Problems of HIRLAM in wintertime stable boundary layer - analysis of forecasts and observations

1
Problems of HIRLAM in wintertime stable boundary
layer - analysis of forecasts and observations

• Timo Vihma, Evgeni Atlaskin, and Laura Rontu

2
Model validation against Sodankylä sounding
data Periods January and March 2005 Model
versions H635E (Gollvik-Rodriques
soil-snow-forest schema), H637 and H640 Model
product validated 24 h forecasts
3
In 2005, January was very mild in
Sodankylä but March was colder than
usually
4
Results for January 2005 focus on the errors in
the air temperature

• - At the heigths of 30 and 90 m, HIRLAM has a
  large positive bias in cold conditions.
• In warm conditions, the bias is often negative
  but much smaller in magnitude
• H635N with the snow schema does not produce
  better results

5
- The largest errors in the lowermost 100 m occur
under conditions of a strong inversion, as
estimated from the temperature difference between
30 and 1100 m
6
- The temperature error at the heights of 30 and
90 m depends much more on ?T(170-30m) than on
?T(1100-170m), i.e., large errors are related to
near-surface-based inversions, but not so much to
clearly elevated inversions.
7
The largest temperature errors at the height of
30 are not associated with saturation (neither
observed nor modelled).
8
Errors in the air specific humidity

• Near the surface, HIRLAM has a positive bias in
  cold conditions and a negative bias in warm
  conditions. This bias in q is qualitatively
  similar to that in T, but now magnitudes of the
  positive and negative bias are approximately
  equal.

9
Errors in q occur also without any temperature
inversion, but in conditions of a large
?T(170-30m), the bias in q(30m) is always positive
10
The largest errors in q(30m) typically occur when
the observed RH(30m) 0.85-0.95. H335N yields
much lower values of RH than H640
11

• March 2005
• Everything told before holds also for March 2005,
  except
• - the error in specific humidity at the heights
  of 30 and 90 m depends neither on T nor on RH
• considering the spoecific humidity, H635N
  performs better than H637

12

• Conclusions
• largest temperature errors occur in cold
  conditions with a large ?T(170-30m)
• - in conditions of a large ?T(170-30m), the bias
  in q(30m) is always large
• - the largest errors in temperature and humidity
  were typically not related to saturation
• the presence of solar radiation does not have a
  large effect on the temperature error in HIRLAM
• Although January 2005 was mild and March 2005
  was cold, the differences in the model
  performance with respect to the air temperature
  were small
• - H635N with the new snow-forest scheme does not
  show improvement, but this may also be related to
  problems in snow analysis and the digital filter
  initialization
• much more analyses are needed next application
  of the tower data,
• then analysis on the relative importance of
  factors controlling T2m (a) in reality and (b) in
  HIRLAM.