AEP and OEP curves from risk results (event based method for 100 ses in 1 year)

[chrisschn...@gmail.com]

Hey community,

just a short question referring to AEP (Aggregate Loss Exceeding Probability) and OEP (Occurrence Loss Exceeding Probability) curves.
Am I right that the table aggregate_loss_curve_data contains my AEP curve?

In term of OEP curve I've just searched (with python script) the rupture with hugest loss per stochastic event set from table event_loss_data .
Afterwards I calculated the recurrence period an then the probability of exceedance.


Best regards,
Chris

[Luigi Panzeri]

Hi Chris,

Yes, you are right! Aggregate Loss curves are stored in that table.

Regards,

Luigi

2013/9/10 <chrisschn...@gmail.com>

--
You received this message because you are subscribed to the Google Groups "OpenQuake Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openquake-use...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Helen Crowley]

Dear Chris,

We probably create a bit of confusion with the use of “aggregate” loss curve as we mean the aggregate (or total) loss across all of the assets in the portfolio. We do not mean the aggregate loss to the assets within a given time period (e.g. a year), which is what I understand AEP refers to in the insurance industry.

We calculate the loss exceedance curve considering all events – we do not aggregate losses within a time period and we do not take the maximum loss within a given time period.

Indeed, for now, it is probably best if you post-process the event loss table yourself to get the type of loss curve that you need. We hope to build in features to extract the different types of loss curves that are used in the insurance industry in future versions of the OpenQuake engine.

I hope this helps.

Thanks,
Helen

---

Da: Luigi Panzeri <defm...@gmail.com>
Risposta: <openqua...@googlegroups.com>
Data: Tue, 10 Sep 2013 14:31:38 +0200
A: <openqua...@googlegroups.com>
Oggetto: Re: AEP and OEP curves from risk results (event based method for 100 ses in 1 year)

To unsubscribe from this group and stop receiving emails from it, send an email to openquake-use...@googlegroups.com <mailto:openquake-users%2Bunsu...@googlegroups.com> .

For more options, visit https://groups.google.com/groups/opt_out.

[chrisschn...@gmail.com]

Hey Helen, hey Luigi,

thanks a lot for the response.

Regards,
Chris

[Eugenio Dissegna]

Hey community,

I write in this topic as a similar issue was already posted here.

I run a risk calculation, event based PSHA, time span 1 year. The stochastic event simulations (ses) are 50, while the number of logic tree samples is just 1 (my hazard area is rather simple). The total value of the assets in this area is around euro 23 000 000.

In the results I get, the event loss table shows me there have been 5 events. I interpret this as: in 50 simulations (the stochastic event sample), 5 times, with a 1-year horizon, an earthquake happened. Is this correct?

Moreover, the highest loss in the table is higher than the total value of the assets in the region (around 50% higher). How is this possible? With 1-year horizon shouldn’t the maximum loss be the total value of the assets (which means, in the worst event all the assets are fully demolished)?

Thank you,

Eugenio 

[Luigi Panzeri]

Hi Eugenio,

What you are experiencing looks strange.

Could you share your input files?

2013/10/2 Eugenio Dissegna <eugenio....@gmail.com>

[Eugenio Dissegna]

Hi Luigi,

I attach you the files I used to run the risk calculation. 

Do let me know if other files can be useful (I can attach you also results of calculations).

 

Thanks a lot,

Eugenio

[Luigi Panzeri]

Hi Eugenio,

I have just noticed that in your exposure file you might have misinterpreted the meaning of the attribute "number" (which stands for "number of units"). It seems that you have used it as an asset ordinal. Bear in mind that "number" can be used to modify the considered replacement cost of an asset depending on the cost conversion you have specified.

More precisely, if you write

<costType name="structural" unit="EUR" type="per_asset"/>

and then

<asset id="a11" taxonomy="B" number="7">

  <location lon="9.18" lat="45.47"/>

  <costs>

    <cost type="structural" value="200000" deductible="10000" insuranceLimit="0.5"/>

  </costs>

</asset>

The considered final replacement cost of asset a11 is 200.000 * 7 = 1.400.000 (the same applies to the deductible and insuranceLimit when absolute values are used).

That could explain why you are getting large loss values as the total value of the considered exposure is actually far larger than 23M. 

Instead, I guess you may want to write:

<costType name="structural" unit="EUR" type="aggregated"/>

and then drop the attribute "number" from each <asset> definition.

Best,

Luigi Panzeri

2013/10/3 Eugenio Dissegna <eugenio....@gmail.com>

[Eugenio Dissegna]

Dear Luigi,

thanks a lot, I really got it wrong! Now results are consistent.

I've also seen that insured loss curves can be produced. In this file, I understand the term "losses" refers to 

the loss ratio of insured assets (insured loss/total value of the asset) for each poEs of the asset. 

Do you know what the values in the "lossRatios" and "averageLoss"cell refer to? They are between 0 and 1 so 

I guess they are loss ratios instead of absolute losses, but I don't get what's the numerator.

Many thanks,

Regards

Eugenio Dissegna 

[Vitor Silva]

Dear Eugenio, 

OpenQuake can indeed calculate the direct losses (also commonly named ground-up losses) and insured losses. The calculation of the insured losses is performed based on a deductible and a limit value, as defined within the exposure model. 

The loss ratios curves contain the various loss ratio fractions and corresponding loss probability of exceedance. You can interpret this curve as the loss curve divided by the associated asset value. The average loss is calculated by summing the product of each loss ratio by the corresponding probability of occurrence, taken from the loss curve. 

For additional information I recommend you to read the OpenQuake book and manual which contain a comprehensive description about the methodologies and assumptions implemented within OpenQuake (available on the OpenQuake website). 

Best regards,

Vitor

[Eugenio Dissegna]

Dear Vitor,

thank you very much for your answer.

Kindest regards,

Eugenio Dissegna