GLASS : activities in 2001

Jan Polcher, Laura Bowling, Bart Nijssen, Nicolas Viovy,Aaron Boone, Luis Bastidas, Randy Koster

WGNE/GMPP, October 2001

• Short description of the structure of the GLASS project.

• New results from PILPS-2e and the planed PILPS-C1 experiment.

• Rhone-AGG is under-way and prepares future global/off-line exper-iments.

• New evidence has highlighted the need for a new approach to forcingland-surface models (LSMs) outside of atmospheric models.

• Two experimental designed for the global/coupled action have beentested.

Next GLASS science panel meeting is set for 7th to 9th November 2001.

1. Organization and managementGLASS will coordinate LSM inter-comparisons fromlocal to globalscales and from off-line forcing to fully coupled.

2. Local-scale/Off-line actionCoordination : Ann Henderson-Sellers, Andy Pitman

2.1. PILPS-2e : L.C. Bowling, B. Nijssen, D.P. Lettenmaier

This experiment aimed at evaluating the ability of land-surface models tosimulate the surface hydrology in northern latitude.

66.0

66.5

67.0

67.5

68.0

68.5

69.0

69.5

66.0

66.5

67.0

67.5

68.0

68.5

69.0

69.5

18 19 20 21 22 23 24 25

18 19 20 21 22 23 24 25

H

K

N

P

A

Ovre Abisko

Kaalasjarvi

Ovre Lansjarv

Pello

Bifurcation

Precipitation Temperature Synoptic stations

The following processes are evaluated:

• Snow regime : Are ablation or sub-limation the main uncertainties ?

• Turbulent fluxes in cold climates :Can we reliably validate surfacefluxes in an off-line environment ?

• Frozen soil and surface storage :Is the representation of these pro-cesses needed for simulating runoffin these latitude ?

21 models have taken part in the PILPS-2e experiment.

Four group of models can be identified :

-15

-10

-5

0

5

10

Ave

rage

ann

ual s

ensi

ble

heat

flux

(W

/m2 )

15 20 25 30 35 40 45

Average annual latent heat flux (W/m2)

AB

C

D

E

F

GH

I

J

K

L

M

N

O

P

Q R S

T

U

Rnet = 16.5 W

/m2

Rnet = 39 W

/m2

Group 1 Group 2 Group 3 Group 4

• The vertical line gives an estimateof the latent heat flux based on anannual water balance.

• Group 4 models (blue) overesti-mate annual mean latent heat flux.

• All other models are within the un-certainty of the estimate.

• A wide range of sensible heatfluxes is observed.

Sublimation is the major cause of uncertainty between models in thisregion of the world.

0

50

100

150

200

250

300

Ave

rage

Feb

ruar

y S

WE

(m

m)

A B C D E F G H I J K L M N O P Q R S T U

-10

-5

0

5

10

15

20

25

30

35

Ave

rage

turb

ulen

t flu

xes

(W/m

2 )

A B C D E F G H I J K L M N O P Q R S T U

Latent heat flux Sensible heat flux

The large differences in the sensible heat flux are explained by the de-coupling of the atmosphere and surface in stable conditions.These are very sensitive situations, especially in un-couped land-surfaceschemes, which can lead to very large near surface temperature gradients.

How do these differences impact the simulated river outflows ?

0

2

4

6

8 A B C D E F

0

2

4

6

8

Ave

rage

run

off (

mm

/day

)

G H I J K

0

2

4

6

8 L M N O

J M M J S N

P

J M M J S N

Q

0

2

4

6

8

J M M J S N

R

J M M J S N

S

J M M J S N

T

J M M J S N

U

Standard dev.

Models which include a representation of lakes and wetlands (H,J) donot stand out.

2.2. PILPS-C1 : Nicolas Viovy

This is the first experiment which will evaluate the ability of land-surfacemodels to simulate carbon fluxes.The selected data is from the Euroflux site at Loobos in the Netherlands.This is a 100 year old forest standing on a sandy soil.

