Unexpected deaggregation results
41 views
Skip to first unread message
francesco....@ingv.it
Nov 7, 2017, 5:31:59 AM11/7/17
to OpenQuake Users
Hi,
we run OQ 2.7 installed using yum, on Centos.
I made a run to calculate deaggregation, OQ run to completion and the output files were generated.
The elaboration used 3 seismicity models and 3 GMPEs (i.e. 9 realizations),
to evaluate 16 points on a regular grid (spaced 0.1°, in the square 13.1-42.2 -- 13.4-42.5)
for some imt (PGA, SA(0.2), SA(1.0), SA(2.0) )
and two poe (0.1, 0.02).
We asked deaggregated on magnitude between 4.75 and 7.75, using intervals of 0.5
and distance between 2.5 and 152.5, using intervals of 5
We were interested only in Mag-Dist deaggregation.
Every output file generated by OQ is relative to one point, one realization, one imt, one poe, one type of disaggregation.
We found 2 unexpected behaviour.
1)
We expected to find the same number of rows for every disaggregation of type Mag_Dist,
i.e. the same number of couples Mag-Dist with the associated deaggregated poe,
but it was not the case.
Even taking, e.g., two file with the same realization, imt, poe, we have difference from point to point.
From the sample files we checked
we found in some files the distance value starts at a higher value (e.g. form 7.5)
or stop at a lower value (e.g. 17.5), or both,
and the magnitude value stop at a lower value (e.g. 6.75)
Also, we considered the files related to points 13.1-42.2 and 13.4-42.2
which "see" the same number of points of elaboration at the same distances,
but again we found differences.
I also note that there are rows for values of minimum magnitude and distance with poe=0
4.75000E+00,2.50000E+00,0.00000E+00
therefore a value of poe=0 is not a reason for not having the associated row.
2)
We expected that the sum of the deaggregated poe within a file is the same of the poe the file refers to.
What we found (at least in one case) is that the file is named
poe-0.02-rlz-8-SA(2.0)-13.4-42.5_Mag_Dist_363.csv
and the sum of the deaggregated poe is 0.00316
i.e. over 50% more then the expected.
Please could you verify the above?
Thank you,
Francesco
I made a run to calculate deaggregation, OQ run to completion and the output files were generated.
The elaboration used 3 seismicity models and 3 GMPEs (i.e. 9 realizations),
to evaluate 16 points on a regular grid (spaced 0.1°, in the square 13.1-42.2 -- 13.4-42.5)
for some imt (PGA, SA(0.2), SA(1.0), SA(2.0) )
and two poe (0.1, 0.02).
We asked deaggregated on magnitude between 4.75 and 7.75, using intervals of 0.5
and distance between 2.5 and 152.5, using intervals of 5
We were interested only in Mag-Dist deaggregation.
Every output file generated by OQ is relative to one point, one realization, one imt, one poe, one type of disaggregation.
We found 2 unexpected behaviour.
1)
We expected to find the same number of rows for every disaggregation of type Mag_Dist,
i.e. the same number of couples Mag-Dist with the associated deaggregated poe,
but it was not the case.
Even taking, e.g., two file with the same realization, imt, poe, we have difference from point to point.
From the sample files we checked
we found in some files the distance value starts at a higher value (e.g. form 7.5)
or stop at a lower value (e.g. 17.5), or both,
and the magnitude value stop at a lower value (e.g. 6.75)
Also, we considered the files related to points 13.1-42.2 and 13.4-42.2
which "see" the same number of points of elaboration at the same distances,
but again we found differences.
I also note that there are rows for values of minimum magnitude and distance with poe=0
4.75000E+00,2.50000E+00,0.00000E+00
therefore a value of poe=0 is not a reason for not having the associated row.
2)
We expected that the sum of the deaggregated poe within a file is the same of the poe the file refers to.
What we found (at least in one case) is that the file is named
poe-0.02-rlz-8-SA(2.0)-13.4-42.5_Mag_Dist_363.csv
and the sum of the deaggregated poe is 0.00316
i.e. over 50% more then the expected.
Please could you verify the above?
Thank you,
Francesco
Michele Simionato
Nov 9, 2017, 3:43:05 AM11/9/17
to OpenQuake Users
For the benefit of others running into the same issue: the problem here was caused by a source model with a rate of occurrence too small, thus producing hazard curves lower than the disaggregation PoE.
The next version of the engine will raise an error message in this case. Fixed in https://github.com/gem/oq-engine/pull/3228.
Reply all
Reply to author
Forward
0 new messages