[Bug 1092056] [NEW] OQ-ENGINE: It should be possible to define a site model exactly on the same position where hazard is computed

[Damiano Monelli]

Public bug reported:

Currently, OQ-Engine allows a user to input a site_model.xml file
containing spatial distribution of site parameters (Vs30, Vs30Type,
...). This information is then used in the hazard calculation to set the
site parameters for the locations for which hazard results (curves,
gmfs) need to be computed - what happens is that for each location the
closest site in the site model is taken and its site parameters are
associated to the location.

Now, an error occur when locations defined in the site model are the
same of the locations defined in the configuration file. It should be
possible however to allow for such a possibility (it's possible to have
informations for sites parameters at the locations where results need to
be computed).

** Affects: openquake
Importance: Undecided
Status: New

--
You received this bug notification because you are subscribed to
OpenQuake.
Matching subscriptions: openquake-bugs
https://bugs.launchpad.net/bugs/1092056

Title:
OQ-ENGINE: It should be possible to define a site model exactly on the
same position where hazard is computed

Status in OpenQuake:
New

Bug description:
Currently, OQ-Engine allows a user to input a site_model.xml file
containing spatial distribution of site parameters (Vs30, Vs30Type,
...). This information is then used in the hazard calculation to set
the site parameters for the locations for which hazard results
(curves, gmfs) need to be computed - what happens is that for each
location the closest site in the site model is taken and its site
parameters are associated to the location.

Now, an error occur when locations defined in the site model are the
same of the locations defined in the configuration file. It should be
possible however to allow for such a possibility (it's possible to
have informations for sites parameters at the locations where results
need to be computed).

To manage notifications about this bug go to:
https://bugs.launchpad.net/openquake/+bug/1092056/+subscriptions

[Giuseppe Vallarelli]

This problem can be seen by running the Test Case 4 for the Scenario Hazard Calculator (see qa_tests/hazard/scenario/case_4).
Solving this issue should make this test green or at least solving the issue with the validation of site model (see .https://github.com/gem/oq-engine/blob/master/openquake/calculators/hazard/general.py#L133)

[Lars Butler]

Just FYI:

The core issue here is not about computing risk on precisely the same
sites as the hazard input data. The issue is that the points of interest
in given calculation all share a common longitude (0.0) and thus, the
points form a perfectly straight line. The error message (which is not
present in this bug description) describes an issue with a shapely
object. The problem occurs here: https://github.com/gem/oq-
engine/blob/f8cebcc359dc8af93b649ea9aeb52fbddecd5767/openquake/calculators/hazard/general.py#L151

Basically, the 'exterior' of a straight line is a just a simplified
straight line, which is a 1-dimensional object. This sort of shape has
no exterior, and thus the failure.