• Which models are able to represent both the biophysical and biogeo-chemical processes ?

• Can models reproduce the observed carbon sinks ?

The first simulations will be un-constrained, this should allow the modelsto run in their GCM configuration and simulate all processes.In a second step parameters will be prescribed to narrow down differ-ences.This type of experiment bring with it new problems : How should thecarbon cycle be spun-up in the models ?It is proposed to run a 100 yearspin-up with data reconstructed from historical observations.

The data and the feasibility of the experiment was tested with OR-CHIDEE at IPSL :

The release of forcing data is set for January 2002.

3. Global-scale/Off-line actionCoordination : Paul Dirmeyer, Taikan Oki

3.1. Rhone-AGG : A. Boone, F. Habets, J. Noilhan

gauging stations

0

50 100

150 200

2500

100

200

300

400

500

gauging stations

0

50 100

150 200

2500

100

200

300

400

500 • How do the various LSMs simulate thedischarge compared to the observationsfor several annual cycles and sub-basins?

• Are the sub-grid runoff and drainage pa-rameterizations scale dependent?

• How do results compare among the vari-ous aggregation methods employed?

• How does soil moisture scale among theschemes?

• What is the impact of grid resolution onthe simulated SWE and snowmelt runoff?

The Rhone-AGG experiment will hold it’s workshop from the 5th to the7th of November 2001.

13 models are participating in the Rhone experiment.

First results for the validation of the snow cover at 24 stations :

0

120 2400.0

0.6

1.2

1.8

0.0

0.6

1.2

1.8

Sno

wD

epth

(m

)

0.0

0.6

1.2

1.8

ObsSim

0

120 240Time (Days since Aug. 1)

Average for 3 years and 24 sites

0

120 240

A

B

C

D

E

F

G

H

I

rms .281

r2 .728

b .021

rms .421

r2 .564

b −.256

rms .836

r2 .087

b −.635

rms .439

r2 .558

b −.260

rms .320

r2 .693

b .007

rms .354

r2 .656

b .158

rms .335

r2 .680

b −.136

rms .360

r2 .645

b .066

rms .269

r2 .740

b .016

0

120 2400.0

0.6

1.2

1.8

0.0

0.6

1.2

1.8

Sno

wD

epth

(m

)

0.0

0.6

1.2

1.8

ObsObsSim

0

120 240Time (Days since Aug. 1)

Average for 3 years and 24 sites

0

120 240

J

K

L

M

N

O

P

Q

R

rms .281

r2 .729

b −.068

rms .865

r2 .561

b .298

rms .290

r2 .725

b .071

rms .584

r2 .451

b −.418

3.2. Planed experiments :Paul Dirmeyer, Taikan Oki

GSWP-2 will use the ISLSCP-II data to perform global off-line simula-tions over a 10 year period.Inter-annual variability of soil moisture willbe the focus.In the meantime GSWP-1.5 will help prepare phase 2(P.I. : Taikan Oki)The following questions will be addressed :

• How Sensitive are LSMs to errors in the forcing data ?

• How well can LSMs be validated at the global scale with remotesensed data ?

• Are the drying out cycles of LSMs comparable ?

• Are theCO2 fluxes simulated at the global scale comparable ?

The added value to GSWP-1 will be :

• Updated and improved forcing data,

• the participation of more schemes and

• bring in the remote sensing and GAIM/IGBP communities.

4. Local-scale/Coupled actionCoordination : Hoshin Gupta, Paul Houser, Pedro Viterbo

4.1. Some evidence of the importance of coupling :Luis Basti-das, Hoshin Gupta

In each case the optimal parameters are determined using a Multi-Criteriamethod.

