temperature moisture & spinup land surface ocean surface
TRANSCRIPT
Workshop on bias estimation, November 2005 1
Model biases
Anton Beljaars (ECMWF)
with contributions by: Antonio Garcia-Mendez, Anna Ghelli, Isabel Trigo, Pedro Viterbo, Xubin Cheng
• Temperature • Moisture & spinup• Land surface• Ocean surface
Workshop on bias estimation, November 2005 2
Systematic errors (differences): Monthly averages of temperature 200508 (fc-an)
Workshop on bias estimation, November 2005 3
Radio sonde statistics of temperature for August 2005 (20N-20S)
Workshop on bias estimation, November 2005 4
-5 -4 -3 -2 -1 0 1 2 3 4 5OBS-FG (C)
1000
925
850
700
500
400
300
250
200
150
100
70
50
30
20
10
5
0H
PA
16384
21960
23915
24786
24743
24653
24551
24493
24377
24156
23723
20650
19244
15986
11721
5532
107
3356
5280
5640
5633
5604
5583
5566
5537
5505
5464
5368
5287
5168
4646
3672
2023
39
9041
12052
12330
12539
12512
12486
12464
12454
12385
12266
12031
11072
10807
9996
8419
5328
103
7658
10563
11141
11395
11374
11329
11283
11264
11235
11094
10914
10290
10034
9102
7177
4310
62
7202
9390
9847
10082
10060
10031
9989
9947
9874
9775
9664
9090
8852
7954
6216
2791
38
JAN 2004 - JUL 2005MEAN OBS-FG TEMPERATURE DIFFERENCES V SOLAR ELEVATION
COMBINED SONDES
GUAN RS VRS80
E< -7.5 -7.5 < E < 7.5 7.5 < E < 27.5 27.5 < E < 47.5 47.5 < E
Workshop on bias estimation, November 2005 5
-5 -4 -3 -2 -1 0 1 2 3 4 5OBS-FG (C)
1000
925
850
700
500
400
300
250
200
150
100
70
50
30
20
10
5
0H
PA
3380
4695
5271
5263
5511
5501
5488
5480
5459
5376
5279
5066
4880
4240
3085
1820
16
412
1520
1514
1510
1591
1590
1583
1571
1564
1544
1500
1440
1320
730
440
187
0
1445
2434
2521
2525
2865
2864
2855
2844
2838
2805
2777
2724
2637
2085
1583
1103
9
2330
3316
3482
3480
3468
3459
3436
3405
3384
3346
3322
3276
3192
2622
1492
766
3
1599
1997
2272
2272
2260
2256
2247
2242
2231
2220
2205
2222
2172
1977
1328
699
0
JAN 2004 - JUL 2005MEAN OBS-FG TEMPERATURE DIFFERENCES V SOLAR ELEVATION
COMBINED SONDES
GUAN RS VRS90-92
E< -7.5 -7.5 < E < 7.5 7.5 < E < 27.5 27.5 < E < 47.5 47.5 < E
Workshop on bias estimation, November 2005 6
Long wave radiation tendencies averaged over August 2002 (ERA40)
Workshop on bias estimation, November 2005 7
Short wave radiation tendencies averaged over August 2002 (ERA40)
Workshop on bias estimation, November 2005 8
Long + short wave radiation tendencies averaged over August 2002 (ERA40)
Workshop on bias estimation, November 2005 9
Systematic differences: Monthly averages 200508 (fc-an)
Relative humidity (%)
Day 1
Day 2 Day 5
Workshop on bias estimation, November 2005 10
Systematic errors (differences): Monthly averages 200508 (fc-an)Relative humidity (%)
Vert. Velocilty(Pa/s)
Day 1
Day 2
Workshop on bias estimation, November 2005 11
Radio sonde statistics of relative humidity for August 2005 (20N-20S)
All VRS80&VRS90 sondes Used VRS80&VRS90 sondes
Workshop on bias estimation, November 2005 12
Systematic errors (differences): Monthly averages 200508 (fc-an)
•xx
Total column water vapour
Total column water vapour (24-hr fc –AN)
Workshop on bias estimation, November 2005 13
Systematic differences: Monthly averages 200508 (fc-an)
Relative humidity difference (%) 700-hPa(24-hr fc – anal)
Relative humidity difference (%) 925-hPa(24-hr fc – anal)
Workshop on bias estimation, November 2005 14
Systematic differences: Monthly averages 200508 (fc-an)
•xx
Total precipitation (mm/day) (0_6 hrs)
Total precipitation difference (mm/day) (48_54-0_6 hrs)
Workshop on bias estimation, November 2005 15
Systematic differences: Monthly averages 200508 (fc-an)
Vertical velocity (Pa/s) 500-hPa(an)
Vertical velocity difference (Pa/s) 500-hPa(24-hr fc-an)
Workshop on bias estimation, November 2005 16
Monthly averaged thermodynamic profiles 200508: (24 Hr fcsts/Sonde)Honolulu Canary Islands
Workshop on bias estimation, November 2005 17
Monthly averaged thermodynamic profiles 200508: (24 Hr fcsts/Sonde)
Dom. Rep. Palau
Workshop on bias estimation, November 2005 18
Radiosondes-land sfc-700 hpa
obs-FGobs-AN
14AUG
2005
15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 1 2 3 4 5 6 7 8 9 10 11 12 13 14SEP
2005
-28-26-24-22-20-18-16-14-12-10
