equinor / webviz-ert

ERT webviz plugins

Web based visualization for ERT

What is Webviz-ert

Webviz-ert is a visualization tool for ERT based on dash and webviz-config.

Download project

The code is hosted on GitHub: https://github.com/equinor/webviz-ert

# From the downloaded project's root folder - to install
pip install .

Running tests

Make sure that you have a browser and driver available for running the tests, e.g. chrome / chromium and the chrome driver, which should be installed / downloaded as binaries and made available in the path.

# From the downloaded project's root folder - to run tests
pip install -r test_requirements.txt
cd tests
pytest

Run Webviz-ert

Webviz-ert connects automatically to a storage server running in ERT. Here are a few steps to get an example version of webviz-ert running.

# Run simulations in ert - <ert-root-folder>/test-data/local/poly_example/
ert ensemble_smoother --target-case smoother poly.ert

# After simulation has finished, start webviz-ert from the same location with
ert vis

# Alternatively, you might have to supply the config file if you're using the
# classic ert storage solution:
ert vis poly.ert

Example

Start up

Plot viewer

Observation Analyzer

Parameter comparison

Response correlation

webviz-ert's Issues

Create Observation Analyzer Viewer

Design a view that should server as an exploration / analyzer for observation related data: misfits, inactive and active observations. Additionally, it can serve as a workbench for manual (user defined) transformations applied on observations: scaling, deactivating.

Write proper README

Update name in README

This was missed in the renaming PR

webviz-ert/README.md

Line 7 in 665cf22

## What is ertviz

Make use of correlation-between-input-parameters-and-responses from webviz subsurface

Make use of what we can based on our data from this page in webviz subsurface
https://webviz-subsurface-example.azurewebsites.net/correlation-between-input-parameters-and-responses

Visualize comparison between responses

Visualize results from MISFIT_PREPROCESSOR

To provide insight into the resulting observation scaling, we should provide a visualization of intermediate clusters, ie. observation groupings. These form the basis for further scaling.

One potential solution could be to visualize correlation matrices (top) and/or matrices with hierarchical clusterings on top (bottom) as demonstrated by @eivindsm

Another solution could be to exploit parallel coordinates for showing clustering results, for instance:

This is dependent of having the MISFIT_PREPROCESSOR job parameters internalized and exposed.

Make ertviz available in komodo bleeding

KDE plot for parameters

Besides histogram, we should provide KDE plot, something like this:

Visual representation of forward models

A relevant idea for webviz-ert: make a visual representation of the forward models; sequence, dependencies, and status during execution. Idea originally submitted by MOD SOUTH

Make pages in webviz-ert persistent

When changing between pages in in webviz-ert, the views are not stored and if the user has set up several plots, they are lost when the user returns. There is a way of persisting this. More info in https://equinor.slack.com/archives/C019REYBD54/p1607000905224600

Improve Ensemble selector

• Default selection?
• Make it possible to hide - takes much space.
• Add hint - how to use it

Authentication issue on webviz

Currently the users getting logged out after a while, and it is not optimal how webviz does authentication. We should poke webviz team and make sure this is getting solved in a proper way by them or making an agreement that we will fix it.

Adapt README to new repo name

Change ertviz -> webviz-ert and fix Travis build status

Clean repository and do a clean setup

rename the current master to poc
create new clean master and make a standard python proj setup

datetime.fromisoformat() not available in python 3.6

The datetime.fromisoformat() function used in ertviz/models/__init__.py is new in python 3.7 and does not run in python 3.6.

Suggestion from @ManInFez: could be something like dateutil.parser.isoparse

Use refactored ERT Storage API (FastAPI)

The API will mutate, however we can begin working on it now to see where the holes are.

Relevant PR in ERT: equinor/ert#1105

Make ert-client an dependency

All data serving the plots should be fetched through the ert-client

Exceptions thrown for some responses in the observation analyzer page

For some of the responses an exception is thrown when using a database from the HM tutorial. The same responses works perfectly fine in the other views.

