v18p5
The hazard calculations for v18p5, which contains all crustal shallow finite faults on NZ is based on the post-processing code results. The post-processing produces a SQLite database that is queried to obtain for a station every value for a given IM. This is quite straightforward as this uses several SQL queries to achieve the result.
v18p6
The SQLite database has been abandoned. We now use the IM_calculation code which produces an IM csv file for each realisation (run name).
After running the IM calculation we end up with the following structure:
<run_group_dir>/<fault_name>/IM_calc/<run_name>/<run_name>.csv
IM aggregation is run on these by running the command:
IM_calculation/aggregation/im_agg.py <run_group_dir>
Which produces the following outputs (one file per IM, im_name):
<run_group_dir>/<fault_name>/IM_agg/<im_name>.csv
Each csv contains:
station, run_name_1, run_name_2, ..., run_name_N station1, value1_1, value2_1, ..., valueN_1 station2, value1_2, value2_2, ..., valueN_2 .... stationM, value1_M, value2_M, ..., valueN_M
Size comparison
I have re-run the IM calculations on v18p5 (North Island faults). The new format produced 3.6G of data, whereas the DB based solution used 9.5GB.