04.08.2013 Views

P. Schmoldt, PhD - MTNet - DIAS

P. Schmoldt, PhD - MTNet - DIAS

P. Schmoldt, PhD - MTNet - DIAS

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

a (horizontal smoothing)<br />

10. Data inversion<br />

Period range: 10<br />

5.0<br />

4.5<br />

a5b1t3<br />

-3 (a) – 10 s<br />

(b)<br />

5.0<br />

4.5<br />

4.0<br />

3.5<br />

3.0<br />

2.5<br />

2.0<br />

1.5<br />

1.0<br />

a3b1t6<br />

a2b1t310<br />

0 1 2 3 4 5 6 7 8 9 10<br />

t (global smoothness)<br />

a (horizontal smoothing)<br />

4.0<br />

3.5<br />

3.0<br />

2.5<br />

2.0<br />

1.5<br />

1.0<br />

Period range: 10 -3 – 100 s<br />

a5b1t3<br />

a3b1t6<br />

a2b1t310<br />

0 1 2 3 4 5 6 7 8 9 10<br />

t (global smoothness)<br />

RMS<br />

misfit<br />

6.0<br />

Fig. 10.1.: RMS misfit for models of the Tajo Basin crustal structures with different combinations of global (τ) and horizontal smoothing<br />

parameters (α), using (a) a period range of 10 −3 −− 10 s and (b) of 10 −3 −− 100 s. Values are obtained through linear interpolation<br />

of 42 α,τ-combinations for each frequency range with applied parameters summarised in Table 10.1. Red-yellow markers, together<br />

with the respective model name, indicate combinations chosen for further investigations.<br />

These six models (three for each period range) are kept for further investigation in subsequent<br />

inversion steps.<br />

10.1.2. Starting model construction<br />

Results of non-stochastic MT inversion are, in general, dependent on the starting model.<br />

For an unsuitable starting model the inversion may yield a model related to a local, rather<br />

than the global, misfit minimum (cf. Sec. 6.3). A uniform halfspace, often 100 Ωm,<br />

is commonly used as a starting model for MT inversion as it is assumed firstly to be<br />

less likely to introduce inversion artefacts, and secondly to be relatively close to average<br />

resistivity values of the Earth at crustal and upper mantle depths (cf. Sec. 5.2). However,<br />

using a priori information about the subsurface, e.g. from 1D inversion of MT data or<br />

results of other geophysical or geological investigations, an enhanced starting model can<br />

(and should be) created that is superior to the halfspace model approach. For the Tajo<br />

Basin, supplementary information about the subsurface is available, in particular, from<br />

seismic reflection and refraction studies, inferring four layers within the crust and a Moho<br />

depth of approximately 31 km (see Table 7.2 for thickness of crustal layers). Presently,<br />

there is no reliable relation between seismic velocity and electric resistivity at crustal<br />

depth that allows for a straightforward construction of a MT starting model from seismic<br />

velocity results. Instead, electric resistivity values for the four crustal layers are inferred<br />

from horizontally averaged resistivities of the six models chosen during the smoothing<br />

parameter analysis (Sec. 6.3.3) using an in-house script developed by Mark Muller. The<br />

script yields mean and standard deviation resistivity values for each row of the model<br />

within the specified horizontal region of the profile.<br />

228<br />

5.5<br />

5.0<br />

4.5<br />

4.0<br />

3.5<br />

3.0<br />

2.5<br />

2.0<br />

1.5<br />

1.0

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!