Ecosystems are characterized by high complexity and are dynamically changing by abiotic and biotic drivers such as climate, vegetation and soil fauna. Due to feedback processes between compartments, ecosystems show an inherent degree of locally heterogeneous properties and structures at multiple spatial and temporal scales.
The constructed catchment “Chicken Creek” represents a unique and outstanding site to study an ecosystem at the initial stage of development. The catchment is located about 20 km south of the city of Cottbus (Germany). The 6-ha catchment area was built in 2005 of coarse-textured quaternary sediments from the adjacent lignite mine and forms a back- and foot-slope that flattens out to a pond. A clay liner of 2-3 m thickness seals the catchment at the bottom, allowing for the formation of a local water body fed by precipitation. After construction the site was left to an unrestricted and unmanaged succession to elucidate the role of structures und processes during the initial establishment of an ecosystem. To study processes of ecosystem development and newly emerging structures the Chicken Creek catchment is equipped with a comprehensive ecological monitoring network. A network of more than 40 wells and two weirs allow for recording groundwater levels, surface runoff and discharge. Meteorological parameters are recorded at three weather stations. Annual aerial photos and vegetation monitoring reveal the floral colonisation of the area. Here, we present the 12 years of diverse and extensive monitoring program which has been launched to measure hydrological, biological, meteorological, and pedological parameters during the ecological development of the catchment area.
A substantial interpretation of electromagnetic induction (EMI) measurements requires quantifying optimal model parameters and uncertainty of a nonlinear inverse problem. For this purpose, an adaptive Bayesian Markov chain Monte Carlo (MCMC) algorithm is used to assess multi-orientation and multi-offset EMI measurements in an agriculture field with non-saline and saline soil. In MCMC the posterior distribution is computed using Bayes’ rule. The electromagnetic forward model based on the full solution of Maxwell’s equations was used to simulate the apparent electrical conductivity measured with the configurations of EMI instrument, the CMD Mini-Explorer. Uncertainty in the parameters for the three-layered earth model are investigated by using synthetic data. Our results show that in the scenario of non-saline soil, the parameters of layer thickness as compared to layers electrical conductivity are not very informative and are therefore difficult to resolve. Application of the proposed MCMC-based inversion to field measurements in a drip irrigation system demonstrates that the parameters of the model can be well estimated for the saline soil as compared
to the non-saline soil, and provides useful insight about parameter uncertainty for the assessment of the model outputs.
High-Resolution Soil Electrical Conductivity Imaging from EMI D Based Probabilistic Inversion
(2018)
Electromagnetic induction (EMI) sensors allow for non-invasive soil characterizations. Proximal soil sensing using EMI hindered due to the problems related to the inversion of apparent electrical conductivity (ECa) data. In this study, I used Bayesian inference to obtain the electrical conductivity layering of the subsurface from multi-configuration EMI data. In this respect, generalized formal likelihood function was used to more accurately describe the sensitivity of the posterior parameter distribution to residual assumptions. Discrete Cosine Transform (DCT) was employed as a model compression technique to reduce the number of unknown parameters in the inversion. I considered apparent electrical conductivity pseudosection as a training image (TI) in multiple-point statistical simulations. Information from TI realizations were utilized to determine dominant DCT coefficients, as well as prior probability density functions for the subsequent probabilistic inversions. The potentiality of the proposed approach was examined through an experimental scenario. The results demonstrated that this methodology allows for soil electrical conductivity imaging with high resolution. This strategy permits to incorporate summary metrics from ensemble of ECa pseudosection realizations in the inversion without resorting to any complimentary source of information. The proposed approach ensures accurate and high resolution characterization of soil conductivity layering from measured ECa values.
Conventional geophysical inversion techniques suffer from several limitations including computational cost, nonlinearity, non-uniqueness and dimensionality of the inverse problem. Successful inversion of geophysical data has been a major challenge for decades. Here, a novel approach based on deep learning (DL) inversion via convolutional neural network (CNN) is proposed to instantaneously estimate subsurface electrical conductivity (σ) layering from electromagnetic induction (EMI) data. In this respect, a fully convolutional network was trained on a large synthetic data set generated based on 1-D EMI forward model. The accuracy of the proposed approach was examined using several synthetic scenarios. Moreover, the trained network was used to find subsurface electromagnetic conductivity images (EMCIs) from EMI data measured along two transects from Chicken Creek catchment (Brandenburg, Germany). Dipole–dipole electrical resistivity tomography data were measured as well to obtain reference subsurface σ distributions down to a 6 m depth. The inversely estimated models were juxtaposed and compared with their counterparts obtained from a spatially constrained deterministic algorithm as a standard code. Theoretical simulations demonstrated a well performance of the algorithm even in the presence of noise in data. Moreover, application of the DL inversion for subsurface imaging from Chicken Creek catchment manifested the accuracy and robustness of the proposed approach for EMI inversion. This approach returns subsurface σ distribution directly from EMI data in a single step without any iterations. The proposed strategy simplifies considerably EMI inversion and allows for rapid and accurate estimation of subsurface EMCI from multiconfiguration EMI data.
