tslist time step

[Ramchandra]

Dear WRF users,
I am running wrf in 25 5 1km nesting with also the output text file in time series using tslist. I could get the output after the setup in tslist but the output time step is in every time_step (as defined in namelist.input file) which is too frequent for my application. What i am interested here is to get the output time series in user defined time interval(1hour, 2hour interval)? is there any way to do this? I look forward for the kind response from users.
regards
Ramchandra

[brian.blaylock]

I've had the same question. Not too familiar with altering the WRF code itself, but the quickest way I deal with this is in post processing: only grab the line in the TS file from the time stamps you want.

Alternatively, you can use the auxhist option and output the variables I want at higher temperal resolution. Of course, this would take a lot more memory to output the variable for the entire domain. http://www2.mmm.ucar.edu/wrf/users/docs ... timeseries

[tanksnr]

I would like to revive this thread.

The timeseries output option is a great addition, however writing every model timestep is incredibly inconvenient considering it was meant to alleviate the need for post-processing. The filesize is also an issue if you're running a long simulation.

From what I can see in the WRF3.8 user guide there isn't any indication that you could control the writing to ts output.

This seems to have been an issue a while back as well.
http://forum.wrfforum.com/viewtopic.php?f=12&t=6015

Anyone with some ideas?

[emilyp]

Did anyone solve this problem?

I am running a high-resolution model for a month. It seems that when I set the ts_buf_size to anything anywhere big enough for the simulation (or even a week) real.exe gets stuck (i.e. crashes such that the node I'm running on has to be reset). Has this happened to anyone else or is it unique to the super computer I'm running on?

I can't really experiment with how big it can be as I am not able to reset the node myself, but ts_buf_size=300 runs OK (~15 min worth of model run), and ts_buf_size=181440 (~1 week of model run) crashes.

Alternatively has anyone modified the code to produce average, instead of instantaneous, values?

[kwthomas]

Hi...

If you are crashing the node, it probably means the node is running out of memory, and critical
processes, such as ROOT tasks, can't run. Most systems have software that detects the
OOM (Out of Memory) case and kills of the offending processes. If not, something running the node ot of memory *can* cause hangs and require a reboot to get the node working.

Try using more nodes and see if this fixes things.

[emilyp]

Hi

Thanks very much for your response. I think you are right about the problem. I have actually run real.exe now, but am having the same problem with wrf.exe. I've tried running on two nodes (64 processors) but this doesn't seem to have helped and I probably can't use many more as it's a shared computer. I may be outputting too many variables as I've added some, so will try cutting down the number of files.

Does tslist write a blank buffer file for each location and variable first rather than just appending it when needed? Does anyone know how to "flush the file to disk" as is mentioned in the README?