Step 1: MT data processing sequence
A standard processing sequence consists in: 1/ collect and back-up the raw data 2/ data format verification (check or fill header correctly) 3/ store data in a database. The synchronization with the reference station is checked (gaps happen) then the impedance tensor can be estimated on the common periods, using statistical and robust error estimator. Here, the BIRRP (Chave, 2004) was applied for frequency < 0.1 Hz. Last, a band averaging is applied to get a fix number of data per decade. The two steps are presented here: “before” (top) and “after” (bottom) band averaging.
This processing was performed in-situ to check data quality (~6h per site) during the survey and decide to repeat some MT soundings if necessary.
Step 2: Data and model fit
Two sounding curves are presented. Site n°5 shows similar trends for the two polarizations (xy and yx) indicating a 1D behavior, whereas the MT site n°26 highlights a clear difference. On both amplitude and phase sounding curves, a low signal to noise ratio is observed around 0.1 Hz and at the lowest frequencies.
Step 3: Overall data quality
After careful check of the tensor components, error estimators, frequency behavior, a quality index is associated to each sounding. The reliability of the inversion results will be affected by the heterogeneity in this map.
Step 4: Topographic effect correction
The Martinique island is surrounded by the sea water with a significant thickness which affects the MT response by comparison with a flat earth response. Two strategies can be applied: to include bathymetry & topography in the model (at the cost of memory and computation time) or to correct the distortion effect before inversion. We have applied the second strategy, and modeled the sea effect. E and B computed at the surface of a “normal model” (horizontal flat earth) are linked to the measured E & B (with sea and bathymetry) via a pair of distortion tensors K & R: Etopo = K-1Enormal and Btopo = R-1Bnormal
Step 5: 3D resistivity model parametrization
The 3D resistivity model relies on a 3D grid: 18x16x12 cells. In red are indicated the MT stations. If the density of 1 MT station per km2 is homogeneous in the center, this density is reduced on the borders (Hautot et al.,2007). That’s why the cells are clustered in coarser cells to keep an almost homogeneous sensitivity for all the inverted cells. Thick lines correspond to constant resistivity (a cluster of initial cells). Cells are bigger at depth as presented for layers 1 to 12 (the deepest).
