3D imaging of snow microstructure for a better knowledge of snow properties and of their time evolution: 2 examples
62A062+62A101
F. Flin
1, N. Calonne
1, 2, C. Geindreau
2, B. Lesaffre
1, X. Wang
1, 3,
C. Carmagnola
1, S. Rolland du Roscoat
2, P. Charrier
2, S. Morin
1, D. Coeurjolly
31
CEN, CNRM - GAME URA 1357, Meteo-France - CNRS, Grenoble, France
2
3S-R UMR 5521, CNRS - Universite Joseph Fourier - Grenoble INP, Grenoble, France
3
LIRIS UMR 5205, CNRS, Lyon, France
3D imaging of snow microstructure for a better knowledge of snow properties and of their time evolution: 2 examples
Computation of snow permeability from 3D images
Calonne et al. 62A101
Evolution of snow properties under temperature gradient conditions
Flin et al. 62A062
EXAMPLE 1
3D Image-based Numerical Computations of the Snow Permeability: Links to Specific Surface Area, Density, and
Microstructural Anisotropy (62A101)
N. Calonne
1, 2, C. Geindreau
2, F. Flin
1, S. Morin
1, B. Lesaffre
1, S. Rolland du Roscoat
2, and P. Charrier
21
CEN, CNRM - GAME URA 1357, Meteo-France - CNRS, Grenoble, France
2
3S-R UMR 5521, CNRS - Universite Joseph Fourier - Grenoble INP, Grenoble, France
where K
*is the normalized (dimensionless) permeability tensor
• Darcy’s law
• One of the main transport properties of snow
• Key variable for several applications (atmospheric-firn chemistry, snow metamorphism, …)
• Related to a characteristic length l (in m), often used for normalization:
p
q K
Intrinsic Permeability Tensor of Snow
Introduction Methods Results & Discussion Conclusions
- K : intrinsic permeability tensor (m
2) - q : discharge per unit area (m s
-1)
- μ : dynamic viscosity of air (kg m
-1s
-1) - : pressure gradient (Pa m p
-1)
l
2K
* K
• Strongly depends on snow density ρ
snow K
*versus ρ
snow l : estimate of the snow grain size
Problems
• Co-existence of models (experimental fits, analytical laws, numerical computations, …) which provide different relationships of K
*vs. ρ
snow.• Challenging measurements: large scatter in K
*vs. ρ
snowobserved in experimental studies Litterature data may differ from several orders of magnitude for the same density
• No or few studies about the anisotropy of K.
Jordan et al., 1999
Shimizu
Carman - Kozeny
Zermatten et al., 2011
Introduction Methods Results & Discussion Conclusions
Issues and Contributions
Our study: Numerical computations of the permeability tensor K from 3D snow microstructure provided by X-ray tomography.
Objectives
1. Study of the relationship between K, l and ρ
snow2. Understand the origin of the scatter observed in litterature:
Comparisons to models
Comparisons to experimental data 3. Study of the anisotropy of K
Introduction Methods Results & Discussion Conclusions
7
Examples of Snow Samples (out of 35)
Precipitation Particles
Decomposing and
Fragmented precipitation particles
Rounded
Grains Depth
Hoar
1 mm 1 mm
1 mm
Melt Forms
1 mm 1 mm
Introduction Methods Results & Discussion Conclusions
L
Snow layer
Macroscopic scale Microscopic scale
α a
a
a
p
q K
pore) ( a
Condition
Separation of scales L >> l
l
Numerical Computation - Homogenisation
Auriault et al. 2009
Periodic homogenisation
K (micros., K
air)
p
q K
Introduction Methods Results & Discussion Conclusions
zz zy
zx
yz yy
yx
xz xy
xx
K K
K
K K
K
K K
K
K
Solving a specific boundary problem arising from the homogenisation process, on a 3D image, by Geodict software.
L
Snow layer
Macroscopic scale Microscopic scale
α a
a
a
p
q K
air) ( a
l
Auriault et al. 2009
Periodic homogenisation
K (micros., K
air)
p
q K
REV
Numerical Computation - Homogenisation
Introduction Methods Results & Discussion Conclusions
Choice of the characteristic length equivalent sphere radius r
es(m)
• directly related to the specific surface area of snow SSA (m
2kg
-1)
• can be measured experimentally (DUFISSS, ASSSAP,…) and numerically from 3D images
• in the followings, we define (dimensionless)
K
*vs. Density – Characteristic Length
2
r
esK
* K
ice
r
es
SSA
3 where ρ
iceis the ice density (kg m
-3)
Introduction Methods Results & Discussion Conclusions
K
*vs. Density - General Evolution
Dimensionless permeability decreases with increasing density
Vertical component
Horizontal component
Introduction Methods Results & Discussion Conclusions
K
*vs. Density - Anisotropy
Anisotropy of K
*= vertical component horizontal component
Anisotropy range:
0.74 – 1.66
Vertical component
Horizontal component
Introduction Methods Results & Discussion Conclusions
K
*vs. Density - Anisotropy
The coefficient of anisotropy is linked to the snow
microstructure
Vertical component
Horizontal component
Introduction Methods Results & Discussion Conclusions
Example of vertically-oriented sample: Depth Hoar
p
Air flow velocity
p
K
*vs. Density - Regression Curve
Exponential regression on 35 mean values of K
*:
) 013 . 0 exp(
94 .
