number of generated events

[hadi]

Hi All,

Some general questions on OQ computational functionality:

is there a limit to the number of events and locations when doing a
run?

is it possible to get any timing information?

Thank you in advance,

Hadi

[Damiano Monelli]

Hi Hadi,

On Oct 5, 2011, at 12:53 AM, hadi wrote:

> Hi All,
>
> Some general questions on OQ computational functionality:
>
> is there a limit to the number of events and locations when doing a
> run?

What type of calculator are you referring to?
If you refer to the event-based PSHA calculator, number of events corresponds to number of earthquakes generated through Monte Carlo sampling of the earthquake rupture forecast (ERF).
The number of events generated depends on the time span and the occurrence rates associated to each source. In general, greater is the time span and the occurrence rates, larger is the number of ruptures
generated in a seismicity history. Now in OpenQuake you can compute as many seismicity histories as you want. There is not hard-coded limit. Larger is the number of seismicity histories, higher will be the stability of your results (e.g loss curves). My suggestion is to perform a convergence analysis. You increase the number of seismicity histories gradually and you stop when you get to stable results.
Regarding number of locations, you can consider as many locations as you want. Again there is not an hard-coded limit.

However, even if OpenQuake does not provide any constrain in the number of seismicity histories and locations, you can have some constrains from the hardware where you are running OpenQuake. For instance, OQ makes an extensive usage of RAM, to store for instance ground motion fields that are then passed to the risk engine, so if you consider many seismicity histories (so many ground motion fields) and many locations (and therefore each ground motion field contains many points) you will probably saturate you RAM (if you do not have enough), and the calculations will get very slow.

Regarding the time, we are working on that. We will first implement a calculation progress report mechanism. As you see currently OQ reports only when the calculation starts and when the calculation finishes but it does not report the calculation progress (so you cannot get an idea about how fast you are computing ground motion fields or hazard curves). Once we will have this mechanism in place we will also estimate the computation time.

Hope it can helps,

Damiano