GNGTS 2014 - Atti del 33° Convegno Nazionale

36 GNGTS 2014 S essione 3.1 inversion is not possible. This induces to an overestimation in retrieving shear wave velocities, that can be dangerous for seismic hazard evaluation. The proposed data processing method, based on refraction microtremors L-shaped arrays (LeMi), can easily and appropriately solve the problem. Refractions Microtremors with L-shaped arrays. The processing of passive data on two dimensional arrays is well known and widely discussed in the literature (Okada, 2003). Some of the processing techniques can be considered extensions of the transform‑basedmethods typically used in active multichannel surface wave testing. As conventional MASW data are transformed fromT-X to F-K (or F-P, F-V), with two dimensional arrays the data are transformed fromT-X-Y into the F-Kx-Ky domain. The transforms can be FKK transforms or other 2D beamformers. In the following we show the results obtained with different 2D arrays in processing non‑uniform wavefield for engineering practice both for synthetic and real cases. The objective of this work is not the evaluation of optimal arrays for passive surface wave testing, but simply suggest 2D array procedure consisting of two linear‑arrays evenly spaced receivers. In our proposal the array geometry consists of two straight branches, to simplify the deployment of receivers and cables and the surveying: the use of evenly spaced arrays is chosen to simplify the processing, and to allow the use of the spectral analysis methods typically available in commercial software. An L-shaped array has the advantage of allowing the acquisition of active multichannel data with a limited extra effort. A far-field linear plane wave propagating across a dual-linear array is detected with two different apparent wavenumbers by the two branches: each identifies the apparent velocity and apparent wavenumber in its direction assuming that a single plane wave is recorded simultaneously. If the two directions are orthogonal, and they correspond to a local reference systems x-y, the true wavenumber can be determined simply as K tru e = sgrt ( kx 2 – ky 2 ) (1) In the case of passive measurements, this allows to overcome the trouble issue of oriented noise source, not easy detectable on site. If a source direction is strongly dominant on the recorded data, then the two averaged measurements represent the same direction and can be combined as in Eq. (1) to retrieve the true wavenumber K true . From true wavenumber we can define the surface wave phase velocity versus frequency dispersion property and then, after the inversion process, the shear wave profile. In the synthetic case of Fig. 1a it is visible how a predominant unknowns orientated source make impossible the common ReMi linear array detection of the true wavenumber (and then of the true seismic velocity). In this peculiar case, considering the y branch linear orientation could bring to an over-estimation of the seismic velocity of 10-20% because the array is almost perpendicular to the main noise source (Fig. 1b). Since orientation noise source is unknown, one operator can potentially consider to use only a linear array as the x branch, leading to a huge misleading over-estimation of the velocity (Fig. 1b) that can have serious consequence in seismic hazard studies. The use of an L-shaped array can on the contrary retrieve the true wavenumber (and then the true velocity) starting from any orientation of the perpendicular branches (Fig. 1b). The combination of the results of two linear arrays can overcome the limitations of ReMi in cases of strongly directional sources. If the sources are stationary, the acquisition of the two arrays could even be done separately, one after the other, processing the arrays separately for the best k resolution. Since common seismograph channels numbers adopted in engineering practice is increasing, it is however recommended to acquire the arrays simultaneously. This ensures the absence of variation of the dominant sources, and allows, if needed, a proper 2D processing as beamformer analysis (Capon, 1969). We do not propose complex theoretical approach but provide a simple modification of the ReMi procedure that can be successfully implemented by practitioners in the field and still be rigorous enough to provide useful estimates of true shear wave velocity. Real data L-shaped array test. The test site was located in Badia Polesine (Ro), N-E Italy, in the southern part of the ‘Po’ river plain. The site was selected for its geological conditions,