SMART - geoelectrical tomographic monitoring of dynamic systems using adaptive self-optimising data acquisition

电阻率层析成像（ERT）是一种快速发展的地下扫描成像技术。它越来越多地用于广泛的地球科学问题，特别是用于对经历动态变化的复杂地球系统的时移监测（例如：污染羽流、盐水侵入、化学相互作用）。近年来，多通道、微处理器控制的仪器系统的设计取得了重大进展。英国地质调查局自己设计了一种世界领先的自动延时电阻率断层扫描（ALERT）仪器，该仪器首次允许使用无线遥测技术“按需”对脆弱地点进行远程实时成像。然而，尽管取得了这些进展，基本的电阻率测量制度仍未改变。对于大多数传统的地下应用，作业者使用预先确定的4点电极排列（例如：Wenner、Schlumberger、Dipole-Dipole）在更大的阵列中测量接地电阻。因此，这些数据往往采样不良，有噪声，并且对地下的特定区域不敏感。即使是最好的反演算法也无法弥补由于采集病态野外数据而造成的空间分辨率的不足。英国地质调查局和其他国际组织的最新进展表明，应该有可能优化自适应采样收集的数据的分辨率。我们提出验证假设，即英国地质调查局设计的ALERT系统可以被编程来收集SMART（灵敏度调制自适应电阻率层析成像）数据。我们将以早期的工作（Wilkinson等人，2006a; 2006b; 2007）为基础，该工作使用了雅可比模型分辨率矩阵的估计，该矩阵由R = (GTG+C)-1GTG给出，其中C是使反演正则化的约束矩阵。R的前导对角线（模型分辨率R）将用于评估现有测量集对地下的成像效果，以及如果添加这些测量值，哪些测量值将对图像分辨率产生最大的改进。然后，优化过程选择几个这样的测量，重新计算R并迭代该过程，直到达到所需的测量总数。在每次迭代中，测量配置将包括一个基本集（在整个图像区域进行优化）和一个附加集（在感兴趣的区域进行优化以提高分辨率）。需要找到方法来存储和检索优化的基集和雅可比元素，而不是依赖于重新计算。算法设计将使用表示简化动态目标的合成数据进行TRL 3测试。结果将与静态配置（标准和优化）的结果进行比较。我们亦会评估噪音对各类监察数据（标准、优化及智能）的影响。在TRL 4的SMART概念测试将使用英国地质调查局的水文地球物理测试设施进行。这包括两个测试细胞，其中第二个将用于监测盐水示踪剂通过静态和自适应ERT测量配置。这些电池具有用于SMART测试的表面和钻孔电极的永久线性2D和3D阵列。模拟钻孔中10厘米深度间隔的多级采样器将为评估新SMART算法的精度和分辨率提供真实值。现有的多电极电阻率测量仪器没有尝试通过调整应用电流分布，使其最适合给定的测量几何形状或现场表面条件，或最好地处理地下的特定（最初未知的）特征，来达到这种程度的背景或目标适应性，从而提高记录数据的质量。如果成功，我们可以预测层析成像图像质量的阶跃变化。

Electrical Resistivity Tomography (ERT) is a rapidly evolving imaging technology for scanning the subsurface. It is increasing used for a wide range of geoscientiific problems particularly for the time-lapse monitoring of complex earth systems undergoing dynamic change (e.g: pollution plumes, saline intrusion, chemical interactions). Major advances have been made in recent years with the design of multi-channel, microprocessor-controlled instrumentation systems. BGS has itself designed a world-leading instrument for Automated time-Lapse Electrical Resistivity Tomography (ALERT) which, for the first time, allows the remote, real-time imaging of vulnerable sites 'on demand' using wireless telemetry. Despite these advances, however, the basic resistivity measurement regime remains unchanged. For most conventional subsurface applications the operator measures ground resistance using a pre-determined 4 point electrode arrangement (e.g: Wenner, Schlumberger, Dipole-Dipole) within a larger array. Consequently the data are often poorly sampled, noisy and insensitive to specific regions of interest in the subsurface. Even the best inversion algorithms cannot compensate for the lack of spatial resolution caused by the collection of ill-conditioned field data. Recent advances at BGS and by other international groups suggest that it should be possible to optimise the resolution of the data collected by adaptive sampling. We propose to test the hypothesis that the BGS-designed ALERT system could be programmed to collect SMART (Sensitivity-Modulated Adaptive Resistivity Tomography) data. We will build on earlier work (Wilkinson et al., 2006a; 2006b; 2007) which uses an estimate of the Jacobian model resolution matrix given by R = (GTG+C)-1GTG, where C is the constraint matrix that regularises the inversion. The leading diagonal of R (the model resolution R) will be used to assess how well the existing set of measurements images the subsurface, and which measurements would produce the best improvements to the image resolution if they were to be added. The optimisation procedure then selects several such measurements, recalculates R and iterates the process until the desired total number of measurements is reached. At each iteration, the measurement configurations will comprise a base set (optimised over the whole image region) and an additional set (optimised to enhance resolution in the region of interest). Ways will need to be found to store and retrieve the optimised base set and Jacobian elements rather than rely on recalculation. The algorithm design will be tested up to TRL 3 using synthetic data representing simplified dynamic targets. The results will be compared with those from static configurations (standard and optimised). The effects of noise on each type of monitoring data (standard, optimised and SMART) will also be assessed. Tests of the SMART concept at TRL 4 will be undertaken using the BGS hydrogeophysical test facility. This comprises two test cells, the second of which will be used to monitor the passage of saline tracers with both static and adaptive ERT measurement configurations. The cells have permanent linear 2D and 3D arrays of surface and borehole electrodes which will be used for the SMART tests. Multi-level samplers at 10 cm depth intervals in the simulated boreholes will provide ground-truth to assess the accuracy and resolution of the new SMART algorithm. No existing multi-electrode resistivity survey instrument attempts to improve the quality of the recorded data by achieving this degree of context or target adaptivity by adjusting the applied current distributions to be those most appropriate for a given survey geometry or site surface conditions, or to best cope with the particular (initially unknown) features of the subsurface. If successful, we can anticipate a step change in tomographic image quality.