search

uafgeotools / lts_array

Least trimmed squares array processing for infrasound and seismic data
MIT License
12 stars 6 forks source link

Major package update #23

Closed jwbishop closed 1 year ago

jwbishop commented 1 year ago

Major Package Update

1) Package internals re-written in an object-oriented framework that uses numba for improved computational speed. 2) Uncertainty estimates for back-azimuth and trace-velocity estimates are now calculated automatically using the slowness ellipse method of Szuberla and Olson. 3) Bug Fixes: Fixes #18, #19, #20, #22 4) Documentation Updates. User notes are now included.

liamtoney commented 1 year ago

@jwbishop can you summarize what the breaking changes are here? Looking at the documentation here it looks like ltsva() function inputs/outputs are unchanged — so I can drop this new code into my existing workflows?

liamtoney commented 1 year ago

Looking at the documentation here it looks like ltsva() function inputs/outputs are unchanged — so I can drop this new code into my existing workflows?

Hmmm, are those docs not up to date? I looked at the new docstring after pulling the code and it looks like rij has been deprecated in favor of lists of lat and lon?

liamtoney commented 1 year ago

Hmmm, are those docs not up to date? I looked at the new docstring after pulling the code and it looks like rij has been deprecated in favor of lists of lat and lon?

I made an issue: #27

jwbishop commented 1 year ago

Hi @liamtoney, thanks for this question. Setup instructions and documentation generation have now been fixed by #28.

There are two key changes that users need to know for this update! 1) Inputs: Latitude and Longitude values are now the preferred inputs. The code will now internally project onto a local planar coordinate system with a vincety inverse. 2) The code now automatically outputs 90% confidence intervals for the trace velocity and back-azimuth, so 2 additional outputs need to be accounted for in current workflows.

jwbishop commented 1 year ago

The code input/output is flexible, so I am open to user feedback.

jwbishop commented 1 year ago

Another note: the least squares estimation (for both trimmed and ordinary least squares) is now largely completed by internal codes compiled by Numba. This results in a massive speedup of the computational time, but it's unclear to me if I can output completion percentages to the external device at the same time. The most expensive part of the code is now the cross-correlations, and output info can be easily added here if users are interested.

liamtoney commented 1 year ago

it's unclear to me if I can output completion percentages to the external device at the same time

Can you explain what you mean by "external device" here?