Working Group leader: Martin Schimmel
“Noise correlation” theorems relating correlation functions and Green's functions can be proved for fully random source distributions or equipartitioned scattered wave fields. Indeed, actual seismic noise does not satisfy these conditions perfectly.
The conditions of emergence of the different types of waves need to be further studied. The fact that the existence of scattering improves the Green's function reconstruction has been verified for short period waves. The influence of scattering at longer periods for long time series has to be studied carefully.
The issue of the role of scattering is of first importance for time dependent imaging since multiple scattered waves are essentially weakly dependent on the noise source fluctuations.
Also, the origin and location of noise sources are not yet completely understood and must be investigated. Quantifying the effects of scattering requires studies from different perspectives. Beyond the analysis of seismological data, laboratory experiments in controlled complex media have to be performed to understand the physics of multiply scattered random fields and their correlation properties.
As far as possible, numerical simulations in models based on the background properties of the Earth (average speeds, layering) are also required. Noise based imaging and monitoring would benefit from a better quantitative description of scattering properties such as mean free paths, even in regions and frequency bands for which scattering is not a problem for direct wave imaging.
Different strategies of signal processing can be adopted for the computation of the longterm averages of the correlations that are identified eventually as the Green's function or components of it. In the light of the mathematical results on which the noisebased imaging and monitoring are based, the processing aims at forcing the stationarity of the initial records by various normalization techniques.
The correlations are computed for limited time windows and then stacked to obtain a satisfactory level of convergence towards the Green's function. Prestack adaptive filtering schemes have been applied based on different approaches. These processing steps have to be evaluated for different contexts and frequency ranges and their performances compared for imaging and monitoring.
New challenges for the processing are to consider simultaneously the multicomponent records of dense arrays and to apply array processing advanced tools. First applications of noisebased monitoring actually rely on spatial averaged properties of the speed of seismic waves. New methods emerged to provide 4D models of the changes, that is, including a step of spatial imaging of the changes.
It was recently shown that other parameters such as the local crosssection of the materials could be monitored as well with coda waves. This measurement is performed from the temporal variation of the coherence of noise cross correlations.
These imaging approaches are based on sensitivity kernels of coda waves computed with diffusion or radiative transfer theories. So far strong hypothesis were made for the computation of the kernels (scalar waves, homogeneous background,..). The development of more realistic kernels is required to improve the time dependent imaging.