Traveltime sensitivity kernels: Banana-doughnuts or just plain bananas?

Conclusions: does it matter?

Practitioners of traveltime tomography typically understand the shortcomings of ray-theory; although they realize using ``fat-rays'' would be better, they smooth the slowness model both explicitly and by regularizing during the inversion procedure. In practice, any shortcomings of traveltime tomography are unlikely to be caused by whether or not the fat-rays are hollow.

However, the null space of seismic tomography problems is typically huge. Smoothing and regularization are often done with very ad hoc procedures. Understanding the effects of finite-frequency through sensitivity kernels may lead to incorporating more physics during the regularization and improve tomography results.