Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

MOSAiC forcing from Xiaochun Wang #488

Open
eclare108213 opened this issue May 22, 2024 · 15 comments
Open

MOSAiC forcing from Xiaochun Wang #488

eclare108213 opened this issue May 22, 2024 · 15 comments
Assignees
@davidclemenssewall
Labels
Forcing

Comments

@eclare108213
Copy link
Contributor

Xiaochun Wang has offered to provide the MOSAiC forcing that he has used for testing Icepack.
@eclare108213
Copy link
Contributor Author

Pinging @xcwang2019
In this issue we can discuss the steps needed to access the forcing data and update the code and scripts to be able to use it.

@xcwang2019
Copy link

Currently the MOSAiC forcing my group used is organized in two periods (Nov. 2019 to May 2020, and June -July 2020) since there is a gap in radiation observation. Please let me know what I should do.

@dabail10
Copy link
Contributor

I would like to loop @davidclemenssewall into this.

@giuliacast
Copy link

I am trying to run Icepack with MOSAiC forcing as well, I would like to be kept in the loop.

@davidclemenssewall
Copy link
Contributor

Hi All,

So sorry that it has taken me this long to get to this (started a new job, field campaign, vacation, etc). Xiaochun and Giulia, I'm thrilled that you're working on using MOSAiC data for Icepack development. There is actually a cross-cutting working group, the MOSAiC Model Forcing Datasets Working Group (MMFD) within the MOSAiC Consortium that is dedicated to developing merged datasets for model development. You and your team members would be more than welcome to join. We hold a monthly virtual meetings (the next is Oct. 16 at 14:30 UTC) and have both a public and a private github repo for sharing the code 'recipes' to produce merged datasets. Speaking of which, we published a first version of gap-filled MOSAiC datasets (which include data in the May-June gap from ASFS observations) last spring: https://arcticdata.io/catalog/view/doi%3A10.18739%2FA2GX44W6J. The datasets are netCDF in the MDF format, which came out of the YOPPSiteMIP project. If you would like to get involved with MMFD please email me at [email protected].

On a related note, I actually have a branch of Icepack that is set up to ingest MDF-formatted forcing data. There are a few things that I had been hoping to add to it still, hence why I hadn't submitted a PR. But it is usable as is and so I will plan to go ahead and submit a PR shortly so others can make use of it.

Cheers,
David

@davidclemenssewall
Copy link
Contributor

Just a quick follow-up here. I've opened a draft PR for the MOSAiC forcing here: #500

@xcwang2019
Copy link

xcwang2019 commented Oct 9, 2024 via email

@eclare108213 eclare108213 mentioned this issue Dec 18, 2024
Draft
16 tasks
@eclare108213
Copy link
Contributor Author

@xcwang2019, apologies for missing your update to this thread. I suggest that we move this forward as follows:

  1. Compare your forcing data with @davidclemenssewall's, to see what is different about it. That will help us decide whether we need both. If the data set is small, you can attach it to this issue. Otherwise, you can post it somewhere online (e.g. zenodo, as you suggested), for us to retrieve.
  2. Create an Icepack test case that uses your forcing data in the manner that you have already used it, and submit the test case in a draft PR. We can help you with implementation decisions, as needed.
  3. As part of the PR process, I think it would be a good idea to cross-compare your test case and @davidclemenssewall's with each other's data, if that's possible. If the data turns out to be equivalent, then we only need one data set but we could keep both test cases.

I welcome everyone's other ideas here, too. Thanks - e

@xcwang2019
Copy link

@luyGithub Mr. Lu Yang, a PhD candidate at Nanjing University of Information Science and Technology, has used MOSAiC forcing for Icepack1.1. Attached is the atmospheric forcing (atm_foecing.txt), oceanic forcing (oceanmixed_daily_3_mosaic.txt）
atm_forcing.txt
and a readme file (mosaic_forcing_readme20241226.docx).
oceanmixed_daily_3_mosaic.txt
mosaic_forcing_readme20241226.docx

I compare our forcing and the forcing from David at a later time. @eclare108213 @davidclemenssewall @dabail10

@davidclemenssewall
Copy link
Contributor

Hi @xcwang2019, thank you for sending these. Do you have a version with a time axis? That would make comparing the forcing data easier.

@xcwang2019
Copy link

xcwang2019 commented Dec 28, 2024 via email

@luyGithub
Copy link

atm_forcing_addtimeaxis.txt
Hi, @davidclemenssewall

I put the forcing file with a time axis here.

Yang Lu

@eclare108213
Copy link
Contributor Author

@xcwang2019 wrote

I compare our forcing and the forcing from David at a later time.

Hi Xiaochun,
We discussed this forcing briefly at our team meeting last week, and we will need to better understand what the similarities and differences are between your data and @davidclemenssewall's. Since his dataset adheres to the MOSAiC Consortium's guidelines including time/location/formatting, it would be easiest for us to link to his data and include your test case using it in our repository. Please compare the forcing data when you get a chance, and then we can decide how to proceed. Thanks!

@xcwang2019
Copy link

xcwang2019 commented Jan 16, 2025 via email

@davidclemenssewall
Copy link
Contributor

Hi All,

I had a chance to take a brief look at comparing the Wang forcing with version 2 of the MOSAiC Model Forcing Dataset (MMFD) working group product. Version 2 has been submitted to the Arctic Data Center and should be public any day now. Below they are plotted on top of one another with the latest MMFD in blue and the Wang forcing in red. Notes: 1) the MMFD forcing cadence is 1 minute, so I have resampled it to 1 hour using pandas.resample.mean() to compare 2) I have assumed that the 'PRECIP' column in the Wang forcing is total (liquid and solid) precipitation because it includes values in July (when I believe the observed precip was all liquid).

Image

Here are the differences between the two datasets (Wang - MMFD) for when they have overlapping data (i.e., Nov. 2019 - early May 2020 and late June - July 2020). Note, based on looking at the Wang forcing I have assumed that 0 represents missing data and for air temperature 273.15 represents missing data (this is a somewhat questionable approach for the shortwave and precipitation data).

Image

Finally, here is a table summarizing the differences between the forcings (Wang - MMFD) when they have overlapping data. The columns are: Shortwave, Longwave, U-Wind, V-Wind, specific humidity, air temperature, total precipitation.

Image

Based on this comparison, here are a couple of quick takeaways:

  • The MMFD forcing covers a broader timespan, notably it includes October measurements at the beginning of the drift, the May-June 2020 leg 3-4 crew transition, and the second drift (Leg 5) from August - Sept. 2020.
  • When there are coincident measurements, there are potentially scientifically significant differences in air temperature, winds, and precipitation:
    • Air temperature: the Wang air temperature is consistently warmer than the MMFD by a median of +0.62 K. The discrepancy is fairly consistent in both the winter and the summer. The impacts are likely to be greater in the summer, when the surface is at 0 C a small change in air temperature can flip the sign of the turbulent fluxes.
    • Wind: the range of wind component speeds in MMFD is reduced relative to the Wang forcing. I suspect that this is because I hourly averaged minutely MMFD data for this comparison, but the comparison is still valid because the Icepack driver would also do similar averaging. I think this raises some questions about how to handle downsampling wind data for Icepack. We can discuss those in a separate issue.
    • Precipitation: precipitation from Wang is about 40% higher (when we have overlapping data) than the ka-radar-derived values in MMFD. There are many uncertainties in precipitation measurements (e.g., https://tc.copernicus.org/articles/16/2373/2022/tc-16-2373-2022.html) so I'm not particularly surprised. For Icepack purposes I think the critical question is how well the snow depth evolution simulated by Icepack matches what was observed.
  • There are a few features of the Wang dataset that confuse me. Notably, on November 17 it shows negative incoming longwave values. Just to make sure I didn't mangle anything when downloading the data, I checked that they were present in the text file directly. In the MMFD dataset we don't have any longwave data for that period because of the measurement issues. Our goal is to fill that measurement gap in the next version of the MMFD dataset. The Wang forcing also has a mid-winter above freezing air temperature spike that did not occur.

I think it is really important for the MOSAiC community to reach consensus on some of these basic parameters. That way, when subsequent modeling studies are done we can be confident that different results are due to different modeling choices, not different forcing. The MOSAiC Model Forcing Dataset working group was founded for this purpose and you (and anyone else) are welcome to join. One of the advantages of the MMFDWG is that many of the folks who collected and processed the original data (e.g., Chris Cox for the Met tower and ASFS and Kiki Schulz for the ocean measurements) participate. Also, I would really like to understand better why the measurements differ. For the MMFD group, we have published the code 'recipes' for how we produced the merged datasets from previously published measurements: https://github.com/SCMStandz/SCMStandz_public/tree/main/src. I think it would be worth understanding how the process of producing the Wang forcing differed from the MMFD process.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@davidclemenssewall davidclemenssewall

Labels
Forcing
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
7 participants
@eclare108213 @apcraig @dabail10 @luyGithub @davidclemenssewall @xcwang2019 @giuliacast