Abstrakt
Aims. Recent studies have shown that time-distance inversions for flows start to be dominated by a random noise at a depth of only a few Mm.
It was proposed that the ensemble averaging might be a solution for learning about the structure of the convective flows, e.g. about the depth structure of supergranulation. Methods.
Time-distance inversion is applied to the statistical sample of similar to 10(4) supergranules, which allows the inversion cost function to be regularised weakly about the random-noise term and thus provides a much better localisation in space. We compare these inversions at four depths (1.9, 2.9, 4.3, and 6.2 Mm) when using different spatio-temporal filtering schemes in order to gain confidence about these inferences.
Results. The flows inferred by using different spatio-temporal filtering schemes are different (even by the sign) even though the formal averaging kernels and the random-noise levels are very similar.
The inverted flows changes its sign several times with depth. I suggest that this is due to the inaccuracies in the forward problem that are possibly amplified by the inversion.
It is also possible that other time-distance inversions are affected by this.