2m RH underestimated

Forum dedicated to older versions of EMS package (WRFEMS v3.2, v3.1 or older). Support is user-to-user based, so please help others if you can.
vonjack
Posts: 16
Joined: Thu Sep 27, 2012 7:14 pm

Re: 2m RH underestimated

Post by vonjack » Sun Dec 09, 2012 10:13 pm

Some news:

my actual setup is
MP WSM6
PBL MYJ

USGS terrain data

Now RH2m prediction follows better the measured values, but in theese days with very low min temperatures t2m is very overestimated (measured -11°C, model output 1°C, in a town on the moutains near here).

vonjack
Posts: 16
Joined: Thu Sep 27, 2012 7:14 pm

Re: 2m RH underestimated

Post by vonjack » Wed Jan 23, 2013 9:31 pm

Update:

I'm monitoring model output on 2 location:
my town (Thiene), that is at the foot of the mountains, and another town (Asiago), that is at 1000m.
After few days of testing, I can say:

T2 is generally underestimated (2-3°C) in my town, all over the day, http://www.avmeteo.it/wxmeteogram_thiene.php
and T2 is overestimated (2-3°C) on the mountais. http://www.avmeteo.it/wxmeteogram_asiago.php

You can see the complete model output here : http://www.avmeteo.it/wxmodel2.php.

For example: today measured max temp 8°C, max temp from the model 3°C!

Have you got any suggestions on how to solve the problem? What can it be related to?
Is there any model variable that can help us to fix the probelm?

As usual, many many thanks!!!


Enrico

PS: the temp bias is noticeable even at the start of simulation, so could it be related to model initialization?

meteoadriatic
Posts: 1587
Joined: Wed Aug 19, 2009 10:05 am

Re: 2m RH underestimated

Post by meteoadriatic » Wed Jan 23, 2013 10:33 pm

What are model heights in comparision with actual heights above sea level of your locations? Are they differ much? Probably yes, especially if terrain is complex and you don't have enough model resolution.

If you say that temperatures in bottom of valleys are underestimated, and temperatures on the moutains are overestimated, it simply looks that there is resolution problem; the terrain height is averaged much more than you might want it to be.

You can do two things here:
1) improve horizontal resolution, or/and
2) add height corrections in postprocessing.

For 2nd idea, I use this approach and temperatures are much closer to measured than without height corrections. The code used is for GrADS and can be applied only for meteograms, numeric output, etc... that is given for fixed lat/lon location with known height:

Code: Select all

'define grad = (tmpprs(lev=850)-tmpprs(lev=1000))/(hgtprs(lev=850)-hgtprs(lev=1000))'
'define vcorr = grad*hgtsfc'
'define temp = tmpprs(lev=1013)+vcorr-273'
'define dz=height-hgtsfc'
'define dt=grad*dz'
'define temp=temp+dt'
If your terrain is on very high terrain you might want to use different layer then 1000/850 hPa to get vertical lapse rate (variable grad in this example is in fact vertical lapse rate in Kelvins/meter).

For 2D maps, unfortunately this cannot be used and you're out of luck there I'm afraid. Only improving resolution can do better job in that case.

You can see from example that I don't use t2m field but temperature for lowest level (1013) and then apply correction to it. You can however achieve similar results using t2m as starting point, in this case something like this:

Code: Select all

'define temp=(t2-273)'
'define dz=height-hgt'
'define grad = (tk(lev=850)-tk(lev=1000))/((geopt(lev=850)/10)-(geopt(lev=1000)/10))'
'define dt=grad*dz'
'define temp=temp+dt'
Those are variables from ARWpost (tk, t2, geopt, hgt, ...) but general idea is the same.

In both cases, variable "height" is actual height above sea level for location, and dz is then difference between actual height and model height.

Hope this helps!

vonjack
Posts: 16
Joined: Thu Sep 27, 2012 7:14 pm

Re: 2m RH underestimated

Post by vonjack » Sun Feb 17, 2013 9:36 pm

Update: after some days of observation/simulation comparation I see that main problem is strong T2 underestimation during daytime (5-6C)
The model height is correct.
I think that the problem is related to snow cover effect.
Model is initialized from Gfs with Snow Cover=1, but there is no snow for real.
So Albedo results higher than real.
Could this be a possible explanation?
A run with ifsnow=0 had brought to more realistic T2...

So the problem seems to be related to low resolution snow cover data from Gsf

Post Reply