scikit linear regression_model, log-scale depths, wet-dry node handling #74

WPringle · 2022-02-16T04:23:49Z

kwarg entry into surrogate from training function
show use of LassoCV model in the example
kwarg entry for conversion from log scale in validation and percentile outputs
kwarg entry for conversion from depths to elevations in validation and percentile outputs
analysis of depths in log scale to preserve positivity in the actual surrogate (instead of post-processing).
handling sometimes wet-sometimes dry nodes (extrapolation of nearby wet elevations to the dry node considering hydraulic friction loss)
subsetting by storm wind swath option added
sparse quadrature and user-defined options for quadrature order and quadrature rule

…ng, and use of LassoCV model in the example

…levations in the output

…ption for surrogate model

…he runs, change normalized bias/mean error to dimensional one

…xample correctly

…ition as this seems to be faster and can handle larger datasets. Switched matrix order for the KL mode to avoid transposes and be consistent with the sklearn output.

… height of surrogte KL fit plot where required

…sfers over element table if present

… geometry

…ties dataset

…d updating plotting for KL prediction

… plasma

…e to a small constant null depth value which was negatively affecting the surrogate generation. Adjusting some validation plotting axis and colormap limits

* use `gartersnake` * update build system * update build system * update build system * update build system * update build system * update Python testing version

WPringle · 2022-03-11T16:03:49Z

@zacharyburnettNOAA This branch is working with storm events v1.2.5 and all the uses that I have (making peturbed tracks and doing the KLPC analysis). Can we merge this in/make a release for this before upgrade to new storm events API?

WPringle · 2022-03-11T16:09:30Z

@zacharyburnett See v0.8.0 draft release: https://github.com/noaa-ocs-modeling/EnsemblePerturbation/releases/tag/untagged-eabd7a6d8d7861fc7e73

ghost · 2022-03-11T16:42:25Z

@WPringle that's a great idea, I agree with that

sorry I made so many API changes, I wanted to include all of the other unused ATCF fields and I got carried away

…tion within the given fraction of the distribution

* use new `StormEvents` API * install wheel first * install PROJ * linting step * build PROJ * build PROJ * build PROJ * build PROJ * fix workflows * fix test files * skip test that hangs * fix original file test * change original track handling (write to file first and then read from threads) * update workflow names * we no longer need to write pickles to a temporary file if the track object already exists in memory * revert import change

ghost · 2022-03-14T14:27:43Z

@WPringle I've made a new branch release/v0.8.x that should be an isolated place to use the old stormevents API until these changes are finalized, and made a pull request #79 that applies the API updates to this branch specifically (for merging these changes into v0.9.x)

…e_stormevents # Conflicts: # ensembleperturbation/perturbation/atcf.py # setup.cfg

WPringle · 2022-03-15T22:52:00Z

@zacharyburnettNOAA Thanks that sounds good. So do you want to try and update this branch with the base branch (v0.8.x) and merge this PR in before then updating main to v0.9.x?

ghost · 2022-03-16T17:33:59Z

yes, I think it's a good idea to freeze what we have here so far (with the old API) to the v0.8.x branch and tag it there, and then merge in the new API changes. I will do that now

ghost · 2022-03-16T17:45:31Z

@WPringle I've frozen these commits to a new tag v0.8.1 in the release https://github.com/noaa-ocs-modeling/EnsemblePerturbation/releases/tag/v0.8.1, that should be backward-compatible with the old API.

Now I will go ahead and merge the new API changes into this branch using #79, unless you have any objections to that

update `stormevents` API to `>=1.3.0`

WPringle · 2022-03-16T17:50:43Z

@zacharyburnettNOAA That's good yep! I'll go ahead and use this new version and see if working OK.
But I think we will need to work on the stormevents a bit more, especially for treating the advisories, breaking them up into individual forecasts instead of having them all squashed together.

codecov-commenter · 2022-03-16T17:55:51Z

Codecov Report

Merging #74 (843b433) into main (57ffd09) will not change coverage.
The diff coverage is 0.00%.

