Drifting temperature 2m

[HJORQUER]

Dear all

I am running WRF4.3.2 (and also WRF4.4) using WUDAPT's local climate zones (LCZ) and detailed land use classification generated using W2W application. I am modeling a large metro area evaluating model performance for wind speed and temperature near surface (U10, V10, T2 variables). I use reanalysis data to conduct retrospective 30-day runs.

I am running WRF with BEP urban physics and BouLac PBLH; first model layers are located near 10, 20 and 30 m above ground (see attached namelist).

T2 values appear comparable to measurements but, after 2 weeks of simulation, modeled T2m (dotted line in attached figure, tmE9 variable) increases its diurnal amplitude and starts to drift away from measurement values (solid line in attached figure), increasing negative biases for minimum T2 values. This same behavior happens for the first layer of modeled temperature (near 10 m above ground), so it is not an issue of diagnostic equations for T2.

Regarding surface wind speeds, they look fine, no artifacts at all.

I would appreciate suggestions on how to find out the error

Regards

Héctor

 

[Ming Chen]

This is a hard question to answer. Usually when WRF is applied for climate simulation, monthly mean or annual mean values are examined to assess the model performance on climate simulation. The real-time comparison between WRF outputs and observations are often conducted for short-term weather simulation. It is not surprising to see the "drifting" of model results from long-term simulation.

 

[HJORQUER]

This is a hard question to answer. Usually when WRF is applied for climate simulation, monthly mean or annual mean values are examined to assess the model performance on climate simulation. The real-time comparison between WRF outputs and observations are often conducted for short-term weather simulation. It is not surprising to see the "drifting" of model results from long-term simulation.

Thank you Ming
I am going to use short-term runs to prevent that drifting,
Regards
Héctor