Rad

iatio

n (S

W +

LW

)

Hum

idity

Pre

cipi

tatio

n

Win

d

Air

tem

pera

ture

Rad

iatio

n (S

W +

LW

)

Hum

idity

Pre

cipi

tatio

n

Win

d

Air

tem

pera

ture

Rad

iatio

n (S

W +

LW

)

Hum

idity

Pre

cipi

tatio

n

Win

d

Air

tem

pera

ture

Lataent heat

flux

Sensible heat

flux

Lataent heat

flux

Sensible heat

flux

Lataent heat

flux

Sensible heat

flux

Off−line LSM Partialy coupled LSM Fully coupled LSM

• Does the coupling to the atmosphere modify the parameters found ?

• Is the convergence toward the optimal parameters modified ?

• Is any of the forcing predominant in the behaviors of the fully cou-pled system ?

0

20

40

60

RM

SE

of

Lat

ent

Hea

t

Independent Sensitivity Analysis, Feedback w/ One Term Only Each Time

offline(no feedback) feedback w/ precip. only feedback w/ specific humid. only feedback w/ air temp. only feedback w/ solar only

Crop Rough. Length Veg. Cover Frac. Porosity Hydraulic Conduc. Soil Thermal Conduc. Soil Heat Capacity

0

10

20

30

40

RM

SE

of

Sen

sib

le H

eat

0.05 0.10

0.5

1

1.5

2

RM

SE

of

Gro

un

d T

emp

.

0.4 0.6 0.8 0.4 0.5 0.6 0.0001 0.01 6 8 10 2 4

x 106

4.2. Forcing LSMs for data assimilation : Paul Houser, PedroViterbo

The various XLDAS projects under way have already shown the value ofusing off-line LSMs for data assimilation.Still a number of problems exist :

• The lack of atmospheric feedback reduces the impact of some obser-vations.

• The near surface atmospheric variable errors can not be used to ex-tract information.

A simplified forcing method is needed to keep the advantage of an off-line system but giving some atmospheric feedback.

A workshop will be held on this topic early in 2002.

5. Global-Scale/Coupled actionCoordination : Peter Cox, Randy Koster

5.1. The poor man’s LDAS : Randy Koster

Currently an experimental design which would apply a flux correction onprecipitation is explored.

Start of forecast: Assume soil moisture is "realistic";

"LDAS-style" phase: soil moisturesguided by observed precipitation

Forecast phase: soil moistures

SOIL MOISTURE INITIALIZATION

are free to evolve and diverge

March April May June July August

moisturesoil

stop use of observed precip.

This experimental set-up allows to determine in mutli-model ensemble :

• The value of improved precipitation for land-surface processes,

• The resulting improvement in the simulated climate,

• And finally the role of improved initial conditions at the surface.

