santisoler
commented
1 year ago Hi @prince-Ha! Opening this issue is not rude at all. The idea of allowing users to open issues or discussions in our forum is to encourage the conversation within the community. Your questions and inputs are very welcomed 🙂.
Answer to point 1
After reading your notebook I noticed that your free-air anomaly data doesn't include the observation height. This height is as important as the observed gravity values: the observed gravity depends on the densities of the bodies that generate it and also on the distance to those bodies. In cell 28 you add a height_geometric_m column to your data and you use these values when computing the terrain effect as the observation heights. Those heights are the same as the topographic heights in continental regions and bathymetry depths in oceanic regions. If we assume that the gravity observations where taken through a ground survey, it makes sense to assume that the observation heights will be approximately the topographic height. But I suspect that in the oceanic regions they weren't taken in the deepest point of the sea, but rather at sea level. So the observation heights you choose might be wrong and this might be the source of this difference between the simple Bouguer correction and the terrain effect. I would also recommend you to plot the Bouguer correction along with the terrain effect so you can see their difference easier rather than seeing them directly in the corrected gravities.
Also, take into account that the simple Bouguer correction doesn't depend on observation height because it assumes an infinite plate. This is why simple Bouguer corrections should be taken carefully, while a proper terrain correction is more suitable for most cases.
Answer to point 2
Your kernel is probably crashing due to lack of memory when trying to fit the deep sources. How many data points are you dealing with? Try to monitor you ram consumption when running that step, it might give a better idea of what's going on.
If you have a lot of data points an alternative would be to use less deep sources: since they are going to be very deep, we don't need too many of them. You can do it by passing a value to the block_size argument of the EquivalentSources. By doing so your region will be divided in blocks of equal size and it will locate a single equivalent source beneath each block. This way you'll end up with less sources than before (when you put one source beneath each data point). Check the EquivalentSources documentation for a complete explanation of this behaviour: https://www.fatiando.org/harmonica/latest/api/generated/harmonica.EquivalentSources.html#harmonica.EquivalentSources
Answer to point 3
I really regret not including an example on how to correctly apply the correction that that Note mentions. I remember having this example somewhere but don't remember exactly where. @leouieda I think you wrote it, do you think where it can be?
BTW, this more complex example should be included in the Harmonica User Guide! (will open an issue for it)
Page that contains the error or needs to be improved (paste link): Hello, I know you are very busy, but I am asking you this question because I need your help. After watching the video on YouTube, I am processing the gravitational data with the fatiando tutorial.
1 . there are many differences between simple Bouguer and complete Bouguer , so I ask you what the problem is.
code : deep_sources.fit((data.easting_m, data.northing_m, data.height_geometric_m), data.gravity_bouguer_mgal)
I'm worried if it's rude to ask.
Thank you for reading this long article.
I'll attach the file together
thanks Description of the problem/suggestion:
gebco_harmonica_sandwell.ipynb.zip