Clarifications on some commands

[Athanasios Papadopoulos]

Dear OQ team,

Just a few questions regarding a couple of OQ commands/options:

-          oq reaggregate: Is it possible to compute aggregate losses for a tag that is present in the exposure database but not given in the “aggregate_by” field of the job.ini file that was run? My problem is that I have 5 tags (different admin levels, occupancy, taxonomy) and, if I enter all of them in the aggregate_by field, the calculation fails saying that the combinations (some hundred thousands) of tags is over the limit. I am not interested in the combinations of tags however, just in the aggregate results per tag.

-      oq engine --reuse-input: If I understand well, this has replaced the reuse-hazard option. What would be the correct syntax of the command for running an ebrisk calculation? It seems that I can’t make it work like the reuse-hazard option. Some information on how it works would also be immensely helpful (e.g. what should be included in the ebrisk job.ini file if the option is used, and what takes precedence, e.g. if I enter the –reuse-input option but specify a different exposure or hazard configuration in the job.ini file).

Many thanks in advance for your help.

Cheers,
Thanasis

[Michele Simionato]

On Tuesday, January 19, 2021 at 9:57:22 AM UTC+1 Athanasios Papadopoulos wrote:

Dear OQ team,

Just a few questions regarding a couple of OQ commands/options:

-          oq reaggregate: Is it possible to compute aggregate losses for a tag that is present in the exposure database but not given in the “aggregate_by” field of the job.ini file that was run? My problem is that I have 5 tags (different admin levels, occupancy, taxonomy) and, if I enter all of them in the aggregate_by field, the calculation fails saying that the combinations (some hundred thousands) of tags is over the limit. I am not interested in the combinations of tags however, just in the aggregate results per tag.

In the released version of the engine you have to run 5 computations, each one with a different tag in the aggregate_by. In the master version of the engine there have been optimizations so that one can actually run a calculation with thousands of combinations and then use reaggregate 5 times, which will be much faster than running 5 complete calculations.

-      oq engine --reuse-input: If I understand well, this has replaced the reuse-hazard option. What would be the correct syntax of the command for running an ebrisk calculation? It seems that I can’t make it work like the reuse-hazard option. Some information on how it works would also be immensely helpful (e.g. what should be included in the ebrisk job.ini file if the option is used, and what takes precedence, e.g. if I enter the –reuse-input option but specify a different exposure or hazard configuration in the job.ini file).

--reuse-input allows to reuse the source model and the exposure, but since most of the time in the ebrisk calculator is spent in generating the GMFs that lives in memory and cannot be reused, there is no much that you can do.

With an event_based_risk calculation the GMFs would be stored and then you could reuse them with the ---hc option. For small calculations this efficient, for large calculations it is better to use the ebrisk calculator instead,

i.e. recomputing the GMFs each time can beat reading them from the filesystem, even it is not intuitive at all.

HTH,

                             Michele

[Athanasios Papadopoulos]

Great! Thanks for the quick reply and info!

Best,

Thanasis