eucp-project / atlas Goto Github PK

After the 0.1.0 release I had expected the atlas to show up in Zenodo, but it doesn't seem to have worked.

In this GH repository --> settings --> webhooks --> zenodo --> recent deliveries, it is visible that three release events were fired on Friday, and one of them failed with error code 429 and more info:

You are being blocked. Please contact us at: [email protected]

429 seems to be related to rate limits.

More info should have a nice summary e.g. like here

Currently ASK constrained maps are not according to raw constrained data. In Atlas, those maps show the unconstrained values. This will be fixed in #38 .

Merging a PR runs the automatic deployment action. However, before that, it would be nice if we can verify that the page looks ok. Is there some way to build it from the PR and serve it somewhere on a preview page?

Are they coming soon (this is what it comes across as...), or is there a technical reason why they are not available? Maybe we could add this:

• to the FAQ
• in a link or brief explanation to the placeholder-figure

About page with FAQ was added in #41, but the content could still use some polishing. See open comments in #41.

Currently, the atlas shows one figure at a time. It would be nice if we could add a "comparison mode" where you can select e.g. multiple methods and see them side by side.

Explain that plus button copies the selection of the previous row.

• Make a python script to plot maps in a uniform way
• Create new figures based on source netcdf data:

variables: pr and tas
season: winter and summer
percentiles: 10, 50, 90 (?)
data: cmip5 and cmip6
methods: climwip and ???
type: constrained, unconstrained, (side-by-side?)

placeholder for unavailable data

Use consistent naming pattern for figure filenames

For instance, the 50th percentile in the atlas includes the lower 50% of the climate projections --> For instance, the 50th temperature percentile shows the highest temperature in the lower 50% of the distribution of temperature changes

When there is more information, address the FIXME in the README.md of the repository.

Pressing + when placeholder / missing data.

For example

Press + will give

A dependency is added by this commit. I found this instruction to install it. Is it the best way? Shall we add how to install this dependency to readme?

• Make the height and weight responsive to the reshaping of the browser
• No scroll bars but always fill the entire page with the available space

As a user comparing multiple methods,
I want to change the value of "temperature" or "constrained" for all rows at once,
so I can more easily compare multiple aspects.

In this screenshot, changing "temperature" to "precipitation" requires changing all 6 dropdowns:

so you can share the exact view with a colleague by copying the link.

In the notebooks (after merging #38), an attrs dict is defined to hold some information about data like description and history. For history, we list the file names of raw data manually, for example, see below. It would be good if we refactor this part.

attrs={"description": f"Contains modified {INSTITUTION_ID} {METHOD} data used for Atlas in EUCP project.",
           "history": f"original {INSTITUTION_ID} {METHOD} data files CNRM_atlas_tas_CMIP6_histssp585_JJA_latlon.nc, CNRM_atlas_tas_CMIP6_histssp585_DJF.nc" }

Include table from deliverable report and rephrase final paragraph:

All methods show improvements in some regions, for temperature projections where as there is less evidence that they improve precipitation projections. One method, CALL, produces projections for the Mediterranean that are markedly worse than just using the raw climate model projections

Currently, the placeholder image shows that we can select ASK and CMIP5. But we don't have cmip5 data for ASK. The image should be fixed.

To review notebook changes it would be helpful if https://www.reviewnb.com/ integrated into this repo.

Now I am using

pip install nbdime
nbdiff-web main python/cleanup_CNRM_KCC_atlas_netcdf.ipynb

ClimWIP and REA are both "weighting methods", meaning that the constrained projections are constructed by assigning a weight to each ensemble member. It would be nice to provide these weights together with the data on Zenodo.

In notebooks (after merging #38), we add seasons as time values together with climatology_bounds, like this example. The preprocessed data can be read with both iris and xarray. However, some operations might return an error, for example:

ds = xr.open_dataset("./AtlasData/preprocess/atlas_EUCP_CNRM_CMIP6_KCC_cons_tas.nc")
ds.to_array()

ValueError: coordinate climatology_bounds has dimensions ('climatology_bounds',), but these are not a subset of the DataArray dimensions ('variable', 'time', 'latitude', 'longitude', 'percentile')

The comparison option (#37) shows multiple selection rows, and multiple maps. There is no explicit link between which row corresponds to which map. Adding letters might help, but in #37 we struggled with their alignment. This issue is for solving that problem.

