search

NOAA-EMC / GSI

Gridpoint Statistical Interpolation
GNU Lesser General Public License v3.0
66 stars 150 forks source link

Add more regression test cases for RRFS #586

Closed hu5970 closed 4 months ago

hu5970 commented 1 year ago

There is only one regression test case for hybrid EnVar with conventional observations in current suite for RRFS.

We need to add more regression tests for RRFS: 1) hybrid EnVar with direct radar reflectivity analysis 2) EnKF with conventional observations 3) EnKF with radar reflectivity observations.

We need to created the case with CONUS 13km grid. Also, the RRFS may combine the two hybrid EnVar analyses for conventional and reflectivity observations into one. If this happens, we need to update current regression test case.

RussTreadon-NOAA commented 11 months ago

@hu5970 , any updates on this issue? We should add the indicated ctests or close this issue and open a new one when ctests are ready to be added.

hu5970 commented 11 months ago

We cannot decide if sdl/vdl can be used for RRFS for now. I will cehck this issue again. If we cannot, I can add test cases based on current RRFS GSI runs.

RussTreadon-NOAA commented 11 months ago

Thank you, @hu5970. Please update this issue once a decision is made.

RussTreadon-NOAA commented 7 months ago

@ShunLiu-NOAA and @hu5970 : Given that Sho's PR #700 is nearing merger into develop, it is time to replace or upgrade rrfs_3denvar_glbens with a RRFS ctest which mimics functionality in the RRFS v1 package being submitted to NCO. We need to ensure that operational and plan operational GSI capabilities are exercised by GSI ctests. This is why we recently updated the GSI global ctest case. GMI assimilation is a possibility for GFS v17. Therefore, the global ctest case was updated to include gmi data.

If possible, let's not add several new rrfs ctests. Ideally we have one rrfs test which includes GSI options to be exercised in RRFS v1.

ShunLiu-NOAA commented 7 months ago

Russ, we will create a new set of test case with using regional GSI ensemble members to practice the functions used in regional DA. However, because we have to use the very different setup than the existing real time parallel to generate such a case, I will discuss with Ming about how to build this test case.

RussTreadon-NOAA commented 7 months ago

I appreciate the challenges in creating computationally inexpensive yet representative GSI regression test cases. To create the updated global case I had to

This is a tedious and, sometimes, problematic undertaking.

hu5970 commented 7 months ago

Yes, we need to create low resolution case. We recent spend all our energy to run large NA 3km and CONUS cases and did not visit low resolution case for a while. I will work with Shun to try a retro run with CONUS 13km grid, which is grid for current GSI case.

ShunLiu-NOAA commented 7 months ago

Russ, Thank you for providing this information. As Ming mentioned, we will revisit 13km resolution test case.

Shun

RussTreadon-NOAA commented 5 months ago

@ShunLiu-NOAA and @hu5970 : what is the status of this issue?

hu5970 commented 5 months ago

I can run CONUS 13 km RRFS case on both Hera now. I need to find a time to create cases for RRFS based on this issue.

RussTreadon-NOAA commented 5 months ago

Thank you @hu5970 for the update. Hera is down today (5/7) for PM.