Title: COSMO Priority Project Tackle deficiencies in quantitative precipitation forecasts
1COSMO Priority Project Tackle deficiencies in
quantitative precipitation forecasts
S. Dierer1, M. Arpagaus1, U. Damrath2, A.
Seifert2, J. Achimowicz8, E. Avgoustoglou7, M.
Baldauf2, R. Dumitrache9, V. Fragkouli7, F.
Grazzini3, P. Louka7, P. Mercogliano6, P.
Mezzasalma3, M. Milelli4, D. Mironov2, A.
Morgillo3, E. Oberto4, A. Parodi5, I.V. Pescaru9,
U. Pflüger2, A. Sanna4, F. Schubiger1, K.
Starosta8, M. S. Tesini3 1MeteoSwiss (CH), 2DWD
(D), 3ARPA-ER (IT), 4ARPA-P (IT), 5Uni Genova
(IT), 6CIRA-CMCC (IT), 7HNMS (GR), 8IMGW (PO),
9NMA (RO)
- 29th EWGLAM Meeting, 9 October 2007, Dubrovnic
2Aim of PP QPF
Good quantitative precipitation forecast is a
challenging task also for the COSMO model
- The aim of PP QPF is improved knowledge about
- most suitable namelist settings or
- parts of the model that need to be reformulated
- to obtain a better QPF at 7 km horizontal grid
size
The project has a focus on model deficiencies
not on errors from e.g. initial and large scale
conditions
3Overview of PP QPF
- Task 1 Selection of test cases representative
for typical QPF deficiencies of COSMO model - Task 2 Definition of sensitivity studies
- Task 3 Run sensitivity studies and draw
conclusions
4List of test cases from all countries
5Forecast errors
- 10 cases of stratiform overestimation (8 from D,
CH and PO) - 4 cases of stratiform underestimation
- 3 cases of convective overestimation
- 7 cases of convective underestimation (6 from I
and GR)
6Sensitivity studies
- 1. Changes of initial conditions
- 2. Changes of numerical methods
- 3.1 Changes of microphysics
- 3.2 Changes of convection schemes
- 3.3 Changes of PBL schemes
7Sensitivity studies initial conditions
- Soil moisture increased/decreased by 20
- Initial humidity increased/decreased by 10
8Sensitivity studies numerical methods
- Halved time step
- Leapfrog, tri-cubic semi-Lagrange advection of QR
and QS - Runge-Kutta, tri-cubic semi-Lagrange advection of
QV, QC, QI, QR and QS - Runge-Kutta, flux-form advection of QV, QC, QI,
QR and QS - Runge-Kutta, flux form advection and T-p
dynamics - increased orography filtering
9Sensitivity studies physics 1 - microphysics
- New warm rain scheme (Seifert and Beheng 2001)
- Strong changes of ice microphysics and new warm
rain scheme - Moderate changes of ice microphysics and new warm
rain scheme
10Sensitivity studies physics 2 convection
- Modified Tiedtke scheme
- Kain-Fritsch/Bechtold scheme
- No parameterization of deep convection
11Sensitivity studies physics 3 PBL
- Decreased/increased scaling factor of height of
laminar boundary layer for heat - Decreased/increased stomatal resistance
- Decreased/increased laminar scaling factor for
heat over sea
12Relative change of 24h area average
precipitation, first forecast day (06 30)
Convection scheme
Snow microphysics
Runge-Kutta
?rel (rrexprrref)/rrref
Initial humidity
Cases
Vertical heat/moisture exchange
Romanian
Polish
Greek
Ital.(EuroLM)
Ital. (LAMI)
Swiss
?rr gt 30 10 lt ?rr lt30 0 lt ?rr lt 10 ?rr
0 0 gt ?rr gt -10 -10 gt ?rr gt -30 ?rr lt -30
German
rlam01
sto50
sea01
ws80
dt20
RKsl
RKtp
micro1
micro2
qv90
conmod
ws120
qv110
RKbott
micro3
conoff
rlam50
sto250
sea40
LFsl
oro
kfb
13Relative change of 24h area average precipitation
RK
RLAM
QV0
MP
CON
?rel (rrexprrref)/rrref
? ?rel
? ?rel
14Change of bias between simulated and measured
area average precipitation
overestimation
underestimation
stratiform
bias gt 200 100 lt bias lt200 bias 100 100 gt
bias gt 50 bias lt 50
convective
convection
Initial humidity
Runge-Kutta
snow microphysics
15Conclusions until now
- Strongest effect (5-40) on area average
precipitation by - Initial humidity
- Runge-Kutta
- microphysics
- convection scheme
- Strong effect for Roman and Greek cases
- Vertical heat/moisture exchange (extreme change
of RLAM) - Runge-Kutta
- reduces mean precipitation in most of the cases
- and has an overall positive effect on the results
- None of the studies completely solves a QPF
problem, but some give a significant improvement
for single cases like - changes of snow microphysics for a case with
overestimation of stratiform precipitation - Kain-Fritsch/Bechtold for underestimated
convective precipitation
16Cross experiments
17Overestimation of humidity? Monthly evaluation of
mixing ratio for COSMO-EU within GOP
January
July
Greifswald
Payerne, Switzerland
Greifswald, Germany
Payerne
18Precipitation bias of COSMO 3.19 vs 4.0
1 much better 11 slightly better 4 no impact 8
slightly worse
19Relative change of area average precipitation in
cross experiments compared to control simulation
?rr gt 30 10 lt ?rr lt30 0 lt ?rr lt 10 ?rr
0 0 gt ?rr gt -10 -10 gt ?rr gt -30 ?rr lt -30
20Bias of reference run and cross experiments
M-Swiss
Italy
HNMS
IMGW
NMA
DWD
21Relative bias of cross experiments
M-Swiss
Italy
HNMS
IMGW
NMA
DWD
22Improvement of area average precipitation in
numbers
- 17 cases improved by one of the studies
COSMO4.0 - KFB 5 cases (3C- / 2S)
- QV90RKTiedtkemod 3 cases
- QV90RKKFB 3 cases
- -/Tiedtkemod/RKQV90 2 cases
- 7 cases hardly affected or worse (5C-)
8S/3S-
23Conclusions
- COSMO4.0
- reduced initial humidity
- modified convection
- Runge-Kutta
- has a positive impact on stratiform
overestimation - little or negative impact on convective
underestimation - Few cases are solved
- COSMO Version 4.0 is a step forward!
- Further improvements expected from Runge-Kutta.
- We should have a closer look at the (initial)
humidity fields. Any improvements in data
assimilation expected? - Convection schemes are the next thing to look at.
- Draft of a final report has been written and will
be revised based on the discussion of PP QPF
sessions in Athens and will be available in the
next weeks - publication of results planned until end of the
year
24Precipitation bias of reference version
25List of sensitivity studies
by Axel Seifert, DWD
26List of sensitivity studies
by Axel Seifert, DWD
27List of sensitivity studies
by Axel Seifert, DWD
28List of sensitivity studies
by Axel Seifert, DWD
29List of sensitivity studies
by Axel Seifert, DWD
30Summer case with underestimated precipitation -
change of 24h area average precipitation sum with
Kain-Fritsch/Bechtold
by Axel Seifert
31Sensitivity studies better or worse agreement
with measured area average precipitation (green
improved red worse)