Study of water tension differences in heterogeneous sandy soils using surface ERT

https://doi.org/10.1016/j.jappgeo.2007.12.007Get rights and content

Abstract

Herbaceous vegetation in the Sahel grows almost exclusively on sandy soils which preferentially retain water through infiltration and storage. The hydrological functioning of these sandy soils during rain cycles is unknown. One way to tackle this issue is to spatialize variations in water content but these are difficult to measure in the vadose zone. We investigated the use of Electrical Resistivity Tomography (ERT) as a technique for spatializing resistivity in a non-destructive manner in order to improve our knowledge of relevant hydrological processes. To achieve this, two approaches were examined. First, we focused on a possible link between water tension (which is much easier to measure in the field by point measurements than water content), and resistivity (spatialized with ERT). Second, because ERT is affected by solution non-uniqueness and reconstruction smoothing, we improved the accuracy of ERT inversion by comparing calculated solutions with in-situ resistivity measurements. We studied a natural microdune during a controlled field experiment with artificial sprinkling which reproduced typical rainfall cycles. We recorded temperature, water tension and resistivity within the microdune and applied surface ERT before and after the 3 rainfall cycles. Soil samples were collected after the experiment to determine soil physical characteristics. An experimental relationship between water tension and water content was also investigated. Our results showed that the raw relationship between calculated ERT resistivity and water tension measurements in sand is highly scattered because of significant spatial variations in porosity. An improved correlation was achieved by using resistivity ratio and water tension differences. The slope of the relationship depends on the soil solution conductivity, as predicted by Archie's law when salted water was used for the rain simulation. We found that determining the variations in electrical resistivity is a sensitive method for spatializing the differences in water tension which are directly linked with infiltration and evaporation/drainage processes in the vadose zone. However, three factors complicate the use of this approach. Firstly, the relation between water tension and water content is generally non-linear and dependent on the water content range. This could limit the use of our site-specific relations for spatializing water content with ERT through tension. Secondly, to achieve the necessary optimization of ERT inversion, we used destructive resistivity measurements in the soil, which renders ERT less attractive. Thirdly, we found that the calculated resistivity is not always accurate because of the smoothing involved in surface ERT data inversion. We conclude that further developments are needed into ERT image reconstruction before water tension (and water content) can be spatialized in heterogeneous sandy soils with the accuracy needed to routinely study their hydrological functioning.

Introduction

The degradation of natural resources in the arid parts of the Sahel, is currently a quite serious problem leading to desertification, loss of biodiversity, increase of surface runoff and soil losses. Within this degraded landscape, sandy deposits are islets of fertility (Thiombiano, 2000) which often take the shape of microdunes (Casenave and Valentin, 1992). Microdunes are very important ecological units where significant infiltration of water takes place (Ribolzi et al., 2000) and are therefore potential starting points for regeneration of the Sahelian environment. Little is known, however, about their hydrological functioning during and after rainfall, processes such as infiltration, evapo-transpiration and drainage which can only be studied using destructive soil moisture and water tension measurements. Since these fragile sandy soils are unstable, they cannot be brought back from the field or reproduced in the laboratory. All experiments must be carried out in the field. To replicate the controlled experimental conditions of the laboratory, we used simulated rainfall to reproduce natural rain/evaporation cycles in the field. The primary objective of our study was to spatialize water content variation during infiltration and evaporation using surface Electrical Resistivity Tomography (ERT), which is a technique highly suitable for groundwater studies.

To achieve this objective, two methodological problems must be overcome. Firstly an experimental relationship between resistivity and water content has to be established. Secondly, a protocol for accurate ERT inversion is necessary to reconstruct actual resistivity distribution in the subsurface. The first problem requires simultaneous water content monitoring and ERT acquisition at several measurements points and/or laboratory experiments on soil samples. These measurements could not be practically undertaken because of the lack of experimental sensors to measure directly water content at several point measurements in a thin dune. Also the soil is too fragile for resistivity measurements on small cores. Therefore, we investigated possible approaches for determining a relationship between resistivity and water tension, which is more easily measured in the soil and a potentially interesting way to study hydrological processes linked with infiltration and evaporation.

