Huge difference in 2m daily temperature daily mean between wrfout and wrfxtm

[Alessandro Delo]

Hi everyone,

I’m working with WRF output (version 4.4.1, coupled with WRF-HYDRO . v 4.2.0) and I’ve noticed something puzzling that I hope someone can clarify.

When I compare the 2-meter temperature (T2) values read from the wrfout files (instantaneous outputs every 6 hours) with the T2MEAN variable from the wrfxtrm files, I see significant differences in some grid points when computing daily averages.

I calculated daily means from the instantaneous T2 values (four per day), while T2MEAN comes directly from the wrfxtrm files. Interestingly, when I look at monthly or annual means, these discrepancies become much smaller — but at the daily scale, they are quite large compared to a benchmark dataset like ERA5-Land.

So, I’d like to better understand:

• What exactly is saved in wrfxtrm for T2MEAN?
• Is it really the average of all model timesteps (in my case, every 30 seconds)?
• Does it make sense to compare T2MEAN with the daily mean computed from wrfout files, given that I only have 4 instantaneous values per day and may not be capturing the full diurnal cycle?

I understand that using only four time steps per day can miss a lot of variability, but I didn’t expect the differences to be this large. Based on the maps I’ve produced (attached 3 examples, 2 random days with daily average and an annual average), it’s hard to decide which dataset — the daily mean from wrfout or the T2MEAN from wrfxtrm — should be considered more reliable.

Has anyone encountered this issue or can explain which of the two is preferable for computing daily mean temperature from WRF outputs?

Thanks a lot in advance for your help!

Alessandro