@@           Coverage Diff           @@
##             main      #74   +/-   ##
=======================================
  Coverage   30.69%   30.69%           
=======================================
  Files          25       25           
  Lines        2926     2926           
=======================================
  Hits          898      898           
  Misses       2028     2028

Impacted Files	Coverage Δ
ensembleperturbation/client/perturb_tracks.py	`0.00% <0.00%> (ø)`
ensembleperturbation/parsing/adcirc.py	`47.54% <0.00%> (ø)`

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 57ffd09...843b433. Read the comment docs.

WPringle · 2022-03-16T20:41:35Z

@zacharyburnettNOAA I'm noticing some changes in the column spacings and the forecast times are not being entered in correctly. The fort.22 columns need to be formatted exactly.

NEW: AL, 06, 2018091106, , BEST, 0,260N, 632W, 115
OLD: AL, 06, 2018091106, , BEST, 0, 260N, 632W, 115

NEW: AL, 06, 2018091106, , BEST, 0 - This zero should update along forecast but is all zeros.

ghost · 2022-03-17T12:42:37Z

-AL, 06, 2018091106, , BEST, 0,260N, 632W, 115
+AL, 06, 2018091106, , BEST, 0, 260N, 632W, 115

Ok, I so latitude should be a 4-digit field (including N / S) and longitude should be 5 digits.

NEW: AL, 06, 2018091106, , BEST, 0 - This zero should update along forecast but is all zeros.

The specification says it should be "0 for best track":

TAU - forecast period: -24 through 240 hours, 0 for best-track, negative taus used for CARQ and WRNG records.

but I think that might be wrong because previously we were outputting positive forecast hours and worked. I will add the hours back in.