-8-6-4-202468
10121416182022242628
hum
idity
%
BIAS OB-FG BIAS OB-AN
14AUG
2005
15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 1 2 3 4 5 6 7 8 9 10 11 12 13 14SEP
2005
-14
-12
-10
-8
-6
-4
-2
0
2
4
6
8
10
12
14hu
mid
ity %
BIAS OB-FG BIAS OB-AN
SYNOP’s-land obs-FGobs-AN
Model 1 to 2 % too moist compared to sondes
Model about 2 % too dry compared to SYNOP’s
Workshop on bias estimation, November 2005 19
xx
200 m above the surface
2 m above the surfaceData: Fred Bosveld, KNMI
Time series of q at Cabauw (Netherlands)
Workshop on bias estimation, November 2005 20
Conclusions on moisture
• Errors are small• The errors with respect to analyses are confirmed by
sonde profiles• Moisture structure is controlled by:
1. Advection (Vertical + horizontal)2. Boundary layer diffusion (including shallow
convection clouds + stratocumulus clouds)3. Deep convection
• Errors in moisture are likely to be related to errors in all 3, but vertical motion (divergent motion) may dominate
• Moisture acts like a tracer: dynamic link between moisture increments and divergent flow is needed (more emphasis on dynamic control variable rather than q to adjust moisture in 4DVAR?)
Workshop on bias estimation, November 2005 21
History of spin-up (20N-20S)
eConvergencEPdt
dTCWV+−−=
Workshop on bias estimation, November 2005 22
History of spin-up
(20N-20S)
29R2: Assim. Rain, limit q-incr
29R1: Moist BL, num. 1st step
28R3: num. cld+cnv, FG from 6+18, No SYNOP-q at night
28R2: Early delivery
28R1: Conv. changes
26R3: AIRS, New q-anal
25R3: Mult. Incr. 4DVAR, SSMI-rad, clds+cnv changes
25R1
24R3: Conv.+supersat
23R4:
List with selection of model changes. In addition there were many changes to the use of satellite data.
Workshop on bias estimation, November 2005 23
History of spin-up
(20N-20S)
29R2: Assim. Rain, limit q-incr
29R1: Moist BL, num. 1st step
28R3: num. cld+cnv, FG from 6+18, No SYNOP-q at night
28R2: Early delivery
28R1: Conv. changes
26R3: AIRS, New q-anal
25R3: Mult. Incr. 4DVAR, SSMI-rad, clds+cnv changes
25R1
24R3: Conv.+supersat
23R4:
List with selection of model changes. In addition there were many changes to the use of satellite data.
Workshop on bias estimation, November 2005 24
History of spin-up (20N-20S)
Workshop on bias estimation, November 2005 25
Conclusions on spin-up
• Spin-up has improved with: 1. A modest reduction in precipitation spindown2. A substantial reduction in TCWV spindown3. A change from increase of evaporation during the
forecast to a decrease of evaporation. (BL has become more dry in analysis).
• It is difficult to make a precise link between model changes and impact on spin-up
• Model changes and data assimilation changes (including use of satellite data) have contributed
• It is impossible to verify TCWV within 1 kg/m2 using radio sonde data.
Workshop on bias estimation, November 2005 26
Surface skin temperature (radiative surface temperature) as seen “from the top of the atmosphere”
• The use of remote sensing channels that peak in the lower troposphere requires an accurate background field of skin temperature
• Current model skin temperatures have large errors over land, underestimating the diurnal cycle, in arid/semi-arid areas
Trigo and Viterbo, 2002
METEOSAT
Clear sky Tb window channel
OBS - model
Workshop on bias estimation, November 2005 27
2m temperature verification (February 2005)
Workshop on bias estimation, November 2005 28
What controls the surface skin temperature?
• Available radiation at the surface• Partitioning of available energy between sensible and latent heat
flux• Aerodynamic coupling of skin to atmosphere (in TESSEL controlled
by land roughness lengths for momentum and heat)• Coupling of skin (vegetation canopy, litter layer, dry top soil layer)
to underlying soil
skT skin layer; no heat capacity
Soil layer 11sT
2sT Soil layer 2
ar
skΛ/1
EλnetQ H
G
T60T
Workshop on bias estimation, November 2005 29
Coupling to the atmosphere• Possible solution for low vegetation tiles
– Bare ground contribution to the sensible heat flux: Introduce a resistance Rs, mimicking the wind shielding effect of the surrounding vegetation.
