Alluvial aquifer thickness and bedrock structure delineation by electromagnetic methods in the highlands of Bolivia

The Challapampa aquifer system belongs to the Poopo enclosed basin, located to the west side of the Cordillera Oriental (eastern mountain range of the Andes), where the highest peaks reach 4600 m above sea level (masl) and the lowest and planar land is at 3700 masl (Fig. 1). In the Cordillera Oriental, the direction of the ridges is from northwest to southeast, which also defines the surrounding landscape and the bedrock relief below the porous aquifers.

The climate in the area is semi-arid, with scarce to non-existing vegetation. The average temperature is about 10 °C. Regarding precipitation, there is a big difference between summer and winter; about 80% of the mean annual 350 mm occurs from December to March (summer). In addition, the annual potential evaporation was estimated to be 1800 mm (SENAMHI 2015), evidencing the aridity of the region. During summer, intense precipitations create floods in the plateau, but the water is quickly evaporated leaving salt in the clayey soils. The land around the study area is mainly used for small-scale agriculture during the rainy season. Sporadic crops can be seen during the rest of the year in sites nearby the few wells used for irrigation in the vicinity. The alluvial fan of River Paria hosts a well field which abstracts water continuously, at a rate of about 300 l/s (SELA 2017), to supply Oruro. This situation has been affecting the piezometric levels in the irrigation wells and created social conflicts due to the limited access to water.

The Challapampa aquifer system is a generic name to refer all the geological formations yielding water in the region. The porous aquifers in the first ~ 100 m below the surface (mbs) are the main groundwater reservoirs near Oruro, characterized by the variety of the existing geological formations and the complexity of settings. The geological formations in the region can be divided into consolidated (rock) and unconsolidated (sediments). Among the first unit of rocks, sedimentary and metamorphic types are the most common in the bedrock (in purple in Fig. 2); also, igneous rocks are present in some volcanic intrusions (in orange in Fig. 2). The unconsolidated unit comprises sediments originated by erosional and depositional processes (in yellow in Fig. 2).

The consolidated rocks around and beneath the study area have folds and faults with northwest–southeast direction. Three Silurian formations have been identified in the bedrock and mountains (in purple in Fig. 2) where sandstones and siltstones are common. These formations are, from the oldest to the youngest, Llallagua, Uncia and Catavi (Suarez Soruco 2000). The Uncia and Catavi formations are exposed nearby the study site and according to the geological information, Uncia’s bedrock underlies the Quaternary sediments. The bedrock might be comprised of green shales with beds of green sandstones and siltstones (GEOBOL and Swedish Geological 1992). Banks et al. (2002) include in their study a bedrock contour map (red lines in Fig. 2), based on drilling logs and geophysical tests like seismic refraction and vertical electrical soundings (VES), which in turn correspond to other previous studies not found in the present investigation. The deepest contact between sediments and bedrock might be located at 3540 masl (~ 160 mbs). Volcanic igneous rocks from the Tertiary are also part of the consolidated geological units; they are the so-called Oruro complex and Escalera volcanics. The first one, close to Oruro city, holds valuable minerals like silver, lead and others (Patureau 2007; Ramos et al. 2011). The second one, to the east of the study area, comprises rhyodacitic and porphyritic lavas (GEOBOL and Swedish Geological 1992). This intrusion might have a strong influence on the regional geothermal activity.

When it comes to the unconsolidated geological units, alluvial Quaternary sediments are the most extensive covering the relief of the bedrock and shaping the flat terrains. The grains in these deposits are mainly comprised of quartz and other silicates. They are largely variable in terms of grain size as well; pebbles, gravels, sands, silts and clays are all chaotically arranged because of the overlapping depositional processes. The well-rounded shape of pebbles and gravels lead to infer that they were transported long distances from the surrounding mountains by strong fluvial and glacial events. The finest sediments might correspond to lacustrine depositions, such is the case of the surficial clay layer (Rigsby et al. 2005). Figure 3 shows an example of how these sediments are arranged in a pit, in the first ~ 12 mbs. The thickness of the unconsolidated sediments of the porous aquifer may be from tenths to hundreds of meters, and the thickness of the surficial clay layer may be from tenths of centimeters to a couple of meters.

The most productive aquifers are the fluvial–lacustrine sediments of diverse thicknesses and settings. A few drilling logs, VES and debris excavations constitute the main sources of stratigraphic information in the study area, where the high vertical and lateral variability between different types of sediments is standing out, as shown in Fig. 3.