WPringle · 2022-03-17T14:10:58Z

```diff
-AL, 06, 2018091106, , BEST, 0,260N, 632W, 115
+AL, 06, 2018091106, , BEST, 0, 260N, 632W, 115
Ok, I so latitude should be a 4-digit field (including N / S) and longitude should be 5 digits.

NEW: AL, 06, 2018091106, , BEST, 0 - This zero should update along forecast but is all zeros.

The specification says it should be "0 for best track":

TAU - forecast period: -24 through 240 hours, 0 for best-track, negative taus used for CARQ and WRNG records.

but I think that might be wrong because previously we were outputting positive forecast hours and worked. I will add the hours back in.

@zacharyburnettNOAA Yes you're right about the BEST track actually should be zero. As long as behavior is like original ATCF similar below then that will be great!

BEST track datetime = valid time, then the forecast hour = 0
OCFL (or other one forecast records), date time = initial time of forecast, then have non-zero forecast hours.

WPringle · 2022-04-27T15:52:19Z

@zacharyburnettNOAA I forgot, can this PR be closed?

ghost · 2022-04-27T15:56:46Z

@zacharyburnettNOAA I forgot, can this PR be closed?

sure thing, I will merge this into main

# Conflicts: # .github/workflows/build.yml # .github/workflows/tests.yml # ensembleperturbation/parsing/adcirc.py # ensembleperturbation/perturbation/atcf.py # ensembleperturbation/plotting/nodes.py # ensembleperturbation/uncertainty_quantification/surrogate.py # setup.cfg # setup.py # tests/data/reference/test_combine_outputs/fort.63.nc # tests/data/reference/test_combine_outputs/fort.64.nc # tests/data/reference/test_combine_outputs/maxele.63.nc # tests/data/reference/test_combine_outputs/maxvel.63.nc # tests/data/reference/test_monovariate_besttrack_ensemble/original.22 # tests/data/reference/test_monovariate_besttrack_ensemble/vortex_1_variable_random_1.22 # tests/data/reference/test_monovariate_besttrack_ensemble/vortex_1_variable_random_2.22 # tests/data/reference/test_monovariate_besttrack_ensemble/vortex_1_variable_random_3.22 # tests/data/reference/test_monovariate_besttrack_ensemble/vortex_1_variable_random_4.22 # tests/data/reference/test_monovariate_besttrack_ensemble/vortex_1_variable_random_5.22 # tests/data/reference/test_monovariate_besttrack_ensemble/vortex_1_variable_random_6.22 # tests/data/reference/test_monovariate_besttrack_ensemble/vortex_1_variable_random_7.22 # tests/data/reference/test_monovariate_besttrack_ensemble/vortex_1_variable_random_8.22 # tests/data/reference/test_multivariate_besttrack_ensemble/original.22 # tests/data/reference/test_multivariate_besttrack_ensemble/vortex_4_variable_random_1.22 # tests/data/reference/test_multivariate_besttrack_ensemble/vortex_4_variable_random_2.22 # tests/data/reference/test_multivariate_besttrack_ensemble/vortex_4_variable_random_3.22 # tests/data/reference/test_spatial_perturbations/original.22 # tests/data/reference/test_spatial_perturbations/original.json # tests/data/reference/test_spatial_perturbations/vortex_1_variable_random_1.22 # tests/data/reference/test_spatial_perturbations/vortex_1_variable_random_1.json # tests/data/reference/test_spatial_perturbations/vortex_1_variable_random_2.22 # tests/data/reference/test_spatial_perturbations/vortex_1_variable_random_2.json # tests/data/reference/test_spatial_perturbations/vortex_1_variable_random_3.22 # tests/data/reference/test_spatial_perturbations/vortex_1_variable_random_3.json # tests/data/reference/test_spatial_perturbations/vortex_1_variable_random_4.22 # tests/data/reference/test_spatial_perturbations/vortex_1_variable_random_4.json # tests/test_track_ensemble.py

ghost · 2022-04-27T17:12:23Z

ok, I've merged the release/v0.8.x branch into mainso are the changes in this branch also important?

WPringle · 2022-04-27T17:13:25Z

@zacharyburnettNOAA Thanks for the merge, of the v0.8.x.
I think we should delete this PR because all the changes in that branch were in here.

WPringle · 2022-04-27T17:17:06Z

@zacharyburnettNOAA I mean whichever is compatible with newer StormEvents according to the two file changes here.

WPringle added 7 commits February 15, 2022 22:22

adding sciki linear regression_model kwarg into surrogate from traini…

528425d

…ng, and use of LassoCV model in the example

add options for analysis in log scale and conversion from depths to e…

f041938

…levations in the output

adding minimum allowable value (for adcirc this is depth of 0.8*h0) o…

4d0f6ad

…ption for surrogate model

adding boxplot showing error distribution of surrogate model across t…

9b525ac

…he runs, change normalized bias/mean error to dimensional one

subsetting by wind swath in example

41aa005

adding ability to handle sometimes wet sometimes dry nodes into the e…

e1d99fb

…xample correctly

adding option to use sklearn PCA tool for the Karhunen-loeve decompos…

7dd7387

…ition as this seems to be faster and can handle larger datasets. Switched matrix order for the KL mode to avoid transposes and be consistent with the sklearn output.

WPringle changed the title ~~choose sciki linear regression_model for surrogate training~~ scikit linear regression_model, log-scale depths, wet-dry node handling Feb 23, 2022

WPringle and others added 22 commits February 22, 2022 20:06

