NS43A-01 INVITED
Data Integration for Interpretation of Near-Surface Geophysical Tomograms
Traditionally, interpretation of geophysical tomograms for geologic structure or engineering properties has been either qualitative, or based on petrophysical or statistical mapping to convert tomograms of the geophysical parameter (e.g., seismic velocity, radar velocity, or electrical conductivity) to some hydraulic parameter or engineering property of interest (e.g., hydraulic conductivity, porosity, or shear strength). Standard approaches to petrophysical and statistical mapping do not account for variable geophysical resolution, and thus it is difficult to obtain reliable, quantitative estimates of hydrologic properties or to characterize hydrologic processes in situ. Recent research to understand the limitations of tomograms for quantitative estimation points to the need for data integration. We divide near-surface geophysical data integration into two categories: ‘inversion-based' and ‘post- inversion' approaches. The first category includes ‘informed-inversion' strategies that integrate complementary information in the form of prior information; constraints; physically-based regularization or parameterization; or coupled inversion. Post-inversion approaches include probabilistic frameworks to map tomograms to models of engineering properties, while accounting for geophysical resolution, survey design, heterogeneity, and physical models for hydrologic processes. Here, we review recent research demonstrating the need for, and advantages of, data integration. We present examples of both inversion-based and post-inversion data integration to reduce uncertainty, improve interpretation of near-surface geophysical results, and produce more reliable predictive models.
NS43A-02 INVITED
Joint inversion of crosshole radar and seismic traveltimes
Joint inversion of different types of geophysical data collected at a common site can improve the determination of lithological boundaries. The structural approach to joint inversion, entailing common boundaries or gradients, offers a flexible way to invert diverse types of surface-based and/or crosshole geophysical data. The cross- gradients function was introduced as a means to construct models in which spatial changes in two models are parallel or anti-parallel. Inversion methods that use such structural constraints also provide estimates of non- linear and non-unique field-scale relationships between model parameters. We have developed an algorithm to invert jointly crosshole seismic and radar traveltimes for structurally similar models using an iterative non-linear traveltime tomography algorithm. A synthetic example demonstrates that the joint inversion better resolve lithological boundaries than the individual inversion. Inversion of field data collected in a saturated and unconsolidated environment show similar results. The robustness, relative simplicity, and apparent generality of the structural approach to joint inversion suggest that it may become a standard approach for jointly inverting geophysical and hydrological data when individually inverted models indicate structural similarity.
NS43A-03
Cross-Gradient Joint 3D Inversion of Geophysical data
In order to understand the complex processes that occur in the deep Earth's interior and dynamic near-surface environments, we require detailed and realistic models of multiple subsurface properties. This has leaded us to develop integrative methodologies for inversion of geophysical data. An emerging philosophy for the integrative analysis of diverse geophysical data sets is the joint inversion technique with cross-gradient constraint, which seeks to enhance the common structure that may control the spatial distribution of two physical properties. In this work, we extend the cross-gradients philosophy for joint inversion to three-dimensional environments and developed a generalized solution procedure based on a statistical formulation. This procedure is applied to potential field data as well as to seismic refraction data. The results on potential field data show that the cross- gradient joint inversion technique succeeds on finding geophysically supported models with major structural concordance between them. The three-dimensional models obtained, also showed an improved definition of the structural features partially overcoming the inherent lack of depth resolution.
NS43A-04 INVITED
Joint Inversion of Hydrological and Geophysical Data: Too Much Information?
Prediction of fluid flow and contaminant transport in unsaturated, highly heterogeneous porous and fractured media requires high-resolution maps that depict spatially variable, process-specific parameters. While geophysical methods have the potential to image subsurface properties with relatively high resolution and spatial coverage, additional information is needed to (a) determine parameters of petrophysical models that relate measured geophysical data to the flow and transport properties of interest, (b) to remove, quantify, or reduce systematic errors and artifacts, and (c) to justify regularization schemes. The often complementary information contained in geophysical and hydrological data makes a joint inversion approach an attractive method to solve the estimation-identification problem in near-surface hydrology; we have implemented this approach by combining (a) a multiphase flow simulator, (b) geophysical forward models, such as ground penetrating radar and electrical resistivity tomography, and (c) nonlinear optimization algorithms. While adding new data types is generally desirable to address the ill-posedness of the inverse problem, it may lead to new challenges regarding parameterization of the problem, potential inconsistencies between the data sets, relative weighting issues, and the introduction of an estimation bias due to increased systematic errors. The value of information added to a joint inversion framework will be discussed from both a fundamental and practical perspective. This work was supported, in part, by the U.S. Dept. of Energy under Contract No. DE-AC02-05CH11231. http://www-esd.lbl.gov/iTOUGH2
NS43A-05
Structure classification methods applied to independent post-inversion physical-property models
The interpretation of geophysical models derived by inversion is a highly subjective part of any study. Our incomplete knowledge of the subsurface, the spatially-varying resolution of the models, and the non-uniqueness of the geophysical inverse problem make it difficult to objectively interpret physical-property models in terms of geologic structure. I present a classification approach to mapping geologic structure by combining independently- derived inverse models from a range of geophysical methods. The method employed is based solely on the statistical correlation of physical properties in a joint parameter space, and is independent of theoretical or empirical relations between the properties. Regions of high correlation between two or more physical properties are identified as separate classes, which can be defined using a variety of techniques and then reexamined in depth section. The spatial distribution of these classes, and the boundaries between them, are able to provide structural information (lithologic, tectonic, and hydrologic boundaries) often not evident in the individual models. This classification method is applied to several case studies on a range of scales including hydrologic assessment of earthen dams and fault characterization at the Dead Sea Transform and San Andreas Faults. The resulting structural sections are in good accordance with ground truth and supporting geophysics.