The Challapampa aquifer system has been studied at some drilling sites to obtain information to extract as much water as possible. However, most of that information was never properly saved and reported, and just a few investigations are available, e.g., Larsson (2016). The rest is partially or completely lost. In the study area, the geometry of the aquifer system was proposed by Banks et al. (2002) based on VES, seismic refraction and drilling logs (Dames & Moore Norge 2000; SCIDE et al. 1996). Unfortunately, just part of the information about stratigraphy is available. There are some drilling protocols around the well field area describing the stratigraphy of the first 100 mbs. These descriptions refer to the content of gravel, sand and clay at different levels. The stratigraphy obtained from those drilling reports is punctual and representative of few meters around the boreholes and not applicable to the rest of the alluvial fan, because of the high lateral and vertical variability in the arrangement of sediments. When it comes to the hydraulic conductivity, this variability ranges from 1.2E−5 to 5.8E−4 m/s in the unconsolidated sediments (Banks et al. 2002). Going deeper, the top, fractured and weathered part of the bedrock has values around K = 1.0E−5 m/s and for the consolidated sedimentary rock, it is about K = 3.0E−8 m/s (Dames & Moore Norge 2000).

According to the regional groundwater gradient (equipotential lines in Fig. 1), the flow seems to run from northeast to southwest in the porous aquifer. Precipitation in the mountains forms streams and rivers running towards the plane, which recharge the aquifers and then flows towards the lowest site of the region, which might be the Lake Poopo, about 20 km south of the study area. However, the annual volumes of water extracted from the well field, about 9.5 million m³ in 2016 (SELA 2017), have changed the natural flow direction and now it goes towards the central part of the alluvial fan, creating a cone of depression with a radius of influence of about 5 km around the well field (Banks et al. 2002).

In addition, hydrochemical and isotopic characteristics were used to propose four flow systems in the middle of the alluvial fan (Gómez et al. 2016), which are summarized in Table 1.

Geothermal activity also exists in the aquifer system, specifically to the east of the alluvial fan, in the site called Kapachos (Fig. 1). In this place, the water coming out from springs is hot and rich in total dissolved solids (high salinity), and the average temperature in the wells extracting from the porous aquifer is 15 °C and in the hot springs is ~ 50 °C. Salinity in terms of electrical conductivity (EC) in the wells extracting from the porous aquifers averages 1.0 mS/cm and in the hot springs is ~ 3.6 mS/cm. There are also boreholes, further away the present study site with thermal influence. For instance, in W10 in the wellfield (Fig. 1), the temperature is 30 °C and EC is 3.5 mS/cm. Flow paths and processes controlling the occurrence of the thermal water in the region remain unknown subjects out of the scope of this study.

TEM is a point-based geophysical prospecting technique where a current flowing in a transmitter antenna generates a primary electromagnetic (EM) field. When that current is cut off, that primary EM field induces eddy currents flowing through the ground. Those eddy currents generate a secondary magnetic field detected by a receiver antenna. The receiver acquires time-dependent decaying voltage signals. During processing, those signals are transformed into electrical resistivity and vary as a function of the type of materials and how they are distributed in the ground (Christiansen et al. 2009; Reynolds 2011).

Water content and the type of water filling voids in the geological formations also determine the resistivity obtained by EM methods. Therefore, it is expected to find a distinction in terms of resistivity between saturated sediments (main aquifers) and the bedrock (see Table 1) to establish the thickness of the saturated alluvial sediments.

The equipment used for the TEM measurements was an ABEM WalkTEM, connected to two batteries of 12 V transmitting dual moment electrical pulses: about 18 A during the high moment (HM) and about 2 A during the low moment (LM). An offset array was used during the tests, where a transmitter loop (50 × 50 m cable AWG #12) is placed next to the equipment. In the centre of the transmitter, the RC-5 receiver (0.5 × 0.5 m) is placed and the RC-200 receiver (10 × 10 m) is 50 m offset from the centre of the transmitter. Although both receivers were able to register both low and high moments, the RC-5 was used for measuring LM for better shallow resolution, and the RC-200 was used for measuring HM for better depth resolution.

Raw TEM data contains sorted information of HM, LM and noise in specific channels for both receivers. Before the inversion, HM channels corresponding to RC-5 and LM corresponding to RC-200 were disabled. Thus, the shallow part of the models was resolved with the RC-5 LM, and the deepest part with the RC-200 HM. Typical data curves, combining LM and HM, need to be trimmed in some specific parts like before intersecting with the noise curves, at the begging of HM and at the end of LM, where they are not overlapping each other within an acceptable displacement defined by error bars, and at any sudden fluctuation out of the shift of the curve.

