1. Summary of the summer project
- task was to get some insight to the failure of day 6 forecast over Europe
- especially, how does convection over N America behave?
- first 1.5 months we were looking at the wrong MCS
- the right day to consider is the 10th of April
- in this project we used OpenIFS resolution T255L91
2. Verification of the bust
- we ordered initial conditions for
days beginning from the 1st and
ending to the 13th of April
- we needed to check that the
bust occurs also in OpenIFS
- we calculated anomaly
correlation coefficient (ACC) for
day six 500hPa geopotential
height over Europe (the blue
square)
3. - the upper one is
from Rodwell et
al. (2013) and the
lower one is the
reproduced figure
- bust occurs also
in OpenIFS
- the overall ACC
is lower
- perhaps lower
resolution might
affect: T1279 vs
T255
4. Synoptic scale evolution
- next three slides including this
one have been plotted from
ERA-Interim reanalysis
- a lifting trough locates over
the Rockies and provides
forcing for development of a
surface low over the Central
US
- substantial wind shear in
southern Nebraska where the
western MCS initiates around
21.00 UTC
5. - strong southerly hot and humid airflow near surface ahead of the surface low
- the northernmost part of the humid surface airmass is going under cooler mid
and high tropospheric airmass
→ build up of CAPE
6. - the CAPE reaches very high values at the left flank of the MCS west of the
Great Lakes
- the maximum value is ~4000 J/kg
- more than enough for strong convection
7. Errors in forecast convection
- next we take a closer look at the control forecast initialised on the 10th
of
April at 00.00 UTC
- this forecast is the worst in terms of ACC of day 6 forecast over
Europe
- we concentrate on type of errors and interactions between errors in
different fields
- CAPE and precipitation fields
8. - forecast CAPE and error of the forecast
- there is severe displacement error and moderate magnitude error in hour 6
forecast already
- error in CAPE leads to error in precipitation 3 hours later
9. - 3 hours after the large error in CAPE field has lead to growing displacement
error in precipitation
- area of heavy precipitation is moving too fast to east in the control forecast
10. - again 3 hours later the MCS is moving too south in the control forecast
11. - model sounding
from Omaha,
Iowa-Nebraska
border 10 Apr 00
UTC
- cloud layer in
the upper
troposphere,
probably anvil
cloud of the MCS
12. - actual sounding from Omaha
- no sign of upper level clouds
- considerably more CAPE than in the model sounding
13. Rossby wave source – What is it?
- first consider nonlinear vorticity equation on one level
- where v is horizontal velocity, ζ is vorticity, D is divergence and F is friction
- as we are looking at the Rossby wave source, we want to get [1] into form of
- where vψ
is rotational wind and S is the Rossby wave source
- this can be done by splitting wind into rotational and divergent components
- then [1] can be written like [2] and the Rossby wave source S is now
(Sardesmukh and Hoskins 1988)
14. - a couple of figures of
the Rossby wave
source
- the upper one is from
the control forecast and
the lower one is from
ERA-Interim reanalysis
- OpenIFS overshoots
the Rossby wave
source
- there is some
overestimation of the
Rossby wave source in
most of the figures
15. - later the overshooting
decreases but does not
end entirely
- note especially the
trough over the east
coast of the US
16. - OpenIFS is still
overshooting the
Rossby wave source
- wave breaking
created an omega
block over N Europe in
the forecast but there
is no sign of blocking in
reanalysis
17. Rossby wave breaking
- as Rossby wave breaking (RWB) is considered to be potential onset for
blocking, we decided to compare the breaking action between the control
forecast and the reanalysis
- RWB can be defined for example so that the gradient of PV has locally
southward pointing component i.e. PV contours are locally overturned or
reversed
- we plotted PV on 315 Kelvin potential temperature surface which actually
penetrates the dynamic tropopause at the midlatitudes
- on the slide 18 there are PV on 315K θ surface (shaded) and 500 hPa
geopotential height from the control forecast and on the slide 19 there are the
counterparts from the reanalysis
- the times are: (a) 10 Apr 00 UTC, (b) 11 Apr 00 UTC, (c) 12 Apr 00 UTC, (d)
13 Apr 00 UTC, (e) 14 Apr 00 UTC, (f) 15 Apr 00 UTC and (g) 16 Apr 00 UTC
respectively in both figures
- the first RWB´s over North America and the Atlantic occur quite similarly
(figs a-c) but in figs d-f there is a large difference in the RWB over the Atlantic
between the control forecast and the reanalysis; the wave breaks cyclonically
in the forecast but anticyclonically in the reanalysis
- the RWB is associated with the same trough which located over the Rockies
at the time of the initialisation of the control forecast
20. 500 hPa geopotential height 16 Apr 2011 00 UTC
The day 6 forecast vs. the reanalysis
- Z500 hPa once again but
now in colours
- the upper one is from the
control forecast and the lower
one is from ERA-Interim
reanalysis
- the control forecast
predicted blocking over
Northern Europe
- no clear sign of blocking in
the reanalysis
- note also the cut-off low
west of the Canary Islands in
the reanalysis
21. Ensemble
- We ran first five members of the
ensemble for the worst initialisation
date of six day forecast
- none of the members performed
well
- the fifth member was the best
- the second member was the
worst
- finally we got some spread
- one problem: there are
perturbations all over the world in
ensemble members so some
contribution may come from
elsewhere than from North America
22. - spaghetti plots of 5400 m
contour of Z500
- green = the control forecast,
black = the reanalysis, other
colours = the ensemble
members
- every member of the ensemble
is forecasting blocking at least to
some degree
- there is a ridge over N Europe
also in reanalysis but every
member over estimated it
- the ensemble forecast and the
reanalysis are remarkably
consistent over the Atlantic and
inconsistent over Europe
23. - often ensemble mean has better skill than a single deterministic control
forecast
- this is not the case with these five members
- ACC of the ensemble mean is worse than of the control forecast (slide 21)
- green = the control forecast, red = the ensemble mean, black = ERA-
Interim reanalysis
- the blocking is only displaced slightly to east
24. Ensemble initial conditions and short forecasts
● we have also looked at the initial conditions. A couple of quick notes:
- characteristics of improving perturbation: amplified Rockies´ trough and
weak surface low
- characteristics of deteriorating perturbation: weak ridge east of the
trough and strong surface low
● CAPE over North America 6 hours after the initialisation
- at least I was a little surprised that difference in CAPE between the
ensemble members and the control forecast is 3-5 times smaller than
between the control forecast and the reanalysis
- none of the CAPE fields is significantly better than in the control forecast
● 3h precipitation 12h after the initialisation
- in most of the members the area of heavy precipitation over the Great
Lakes is moving even faster east than in the control forecast
● OpenIFS was not especially sensitive to the perturbations in these five
ensemble members
25. Sources
Rodwell, M.J. and Coauthors, (2013). Characteristics of occasional poor medium-
range weather forecasts for europe. Bull. Am. Meteorol. Soc., 94(9) 1393–1405.
Sardesmukh, P. D. and Hoskins, B.J., (1988). The Generation of Global Rotational
Flow by Steady Idealized Tropical Divergence. J. Atmos. Sci., 45(7), 1228–1251.