Localisation: Added option to specify GEN_OBS nodes in the form nodename:index in localisation config file #363
Conversation
oddvarlia
changed the title
oddvarlia
force-pushed
the
use_active_obs
branch
from
f704553 to
ed3ea68
Compare
oddvarlia
force-pushed
the
use_active_obs
branch
3 times, most recently
from
fe66e39 to
2e081a1
Compare
oddvarlia
force-pushed
the
use_active_obs
branch
6 times, most recently
from
360dbdf to
a8199cf
Compare
oddvarlia
force-pushed
the
use_active_obs
branch
2 times, most recently
from
66aff11 to
b3c5738
Compare
oddvarlia
force-pushed
the
use_active_obs
branch
9 times, most recently
from
cc01eb3 to
fbaa8bc
Compare
oddvarlia
force-pushed
the
use_active_obs
branch
4 times, most recently
from
a6a1410 to
64d282e
Compare
| @@ -853,3 +878,31 @@ def write_qc_parameter( | |||
| grid.write_grdecl(scaling_kw, file) | |||
| # Increase parameter number to define unique parameter name | |||
| cls.scaling_param_number = cls.scaling_param_number + 1 | |||
|
|
|||
|
|
|||
| def get_obs_from_ert(ert, expand_gen_obs_max_size): | |||
There was a problem hiding this comment.
Is it possible to avoid the special casing of GEN_OBS?
There was a problem hiding this comment.
If not we should add some testing of it
There was a problem hiding this comment.
The reason why I defined this 'expand_gen_obs_max_size' was to avoid having to expand a GEN_OBS node into the list of type nodename: index for index from 0 to size where the size sometimes can be very large (seismic data with 100.000's of values). The use of individual GEN_OBS observations is primarily meant for GEN_OBS nodes of moderate size. Examples I was thinking about was e.g geophysical data like gravimetry measurements or other cases where the number of observations in a GEN_OBS node is < 100. In this case it can be meaningful to setup localisations between individual observations and model parameters. For seismic data, the obvious way of handling that is to split the seismic data into a moderate number of groups (e.g one per fault block or segment) and define one GEN_OBS node for each of these and use these without referring to individual observations within the GEN_OBS node. Sometimes, like for isolated segments, it can be relevant to setup localisation where a GEN_OBS node for a segment is used together with the model parameter values related to the same segment. I thought at least that there might be unnecessary overhead to work with lists of 100.000's of observations in the localisation script and introduced this expand_gen_obs_max_size parameter to handle ERT models where there exists GEN_OBS nodes having large number of observations and GEN_OBS nodes having moderate number or small number of observations ( < expand_gen_obs_max_size). Do you have any better idea here?
| if len(obs_index_list) > 0: | ||
| active_obs_list = obs_group.getActiveList(obs_node_name) | ||
| if len(obs_index_list) > 50: | ||
| debug_print( |
There was a problem hiding this comment.
I would still argue that this should be a documentation issue instead.
There was a problem hiding this comment.
You mean that the user should set the number, hardcoded to 50, or that I should drop the debug print out here? I can of course write a sentence in the user doc, but I don't think the users care much about this as long as the output is not too large. Maybe I have misunderstood something?
|
This PR is waiting for some use cases before we merge it into main branch |
oddvarlia
force-pushed
the
use_active_obs
branch
2 times, most recently
from
cc0df89 to
459bca5
Compare
oddvarlia
force-pushed
the
use_active_obs
branch
from
459bca5 to
a120ca6
Compare
oddvarlia
force-pushed
the
use_active_obs
branch
2 times, most recently
from
5eb9b80 to
241ac7e
Compare
|
Is this a draft or is it ready for review? |
oddvarlia
force-pushed
the
use_active_obs
branch
from
241ac7e to
9f94439
Compare
oddvarlia
force-pushed
the
use_active_obs
branch
2 times, most recently
from
105c6f3 to
432ad85
Compare
This makes it possible to specify individual observations from an GEN_OBS node in localisation. Add tests to verify active param and active obs in ministeps Updated integration tests with consistency check for active obs and param Fix time and date format in testdata Updated documentation of localisation using individual observations from GEN_OBS nodes.
|
Closing due to no traction. Feel free to reopen whenever there is something new to add to this. |
To be able to specify localisation were individual observations within an obs node of type GEN_OBS is used, it can be specified in the localisation config file in the same way as model parameters of type GEN_PARAM with nodename:index
To avoid problems mixing observation nodes of type GEN_OBS with notation: nodename and nodename:index a keyword is defined to specify if individual GEN_OBS observations is allowed to be specified or not. Since a GEN_OBS node specified in the ERT config file can be expanded into N individual observations, a GEN_OBS node is expanded into a list of N observation names: [ nodename:0, nodename:1, ....,nodename:N-1].
The user will then have to specify observations of type GEN_OBS in the form nodename:index.
But for some nodes of type GEN_OBS, like seismic data, the number N can be very large and for those cases it is best not to specify individual observations, hence, the form nodename is used, not expanded into individual observations.