adding correlation coefficient for wet/dry logical switch, increasing…

83f479a

… height of surrogte KL fit plot where required

adding subset_dataset option to adcirc parser that renumbers and tran…

3d6e756

…sfers over element table if present

correcting subsetting with elements inside, adding surface plotter to…

cfd416a

… geometry

adding KL surface plot option and adding element table into sensitivi…

f85f749

…ties dataset

adding element_table into surrogate output netcdf/xarray.Datasets, an…

fa1c449

…d updating plotting for KL prediction

adding surface plot option for sensitivities and changing colorbar to…

9f2dab2

… plasma

redefining placement of colorbar for sensitivities plot

2a33527

adding plot surface ability for validation and percentiles

df56220

change colormap extent to only consider model source

a1be127

Adding an elevation extrapolator over dry nodes to avoid setting thes…

c77115b

…e to a small constant null depth value which was negatively affecting the surrogate generation. Adjusting some validation plotting axis and colormap limits

send kwargs to coastline plot function

0ca5a1b

build on latest Python

f4ff276

bump adcircpy version (#69)

103580b

change storm_id to nhc_code

358706b

change sample_rule to sample-rule

2829260

change sample_rule to sample-rule

87cbbad

internal refactoring

83bf553

remove unused dependencies

6f1d053

tables is not named the same on Anaconda

9feeb94

formatting

6e27966

update setup and package installation (#75)

130be30

* use `gartersnake` * update build system * update build system * update build system * update build system * update build system * update Python testing version

update workflows

21bef40

bumping stormevents to 1.2.5 version for working version with old API

257f837

WPringle and others added 2 commits March 13, 2022 15:49

introducing inhouse sampling method that equally divides the distribu…

7912f5c

…tion within the given fraction of the distribution

ghost changed the base branch from main to release/v0.8.x March 14, 2022 14:26

zacharyburnett added 2 commits March 14, 2022 10:41

Merge remote-tracking branch 'origin/release/v0.9.x' into merge/updat…

a66983d

…e_stormevents # Conflicts: # ensembleperturbation/perturbation/atcf.py # setup.cfg

formatting with oitnb

e531270

ghost changed the base branch from release/v0.8.x to main March 14, 2022 14:57

better handling of input fort.22 file

f1ce14b

found an extranneous usage of .from_fort22()

4c86d52

Merge pull request #79 from noaa-ocs-modeling/merge/update_stormevents

9ce16db

update `stormevents` API to `>=1.3.0`

ghost mentioned this pull request Mar 17, 2022

ATCF fields are incorrect lengths when written to file oceanmodeling/StormEvents#24

Closed

ghost closed this Apr 27, 2022

This pull request was closed.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

scikit linear regression_model, log-scale depths, wet-dry node handling #74

scikit linear regression_model, log-scale depths, wet-dry node handling #74

WPringle commented Feb 16, 2022 •

edited

Loading

WPringle commented Mar 11, 2022

WPringle commented Mar 11, 2022 •

edited

Loading

ghost commented Mar 11, 2022 •

edited by ghost

Loading

ghost commented Mar 14, 2022 •

edited by ghost

Loading

WPringle commented Mar 15, 2022

ghost commented Mar 16, 2022

ghost commented Mar 16, 2022

WPringle commented Mar 16, 2022 •

edited

Loading

codecov-commenter commented Mar 16, 2022 •

edited

Loading

WPringle commented Mar 16, 2022 •

edited

Loading

ghost commented Mar 17, 2022 •

edited by ghost

Loading

WPringle commented Mar 17, 2022

WPringle commented Apr 27, 2022

ghost commented Apr 27, 2022

ghost commented Apr 27, 2022

WPringle commented Apr 27, 2022

WPringle commented Apr 27, 2022 •

edited

Loading

scikit linear regression_model, log-scale depths, wet-dry node handling #74

scikit linear regression_model, log-scale depths, wet-dry node handling #74

Conversation

WPringle commented Feb 16, 2022 • edited Loading

WPringle commented Mar 11, 2022

WPringle commented Mar 11, 2022 • edited Loading

ghost commented Mar 11, 2022 • edited by ghost Loading

ghost commented Mar 14, 2022 • edited by ghost Loading

WPringle commented Mar 15, 2022

ghost commented Mar 16, 2022

ghost commented Mar 16, 2022

WPringle commented Mar 16, 2022 • edited Loading

codecov-commenter commented Mar 16, 2022 • edited Loading

Codecov Report

WPringle commented Mar 16, 2022 • edited Loading

ghost commented Mar 17, 2022 • edited by ghost Loading

WPringle commented Mar 17, 2022

WPringle commented Apr 27, 2022

ghost commented Apr 27, 2022

ghost commented Apr 27, 2022

WPringle commented Apr 27, 2022

WPringle commented Apr 27, 2022 • edited Loading

WPringle commented Feb 16, 2022 •

edited

Loading

WPringle commented Mar 11, 2022 •

edited

Loading

ghost commented Mar 11, 2022 •

edited by ghost

Loading

ghost commented Mar 14, 2022 •

edited by ghost

Loading

WPringle commented Mar 16, 2022 •

edited

Loading

codecov-commenter commented Mar 16, 2022 •

edited

Loading

WPringle commented Mar 16, 2022 •

edited

Loading

ghost commented Mar 17, 2022 •

edited by ghost

Loading

WPringle commented Apr 27, 2022 •

edited

Loading