NS43A-06 INVITED
Conditional stochastic simulation using high-resolution geophysical data for the local-scale characterization of heterogeneous aquifers
Simulated annealing (SA) provides a flexible means of integrating diverse types of data for the purpose of conditional stochastic simulation. Although the SA technique has been widely used in reservoir characterization studies, relatively little work has been done on the application of this method to hydrogeophysical problems. We present research that builds on previous work involving the use of simulated annealing for near-surface geophysical and hydrological data integration. Here, we introduce a new SA algorithm that provides a significant advancement in the way that large-scale structural information from a geophysical experiment is incorporated into the output realizations. Our SA algorithm contains two key features. First, model perturbations in the annealing procedure are made by drawing from a probability distribution for the target parameter, conditioned to the available geophysical data. This is the only place where geophysical information is utilized in our algorithm, and is in contrast to more traditional SA approaches where model perturbations are made through the swapping of values in the simulation grid, and agreement with soft data is enforced through a correlation coefficient constraint. The second major feature of our SA algorithm is the way in which stochastic information is introduced into the output realizations. Instead of constraining realizations to match a target covariance model at a wide range of spatial lags, we let the perturbation approach of drawing from a conditional distribution control the large-scale subsurface structure, and we stochastically constrain the output realizations only at smaller lags where the available data cannot provide enough information. With this strategy, we allow the geophysical data to have more due control over the output realizations in comparison with previously published SA algorithms. In addition, since the only objective function required in our approach is a covariance constraint at small lags, the algorithm has improved convergence and increased computational speed over more traditional SA methods. We show the results of using our procedure to integrate porosity log and crosshole georadar data to generate realizations of the subsurface porosity field. We do this for a synthetic example, and then for a field data set collected at the Boise Hydrogeophysical Research Site.
NS43A-07
Joint and Cooperative Inversion Strategies for Mineral Exploration
Geophysical inversion for mineral exploration typically involves a single type of data sensitive to a single physical property. Combining several complimentary types of geophysical data collected over the same Earth region reduces ambiguity and can enhance inversion results. This is important whenever the Earth can not be adequately resolved by any one type of data. Combining different types of data into an inversion is important when a structural or stochastic relationship is thought to exist between the different physical property distributions. By inverting each data set individually, the recovered physical property models may be inconsistent with the geologic knowledge regarding structure or property relationships. Cooperative and joint inversion strategies can be employed to ensure consistency between the different models. A conventional cooperative inversion strategy is to invert one set of data independently and use that result to constrain a subsequent independent inversion of the second set. However, the models obtained through this simple procedure are commonly biased towards the result of the first inversion and/or towards the survey with greater sensitivity. We have developed more appropriate cooperative inversion strategies that exploit the smoothness weighting functionality of the UBC-GIF inversion codes. Another approach is to fit the data sets simultaneously in what is termed a joint inversion. Many authors have performed simultaneous inversions of data from different surveys sensitive to the same physical property. Others have jointly inverted data sets responsive to different physical properties between which there is an established analytic relationship. However, little work has focused on joint inversion of disparate data sets (from surveys responsive to different physical properties) when there is no analytic relationship available between the properties. To do so, we enforce the structural or compositional similarity between the property models. We use model-gradient dot- and cross-products to measure structural similarity (e.g. coincident boundaries of geologic bodies). We use statistical cross-correlation to measure compositional similarity, which assumes some linear relationship between the physical properties. Our joint and cooperative strategies allow inversion of disparate geophysical data sets. They ensure consistency between the recovered physical property models when stipulated by the available geologic information. These property models can then be better interpreted in concert with the geologic structural and petrophysical information.