ME (Table A1). The magnetic susceptibility measurements (Fig. A1) reveal very low susceptibility in the catchment with susceptibility values, all smaller than 1×10 -5 SI units with an average of 3×10 -6 SI units. A previous study from Chen et al. (1999) also reported low magnetic susceptibility of the RH core ( Fig. 6c) with 120 m length located 40 km east of the study area in Ruoergai Basin.
The low magnetic susceptibility ensured the suitability of applying MRS in the study area. 645 Figure A1. Magnetic susceptibility measurements (10 -3 SI Units) ensured the suitability of applying MRS in the study area.

A3 ERT
Detailed information on ERT profiles and inversion parameters are listed in Table A4 and Table A5, respectively. For a specific pseudo depth, the values between adjacent points generally vary smoothly. Bad data points can be easily identified as they appeared as outlier points in the pseudosection plot in RES2DINV due to their too high/low apparent resistivity values. The 650 bad data points were filtered out based on the following criteria: (i) having negative apparent resistivity or small apparent resistivity close to 0 Ω m; (ii) having negative/positive pulse amplitude ratios < 0.75 or > 1.33 (a measure of waveform symmetry) (Slater et al., 2010;Wilkinson et al., 2016). After filtering, the least square method was used for the inversion.
Results are shown in Fig. A2, with the root mean square (RMS) error less than 5%. A pattern of roughly regular parallel to surface electrostratigraphy is observed in all ERT profiles, except 0 m -310 m of profile ERT5, where the pattern is dipping 655 relatively to surface. This means that strata are likely to be stratified in most parts of the study area. For ERT2, ERT3, ERT5, and ERT6, three electrostratigraphic layers can be identified: the first layer with the highest resistivity, the second layer with the lowest resistivity, and the third layer with a medium resistivity. The second layer is likely to represent an aquifer. However, considering ERT4 and ERT7, there is a lack of marker electrostratum, i.e., a layer with high resistivity does not exist at the ground surface. This is probably due to high water content near the ground surface in the mountainous area where ERT4 and 660 ERT7 were located. As for ERT1, rainfall occurred during the field measurement. Rainwater accumulations occurred next to some of the electrodes, causing abnormal current distribution during the ERT measurements and about half of the data were filtered out. The ERT1 inversion results show a threelayers pattern similar to the one observed along the ERT2, ERT3, ERT5, and ERT6 profiles. One or more short wavelength anomalies (< 200 m) are observed along all profiles but particularly in the case of ERT1, ERT3 and ERT6. Short wavelength anomalies along ERT1 may be due to data acquisition made during 665 rainfall, while in the case of the other profiles, localized changes in water content or lithology variations are suspected.
Compared to the Dipole-Dipole configuration, the investigation depth of the Wenner configuration is deeper. So resistivity values obtained from Wenner configuration were used to establish geoelectrical models for MRS inversion. For ERT2, ERT3, ERT5, and ERT6, three-layer geoelectrical models were extracted, while for ERT4 and ERT7, two-layer geoelectrical models were extracted. ERT1 was neglected due to the influence of rainfall. For ERT5, from 0 m to 310 m, there's a topographic 670 change, the ground surface elevation decreases from 3395 m a.s.l and stabilizes at around 3390 m a.s.l. Ground surface with low resistivity exists along this 310 m transect. Since the MRS soundings were conducted in flat areas, so only resistivity from 310 m to 810 m was used for the first layer of the geoelectrical model. The geoelectrical models and corresponding MRS measurements are shown in Table A6. The depths of the last layer of geoelectrical models are extended to 1.5 times of the MRS investigation depth since signal distortion due to subsurface resistivity is calculated down to that depth while making the 675 MRS linear filter. In this particular version, MRS investigation depth was considered to be the MRS loop size, i.e., 150 m and 100 m. Nevertheless, like other geophysical methods, ERT has equivalence problems, i.e., non-uniqueness of inversion results.
Despite equivalence problems, the ERT method still provides important subsurface information in Maqu catchment where we have little fundamental information. This is a very first investigation in this area, when more lithology information becomes available later, ERT can be better constrained to reflect the subsurface lithology. 680 The thickness of the first layer 0.5 m Factor to increase thickness layer with depth 1.       Table  YRSR Yellow River Source Region