GNGTS 2021 - Atti del 39° Convegno Nazionale

GNGTS 2021 S essione 3.1 390 ATEM is ongoing: the development of an effective training dataset and the associate NN can be performed before the survey or can be based on the outcomes from the first flight(s) of the survey if the area is assumed to be relatively “stationary”, and, once the NN is available, reliable results can be instantaneously obtained just after each flight. This can lead to real-time rearrangements of the original tentative survey plans in order to maximize the VoI (Value of Information) of the measurements to be further collected. 2. In a more conservative perspective, the NN speed can be useful for effective Quality Check (QC) of the data during the survey. 3. The availability of a good starting model (derived from the NN inversion) can be used to speed-up the 1D deterministic inversion by reducing the number of iterations. Of course, if, for producing the final results, post-processing analyses are necessary (e.g., in 3D environments), the same will be true also when adopting the proposed NN strategy: NN based on a 1D forward modelling approach cannot generate better results compared with the deterministic inversion; NN can only provide solutions of similar quality (but in a fraction of the time and by using cheaper and more flexible computational resources). Conclusions We present a novel approach to the inversion of ATEM data based on NNs. We show the effectiveness of the NN inversion strategy by testing it on both synthetic and field data; based on the outcomes, we conclude that the proposed approach is capable of retrieving the conductivity distribution of the subsurface from the measurements collected by the airborne geophysical systemwithanaccuracy that is comparablewith themost commonly used (in the academia and in the industry) inversion strategies and that relies on 1Ddeterministic inversion approaches. These results are particularly noticeable as the NN inversion takes only a fraction of the time required for the deterministic inversion (a few seconds on a standard laptop versus hours on a computational server). Fig. 2 - (a) The NN inversion of the field data. (b) The 1D deterministic inversion of the field data.