Verification Verification with SYNOP, TEMP, and GPS data P. Kaufmann, M. Arpagaus, MeteoSwiss P. Emiliani., E. Veccia., A. Galliani., UGM U. Pflüger, DWD.

Verification

Verification with SYNOP, TEMP,and GPS data

P. Kaufmann, M. Arpagaus, MeteoSwissP. Emiliani., E. Veccia., A. Galliani., UGM

U. Pflüger, DWD

Verification

Surface Verification with SYNOP

Verification scores neglecting a diurnal cycle

Verification

3

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

Availability of SYNOP data on GTS

Red dots: all “main” SYNOPS (6-hourly) availableGB: Sea level pressure but no pressure at station heightSpain, Italy, Balkans: Many stations do not report during the night (not even the automatically measured parameters)

PMSL

PS

Verification

4

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

SYNOP Pressure: Standard Deviation of Errors

Standard deviation of errors (STDE) is considerably larger in winter than in summerIn winter much larger in the north than in the south

Winter 2003/4, 42 – 48 h

Summer 2004, 42 – 48 h

Verification

5

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

SYNOP Pressure

Dependence of mean error (ME) on forecast rangeWinter: IncreaseSummer: Decrease

Winter 2003/4

Summer 2003

but …

Verification

6

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

SYNOP Pressure

… decrease in summer disappeared in summer 2004Effect of ECMWF-IFS boundary conditions?Supported by results of IFS test chain summer 2003

Summer 2004

Verification

7

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

Model Grid Points over Water on Coasts and Lakes

Winter 2003/4, forecast range 42 – 48 hCoasts and lakes: Large bias if model grid points over water are used!but: small bias for small islands and oil platforms – not all water points need to be eliminated

Nearest grid point regardless of surface type

Land points preferred

Verification

8

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

SYNOP Temperature

Strong cold bias in winter (see verification of daily cycle), pronounced in AlpsWeak cold bias in summerStandard deviation high in complex terrain

Winter 2003/4, 42 – 48 h

STDE

ME

Summer 2004, 42 – 48 h

ME

Verification

Verification with TEMPs

Vertical profiles

Verification

10

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

TEMP Geopotential Summer 2004 (JJA)

Sharp increase of mean error in stratosphere, reason unclearStandard deviation largest around tropopause LM only: negative bias below 600 hPa at +48 h

LM ME aLMo ME STDE

Verification

11

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

TEMP Geopotential aLMo – LAMI

“South of the Alps”aLMo 6 stationsLAMI 7 stations

LAMI only July, AugustBias differences to aLMo below 500 hPa:

Bias at +12 h more positiveNegative bias at +48 hLAMI and LM both driven by GME, aLMo by IFS

aLMo ME

aLMo ME

Verification

12

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

TEMP Temperature Summer 2004 (JJA)

Cold bias up to 0.7 K between 1000 and 750 hPaWarm bias 750 – 500 hPa, increase with forecast time up to 0.6 KWeak cold (LM: no) bias 500 – 250 hPaaLMo only: Strong cold bias in the stratosphere (150 hPa)Analysis (+00) h far better than all others – effect of nudgingStandard deviation largest spread around tropopause (200 hPa)

LM ME aLMo ME STDE

Verification

13

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

TEMP Temperature aLMo – LAMI

LAMI only July, AugustSmaller cold bias near surfaceWarm bias extends higher up, up to 200 hPa level

aLMo ME

Verification

14

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

TEMP Relative humidity Summer 2004 (JJA)

Bias of varying sign, 5 – 8%, related to bias of temperature?Moist below 850 hPaDry between 850 hPa and 500 hPaMoist between 500 hPa and 200 hPaDry above 200 hPaClear distinction between analysis and forecastsStandard deviation nearly uniform over large part of troposphere

LM ME aLMo ME STDE

Verification

15

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

TEMP Humidity aLMo – LAMI

Relative humidity and dew point temperature are not directly comparable

?

aLMo ME

Relative humidity bias Dew point bias

Verification

16

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

Long term evolution of RH RMSE

Decrease of RMSE for LM in 2004Effect of prognostic ice scheme and assimilation of pseudo-TEMPS from ECMWF into GME

LM RMSE of relative humidity

Verification

17

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

TEMP Wind speed aLMo Summer 2004 (JJA)

Mean error well below 1 m/sSlightly increasing with forecast time Largest in boundary layer and near or above tropopause

Boundary layer: Different sign of bias for LM and aLMoStandard deviation largest at tropopause, consistent with strongest winds

LM ME aLMo ME STDE

Verification

18

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

TEMP Wind speed aLMo – LAMI

Only stations south of the AlpsLargest bias at 700 hPa

Verification

Verification with GPS

Vertically Integrated Water Vapour

Verification

20

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

GPS Integrated Water Vapour

Path delay (W – W0) caused by refraction

Error for geodesySignal for meteorology

Refractivity index depends onDry-air density (hydrostatic delay)Humidity (wet delay)

Wet delay = total delay – hydrostatic delayIntegrated water vapour:

Need of surface pressure and temperature from nearby SYNOPCurrently calculated by KNMI

Verification

21

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

GPS data availability

Three processing centers deemed most reliable

Each covers primarily its home countryTime series have many gapsProcessing centers are not WMO members

LPT Switzerland

GFZ Germany GOP Czechia

Verification

22

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

GPS IWV BIAS

Weak dry bias outside complex topographyBias (or mean error) is largest

In summerIn the mountains

Possible explanationsLarger specific humidity contents in summer than in winterHeight differences GPS antenna – SYNOP station – model surface are largest in mountainous terrain Summer 2003

Winter 2002/3

Verification

23

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

GPS IWV Standard Deviation

STDE largest in summerHomogeneous in whole model domain

Problem in mountains only affect bias, i. e. values are shifted by constant offset

Summer 2003

Winter 2002/3

Verification

24

Pir

min

.Kaufm

ann@

Mete

oSw

iss.

ch6

th C

OSM

O G

enera

l M

eeti

ng,

22

-24

Sept

20

04

, Li

nate

Summary and Outlook

Model verification often stops at national boundaries: Model comparison difficultSurface verification with SYNOP

Pressure increase/decrease with forecast rangeStrong cold bias in winterNeed for continuous, quality-controlled data for the whole of Europe!

Model comparison even for TEMP difficult. Differences:Set of stationsTime (definition of a season)Statistical measures used: Only two out of ME, MAE, STDE, RMSE

GPSData availability problem (currently only available through European project; each centre provides subset only )Zenith total delay (ZTD) verification in the future to avoid interpolation problems between SYNOP and GTS locations