As with resistivity, soil water tension is strongly related to water content. Soil water tension measurements are well suited for determining soil water status in the field (Richards, 1931). Water tension (or pressure head, or matrix potential) is generally expressed as a negative value that reaches zero when the soil is saturated. Water tension data can be considered in two ways: first, there is a direct relationship between water tension and water content, as studied by Van Genuchten (1980) for example. This relationship is generally non linear. Second, the difference between water tension at the final state and initial states, provides important information on the movement of water: a positive difference shows an increase in water content, whereas a negative difference indicates a decrease. A difference of zero suggests zero flux with drainage (below) and evaporation (above). Hence, by spatializing water tension differences we can get a better understanding of water movement in the vadose zone. Therefore, the development of a spatialized image of water tension differences during natural or artificial infiltration (and evaporation) experiments would be of strong interest. But the number of sensors (small ceramic cups) that can be used remains limited because if too many are inserted the medium will be destroyed. This makes it difficult to spatialize the results laterally when the soil is heterogeneous. Thus, if a relationship between water tension and resistivity can be established, ERT offers attractive possibilities for spatializing water tension. In this study, we investigated both aspects of tension (its actual value and its differences from one state to another) in relation to resistivity. This was done using experimental data collected in the field during artificial rainfall cycles. A possible extension of this approach was also considered: we attempted to establish an experimental relationship between water tension and water content with the aim of using ERT to spatialize water content. Because of scattered results due to unexpectedly heterogeneous soil, this final objective was not achieved.

The second issue we addressed was how to use ERT to reconstruct spatial and temporal variations in resistivity from one hydric state to another within a dune. In the vadose zone, electrical resistivity mainly depends on 3 parameters: water content, water conductivity and porosity. The empirical Archie's law (Archie, 1942), which is applicable for studies on sandy soils, integrates these three parameters. Temperature and clay content can also modify resistivity values (Telford et al., 1990). For saturated sandy soils, Archie's law is convenient for monitoring variations in water conductivity, when water content and porosity remain constant. Singha and Gorelick (2006) investigated the use of Archie's law for monitoring tracer concentration. They found that water conductivity values derived from Archie's law using ERT agreed with experimental data only if the formation factor (the porosity-dependent parameter of Archie's law) was varied in space and time. In the present study, the effect of spatial variations in porosity inside the dune was taken into account when interpreting our results. In unsaturated soils, such as those in our study, it is more difficult to derive resistivity variations into hydrological parameters because water content has also a significant influence on resistivity values. On the other hand, the sensitive dependence of resistivity on water content is an advantage for tracking hydrologic processes, especially if we consider that water tension is also sensitive to water content. French et al. (2002) found that ERT is suitable for localizing infiltration zones, even if resistivity is affected by changes in saturation values and tracer concentrations. In this study, we tried to minimize the extent of water conductivity variability by using demineralized water for the 2 first experimental rainfall events. We also investigated the use of salted rain to evaluate the effect of water conductivity, during the third (and final) experimental rainfall.

Despite the methodological difficulties involved, resistivity remains an attractive parameter because it offers the advantage of being easy to measure using non-destructive ERT surface measurements. Moreover ERT can be applied at different scales of investigation and is well adapted for plot studies. For example, Michot et al. (2003) used ERT and additional measurements to study water uptake by plants by successfully monitoring corn crops growing under irrigation. In our study, we specifically adapted miniaturized arrays to the scale of the dune, following the concept proposed by Depountis et al. (2001) in their ERT centrifuge modeling experiment. In addition, ERT can be used to study 2 or 3D geometry which is a promising technique/method for reconstructing complex resistivity distribution patterns in the subsurface. Although, in many studies, cross bore hole ERT was considered as an efficient method for resistivity imaging (see the pioneering studies by Daily et al., 1992 as well as work by Slater et al., 2000, or Kemna et al., 2002), we did not use cross bole hole imaging in this study because our main methodological objective was to evaluate the efficiency of non destructive surface ERT.

Several authors reported that ERT resolution is limited due to the spatial smoothing of inversion algorithms (see Kemna et al., 2002, Singha and Gorelick, 2006), thus in this paper we also focus on the following issues: does ERT give a reliable estimate of resistivity? Finally, is ERT monitoring a useful technique for determining water tension distribution in sandy soils? To address these questions, we performed a field experiment with artificial rainfall. ERT was applied before and after each rainfall. In a first step, to evaluate ERT accuracy, ERT inversion was optimized by comparing the calculated resistivity with in-situ resistivity. In a second step, we investigated the experimental relationship between water tension and resistivity before and after rain. The relationship is then discussed and enlightened with Archie's law. In conclusion we discuss practical considerations for using surface ERT for spatializing water tension in heterogeneous sandy soils.

