key considerations for simulating arctic weather and climate with limited area models nicole...

17
Key considerations for simulating Arctic weather and climate with limited area models Nicole Mölders University of Alaska Fairbanks, Geophysical Institute and College of Natural Sciences, and Mathematics, Atmospheric Science Program

Post on 20-Dec-2015

217 views

Category:

Documents


1 download

TRANSCRIPT

Key considerations for simulating Arctic weather and climatewith limited area models

Nicole Mölders

University of Alaska Fairbanks, Geophysical Institute and College of Natural Sciences, and Mathematics,

Atmospheric Science Program

Various scales have to be bridged

Step 1:Observe processat laboratory/field

scale

Step 2:Generalized

constitutive laws

Step 3:Conservation laws to derive partial

differential equation

Step 4:Predict system behavior

at different scalesthan lab/field scale

Small scales imbedded in large scales

2T

2T

1

t

T '''

TPTP '

stochasticquantity

Scales in atmosphere, soil, ocean differ

Ekin ≈ v2/2 ≈Lp2/Tp

2

Modified after Mölders 1999

Methods for scaling of processes at the

atmosphere-surface interfaceAggregation (upscaling) distribution to mean (lumping of info)

Disaggregation (downscaling) mean to distribution!!!! (transfer to more detail)

n

1iin321 n

1,,

)),t(BIAS,t,z,y,x,c,b,a,(f???,?, in321 ?

atmospheric grid cell at the surface

Schematic view of aggregation methods

M

1m

kmj

kj F

M

1F

explicit subgrid scheme(e.g. Seth et al. 1994)GESIMA

kdom,j

kj FF

several km

strategy of dominance

e.g. MM5, WRF

mosaic approach (e.g. Avissar & Pielke1989)RAMS, GESIMA,CCSM, most GCMs

n

1iij

kij

n

1iij

kj 1awithFaF

Aggregation method may affect results

From Mölders 2001

mosaic approach

dominance strategy

explicit subgrid

>0.0 0.01 0.03 0.05 0.07 0.09

Scales problematic in coupling

Modified after Mölders et al. 1999

>0.0 0.01 0.03 0.05 0.07 0.09

No feedback (one-way) With feedback (two-way)

1mm/h=106l/h=103m3/h

riverdischarge

runo

ff g

ener

atio

n

Hydrological approach

lower saturated zone

upper unsaturated zone

soil moisture

gravity

gravity

lakes

snow

ETinterception

Research interest-specific modeling approaches

gravity

gravity

runoff generation

ETinterception

Meteorological approach

snowsoil moisture 1

soil moisture 2

soil moisture n

diffusivity

diffusivity

runo

ff g

ener

atio

n

Model inconsistency potential error source

0 1 2 3

From Mölders et al. 1994

Cloud properties differ in meteorological & chemistry part of EURAD

Gas phase concentrations after cloud event differ for all species affected by cloud processes

Consistency required

Investigate uncertainty range resulting from parameterizations

Fulfill six important evaluation criteria for scientific credibility

Comparison to known analytical solutions

Determination of mass and energy budgets to determine conservations of these quantities

Comparison of model results with those of other models (model inter-comparison)

Comparison of model results to observations

Publication of model description/parts/modules in peer-reviewed journals

Code must be available on request

Sparse data, network design/density aggravate evaluationHistoric Network

Modified after PaiMazumder & Mölders 2007

100 sites network

400 sites network

Arctic networks along haul waysLess 1st class sites for precip than WMO recommends

Is error within uncertainty range of observations?Right or wrong for what reasons?

Lateral boundary & initial conditions introduce errors

Skill scores, methods for identification of error sources

-4

-3

-2

-1

0

1

2

3

4

5

152

154

156

158

160

162

164

166

168

170

172

174

176

178

180

DOY

erro

rs T

mea

n (

K)

BIASRMSESDEcorrelation

Modified after Mölders 2007

External forcing may introduce errors

Develop evaluation strategies

Which errors are due to external forcing?

Which errors are due to the limited area model?

Modified after Brown & Mölders 2007

Key considerations

Identify state-of-the-art and work/start from thereBridging of scales

Couple where necessary, not everywhere you couldCheck whether one or two-way coupling is requiredDefine data exchange (bottleneck in parallel processing!)Consistency within the model

Evaluation and analysis strategies6 criteriaIdentification of external error sourcesIdentification of “imported errors” from driving modelDetermination/definition of “investigation area”Identify & reduce uncertaintyHeuristic/indirect evaluations may be a chance

AcknowledgementsI thank

The IARC scientists for inviting me

M.E. Brown, D. Henderson, T. Fathauer, Z. Li, D. PaiMazumder, and G. Kramm for collaboration

NSF for support under contract OPP0327664

You for your attention