Coda-Q in the 2.5s -20s period band from seismic noise - Application to the greater Alpine area

International audienceCoda-Q is used to estimate the attenuation and scattering properties of the Earth (Aki & Chouet 1975). So far focus has been on earthquake data at frequencies above 1 Hz, as the high noise level in the first and second microseismic peak, and possibly lower scattering coefficient, hinder stable measurements at lower frequencies. In this work, we measure and map coda-Q in the period bands 2.5 s-5 s, 5 s-10 s and 10 s-20 s in the greater Alpine region using noise cross-correlations between station pairs, based on data from permanent seismic stations and from the temporary AlpArray experiment. The observed coda-Q for short interstation distances is independent of azimuth so there is no indication of influence of the directivity of the incoming noise field on our measurements. In the 2.5 s-5 s and 5 s-10 s period bands, our measurements are self-consistent, and we observe stable geographic patterns of low and high coda-Q in the period bands 2.5 s-5 s and 5 s-10 s. In the period band 10 s-20 s, the dispersion of our measurements increases and geographic patterns become speculative. The coda-Q maps show that major features are observed with high resolution, with a very good geographical resolution of for example low coda-Q in the Po Plain. There is a sharp contrast between the Po Plain and the Alps and Apennines where coda-Q is high, with the exception a small area in the Swiss Alps which may be contaminated by the low coda-Q of the Po Plain. The coda of the correlations is too short to make independent measurements at different times within the coda, so we cannot distinguish between intrinsic and scattering Q. Measurements on more severely selected datasets and longer timeseries result in identical geographical patterns but lower numerical values. Therefore, high coda-Q values may be overestimated, but the geographic distribution between high and low coda-Q areas is respected. Our results demonstrate that noise correlations are a promising tool for extending coda-Q measurements to frequencies lower than those analysed with earthquake data