Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE credits
In August 2008 and June 2009 two sets of airborne measurements were made in Falun and Borlänge, respectively, by the Geological Survey of Sweden (SGU). The purpose of these campaigns was to test the new Multi Frequency Receiver (MFR) instrument called ADU07 for the collection of airborne data in the VLF and LF bands. This system was designed by Metronix and adapted by Uppsala University together with the SGU in the frame of a joint research project.
The SGU in its bedrock mapping program routinely records VLF signals from only two transmitters in the frequency band of 14‐30 kHz. The RMT technique also makes use of electromagnetic signals in both the VLF and LF bands in the frequency range of 10‐250 kHz. By measuring all the three magnetic field components in this broader band, the data acquired by the new MFR system can provide high lateral and vertical resolution compared to the VLF data. This can be done by applying the concepts used for the EnviroMT. The joint research therefore aims at extending the VLF technique currently used by the SGU for geophysical investigations and whereby generating improved and more detailed anomaly maps.
The airborne measurements with the ADU07 system were performed by continuously recording the three magnetic field components with a sampling frequency of 512 kHz in three channels. The prior evaluation of the data gave good results in the beginning. However, later tests showed that there were some near field sources onboard the aeroplane that contaminated the data and highly affected the estimation of transfer functions from the radio transmitters’ signals. The noise was basically generated by other measuring instrumentation and the common power supply used to feed both the ADU07 and the PC controller. The main aim of the present project is to develop a processing method that identifies frequencies from these near field sources and filters them out from the spectral ADU07 data. This work has been carried out by writing MATLAB routines. After the filtering, more reliable transfer functions that provide relevant information about the Earth’s resistivity structure can be estimated.
Different methods were applied in order to detect the noise in the data. The mean value of the real part of the vertical magnetic field component (Hz) and the scalar tippers were firstly calculated along the profiles. These values should normally be close to zero. These methods did not give any valuable information since no patterns could be seen in the results. Afterwards, the vertical signal‐to‐noise ratio (VSNR) was calculated for every frequency at each station. This criterion showed that, for the first campaign, there were practically two sets of noise frequencies in the spectra: the first group corresponds to the even and odd harmonics of frequency 8 kHz, and the second group to frequency 23.47 kHz and its harmonics. For the last campaign, frequency 10.28 kHz and its harmonics were identified. The band averaging technique that splits the main frequency band into 9 overlapping sub‐bands with 1 octave of width was used. Finally, Prediction Errors (PE’s) were estimated to detect the remaining noise. A threshold was then chosen in order to remove from the spectra those frequencies with a PE above 3 and up to 20% of the number of transmitters in the sub‐bands. These processing steps improved considerably the tipper behaviour for the VLF band along the profiles, although some noise was also added. For the LF band, the filtering steps seem to have worsened the data quality and therefore the tipper estimation.
The removal of important frequencies that were hidden in the high noise levels and the useof some other instruments during the data collection could be the causes of these responses.
2009. , 103 p.