I have the database-file if needed for investigation. Stacktrace below

Traceback (most recent call last):
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/flask/app.py", line 2447, in wsgi_app
    response = self.full_dispatch_request()
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/flask/app.py", line 1952, in full_dispatch_request
    rv = self.handle_user_exception(e)
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/flask/app.py", line 1821, in handle_user_exception
    reraise(exc_type, exc_value, tb)
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/flask/_compat.py", line 39, in reraise
    raise value
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/flask/app.py", line 1950, in full_dispatch_request
    rv = self.dispatch_request()
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/flask/app.py", line 1936, in dispatch_request
    return self.view_functions[rule.endpoint](**req.view_args)
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/dash/dash.py", line 1059, in dispatch
    response.set_data(func(*args, outputs_list=outputs_list))
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/dash/dash.py", line 994, in add_context
    output_value = func(*args, **kwargs)  # %% callback invoked %%
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/ertviz/controllers/observation_response_controller.py", line 121, in _update_graph
    response_plots = [
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/ertviz/controllers/observation_response_controller.py", line 122, in <listcomp>
    _generate_plot(ensemble_id, data["color"])
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/ertviz/controllers/observation_response_controller.py", line 118, in _generate_plot
    plot = _create_misfits_plot(ensemble.responses[response], None, color)
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/ertviz/controllers/observation_response_controller.py", line 33, in _create_misfits_plot
    response.univariate_misfits_df(selected_realizations),
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/ertviz/models/response.py", line 66, in univariate_misfits_df
    return misfits_df.astype("float64")
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/pandas/core/generic.py", line 5546, in astype
    new_data = self._mgr.astype(dtype=dtype, copy=copy, errors=errors,)
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/pandas/core/internals/managers.py", line 595, in astype
    return self.apply("astype", dtype=dtype, copy=copy, errors=errors)
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/pandas/core/internals/managers.py", line 406, in apply
    applied = getattr(b, f)(**kwargs)
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/pandas/core/internals/blocks.py", line 2100, in astype
    return super().astype(dtype=dtype, copy=copy, errors=errors)
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/pandas/core/internals/blocks.py", line 595, in astype
    values = astype_nansafe(vals1d, dtype, copy=True)
  File "/Users/OOH/code/virtualenvs/ert/lib/python3.8/site-packages/pandas/core/dtypes/cast.py", line 941, in astype_nansafe
    raise TypeError(f"cannot astype a datetimelike from [{arr.dtype}] to [{dtype}]")
TypeError: cannot astype a datetimelike from [datetime64[ns]] to [float64]

Functionality for alternative filtering when AHM is not working properly

Options:

• Filter on overall objective value / misfit
• Filter on misfit for key observations relevant for the specific use of the predictions (eg rates & pressure & "flooding information" in an area where a well is considered
• The above combined with realizations capturing the key uncertainties for the specific use (i.e. ensure capturing the relevant uncertainty span of the key parameters having effect on eg the deltaproduction of adding a new well)

Split parameters / responses into two pages

Try to split into two pages and see how that looks compare to how it is today

Display gen data names instead of general observation names

Create DB file for webviz azure demo

The demo needs a db which is updated and supports the latest version of webviz-ert / ert api.
The file and the data needs to be open source / not closed data.

A suggestion would be dockerfile with ert api serving the data from the database in a bundle handed over to the webviz-config team

Parameter plot should be created from Dataframe

Due to easier testing and making it less dependent on the current rest API, we should provide initialization of parameter plots / models through pandas dataframe instead.

Dropdown boxes acting as comboboxes should look like and behave like a combobox

Create graph layout of ensembles

The layout of the ensemble graph in the ensemble-selector should be calculated to a reasonable tree layout. Currently we use predefined layouts from cytoscape, but this should be something that makes sense for multiple trees.

