The new multilayer land surface scheme TERRA-ML

Validation and improvement with the Rhone-AGG dataset

M. Lange, Bodo Ritter, German Weather Service DWDemail: martin.lange@dwd.de

Outline

• The Rhone-Agg experiment• Soil model TERRA-ML• Setup for the validation experiment• Model results TERRA-ML vs. others• Summary and outlook

Rhone-Agg experimentwww.cnrm.meteo.fr/mc2/projects/rhoneagg/index.html

Validation domain

Physics of TERRA-ML

Physics of TERRA-ML

Experimental setup

• Run of soil model as stand alone scheme with external atmospheric forcing and prescribed soil and vegetation parameter.

• 4 years period 1.08.1985 -31.07.1989, 1 year spin up, 3 years validation

• Initialisation of soil moisture with July 89 results from prerun.

Model input data: 3 hour atmospheric forcing parameter from SAFRAN model

Variable Description UnitsRainf Liquid precipitation rate kg m-2 s-1

Snowf Liquid equivalent solid precipitation rate kg m-2 s-1

Tair Air temperature at 2 m KPsurf Surface pressure PaQair Specific humidity at 2 m kg kg-1

SWdown Incident shortwave (solar) radiation W m-2

LWdown Incident longwave (atmospheric) radiation W m-2

Wind Average wind speed at 10 m m s-1

Model input data: Soil properties

Variable Description Time step UnitsGrid box mean altitude - mLatitude and Longitude at cell center - degreesSand and Clay fractions -Total Soil Depth - mRooting Depth - mField Capacity volumetric water content - m3 m-3

Wilting Point volumetric water content m3 m-3

Hydraulic conductivity - m s-1

Soil porosity - m-3

Matric potential at saturation - mSlope of the retention curve - -Snow-Free Surface Albedo - -Minimum Stomatal Resistance - s m-1

Sub-grid surface-type weights - -LAI: Leaf Area Index monthly m2 m-2

Snow-Free Surface Roughness monthly mVegetation Cover fraction monthly -

ResultsResults from TERRA-ML are discussed as area averaged monthly mean values for the validation period (Aug. 1986 - July 1989) and compared with model results from ISBA (CNRM), ECMWF and NOAA.• 4 different runs are presented

– 1. initial test run

– 2. adjusted transpiration from vegetation

– 3. Snow aging by decrease of albedo

– 4. Snow melting from upper side

• Summary of different runs for TERRA-ML

TERRA-ML initial test runTERRA-ML ECMWF NOAH ISBA

TERRA-ML initial test runTERRA-ML ECMWF NOAH ISBA

Conclusion from 1st run

• Total evapotranspiration is much lower than in other models. This is due to a bug in transpiration from vegetation.– bare soil evaporation too large

– soil moisture too large

• 3D snow water equivalent increases– accumulates continuously, snow does not melt at high altitude

gridpoints during summer

2nd Run: adjusted transpiration from vegetation

• Model physics has been reviewed and transpiration from vegetation has been corrected (R. Schrodin).

Adjusted transpiration from vegetationTERRA-ML ECMWF NOAH ISBA

Adjusted transpiration from vegetationTERRA-ML ECMWF NOAH ISBA

Adjusted transpiration from vegetationTERRA-ML ECMWF NOAH ISBA

Adjusted transpiration from vegetationTERRA-ML ECMWF NOAH ISBA

Conclusion from 2nd run

• Evaporation components adjust according to correct transpiration from vegetation– Bare soil evaporation decreases

– Total evapotranspiration and corresponding latent heat flux fit to other model results,

– During summer latent heat flux slightly stronger, sensible heat flux weaker

• Soil moisture is at the lower edge of different models, variation is similar– Stronger transpiration dehumidifies the Soil

• 3D snow water equivalent still too large– Change could not be expected from modification of TVeg

3rd run: Aging of snow

• Snow melting is too weak – resulting from slow heating of thick snow layer

• Aging of snow leads to decrease of albedo. Enhanced absorption of solar radiation increase snow melting during summer.

• New process with continuous variation of snow albedo between maximum of 0.7 for fresh snow and minimum of 0.4 for old snow (B. Ritter).– Fresh snow is given after snowfall of 5 mm snow water

equivalent within less than one day.– Old snow is given after one week without snowfall.

Decrease of albedo by aging of snowTERRA-ML ECMWF NOAH ISBA

Decrease of albedo by aging of snowTERRA-ML ECMWF NOAH ISBA

Decrease of albedo by aging of snowTERRA-ML ECMWF NOAH ISBA

Conclusion from 3rd run• Evaporation and soil water components do not

change significantly.• Runoff does not change strongly although snow

melting increases during summer.• Increased snow melting from aging of snow

corrects accumulation of snow. – At some high altitude gridpoints snow does not vanish

totally in summer. Minimum snow albedo of 0.2 improves the result but is not realistic.

4th Run: Snow melting from top side• In previous simulations snow was melted from the

bottom side in the sense that no melting occurred before the top soil layer was heated up to Tmelt.

• In fact melting especially of glaciers occurs in far the most cases from the top side due to solar radiation or liquid precipitation.

• Therefore snow melting from the top side is included (B. Ritter, E. Heise), i.e. for the case that average snow temperature exceeds Tmelt and the temperature of top soil layer is below Tmelt.

Melting of snow from the top sideTERRA-ML ECMWF NOAH ISBA

Conclusion from 4th run• Snow water equivalent does not change

significantly, snow accumulation increases slightly.

• Evaporation and soil water components do not change significantly

• Surface runoff increases, sub surface runoff decreases during summer.

– Melted snow contributes directly to surface runoff. In earlier version surface runoff required saturation of top soil level up to field capacity.

Comparison of TERRA-ML simulations MELTING TOP SIDE

TVEG KORR

1. RUNAGING SNOW

MELTING TOP SIDE

TVEG KORR

1. RUNAGING SNOW

Comparison of TERRA-ML simulations

Comparison of TERRA-ML simulations MELTING TOP SIDE

TVEG KORR

1. RUNAGING SNOW

Comparison of TERRA-ML simulations MELTING TOP SIDE

TVEG KORR

1. RUNAGING SNOW

Summary• The Rhone-Agg dataset has been proved as a

powerfull tool for validation and improvement of TERRA-ML.

• The different simulations have discovered a bug in calculation of transpiration rate and deficiencies in the physical parameterization of snow melting at high altitude regions during summer.

• Problems have been fixed, some accumulation of snow still occurs. – In a 1 layer snow scheme melting is slowly for deep

snow since average temperature of the whole snowpack must exceed Tmelt.

Summary• TERRA-ML is a modern state of the art multilayer

LSS that compares with its main results to other prominent LSS.

• It has to be kept in mind that all results are „model results“ and for most components it is not clear what the „best“ result is.

• The comparison of river discharge calculated from the hydrological routing model MODCOU with observations is a part of CNRM work. – Presented?

Future tasks• Eldas project is designed to develop a land data

assimilation system. Soil moisture is optimized with respect to– 2m screen level temperature (1D Var data assimilation

Hess, 2001)

– 2m relative humidity

– surface radiation (Uni Bonn)

Acknowledgement• Piedro Viterbo from ECMWF Aaron Boone from

CNRM and Dag Lohmann from NOAA center to show their model results.

• Special thanks to Aaron Boone at CNRM who coordinates the Rhone-Agg experiment in the group of Joel Noilhan, and provided the results from different participants in the Rhone-Agg Experiment.

• Physics section at DWD worked very efficiently to eliminate deficiencies.