Low-frequency loop–loop electromagnetic induction (EMI) offers several key advantages over many other geophysical techniques for proximal soil sensing. Yet, because of problems with the inversion of measured apparent electrical conductivity (ECa) data, application of EMI for geophysical imaging and interpretation is limited. In this study, a Bayesian inference was used to obtain electromagnetic conductivity images (EMCIs) from multiconfiguration ECa data. This approach allows analysis of highly nonlinear problems and renders an ensemble of models obtained from the posterior distribution that can be used to explore parameter uncertainty. In this respect, generalized formal likelihood function was used to more accurately describe the sensitivity of the posterior distribution to residual assumptions. Discrete cosine transform (DCT) was employed as a model compression technique to reduce the number of unknown parameters in the inversion. The DCT parameterization was performed using training image (TI)-based geostatistical simulations considering the ECa data pseudosection as a TI. The potential of the proposed approach was examined through different theoretical scenarios. The estimated subsurface EMCI shows excellent agreement with the original synthetic models subject to the appropriate choice of prior information. Moreover, DCT parameterization reduces the number of unknown parameters, increasing accuracy of the inversion with the Bayesian procedure. The proposed approach ensures accurate and high-resolution characterization of subsurface conductivity layering from measured ECa values.
Machine learning to estimate soil moisture from geophysical measurements of electrical conductivity
(2019)
Soil water content (θ) is a key variable in different earth science disciplines since it mediates the water and energy exchange between the surface and atmosphere. Electrical and electromagnetic geophysical techniques have been widely used to estimate soil electrical conductivity (σ) and soil moisture. However, obtaining the σ − θ relationship is not straightforward due to the non-linearity and also dependency on many different soil and environmental properties. The purpose of this paper is to determine if artificial neural network is an appropriate machine learning technique for relating electrical conductivity to soil water content. In this respect, time-lapse electrical resistivity tomography measurements were carried out along a transect in the Chicken Creek catchment (Brandenburg, Germany). To ensure proper retrieval of the σ and θ, reference values were measured near the beginning of the transect via an excavated pit using 5TE capacitance sensors installed at different depths. We explored robustness and pertinence of the artificial neural network approach in comparison with Rhoades model (as a commonly used petrophysical relationship) to convert the inversely estimated σ from electrical resistivity tomography to the θ. The proposed approach was successfully validated and benchmarked by comparing the estimated values with the reference data. This study showed the superiority of the artificial neural network approach to the Rhoades model to obtain σ − θ relationship. In particular, artificial neural network allowed for more accurate estimation of the temporal wetting front than the petrophysical model. The proposed methodology thus offers a great promise for deriving spatiotemporal soil moisture patterns from geophysical data and obtaining the in situ σ − θ relationship, taking into account the non-linear variations of the soil moisture.
In recent years, probabilistic solution to the inversion of electromagnetic induction (EMI) data has been progressively developed for non-invasive subsurface characterization. However, Bayesian inversion of EMI data using forward solvers based on full solution ofMaxwell's equation is associated with computationally expensive modelings, particularly for large-scale surveys. Here, we incorporated artificial neural network (ANN) with Bayesian inference to obtain subsurface electromagnetic conductivity image (EMCI) fromEMI data down to 10mdepth. In this respect, a complex EMI forward model was replaced by a trained neural network (ANN proxy forward function) that can be evaluated comparably rapidly. The accuracy of the ANN-based forward solver was examined using different synthetic subsurface models. The proposed methodology was applied on EMI data measured with a DUALEM-421 s sensor from 10 ha study site in the Alken Enge area of Denmark. We compared the inversely estimated EMCI with the counterpart obtained from a quasi-three-dimensional (quasi-3D) spatiallyconstrained deterministic algorithm as a standard code. The network training procedure was performed within fewminutes, and once itwas trained, the ANN-based forward solver returned roughly 150,000 model responses per second. This value for the EMI forward solverwas around 400, demonstrating the computational efficiency of the ANN proxy forward function. The theoretical simulations demonstrated that the ANN-based forward solver accurately mimics the EMI response within the training range. Moreover, the proposed inversion strategy successfully delineated the subsurface EMCI from Alken Enge area. This approach thus facilitates rapid and accurate subsurface conductivity imaging using Bayesian inversion of multi-configuration EMI data, which is particularly pertinent for large-scale measurements.
Monitoring spatiotemporal variations of soil water content (θ) is important across a range of research fields, including agricultural engineering, hydrology, meteorology and climatology. Low frequency electromagnetic induction (EMI) systems have proven to be useful tools in mapping soil apparent electrical conductivity (σa) and soil moisture. However, obtaining depth profile water content is an area that has not been fully explored using EMI. To examine this, we performed time-lapse EMI measurements using a CMD mini-Explorer sensor along a 10m transect of a maize field over a 6 day period. Reference data were measured at the end of the profile via an excavated pit using 5TE capacitance sensors. In order to derive a time-lapse, depth-specific subsurface image of electrical conductivity (σ), we applied a probabilistic sampling approach, DREAM(ZS), on the measured EMI data. The inversely estimated σ values were subsequently converted to θ using the Rhoades et al. (1976) petrophysical relationship. The uncertainties in measured σa, as well as inaccuracies in the inverted data, introduced some discrepancies between estimated σ and reference values in time and space. Moreover, the disparity between the measurement footprints of the 5TE and CMD Mini-Explorer sensors also led to differences. The obtained θ permitted an accurate monitoring of the spatiotemporal distribution and variation of soil water content due to root water uptake and evaporation. The proposed EMI measurement and modeling technique also allowed for detecting temporal root zone soil moisture variations. The time-lapse θ monitoring approach developed using
DREAM(ZS) thus appears to be a useful technique to understand spatiotemporal patterns of soil water content and provide insights into linked soil moisture vegetation processes and the dynamics of soil moisture/infiltration processes.