Visualize in the dropdown which responses having observations

We should visualize or make the user in some way aware of which keys containing observations

Should be done in the same way as in the legacy plotter with highlighting the background

Recreate the different plots from old solution

• statistics view with all options
• response plot
• parameter histogram

Allow for custom bin-size in the histogram

Make use of assisted-history-matching-parameters-update-vs-observations from webviz subsurface

https://webviz-subsurface-example.azurewebsites.net/assisted-history-matching-parameters-update-vs-observations

Investigate sharing a database file among multiple clients

Should it be possible running multiple API's on top of the same databasefile?
Running "ert vis --project" and point towards one path with the databasefile should allow for more clients towards the same running API

Running ERT when an API already runs causing error

Needs to be handled gracefully

Use CSS for ertviz

There is currently many views in ertviz with styles defined inline. The styles should move out of the code and into a CSS file. This CSS file should reflect the style of equinor pages, see https://webviz-subsurface-example.azurewebsites.net/ for an example

Add parameter histogram to ensemble comparison view

We should provide showing parameters distributions of several ensembles similarly to response plots.

Make data loader configurable

Webviz-ert is now automatically connecting to a ert storage using the information registered in the storage_server.json lnked in a tmp folder, using the ert_shared.storage.connection module. There should be a way for the user to specify the folder containing the storage_server.json. If none is specified it would search in the current working directory, and lastly try the "registered" ert server. If it fails to find any storage server it should fail gracefully.

Make a better formatting of x_axis labels for missfits

As show here, the labels take a lot of space:

We need to format it more explanatory. Also this might be point for discussion as when we are to introduce labels (see #60) the label should describe the group rather than the missfit instead.

Make x-axis for log-priors logaritmic

Write dash tests

Make use of parameter-parallel-coordinates from webviz subsurface

https://webviz-subsurface-example.azurewebsites.net/parameter-parallel-coordinates

Test browsing the different views across different tabs in the browser

If webviz does not support - have some discussions with webviz-team.

Check multipe selections on ensemble selector

For håvard, it didn't work. Test in different browsers and versions

Discussion: caching vs data simplification for bigger datasets

Was running one iteration on snake_oil example and interaction with plots becomes unbearable slow.
Just selecting a new response (ie. populating the list box and selecting the response) takes several seconds. I know that we decided not to focus on the optimization issue yet, but still we should have it noted somewhere at least :)

Displaying metadata on observations

Displaying metadata such as values, dates, comments eventually by hovering over observation points.

Use async requests

...will have to wait probably until ert is using a prod server or some framework supporting multiple requests simultaneously.

Introduce groupings / labels into observations and responses

While working on misifts visualization an issue came up relating to how to visualize univariate misfits for a large number of observations. Currently each observation provides misfits stats, which in case of tens of thousands would be unreadable. One potential solution is to group observations into clusters (groups) and visualize misfits statistics of such a group instead (by aggregating misfits over the group).

Here, we can employ clustering results from misfits_preprocessor job for instance. The labels are not persisted yet though.

Another discussion relates to grouping of responses, which would provide additional hierarchical aspect, eg. show cumulative misfits statistics from a given response group. This reflects also the upcoming MultiResponseView plugin, which would automatically unroll all response plots from a given response group.

Observations seems to be persisted between different responses when navigating back and forth

Seems to happen sometimes when switching back and forth. Needs to be investigated.

Historical vectors should be omitted from the dropdown

Should just be default added to the summary vector

Create tree view for Ensemble relationships

An ensamble may have a parent and sevaral children. A tree view would be a good way of obtaining an overview of how these are connected.

Plotting mulitple ensembles with the same name in parameter plots

Since name is used as an identifier to ensembles in parameter plots, only one ensemble is shown if two ensembles with the same name is selected. Preferably they should be seperated using the ensemble_id in order to remedy this.

We need to keep track of and store the relations between ensembles in ERT

Add realisation number to the hovering of the data in parameter plot