Difference between revisions of "UCVM Development"
Line 1: | Line 1: | ||
+ | == Prototype Description == | ||
+ | *[http://hypocenter.usc.edu/research/UCVM/Gill_ucvm_19Aug2016.pptx UCVM Prototype Overview - David Gill (pptx)] | ||
+ | |||
== Development Plans == | == Development Plans == | ||
We are defining a minimal viable product (MVP) version of UCVM. The expected capabilities include: | We are defining a minimal viable product (MVP) version of UCVM. The expected capabilities include: |
Revision as of 22:54, 26 August 2016
Contents
Prototype Description
Development Plans
We are defining a minimal viable product (MVP) version of UCVM. The expected capabilities include:
- Support for these models:
- CVM-S4 (already in prototype)
- CVM-S4.26
- CVM-S4.26.M01
- CVM-H
- CenCal
- CCA (Po and En-Jui’s model)
- Imported DEM (part of DEM already in prototype), Vs30 data
- ucvm_query capability (already in prototype)
- UCVM meshing capabilities
- multi-core support, single-core, MPI
- Write/read from AWP-ODC style meshes (binary floats)
- Write/read from E-tree format (Hercules support)
- Plotting capabilities
- Easy installation support with pip/easy_install
- Significantly improved documentation and references
- E-tree meshing for UCVM
- Plotting utilities
- Possibly SSH
- Unit Test suite
- Acceptance test suite
- Release notes and documentation
July 18, 2016
Produced an example script showing how to query a velocity model (CVM-S4.26 in this case) and generate a suite of 1D profiles, including just Vs, just Vp, just density, and all material properties on one. Also, this extracts the raw data and provides it in text format for the user.
The depth profiles unfortunately do not have a nice command-line interface for generating them. This was a capability that was originally intended for UCVM 14.7.0. It’s in the API, hence why it is accessible via Python. The simplest way to generate a profile is:
from pycvm import DepthProfile, Point depth_profile = DepthProfile(Point(-118, 34, 0), 1000, 100, “cvms") depth_profile.plot(“vs”)
That will produce a 1D Vs plot from 0 (surface) to 1000m depth with interval spacing of 100m from CVM-S4 at point (-118, 34). The example scripts break things down into variables that are easy to edit.
July 12, 2016
Generated CVM-H 1D profile showing 100 meter voxel spacing:
July 11, 2016
Updated the wiki to detail the final 1D Central California velocity model, including how to get the script to re-generate the 1D model and the small modification needed to UCVM to run it:
July 1, 2016
Added top 5km plots to CCA model 1D page:
Started UCVM Prototype page:
June 30, 2016
1D Plots for CCA model revisions:
June 27, 2016
Plots showing a similar color scheme and axis style as the SRL paper have been generated:
These plots use a new plotting code that is more robust and will help to form the basis of a new UCVM. This code uses:
- Python unit tests for testing (including coverage tool)
- Sphinx for Documentation
June 20, 2016
1D profiles for the two 1D models (the SCEC CVM-S4 modified HK model and the Whittier-Narrows BBP model) have been generated.
EL-style plots:
June 17, 2016
1D profiles for the entire CCA region, the CCA region minus any grid point in water, and the CyberShake box minus any grid point in water, are now complete.
June 13, 2016
Average slowness for Vp, Vs, and average density computed for both CCA Iteration 0 and CCA Iteration 6. Note these plots are slightly mislabeled - they should say depth in meters, not km.
Plots for CVM-S4.26 and CVM-S4.26.M01 are completed.
June 8, 2016
The same plots as May 26 have been done with just the CCA model as opposed to the CVM-S4.26, Cencal, 1D model before:
For 70km depth as before:
June 6, 2016
Tests for UCVM have been completed. The results of the tests, including what was actually tested, is located here:
May 26, 2016
Verification that we can reproduce RG plots was completed: