Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

How to Make SST Data Match the Coastline in WRF Simulation?

bowen

New member
Hello everyone,
I encountered some issues after adding SST data into my WRF simulation. The SST data I used is from the ERA5 reanalysis dataset with a resolution of 27 km. I am using WRF version 4.3.2. The two images below show the visualization of SST and LANDMASK from the wrfout file. As you can see, the SST data still maintains the same resolution as the input dataset. This causes a mismatch between the SST and the coastline, resulting in a coarse grid-like jagged error near the coast.I would like to ask: How can I make the SST data better align with the coastline during the simulation and avoid this jagged grid error? I have attached the visualizations of SST and LANDMASK, as well as my namelist.wps and namelist.input files for reference.
Thank you!
 

Attachments

  • SST.png
    SST.png
    26.7 KB · Views: 7
  • LANDMASK.png
    LANDMASK.png
    81.9 KB · Views: 7
  • namelist.wps
    1.5 KB · Views: 3
  • namelist.input
    4.9 KB · Views: 4
I guess you use ERA5 data to drive WRF run? Please let me know if I am wrong.

If this is the case, it is apparently because you missed landmask for ERA5 SST. We have recently addressed the issue, following the data hanges in NCAR RDA.

Please download ERA5 data from this website:
NCAR RDA Dataset d633000

We provide a python package to process the ERA5 data:
GitHub - NCAR/era5_to_int: A simple Python script for converting ERA5 model-level netCDF files to the WPS intermediate format

This package will yield intermediate data files for WRF. You can simply skip ungrib.exe and move on to run metgrid.exe. The landsea mask issue is well addressed by the package.

Please try and let me know if you have any issues.
 
Thank you for your response! Previously, I used the 6-hour SST data from the ECMWF ERA5 dataset provided by Catalogue — Climate Data Store , and for meteorological data, I used the 6-hour FNL data from NCAR RDA Dataset d083002 . Then, I ran WRF and encountered the issue described in the post.

I truly appreciate your suggestions! I will follow your advice and use the data from NCAR RDA Dataset d633000 as surface and atmospheric forcing data for a new simulation. Thanks again!
I guess you use ERA5 data to drive WRF run? Please let me know if I am wrong.

If this is the case, it is apparently because you missed landmask for ERA5 SST. We have recently addressed the issue, following the data hanges in NCAR RDA.

Please download ERA5 data from this website:
NCAR RDA Dataset d633000

We provide a python package to process the ERA5 data:
GitHub - NCAR/era5_to_int: A simple Python script for converting ERA5 model-level netCDF files to the WPS intermediate format

This package will yield intermediate data files for WRF. You can simply skip ungrib.exe and move on to run metgrid.exe. The landsea mask issue is well addressed by the package.

Please try and let me know if you have any issues.
 
Hi all, I am experiencing similar issue with the SST when running WPS metgrid and using ERA5 as input. I posted my problem here: ERA5 as input - SST interpolation issues . I am running WPS on Casper at CISL. I am confused on what data to use. The script era5_to_int.py by default access the d633000 dataset. In my post the moderator refers to the AWS curated dataset for ERA5 (NSF NCAR Curated ECMWF Reanalysis 5 (ERA5) - Registry of Open Data on AWS). Are the AWS and d633000 datasets the same? In spite of using era5_to_int.py and the default ds633000 dataset, I am still having issue for the way SST is interpolated. I thought of not including SST in the input data and only using SKINTEMP, but can't find a way to do it with the era5_to_int.py script. Should I manually list all the needed variables minus SST? Thank you anyone for any suggestion. I am badly stuck on this problem.
 
Hello everyone,
I’ve faced a similar issue in my WRF simulations when trying to align SST data from reanalysis datasets with the WRF land-sea mask. Specifically, when using ERA5 SST data at ~27 km resolution with a higher-resolution WRF domain, the SST field doesn't perfectly follow the coastline, and this results in jagged or unrealistic temperature gradients at land-sea boundaries.
 
@tiziana @USA_PharmaStore
I believe the SST issue has been addressed in the package we provide. Detailed information how to access ERA5 and how to apply the Python tool is provided in my post on 25 March (see the thread above). Can you try and let me know if you still have the issue?
 
Hi, I tried it, and followed both your post and the instruction for era5_to_int.py but couldn't solve the issue. I can eventually create the wrf input, but I am not sure what goes into them as far as SST and SKINTEMP are concerned. I ended up using output=no in the METGRID.TBL for SST. SKINTEMP looks averaged (like in ERA5 as input - SST interpolation issues) and seems masked not by LANDMASK but the ERA5 LANDSEA. Question on the following: The era5_to_int.py script on Casper fishes ERA5 locally. Could you please confirm that this script, without a specified path as argument uses the correct dataset for high res runs with WRF/WPS? Is this the curated dataset also posted on AWS? Thank you.
 
@tiziana

You are right that, if you run era5_to_int.py in Derecho/Casper, it uses correct ERA5 dataset on AWS to create intermediate files for WPS/WRF and MPAS.

We have tested this script days ago and I believe it works as expected.

Hi, I tried it, and followed both your post and the instruction for era5_to_int.py but couldn't solve the issue. I can eventually create the wrf input, but I am not sure what goes into them as far as SST and SKINTEMP are concerned. I ended up using output=no in the METGRID.TBL for SST. SKINTEMP looks averaged (like in ERA5 as input - SST interpolation issues) and seems masked not by LANDMASK but the ERA5 LANDSEA. Question on the following: The era5_to_int.py script on Casper fishes ERA5 locally. Could you please confirm that this script, without a specified path as argument uses the correct dataset for high res runs with WRF/WPS? Is this the curated dataset also posted on AWS? Thank you.
 
@Ming Chen. Thank you for your reply. Is there any chance I could check your METGRID.TBL set up your met_em output files to see if you get similar results to mine? Perhaps you have it on a shared directory on the glade system? My particular case might be trickier just because Hawaii is composed of very small islands, for a 27km resolution background data. Still I have no seen this issues when I use the GFS as input. Thank you so much for all this help.
 
@tiziana
I just use the default METGRID.TBL in WPSV4.6. My test is over CONUS and I agree that the study area over Hawaii may induce certain unexpected features. Can you send me your namelist.wps for me to take a look? Thanks.
 
Top