Section snippets

Experimental setup

The site is located in northern Burkina Faso (Fig. 1). It is a small degraded 82 ha catchment overgrazed by livestock. The climate is Sahelian, with only one rainy season (June to September). The average annual rainfall is 512 mm and mean annual potential evapotranspiration is 2396 mm. Sandy soils are formed by aeolian and/or runoff deposits. They are thin (0.3 to 1 m thick) and form more than one third (1/3) of the landscape surface. The microdune (5.3 m2) shows a typical crescent form (Fig. 1

Resistivity logging

An example of the results logged with the electrical probe on the windward side is shown in Fig. 3. The trend was similar on the leeward side. We corrected the data using temperature (at 25 °C reference). The result shown was obtained for Rain 2 (demineralized water) and Rain 3 (salted water).

Prior to Rain 2, resistivity varied between 400 Ω m (at 7–11 cm) to more than 700 Ω m (at 27 cm, and deeper). A sharp variation in resistivity was noted between 15 and 21 cm. Just after Rain 2, the

Relationship between ERT resistivity and water tension

The relationship between ERT resistivity and water tension measurements is shown in Fig. 10, taking Rain 2 as an example. 3 points were drawn for each tensiometer. They correspond to the values of the resistivity–water tension couple taken before and just after Rain 2, and following 1 h of evaporation. The results show that there is no correlation between resistivity and water tension, with a scattered relationship. Data from tensiometers 1, 2, 3, 6, 7 located near the surface show the same

Conclusion

In this field study, we compared the soil electrical resistivity with water tension measurements under a series of 3 simulated rainfall events, with the aim of using electrical resistivity as a parameter to facilitate spatialization of hydrological processes (infiltration, evaporation) within a typical sahelian microdune. For this, the usefulness of surface ERT for examining the distribution of soil resistivity in the microdune, with heterogeneous sandy soils, was evaluated.

Our first set of

Acknowledgements

This work was funded and conducted by Unités de Recherche 027 and 049 of Institut de Recherche pour le Développement (IRD), and INSU Programme National Sol-Erosion (PNSE) project no. 99/44. We greatly acknowledge Institut National de l'Environnement et de la Recherche Agricole (INERA) of Burkina Faso for providing access to the site. We also thank the team of the hydrological laboratory of IRD at Ouagadougou, with a special mention of Moussa Barry, Yves Dzouali, Harouna Karambiri, Dial Niang,

References (31)

  • DailyW. et al.

    Electrical resistivity tomography of vadose water movement

    Water Resources Research

    (1992)
  • Day-LewisF.D. et al.

    Applying petrophysical models to radar travel time and electrical resistivity tomograms: resolution-dependent limitations

    Journal of Geophysical Research

    (2005)
  • KellerG.V.

    Rock and mineral properties

  • KellerG.V. et al.

    Electrical methods in geophysical prospecting

  • KnightR.

    Hysteresis in the electrical resistivity of partially saturated sandstones

    Geophysics

    (1991)
  • Cited by (32)

    • Canal dike characterization by means of electrical resistivity, shear wave velocity and particle size data fusion

      2022, Journal of Applied Geophysics
      Citation Excerpt :

      The inversion was carried out over 4 iterations, yielding a final RMS error of 5.3%. A greater number of iterations tend to overstructure, geologically speaking, our imaging result, as explained by Descloitres et al. (2008). Fig. 3 highlights the strong correlation existing in our study between measured and calculated apparent resistivity values.

    • Time-lapse electrical surveys to locate infiltration zones in weathered hard rock tropical areas

      2017, Journal of Applied Geophysics
      Citation Excerpt :

      Time-lapse resistivity variations have been already investigated in several studies exploiting this behavior to describe various hydrological processes. For very shallow processes (i.e. less than 1 m), Descloitres et al. (2007) investigated rain infiltration in sandy soils and found that electrical resistivity ratio observed before and after the rain can reveal water tension variations. They used in-depth sensor data to constrain data interpretation.

    • Geophysical demonstration of the absence of correlation between lineaments and hydrogeologically usefull fractures: Case study of the Sanon hard rock aquifer (central northern Burkina Faso)

      2017, Journal of African Earth Sciences
      Citation Excerpt :

      Chi2 can thus be a better estimator of the model’s accuracy, since it provides a representative estimation of the errors due to resistivity measurements (Günther, 2004). However, as it has been shown that low values of RMS or Chi2 do not guarantee that the model provides an accurate representation of the subsurface (Descloitres et al., 2008a), it is nevertheless important to refer to lithologs for geophysical data calibration. The geophysical models obtained by inversion were clustered (e.g. Chaudhuri et al., 2013), according to the resistivity ranges corresponding to the “saprolite”, “fissured” and “fresh rock” layers obtained from the electrical resistivity logs.

    View all citing articles on Scopus
    View full text