Add hover info over each of the dropdown menu's to explain them in more detail like e.g. this one:

The maps show the “differences between 2041-2060 mean conditions, with respect to the 1995-2014 present day baseline”. It would be good to clarify this in the “What do these maps show?” about section, and perhaps also reflect this in the individual figure labels.

Preprocessed data lives here. The documentation should be added to the "description" field on Zenodo. Here is the documentation:

Preprocessed data of Atlas in EUCP-WP2

We provide some notebooks that check the original/raw data, fix/add the metadata
using CF-conventions https://cfconventions.org/Data/cf-conventions/cf-conventions-1.9/cf-conventions.html
and save data in a NetCDF format. See https://github.com/eucp-project/atlas/blob/main/python/README.md.

The pre-processed data follows the following standards:

coordinates

• climatology_bounds (climatology_bounds) datetime64[ns] ['2050-06-01', '2050-09-01', '2050-12-01', '2051-03-01']
• time (time) (datetime64[ns]) [2050-07-16 2051-01-16] # "JJA", "DJF"
• latitude (lat) (float64) [30, ..., 75]
• longitude (lon) (float64) [-10, ..., 40]
• percentile (percentile) (int64) [10, 25, 50, 75, 90]

variables

• tas (time, latitude, longitude, percentile) (float64)
• pr (time, latitude, longitude, percentile) (float64)

attributes

The attributes of variables and coordinates are defined as:

• "tas": {
  "description": "Change in Air Temperature",
  "standard_name": "Change in Air Temperature",
  "long_name": "Change in Near-Surface Air Temperature",
  "units": "K",
  "cell_methods": "time: mean changes over 20 years 2041-2060 vs 1995-2014",
  },
• "pr": {
  "description": "Relative precipitation",
  "standard_name": "Relative precipitation",
  "long_name": "Relative precipitation",
  "units": "%",
  "cell_methods": "time: mean changes over 20 years 2041-2060 vs 1995-2014",
  },
• "latitude": {"units": "degrees_north", "long_name": "latitude", "axis": "Y"},
• "longitude": {"units": "degrees_east", "long_name": "longitude", "axis": "X"},
• "time": {
  "climatology": "climatology_bounds",
  "long_name": "time",
  "axis": "T",
  "climatology_bounds": ["2050-6-1", "2050-9-1", "2050-12-1", "2051-3-1"],
  "description": "mean changes over 20 years 2041-2060 vs 1995-2014. The mid point 2050 is chosen as the representative time.",
  },
• "percentile": {"units": "%", "long_name": "percentile", "axis": "Z"},

The attributes of the data is defined as:

• "description": "Contains modified institute method data used for Atlas in EUCP project.",
• "history": "original institute method data files ...",

output file names

output_file_name = prefix_activity_institution-id_source_method_sub-method_cmor-var

example: atlas_EUCP_CNRM_CMIP6_KCC_cons_tas.nc

Reference:

https://eucp-project.github.io/atlas/about

• Project organization (+ brief summaries) in the main README
  • Link to contribution guide
  • Citation info
• (basic) Information about current notebooks into the python/README
• Contributor information:
  • Generate a new script: where and how
  • Where to put the output image
  • Make internal changes to the project (i.e. what to update when adding a new script)
  • Rendering the atlas locally
  • Adding/changing javascript pages

• Address open comments in #14
• Include CMIP5 figures if available
• Add placeholder for unavailable figures or make it impossible to select them
• ...

It can be hard to find things on the about page, because you have to expand the question title to see the answer body. Would be nice to add search functionality using https://content.nuxtjs.org/fetching#searchfield-value

Another way is to add expand/collapse all buttons so you can use the browsers search function.

Add a licence (apache for code added to repo, CC-by for content in footer?)
Improve description in README (add that this is work in progress and will be unstable)
Look into possibilities for testing
Perhaps add a code quality checker
...

Also see https://guide.esciencecenter.nl/#/best_practices/language_guides/javascript?id=testing

Running locally on wtih npm run dev shows the constraint dropdown like

However on https://eucp-project.github.io/atlas/ the contraint dropdown is empty, see

The plot shown is constraint.

Should only deploy once PR is approved and merged into main.

Currently, we add a header including logo and page title to each page separately. To make it consistent, we might use the layout file.