Home

Quick Links

Search

 
Accurate processing and inversion as the ultimate QC of AEM dataGold Open Access

Author: Andrea Viezzoli
Event name: 6th International AEM Conference & Exhibition
Session: Data processing
Publication date: 10 October 2013
Language: English
Info: Abstract, PDF ( 77.49Kb )

Summary:
e ground gives higher signal, especially at early times/high frequencies , even over the same piece of ground. The readings in the Rx are the convolution of ground response and system transfer response (STF). In re-flights, what needs to be repeatable –hence precise- is the ground component part of the readings. The only way to assess precision is therefore to invert the data, uncoupling, from the measured signal, system STF and ground response. Another concept, rarely directly used in QC, but sometimes hinted at, and sometimes confused with precision, is that of accuracy. The latter is the measurement of closeness of measurement of a quantity to its actual (true) value. It should be evident that precision (repeatability) is not that useful per se, if the measured value we can repeat to a satisfying degree is not accurate (i.e., it is repeatedly far from the true value). An AEM system is well calibrated, if it is accurate. It is obvious that also in order to assess the accuracy (calibration) of an AEM system in rendering the actual distribution of the electrical resistivity, we need to work in the model space. The first approach is to invert the AEM data and compare the outcome with a relevant resistivity reference model (obtained from DC/borehole data, other EM data). The second is to forward model the reference resistivity model with the STF of the particular system in the condition of acquisition, and compare it with the actual observations of the system over it. For different reasons, the first approach tends to be the most tempting. It must however be stressed that an inversion output is the end result of a series of steps. From survey design, to data acquisition, pre-processing, post-processing, data integration (another issue that only makes sense contemplating the model space), and inversion, they all contribute to the output. Inaccurate processing and inversion can jeopardize the recovered models, and in turn also the ultimate assessment on AEM data precision and accuracy. If a problem of inaccuracy/poor calibration of an AEM system has been detected, it is again through inversions that, in some cases, the accuracy/calibration of that dataset might be improved. We will present examples that illustrate all the aspects mentioned above.

Download
Back to the article list