Stochastic Event-based Risk Analysis - ELT vs YLT

[houman...@gmail.com]

Dear users and developers,

As you may know there are two types of loss table results used in insurance catastrophe modelling: ELT and YLT (samples are attached). These are my qustions:

1. Which one is actually generated by OQ? The aggregate loss table produced by OQ does not contain Mean and SD values (as in ELT). There is also no time tags (as in YLT) in it.

2. How can I make OQ simulate say 10,000 consecutive years of events (containing 0, 1 or more events in each year as in YLT sample)? As far as I understood ses_per_logic_tree_path and investigation_time parameters need to be determined in job.ini. The qustion is: what are the exact values?

I appreciate your help in advance,

Best regards,

Hooman

[anirudh.rao]

Dear Hooman,

To answer your second question first: in order to simulate 10,000 years of events, you would need to set:

investigation_time = 1

ses_per_logic_tree_path = 10000

With these values, OpenQuake will simulate 10,000 "event-sets", with each "event-set" spanning one year, and where each "event-set" could contain 0, 1, or more events.

Below is an example of an aggregate loss table produced by OpenQuake for a calculation using

ses_per_logic_tree_path = 100000 and investigation_time = 1

Event #	Tectonic Region	Event Set	Source ID	Rupture ID	Portfolio Loss
1	trt=00	ses=00005	src=PS2950	rup=122892-01	2.57E+07
2	trt=00	ses=00010	src=SFS5223	rup=717079-01	4.26E+09
3	trt=00	ses=00011	src=PS4972	rup=399296-01	1.46E+08
4	trt=01	ses=00013	src=PS4393	rup=321882-01	5.17E+09
5	trt=00	ses=00014	src=PS4231	rup=299323-01	1.62E+07
6	trt=00	ses=00014	src=SFS7094-0	rup=719606-01	1.97E+08
7	trt=00	ses=00015	src=SFS7094-3	rup=720039-01	2.35E+07
8	trt=00	ses=00015	src=PS5174	rup=424020-01	5.90E+08
9	trt=00	ses=00023	src=PS4313	rup=310753-01	7.22E+07
…	…	…	…	…	…
43924	trt=00	ses=99996	src=PS5067	rup=410979-01	1.84E+09

In this case, since each event-set spans one year, the Event Set tags are identical to the simulation year (eg. ses=00023 corresponds to year 23). The event-sets for which no events are generated are not listed in the table. You will notice that years 14 and 15 both generated two loss producing events. Hopefully, this should also answer your first question.

Best regards,

Anirudh Rao

[Hooman M]

Dear Anirudh,

I hope this message finds you well.

Since I am performing Probabilistic Event-Based Risk Analysis for the entire country (Iran), I had to reduce the number_of_ses_per_logic_tree_path to 1000 and number_of_logic_tree_samples to 1. This is the relevant part of the job.ini I use:

--------------------------------------------------------------------

[hazard calculation]

random_seed = 24

truncation_level = 3

maximum_distance = 200

investigation_time = 1

number_of_logic_tree_samples = 1

ses_per_logic_tree_path = 1000

minimum_intensity = 0.1

[risk calculation]

master_seed = 42

risk_investigation_time = 1

[risk output]

avg_loss = true

insured_loss = false

quantile_loss_curve = 0.995

conditional_loss_poes = 0.0021

-------------------------------------------------------------------

The problem is that all I get is one aggregate loss of all my assets for a limited number of return periods (agg_loss-rlz-000.csv) or one average loss for each asset (loss_maps-rlz-000). I would appreciate if you tell me how I can generate a detailed event loss table containing columns for eid, sid, lon, lat, structural_loss and rows corresponding to each realisation (asset event loss table).

Best regards,

Hooman