– Parameters a and b (fixed for each grid point) in Rs are estimated by fitting the modelled daily amplitudes of Tb_BG to Meteosat clear sky observations, for the window channel.
TESSEL
2*
1sR
a b u=
+
RaRa
Tc Tsk
Improved Scheme
Rs
RaRa
Tc
Tsk
2*
1=
+sRa bu
Workshop on bias estimation, November 2005 30
Development and validation• SCM + (radiative transfer model) applied to points in semi-arid
areas. Triple collocation of model, point observations and METEOSAT ensures that the improvements to skin temperature do not degrade two-metre temperature
• Preliminary results (Western Sahara, one point). Amplitude of Tb diurnal cycle:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 155
10
15
20
25
30
35
40
Tb-Am
p (
K)
1-15 FEB 2001;26N08E
ATbobsATbctrlATbnew
0 2 4 6 815
20
25
30
35
40
Tb-Am
p (
K)
Wind Speed 10m (ms-1)
1-15 FEB 2001;26N08E
OBSCTRLNEW
Calibration period: 2001
10m wind speed (ms-1)Day of the month
1 2 3 4 5 6 7 8 9 10 11 12 13 14 155
10
15
20
25
30
35
40
Tb-Am
p (
K)
1-15 FEB 2002;26N08E
ATbobsATbctrlATbnew
0 2 4 6 8 1010
15
20
25
30
35
Tb-Am
p (
K)
Wind Speed 10m (ms-1)
1-15 FEB 2002;26N08E
OBSCTRLNEW
Validation period: 2002
10m wind speed (ms-1)Day of the month
Workshop on bias estimation, November 2005 31
Skin temperature: Coupling to the soil• Model changes
– Turbulent orographic form drag (TOFD) parameterisation
– Tested additional resistance to heat (Rs)– Reduce coupling from skin to soil (Λ)
• Future– Test the changes above in data assimilation, assess Tsk against
GOES/METEOSAT
286325290309TOFD+Rs+Λ
322
320
DayTskin
292
293
Night
309
310
DayT2Cycle
289TOFD
29026r3
Night
TESSEL
2*
1sR
a b u=
+Ra
Ra Tc Ts
k
Improved Scheme
Rs
Ra Tc Ts
k
Workshop on bias estimation, November 2005 32
1-15 Feb 2001 1-15 Feb 2002Control -8.9 K -8.5 KRevised -0.4 K -0.1 K
Tb amplitude: Model - Observations
Summary Sahara
•The revised formulation is capable of removing the very large model bias in the diurnal amplitude
•Parameters calibrated in 2001, are used unchanged in 2002, with equal success
Workshop on bias estimation, November 2005 33
History of 2m T-errors over Europe in the ECMWF model (step60/72)
•Model temperature errors are influenced by many processes. •Observations at process level are needed to disentangle their effect
Bias (day/night)
RMS (day/night)
Workshop on bias estimation, November 2005 34
SYNOP (2m) humidity verification over Europe (monthly)
Workshop on bias estimation, November 2005 35
Soil temperature verification over Germany (layer 1)
Workshop on bias estimation, November 2005 36
Soil temperature verification over Germany (layer 2)
Workshop on bias estimation, November 2005 37
Ocean warm layer effect (diurnal cycle)
Fairall et al.
TOGA/COARE data (19921111-19921203)
Workshop on bias estimation, November 2005 38
Ocean warm layer model
)()/(
)1()1/(
*ds
w
ww
dsds TT
Lddku
cdRRQTT −−
− −+
−+
−+=−
φν
ννρ
velocityfrictionumscaledepthd
profiletempforparamshapevREHQ
w
LW
===
++=
*
)3()3.0(..
λ
Cheng and Beljaars 2005
.)( tempsurfaceoceanTs
sRQ
dR−
.)( tempbulkoceanT d−
Workshop on bias estimation, November 2005 39
Ocean warm layer model
Observations during TOGA/COARE (December 1992)
Workshop on bias estimation, November 2005 40
Diurnal cycle of ocean surface temperature
Ts difference between 12 UTC and 20 UTC (model 20-22 may 1998)
Wind speed
Derived from GOES (Wu et al. 1999)
Workshop on bias estimation, November 2005 41
Concluding remarks
• Temperature errors are small except in stratosphere. Radio sondes give info up to 10 hPa.
• Moisture structure near boundary layer top shows errors related to BL cloud processes.
• Day time land humidity shows dry bias in summer (shallow convection).
• Spin-up is very subtle and moisture observations are not calibrated well enough to distinguish between model and observation biases.
• SYNOP’s and radio sondes show systematic differences. • Land “radiative” surface temperature shows large biases
related to unknown coupling coefficients between air, surface skin and soil. These coefficients depend on: location, vegetation, surface condition (e.g. wetness), flow, solar elevation.
• Diurnal cycles in ocean surface temperature may be relevant for data assimilation.