OR/13/003 Merging the 4 data sets (grids)
|Beamish, D. 2013. The construction of a merged EM (conductivity) database using Tellus and Tellus Border airborne geophysical data. (Land Use, Planning and Development Programme). British Geological Survey Internal Report, OR/13/003.|
The general merging procedure adopted was to move the data sets ‘forward in time’ so that the latest data set (TB) remained static and previous survey data (TEL-05, TEL-06 and Cavan) were sequentially adjusted, as appropriate. The contractor delivered coupling ratios (P, Q) and radar altitude (RALT) formed the basis of the procedure. The full procedure involved:
- Inversion of the (P, Q, RALT) data as described previously, for each of the surveys. This procedure provided estimates (a model) of half-space conductivity, the L1 inversion misfit error and the thickness of an at-surface resistive zone, overlying the half-space.
- Examination of the L1 misfit errors (ERR) and high-fly zones. A standard combined rejection criteria of ERR>100% and RALT>180 m was adopted to cull poor-quality estimates of the half-space conductivity. The procedure produced a revised, reduced data set containing data gaps from single points, clusters of points and zones, as described previously.
- The apparent resistivity estimates from the reduced data set were subjected to a FMD method of levelling involving a window length of 1000 m (along line) and a 500 m length/radius across lines (i.e. using 2 lines adjacent to each central line, a total of 5 lines). The procedure was applied in a N–S rotated coordinate system. The procedure adopted ensured that the poor-quality model estimates did not contribute to the levelling of the data.
- The levelled data (with gaps) were used to create a 50 m grid using a natural neighbour (NN) algorithm. This gridding procedure accounts for edge effects in the data and returns values precisely within the bounds of the observed data (no negatives and excessive positive values due to under and over shoots). Prior to NN gridding, apparent conductivity values were clipped to a maximum value of 1000 mS/m. This thresholding procedure restricts both offshore conductivities and isolated outliers (spikes) and allows for more accurate gridding/imaging. The NN grids are unrestricted at low values of apparent conductivity
- The NN procedure provides an interpolation across all the gaps within the data set. The interpolation is effective when the gaps are small and close natural neighbours exist. When the gaps are extensive and no close natural neighbours exist, the interpolation becomes exotic (unrealistic). The levelled NN grids form the basis for the merging procedure.
Merging TEL-05 and TEL-06
As noted previously, different frequencies were employed at LF and HF during the Tellus surveys. In order to estimate an appropriate offset, 2 repeated survey lines (L1214 and L1215) were employed. The inverted model data from the two repeat lines were used to assess the DC level adjustments required to adjust the TEL-05 model parameters to fit those of the TEL-06 parameters. The adjustments were found to be as follows:
- Apparent conductivity: LF. A value of 2.47 mS/m was added to the TEL-05 LF data.
- Apparent conductivity: HF. Although the analysis indicated a subtraction of 2.50 mS/m to the TEL-05 results, this was found to produce extensive areas of negative values. No adjustment to the TEL-05 HF data was performed.
- There is no adjustment applied to the ERR or THK model parameter.
Merging of the TEL-05 and TEL-06 grid results was undertaken using the Geosoft extension GridKnit module. An automatic suture method was applied using a no static and no trend adjustment condition. The resulting merged grids (LF and HF) are referred to as TEL-05-06.
Comparisons were undertaken between existing transform and inversion estimates of the LF and HF apparent conductivities. The existing ‘whole Tellus’ estimates were obtained from the contractor supplied Version 2 data file (emap_v2.xyz, 2006). Figure 22 shows a comparison across the northern coastal area centred on Magilligan Point and part of the TEL-05 data set. The images shown are LF apparent conductivity (NN grids at 50 m) obtained by the transform (Figure 22a) and by the inversion method (Figure 22b) considered here. Both data sets are clipped to the coast and use the same linear colour scale restricted to a high value of 100 mS/m.
It can be seen that that the two sets of results are broadly similar with the same hard-edges (rapid changes in gradient) identified. More subtle differences are however evident relating to the amplitudes of the variations. Two methods of identifying high-fly zones are demonstrated. Figure 22a uses an overlay polygon mask of towns, while (Figure 21b) identifies (with blanks) only those zones with RALT>175 m. The latter is obviously a more accurate description of the actual flying behaviour.
Merging CAV and TB
No overlap exists between the merged TEL-05-06 data sets and the CAV data. The CAV data were therefore merged with the TB data set which was designed to provide an adequate overlap with the CAV survey area. The CAV area is effectively a hole within the larger TB area. Merging of the CAV and TB grid results was undertaken using the Geosoft extension GridKnit module. An automatic suture method was applied using a static correction to the CAV grid only and a no trend adjustment condition. The merging result was found to be adequate and the resulting grids (LF and HF) are referred to as CAVTB.
Merging TEL-05-06 and CAVTB
The TB survey was designed to provide an adequate overlap with the existing TEL-05-06 survey area. In principle, the merged TEL-05-06 data has been adjusted to fit the actual frequencies (3005 Hz and 11962 Hz) used in the TB survey. Merging of the TEL-05-06 and TB grid results was undertaken using the Geosoft extension GridKnit module. An automatic suture method was applied using no static corrections and no trend adjustment, since this method proved the most effective.
The final merged result is a NN 50 m grid that contains a small number of negative values introduced during the merging procedure. The final merged grid obtained at LF is shown in Figure 23 with the suture lines from grid merging displayed in white. It can seen that the merging is broadly satisfactory. The equivalent result obtained at HF is shown in Figure 24.
Most of the problems of the merging process are associated with low values of conductivity and this is because of the reduced signal/noise ratio in the resulting observed coupling ratios. The effect increases with decreasing frequency (Appendix 1). EM data obtained across resistive zones are difficult to level, and this may result in residual effects (small amplitude undulations) that can observed in all the data along the flight line direction of 345°. These larger scale effects are associated with the resistive aperture of the EM system employed, as discussed in Appendix 1.