The success of geophysical investigations associated with environmental problems is to a large<br>extent determined by the density of the measurements and the quality of the interpretations. Many case<br>studies of mapping of raw materials and hydrogeological investigations have confirmed that dense<br>measurements with traditional methods of geoelectrical and transient (TEM) soundings as well as novel<br>methods of continuous electrical profiling (CEP) and El-logging enhance the data reliability and the<br>possibilities of finding a trustworthy interpretation of the complexities of the geological settings<br>(Christensen and Sorensen 1994).<br>TEM soundings have proven very valuable in delineating the geometry of aquifers in<br>hydrogeological investigations in the Quaternary deposits in Denmark, and they are now routinely used<br>for the purpose of mapping the dept,h to a good conductor, which may be well conducting heavy clays or<br>salt water horizons, both being the effective bottom of an aquifer. A hydrogeological investigation may<br>contain 100-500 TEM soundings, often along profile lines, and it is of importance that the interpreter as<br>well as the field crew get an overview of the results as quickly as possible to adjust the strategy of data<br>aquisition.<br>An ordinary 1D least squares iterative inversion of TEM sounding data require that the<br>interpreter supply an initial model, and the computation time is usually between 10 and 30 minutes on a<br>PC. With a daily production of more than 15 soundings this procedure is slow. and there is need for fast<br>approximate ways of interpretation. The newly developed pulled array transient electromagnetic method<br>(PA-TEM) (Sorensen 1995), where a transient equipment is towed behind a small vehicle while<br>measuring, produces huge amounts of data to be interpreted. It would be an impossible task to interpret<br>the resulting number of TEM soundings with ordinary least squares inversion procedures.<br>An algorithm for imaging of TEICl soundings based on the Frechet kernel is presented, where the<br>computation time is appr. 0.5s/sounding / Mflop. The imaging produces models with 20-40 layers, which<br>fit the original data typically within 510%. No initial model is required, and the algorithm is therefore<br>well suited for automatic inversion. The algorithm makes it possible to see the results of a days work in a<br>matter of minutes and to implement on-line inversion simultaneous with the measurements.


Article metrics loading...

Loading full text...

Full text loading...

This is a required field
Please enter a valid email address
Approval was a Success
Invalid data
An Error Occurred
Approval was partially successful, following selected items could not be processed due to error