Thanks for the follow up. For context, here is my discussion with Jack Baker regarding the benefit of a variable DELTA_MAG value. At this point in time, I do not foresee pursuing this any further.
With regards to looking up the parameters for other ERFs, I may return to the question of mag-area relationships later. Thank you for the tip.
Anne:
I wanted to follow up on my question from the presentation regarding whether each scenario is sampled multiple times.
I’m wondering what would be lost if you don’t use multiple samples. My gut reaction is that if you have one rupture that controls the lower tail of the hazard curve, you would have trouble if you end up taking a single realization that is either really low or really high. Taking many samples would do a better job of P(Y>y|scenario). The counter to that concern would be that if you really have all the scenarios represented, you should have pretty similar scenarios that even out the problem of using only one sample per scenario. (E.g. 7.8 and 7.9Mw on San Andreas)
Did Mahalia find that using a single sample was comparable to using multiple samples?
Baker:
Good question, and it seems like you have a pretty good handle on it. In general, I would say that there's not a lot of benefit to generating multiple ground motion samples per rupture. More samples can give better resolution of the tail. But I you want more samples, I would prefer to sample more ruptures as well.
That said, the OpenSHA rupture generator actually discretizes the possible ruptures instead of doing a traditional Monte Carlo sampling. So getting more ruptures entails changing some of the OpenSHA parameters (e.g., sampling magnitudes in intervals of 0.05 instead of 0.1). So Mahalia decided that a convenient way to get more samples was just to keep her basic set of ruptures and sample multiple ground motions per rupture. It was a fine thing to do, but I would call it a pragmatic choice rather than a theoretically necessary choice.