After conducting measurements and conditioning the TEM data, the next step is to find a model that might reproduce the obtained data and be representative of the resistivity distribution in the test site. The process to find that model is called inverse numerical modeling, and for TEM it is based on curve matching aided by mathematical relationships and algorithms to solve an inverse scattering problem (Nabighian 1991). Data residual is an indicator of how small the differences are between the measured and the modeled data after a curve matching process. Two types of models are used to match the data curves; a layered model, where mean resistivity values correspond to a few layers, and a smooth model where gradual transitions in resistivity correspond to a fixed number of layers increasing in thickness by depth (Christiansen et al. 2009). The limit at which the resistivity structure of a model is reliable is the depth of investigation (DOI), below that limit the geological interpretations might not be true (Christiansen and Auken 2012). Likewise, the method has a minimum depth of resolution depending on the earliest acquisition times and shallow resistivity, which means that the top part of subsoil cannot be resolved with TEM (Spies 1989).

The inverse numerical modeling requires a lot of computing capacity for TEM, so accurate electronic components are also needed to get signals from a wide dynamic range. These issues became achievable during the last decades, making this technique relatively young in groundwater mapping applications (Christiansen et al. 2009). The TEM data were processed with SPIA (Aarhus GeoSoftware 2017a), which runs one-dimensional inversions of individual soundings. Finally, the creation of maps and cross-sections based on inverted TEM data was done with Aarhus Workbench (Aarhus GeoSoftware 2017b), which carry out interpolations with the closest one-dimensional models by the Kriging method.

Resistivity, as a physical property of geological formations was also evaluated with the ERT technique. In principle, the resistivity distribution in the ground corresponding to the same site from TEM and ERT is expected to be similar, although each method has its own characteristics making them sensitive to external factors (3-D effects, DOI, high conductive structures and others). For this study, resistivity obtained from TEM and ERT is assumed equally valid. A thorough analysis of the differences between both techniques is out of the limits of this study. Complementary ERT measurements were performed to obtain two-dimensional resistivity profiles. This technique is based on the DC methods in which the apparent resistivity of the subsoil is estimated by injecting electrical currents through a pair of electrodes and measuring the potential difference between another pair of electrodes. The geometry and type of arrangement of the electrodes determine the measured value of the apparent resistivity, as well as the intensity of current injected during the tests. Modern multi-electrode systems and automatized acquisition systems improved the efficiency of this technique (Dahlin 2001; Dahlin and Zhou 2006). The ERT measurements were conducted using two ABEM Terrameter LS equipment, each one together with four cables of 21 takeouts and stainless steel electrodes and connectors. The type of array selected for the measurements was multiple gradient due to its high speed data acquisition, good signal-to-noise ratio and high data density (Dahlin and Zhou 2006). The roll-along technique was used to reach distances of 2400 and 2000 m to the north and to the south, respectively. Likewise, the current injected was 500 and 200 mA in the northern and southern profiles. Similarly to the TEM data, the ERT data need to be processed and inverted. The process consists of minimizing the difference between the pseudo-section (apparent resistivity from the measurements) and the resultant resistivity from a synthetic model by refining this later in each iteration round. The final and refined synthetic model is assumed as the most likely representation of the resistivity distribution in the ground (Dahlin 2001; Loke et al. 2003). The software used to invert the ERT data in this study was Aarhus Workbench (Aarhus GeoSoftware 2017b), 2-D inversions were selected applying the L1-norm (robust constrain inversion), STD 1.3 for lateral constrain and STD 2.0 for vertical constrain to obtain the resistivity models.

The TEM data include 145 soundings, conducted from May to July 2016 in the study area. Those tests were distributed in a rectangular grid (about 6 km × 1.5 km) with ~ 250 m separation between soundings. Not all of them were uniformly separated, because potential noise sources like power lines and houses were avoided. In addition, a number of soundings are aligned in two cross-sections, one of them overlapping both ERT profiles (See Fig. 4), which were conducted in March 2015.

The majority of the TEM soundings were conducted in the flat land, where sediments overlie the bedrock. In addition, two single TEM tests were conducted close to the bedrock outcrops to get the distinctive resistivity of those rocks. Figure 5 shows models corresponding to the consolidated rock where values ≥ 100 Ω-m might be the characteristic of this type of geological unit.

The TEM soundings in the grid and nearby were similarly modeled by single one-dimensional inversions, which later were interpolated using the Kriging method to obtain horizontal resistivity maps and resistivity cross-sections. During the inversion process, a few soundings reached a DOI of 800 m which are assumed as outliers, while most of the soundings reached a depth between 100 and 400 m (see Fig. 6). Consequently, resistivity maps showed more blanks as they progress in depth (see Fig. 7).

