IPILPS Workshop ANSTO 18-22 April 2005 IPILPS Forcing & REMOiso Performance by Dr Matt Fischer and Kristof Sturm.

IPILPS Workshop

ANSTO 18-22 April 2005

IPILPS Forcing & REMOiso Performance

by Dr Matt Fischer and Kristof Sturm

Topics Topics • History

• ECHAM

• Forcing Variables

• Performance - Europe - South America - Australia

• FAR interpolation

HistoryHistory• PILPS, 8 forcing variables: 2 radiation,

pressure, 2 wind, temp., r’fall, humidity

• IPILPS, where to get isotope var’s from?

• Observations?: high resolution data rare, & no information on spatial variance

• Models? - MUGCM - GISS - ECHAM - REMOiso

•

• Resolution ?

• Climate ?

• Isotopes ?

• Locations ?

ECHAMECHAM• ECMWF model, aimed

at climate simulations• ECHAMiso: Spectral

resolution : T30, i.e. 3.75º (~ 450 km)

• Vertical resolution: 19 levels

• REMOiso - RCM nested in ECHAMiso, run in Europe & South America

• List

Fo

rcin

g V

aria

ble

sF

orc

ing

Var

iab

les

Tumbarumba EQY1: RadiationTumbarumba EQY1: Radiation

• List

Tumbarumba EQY1: RainfallTumbarumba EQY1: Rainfall

Tumbarumba BC24: varianceTumbarumba BC24: variance

Tumbarumba BC24: varianceTumbarumba BC24: variance

• REMOiso was run for 4 years (except Manaus), at =< 5 minute resolution

• EQY1 Equilibrium experiment

• BC24 experiment

• FAR interpolation (for Manaus)

REMOisoREMOiso

Topics Topics • History

• ECHAM

• Forcing Variables• Performance

- Europe - South America - Australia

• FAR interpolation

LocationsLocations• Mid-latitude deciduous forest eg. Munich

• 48°N 16°E

• Tropical rainforest eg. Manaus

• 3°S 60°W

• Mid-latitude eucalypt forest eg.Tumbarumba

• 35°S 148°E

Nordeney 18O from April 97 to Feb 99

18O

18O as good, or bad, as precipitation amounts (Sturm et

al., 2004)

18O

Obs

REMO

EuropeEurope

EuropeEurope

South AmericaSouth America

• 5 minute simulations, written for 25 cells:

• Manaus x9

• Zongo x9

• Rocafuerte x9

• Belem x1

• Izobamba x1

1818O in precipitationO in precipitation

ManausManaus

AustraliaAustralia

• First experiment for this domain

• GNIP Stations:Perth, Darwin, Alice Springs, Brisbane, Adelaide, Cape Grim, Melbourne

• Tumbarumba x9

• Padthaway x9

18O Australia

January

July

18O Australia

TemperatureTemperature

RainfallRainfall

RainfallRainfall

1818OO

D-excessD-excess

Summary so far ...Summary so far ...

• REMOiso simulations of Europe and South America compare with data from GNIP stations and general climatology

• REMOiso simulations of Australia are too wet or dry for some sites, but generally compare with GNIP data, excpet perhaps for D-excess

Topics Topics • History

• ECHAM

• Forcing Variables

• Performance - Europe - South America - Australia

• FAR interpolation

Functional autoregression (FAR)Functional autoregression (FAR)• yt = yt-1 + xt + t

• but, we can replace x & y by orthonormal vector subsets.

• y is 5 minute data, x is 6 hour data

• Estimation: of the parameters , , (, ) using method of Damon & Guillas 2001

• Interpolation: Eigenvectors of x & y are ‘stretched & pulled’ over a new set of 6 hour data to form a new set of 5 min data.

• Applications: NCEP

Temperature & WindTemperature & Wind

FAR 5 min.

Linear 5 m.

Original 5 m.

Temperature & WindTemperature & Wind

Temperature & WindTemperature & Wind

FAR 5 min.

Linear 5 m.

Original 5 m.

RainfallRainfall

FAR 5 min.

Linear 5 m.

Original 5 m.

RainfallRainfall

RainfallRainfall

FAR 5 min.

Linear 5 m.

Original 5 m.

Future : more comparisons!Future : more comparisons!REMOiso comparison of:

• the stochastic properties of storms (duration, intensity, rainout) with Australian BOM data (40 yrs of 6 minute rainfall data, Capital cities only)

• Isotopic rainout of individual storms,for BOM data this is an inverse problem using storm data + monthly GNIP data

ConclusionsConclusions• REMOiso, a regional climate model,

produces monthly isotope values in 3 domains which compare with GNIP data, except, perhaps for D-excess

• FAR can be used to downscale observational data; and is based on functional relationships between different timescales, calibrated with an isotope model.

model

FARFAR

FAR 5 min.

Linear 5 m.

Original 5 m.

FARFAR

FAR 5 min.

Linear 5 m.

Original 5 m.

‘‘

• List