CCA06 Projection Review
Information about the CCA06 Projection Correction.
Contents
Issue documentation
A description of the issue and the fixes are collected here: https://scec.usc.edu/scecpedia/CVM_Projection_Issue_Discussion
Current resolution
Three CCA06 data config that contain slightly different lat/lon points for the four corners of the mesh. The results below show how ucvm_query results differ,using each data config file for the CCA model.
- The original one that came from CCA github repository.
- The fixed one that came from UCVM github repo under model name CCA06.
- A different from that is supplied from Scott that he has used.
All three of them are different
Result for CCA 06
| Algorithm | orig config | fixed config | Scott's config |
|---|---|---|---|
| First Crossing | data | data | data |
| First Or Second Crossing | data | data | data |
commands used:
./plot_z25_map.py -b 33.35,-123.0 -u 39.35,-115.25 -c cca -s 0.005 -f cca_z2.5.first -o cca_z2.5_first_map.png -x "(first crossing)"
./plot_z25_map.py -b 33.35,-123.0 -u 39.35,-115.25 -c cca -s 0.005 -f cca_z2.5.firstOrSecond -o cca_z2.5_firstOrSecond_map.png -x "(first or second crossing)"
and
mpirun -np 20 ./basin_query_mpi -b ${TEST_TOP_PATH}/RESULT/basin_query_mpi_cca_z2.5_n20.first, ${TEST_TOP_PATH}/RESULT/basin_query_mpi_cca_z2.5_n20.firstOrSecond, ${TEST_TOP_PATH}/RESULT/basin_query_mpi_cca_z2.5_n20.last, ${TEST_TOP_PATH}/RESULT/basin_query_mpi_cca_z2.5_n20.secondOnly, ${TEST_TOP_PATH}/RESULT/basin_query_mpi_cca_z2.5_n20.threeLast -o ${TEST_TOP_PATH}/RESULT/basin_query_mpi_cca_z2.5_n20.result -f ucvm.conf -m cca -i 20 -v 2500 -l 33.35,-123.0 -s 0.005 -x 1551 -y 1201
Results for CVM-S4.26
Several CyberShake hazard models were calculated using CVM-S4.26 (and CVM-S4.26-M01). These were done before the projection issue was identified and fixed, so the expectation is that all these CyberShake models were based on the "uncorrected" CVM-S4.26.
However, a projection correction for CVM-S4.26 was calculated, and it is included in the two UCVM releases made in 2017, including the c-language UCVM v17.1 and python language UCVM v17.3. This could cause small changes if we try to reproduce previous CyberShake hazard models, with new versions of UCVM.
