Previously,

In the sea-ice group seminars…

November 17

IntroductionRole of resolution and complexity on model

performance

December 1

Data assimilation in NEMO-LIM2

Ocean and sea-ice modelling in the Southern Ocean

December 8

Finite elements methods for sea ice

New sea-ice rheology

Today, Episodes 7 & 8:

Role of snow physics Sea-ice ecosystem modelling

Season 1 – episode 7:

On the improvement of the snow component in large-scale sea ice models

Olivier Lecomte (1), Martin Vancoppenolle (1), Thierry Fichefet (1), Hugues Goosse (1), Hubert Gallée (2).

(1) UCL TECLIM, ELIC.(2) LGGE, Laboratoire de Glaciologie et de Géophysique de

l’environnement, Grenoble.

Introduction

Importance of sea ice

• Strong reflector

• Heat insulator

• Ocean circulation

• Feedbacks

Introduction

Synopsis

• Once upon a time, a snowflake…

• Sea-ice snow cover macroscale properties

• Moral of the story (for large scale sea-ice modelling)

• Snow modelling issues in LIM1D/LIM3

• Conclusions

Metamorphism

Pictures

Schematic from Sturm and Massom 2009.

Once upon a time, a snowflake…

Snow Layering / Stratigraphy

“With respect to the sea ice snow cover, it hardly matters which class or type of snowflake falls. What is important are: (1) the amount that falls, (2) the rate of snow accumulation and (3) whether the snow falls with or without wind.” Sturm & Massom, 2009.

Þ Stratified Snow pack

Sea-ice snow cover macroscale properties

From Massom et al. 2001

+ importance of sea-ice thickness

Sea-ice snow cover heterogeneity

• Wind transport (blowing snow)• Sea-ice thickness distribution• Sea-ice dynamics

Sea-ice snow cover macroscale properties

Snow depth distribution on the ice of the Chukchi Sea, Arctic (from SHEBA campaign, Sturm et al. 2002a).

Snow Pack microscale & macroscale properties

Met

amor

phis

m

Ice Growth and Melt

ATMOSPHERE

OCEAN

Sea-ice snow cover macroscale properties

Feedbacks

Moral of the story (for large scale sea-ice modelling)

Moral :Snow on sea ice is a Holy Mess.

Including snow components in large scale sea-ice/ocean models implies choosing the most important snow processes, on the basis of :

• What is bound to affect large scale snow/sea-ice properties (evaluate mean state) on relatively long periods (capture trends)

• Models’ structure

• Models’ sensitivity to snow parameterizations

Snow modelling issues in LIM3

Snow representation in LIM3

Current state : one layer, constant physical properties.

Sensitive to :

- Snow depth (1)

- Snow radiative properties (2)

- Snow thermal conductivity (3)

(1) => Driven by snowfall and wind transport (blowing snow)(2) => Albedo, snow scattering properties

(3) => Function of density => density stratification

Snow modelling issues in LIM3

How to improve things?

- Snow depth (1)

- Snow radiative properties (2)

- Snow thermal conductivity (3)

(1) => Blowing snow parameterization(2) => Improve radiative scheme(3) => Parameterize snow thermal conductivity as a function of

density and try to better represent the snow stratigraphy.

Impact on vertical heat conduction in the snow / sea-ice pack -> 1D process

Snow modelling issues in LIM3

Multi-layer snow scheme in LIM1D

Lecomte et al., 2010

Snowfall density parameterized as a

function of wind speed (linear relationship)

-Assumption from Jordan et al., 1999

Date (mm/dd)

01/16 03/12 06/1604/30

Snow

dep

th &

Ice

thic

knes

s –[

m]

0,2

0

-0,2

-0,6

-1

-1,4

Temperature – [°C]

0

-5

-15

-25

-35

Snow modelling issues in LIM3

Validation at Point Barrow (Alaska)- Seasonal Landfast sea-ice- Comparison of model results with observed snow/sea-ice temperature profiles and thickness measurements

• Ice thickness average deviation : - 2,2 cm• Correlation between observed and simulated snow temperature profiles 27% better compared to reference run• 3 layers = required minimum

Snow modelling issues in LIM3

Sensitivity experiments on snow density

1 layer, snow density = 290 kg.m-3

10 layers, snow density = 290 kg.m-3

10 layers, snow density gradient = 450 kg.m-4 upwards, mean = 290 kg.m-3

~12% thicker sea-ice

10 layers, snow density gradient = 450 kg.m-4 downwards, mean = 290 kg.m-3

Þ ~12% thinner sea-ice

Snow modelling issues in LIM3

Sensitivity experiments on snow density 10 layers, snow density = 290 kg.m-3

Density of layers 4,5,6 = 350 kg.m-3

Þ ~8% thicker sea-ice

10 layers, snow density = 290 kg.m-3

Surface density (layers 1,2,3) = 350 kg.m-3

Þ ~10% thicker sea-ice

Model validation runSurface density (snowfall) as a function of

wind speed

Model validation runPrescribed snowfall density : 330 kg.m-3

Þ Impact on both snow depth & conductive heat fluxes

Snow modelling issues in LIM3

Ongoing work:Implementation of a multi-layer snow scheme in NEMO-LIM3 with :- Prescribed vertical density profile

- Surface layer density adjusted with respect to wind speed

- Refinement of vertical grid in thermodynamical routines

DEBUG STAGE

Programming Compiling Debugging

Crashed

Time line

Conclusions

IF (ever) you want to remember something…

• Sea-ice snow cover is an important component to account for in climate simulations

• It’s damn complicated

• Snow representation in global scale sea-ice models is so simple that improvement can be done with respect to real life snow physics

• Levers on which we can play : depth – albedo – Stratigraphy

ÞImportance of density profile, special importance of the surface layer properties.

• Ongoing work on these aspects with LIM3

What about the density of

this one, MAAAAN!!

Thanks!

Questions

Nicolaus et al., 2009

Questions

Snow in LIM1D

Questions

xxx