A progressive resistivity visualization by depth is shown in Fig. 7. At the top slice, the map of resistivity distribution shows two distinctive parts; to the west, values about 30 Ω-m suggest dry sediments, although in the top 20 m from the surface, clay is more abundant, but the water content was small, because the tests were conducted from May to July (dry season). To the east, values between 5 and 10 Ω-m are in concordance with the presence of thermal springs and saline water disposed in this area. The next resistivity slice, at 3680 masl, is the most homogenous one, with values around 10 Ω-m, interpreted as fully saturated sediments. In addition, the most eastern part still exposes lower values linked with the hot springs. The next slice, at 3640 masl, starts showing higher values, which might indicate the relief of the bedrock exposing folds with southeast–northwest direction (white dashed line). For the rest of the slices below 3600 m, resistivity values remain almost constant, assumed as an indication of the DOI for most of the soundings not reaching these levels. Resistivity values ≥ 100 Ω-m in the western and central part of the grid, are characteristic of consolidated rock in this area. However, the most eastern part still shows lower values, suggesting in turn, the presence of ion-rich water. This trend is repeated in all the slices (black dashed line) which might be an indication of a geological structure, most likely a fault, where saline water is saturating the openings. Major faults in the region have the same direction, which lead to consider a possible connection.

The resistivity sections from the ERT tests are shown in Fig. 8. Profiles 1 and 2 (ERT1 and ERT2) show a thin surficial layer of a couple of meters with values around 80 Ω-m, this might indicate unsaturated sediments on top. Underlying the latter, lower values of around 20 Ω-m might indicate saturated sediments with a thickness varying from tens to hundreds of meters, following the relief of the bedrock. The bedrock has higher resistivity values ≥ 100 Ω-m. This interpretation is similar to those extracted from TEM results. However, ERT1 and ERT2 expose more detailed features, especially within the unconsolidated unit, because this technique generates a larger density of data for the tested profile in comparison to the TEM interpolations. For example, lenses of lower resistivity (~ 10 Ω-m), might indicate high clay content as it was seen in some pits in the area (Fig. 3).

The bedrock in ERT2 seems to sink into a depression to the northeast of the profile. To compare the resistivity distributions obtained from the ERT profiles, interpolations of TEM soundings are reported as cross-sections (TEM1 and TEM2). Figure 8 also shows two resistivity cross-sections based on smooth models and interpolated with the Kriging method, where it is possible to observe the uneven contact between basement rock and sediments. These cross-sections are perpendicular to the possible lengthening of some geological structures. The cross-section TEM1 overlaps the profiles ERT1 and ERT2 in the Fig. 8.

The same resistive layer on top indicated in ERT1 and ERT2, is also shown in TEM1 and TEM2 as a thin layer. Values about 50 Ω-m might correspond to the surficial unsaturated sediments (as seen in Fig. 3). Just a few meters below the latter, the resistivity decreases to about 10 Ω-m, which might be an indication of saturated sediments. The thickness of these sediments varies according to the bedrock relief, from few meters to about 100 m. In the southwesternmost part of TEM2, the layer of saturated sediments is just a couple of meters, because the Silurian bedrock is outcropping nearby (see Fig. 2). Both TEM cross-sections show distinctive zones of high resistivity (≥ 500 Ω-m in purple), this feature might be an indication of consolidated bedrock with small to non-existent porosity and therefore low water storage. The big purple zones in TEM1 and TEM2 might correspond to the same structure, probably a buried ridge, which has similar direction (northwest–southeast) to the surficial outcrops, ridges and mountains in the region.

The settings and distribution of sediments (gravel, sand and clay) showed in the stratigraphic cross-section of Fig. 8 are difficult to distinguish just looking at the resistivity distribution in cross-sections ERT1, ERT2 and TEM1. The factors impeding a clear distinction might be the scale of the resistivity models and most important, the porosity and saturation degree of the sedimentary package, which makes the apparent resistivity being in the same range despite the difference in grain size distribution. Only the top clay layer differentiates itself from the rest of the unconsolidated materials due to its dryness.

Banks et al. (2002) reported the contact between sediments and bedrock located at 3580 masl in the northeast part of TEM1. In this cross-section, the resistivity corresponding to saturated sediments is still present even at 3300 m. Similarly, in a notorious part of TEM2, low resistivity might suggest the existence of deep valleys filled with sediments containing water. However, these hypothetical deep features do not coincide with the bedrock contour map (Banks et al. 2002), nor they do appear in the regional geological cross-section (Fig. 2b). The answer might be found in some structures and faults of the consolidated hard rock with southeast–northwest direction, which probably continue even below the unconsolidated sediments. Figure 9 shows the resistivity at 3650 m where the aligned distribution of low values coincides with the direction of two faults located to the southeast of the study area.