Ray path inaccuracies

[ncvangilse]

I am using skfmm to simulate seismic ray paths and have discovered some inaccuracies when I back trace the rays from seismic receivers. I issued it as a bug at the github yesterday and now I found this forum, so I thought I would just post a link to the issue in here as well. 

I suspect that it is due to the zero-contour is not a point source?

https://github.com/scikit-fmm/scikit-fmm/issues/80