Sunday, April 17, 2011

A Compendium of WRF Errors

As an atmospheric science grad student, I often run the WRF model. The errors it throws can be somewhat inscrutable, so here is a compendium of errors that I have frequently run into:

  1. X  points exceeded cfl=2 in domain d01 at time Y
    1. Meaning: Your time step is too long for the waves that exist within the model; it cannot solve the PDE's.  The CFL condition is explained here. Some may recall it from their dynamics courses. 
    2. Effect: If the model becomes sufficiently unstable it will fail.
    3. Solution: Either lower your time step or let the model damp out the instability. It seems to happen particularly at the upper boundary of the model, which is normally in the stratosphere and thus not a huge concern.
  2. Flerchinger USEd in NEW version. Iterations= 10
    1. Meaning: This condition (technically not an error) is thrown by the NOAH Land Surface Model module. Apparently, it involves the amount of supercooled water available if the ground is frozen. Often, this is because there is very low soil temperature in your model run.
    2. Effect: Because this is an explicit method of solving for the supercooled water available, the model slows down by several orders of magnitude. 
    3. Solution: Unclear, but changing from the NOAH land surface model suppresses the error.
  3. glibc detected *** free(): invalid pointer: 0xb7f4fc90
    1. Meaning: Invalid pointer means a reference to memory returned empty. In my case, this was caused by attempting to use the Zhang-McFarlane convection scheme in WRF 3.3.
    2. Effect: Model halts, often before the first time step.
    3. Solution: None that I know of, but it appears the CAM ZM module requires very specific namelist.input choices. These include only using the MYJ or Brotherton PBL schemes and including a shallow convection parameterization. As of now, I have not been able to use the module.
  4. Avost in Vilka
    1. Meaning: Cryptically, this is Russian for 'Crash in Fork', and is related to the land surface model.
    2. Effect: Model crashes, and this seems to be related to bad soil temperature values again.
    3. Solution: Unclear, sometimes my model runs go through, others grind to a halt with no change in namelist.input.
If you have any insight into these problems, feel free to post in the comments.  The WRF model is a wonderful tool, but its errors can be confounding. Often, I have found looking directly in the code (particularly if the errant module is named) can at least give you some sense of where the problem lies.

3 comments:

  1. I'm a meteo grad student as well, and I run WRF ensembles. The "AVOST IN VILKA" error is quite vexing (as a clarification, I've only seen it occur while using the RUC LSM). Your post gave me the idea to try running my case on fewer processors, and when I dropped all the way down to 4 processors WRF finally ran to completion without errors. However, for a different case day, I still got "AVOST IN VILKA" errors even with just 4 processors. What's also odd is that these cases ran to completion with WRF v3.2.1, but in WRF v3.3 I'm getting these errors. I've started a thread about it in the WRF Forum here:

    http://forum.wrfforum.com/viewtopic.php?f=6&t=3184

    Thanks for putting together this list of common problems and solutions. I've run into most of these at some point as well.

    Jared

    ReplyDelete
  2. Hi,
    We face the same "Flerchinger USEd in NEW version. Iterations= 10" issue. Could you elaborate a bit more when you say "changing from the NOAH land surface model suppresses the error" ?
    Thanks for posting feedback on you experience !
    Colin

    ReplyDelete
  3. Hi,

    I found out that a new module has been added to v3.3 called: module_sf_noahlsm.F . And actually the error message comes from this program. I added some lines that is suggested in sea-breeze code to include NOAH (I am running idealized 2D hill). While there is no error in compiling the program, but the error message still exists. I appreciate it if you let me know how to overcome this problem that happens for Flerchinger eq.

    ReplyDelete