120W 100W 80W

30N

40N

50N

P anomaly (pred-amip):1999 June

****

-5.0

-3.0

-2.0

-1.0

-0.7

-0.5

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.5

0.7

1.0

2.0

3.0

5.0

10.0

120W 100W 80W

30N

40N

50N

w anomaly (pred-amip):1999 June

-0.25

-0.22

-0.20

-0.17

-0.15

-0.12

-0.10

-0.08

-0.04

-0.01

0.00

0.01

0.04

0.07

0.10

0.12

0.15

0.18

0.20

0.22

0.25

120W 100W 80W

30N

40N

50N

P anomaly (pred-amip):1999 July

****

-5.0

-3.0

-2.0

-1.0

-0.7

-0.5

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.5

0.7

1.0

2.0

3.0

5.0

10.0

120W 100W 80W

30N

40N

50N

T anomaly (pred-amip):1999 June

****

-9.0

-8.0

-7.0

-6.0

-5.0

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

120W 100W 80W

30N

40N

50N

w anomaly (pred-amip):1999 July

-0.25

-0.22

-0.20

-0.17

-0.15

-0.12

-0.10

-0.08

-0.04

-0.01

0.00

0.01

0.04

0.07

0.10

0.12

0.15

0.18

0.20

0.22

0.25

120W 100W 80W

30N

40N

50N

T anomaly (pred-amip):1999 July

****

-9.0

-8.0

-7.0

-6.0

-5.0

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

120W 100W 80W

30N

40N

50N

T anomaly (pred-amip):1999 Aug

****

-9.0

-8.0

-7.0

-6.0

-5.0

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

120W 100W 80W

30N

40N

50N

w anomaly (pred-amip):1999 Aug

-0.25

-0.22

-0.20

-0.17

-0.15

-0.12

-0.10

-0.08

-0.04

-0.01

0.00

0.01

0.04

0.07

0.10

0.12

0.15

0.18

0.20

0.22

0.25

120W 100W 80W

30N

40N

50N

P anomaly (pred-amip):1999 Aug

****

-5.0

-3.0

-2.0

-1.0

-0.7

-0.5

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.5

0.7

1.0

2.0

3.0

5.0

10.0

1999

ROOT ZONEMOISTURE

"Initialization" ensemble minus "AMIP" ensemble.(Differences shown only where significant at 5% level.)

ANOMALY DIFFERENCES DUE TO SOIL MOISTURE INITIALIZATION

TEMPERATUREPRECIPITATION

June,

July,

Aug.,

1999

1999

An alternate experimental designwould be to restore in an ensembleof simulations the surface to the output of a previous experiment (La-beled R). This is then be compared to an ensemble of the free runningGCM (Labeled F).

0

5 10 15 20 25 30DAYS

0

5

10

15

20

P (

mm

/day

)

Omega[R]=0.85

0

5 10 15 20 25 30DAYS

0

5

10

15

20

P (

mm

/day

)

Omega[R]=0.07

Each ensemble is characterized bythe ratio of signal variance to totalvariance :ΩP .The two ensemble in this experi-ment can then be compared by thedifference inΩP :

ΩP (R)− ΩP (F ) small :Restoring the surface ateach time step will not havestrongly modified the internalvariability.

ΩP (R)− ΩP (F ) large :The constraint on the surfacehas strongly reduced theinternal variability of themodel.

Models show a wide range of sensitivity in land-atmosphere interactionsΩP (R)− ΩP (F ) for precipitation

180 120W 60W 0 60E 120E 180

90S

60S

30S

EQ

30N

60N

90N

0.000.050.080.110.150.200.250.300.350.400.450.500.550.600.650.700.800.901.00

NSIPP

ΩP[R] - ΩP[W]: Impact of Land BCs on 3-Day Precipitation

180 120W 60W 0 60E 120E 180

90S

60S

30S

EQ

30N

60N

90N

0.000.050.080.110.150.200.250.300.350.400.450.500.550.600.650.700.800.901.00

CCM3/BATS

180 120W 60W 0 60E 120E 180

90S

60S

30S

EQ

30N

60N

90N

0.000.050.080.110.150.200.250.300.350.400.450.500.550.600.650.700.800.901.00

COLA

180 120W 60W 0 60E 120E 180

90S

60S

30S

EQ

30N

60N

90N

0.000.050.080.110.150.200.250.300.350.400.450.500.550.600.650.700.800.901.00

HadAM3

6. conclusions• All actions are progressing well, even if some delays have occurred.

• The infrastructure action ALMA has already proved it’s usefulness.

• The local-coupled action needs to intensify it’s activity. It is going tobe key for the interactions of GLASS with :

1. GCSS for an improved land-surface treatment in cloud experi-ments.

2. GABLS to look for a better way of forcing land-surface schemesoutside of atmospheric models.

3. XLDAS to improve LSM setups for data assimilation.

Items on next weeks meeting of the panel are :

• Review of current and planed experiments.

• GSWP-1.5 is late but key to the interaction with the remote-sensingcommunity.

• Discuss the request from GCSS for a better solution than prescribingsurface fluxes in cloud resolving models.

• Discuss the request of CLIVAR-AAMP for experiments to evaluatethe role of surface processes in monsoonal systems.