2
2
sr
esK
Strong relationship between K, SSA and ρ
Introduction Methods Results & Discussion Conclusions
R² = 0.985
K
*vs. Density - Comparison to Models
Introduction Methods Results & Discussion Conclusions
• good agreement between models and computations, in overall
• Shimizu (experimental
fit, 1970): due to the
definition of the grain
size estimates?
K
*vs. Density - Comparison to Experimental Data
• Our computations are consistent with values from others studies
• experimental values are more scattered:
due to the difficulty of measurements?
Introduction Methods Results & Discussion Conclusions
• Permeability is strongly correlated with density and SSA (r
es)
l = r
escan be good choice
• Anisotropy of K ranges between 0.74 - 1.66 and is linked to snow microstructure.
• In overall, our computations are consistent with models and datasets from litterature
• An average value of K can be reasonnably inferred from SSA and density, using the proposed regression equation:
Conclusion
Introduction Methods Results & Discussion Conclusions
) 013 . 0 exp(
94 .
2
2
sr
esK
Discussion paper:
Calonne et al., TCD, 2012
EXAMPLE 2
3D Characterization of Snow Evolution
under Temperature Gradient Conditions (62A062)
F. Flin
1, N. Calonne
1, B. Lesaffre
1, X. Wang
1, 3, C. Carmagnola
1,
S. Rolland du Roscoat
2, P. Charrier
2, S. Morin
1, D. Coeurjolly
3, C. Geindreau
21
Meteo-France - CNRS, CNRM-GAME URA 1357, CEN, Grenoble, France
23S-R CNRS UMR 5521, Universite Joseph Fourier - Grenoble INP, Grenoble, France
3
LIRIS UMR 5205, CNRS, Lyon, France
t = 0 h t = 313 h t = 500 h
• Common metamorphism regime in natural conditions
• Complicated phenomena involving various physical processes (heat and mass transfer, facetting, …)
• Complicated impact on snow mechanical and physical properties
Snow Metamorphism under Temperature Gradient (TG)
Atmosphere, T
1
zT
Ground, T
2Introduction Methods Results & Discussion Conclusions
vapor
growth direction
Yosida et al., 1955
Colbeck 1983
• grain growth
• neck evolution
• facetting effects
• evolution of physical properties
• anisotropic behavior
• Relationship between physical properties (K, k
eff, …) and snow microstructure poorly understood (anisotropic effects)
Problems
• How quantifying the snow evolution?
• Little complete 3D datasets to fully quantify the snow evolution under TG metamorphism
Introduction Methods Results & Discussion Conclusions
Issues and Contributions
Our study: 3D characterization of the snow evolution submitted to TG from tomographic images.
Objectives
1. Obtain precise measurements to characterize the
morphological evolutions occurring under TG metamorphism 2. Study the relationship between microstructure and physical
parameters (anisotropy)
Introduction Methods Results & Discussion Conclusions
Cold-room Experiment
Insulating polystyrene plate Copper plate
Snow slab obtained by fresh snow sieving
Thermal regulation of the copper plates by a fluid
circulation
T = - 1 °C T = - 7 °C
|| T|| = 43 K m-1 1 m
14 cm
Introduction Methods Results & Discussion Conclusions
Temperature conditions during 3 weeks
Density = 300 kg m
-31.6 cm 5.35 cm
1. Sampling 4. Machining
3. Freezing 2. Impregnation
Numerical Computation - Acquisition of 3D Images
5. X-ray tomography 6. Image processing
1 mm
- Image size: 6 - 10 mm - Voxel size: 7 - 9 µm
Introduction Methods Results & Discussion Conclusions
Gaussian and Mean Curvatures, Related Histograms
Gaussian curvature:
1 2
1 1
.
G R R [L
-2
]
1 2
1 1 1
C 2
R R
[L
-1]
Mean curvature:
convex 0 concave
Introduction Methods Results & Discussion Conclusions
• Idea: to measure the degree of convexity of the ice-air interface
• Methods:
Flin et al, Ann. Glaciol., 2004 Brzoska et al, PCI, 2007
Curvature histograms:
Count the amount of values having the same curvature on the whole
surface
Microstructural Anisotropy and Related Histograms
• Idea: to measure the anisotropy of the normal field of the ice-air interface
• Method:
• compute outward unit normal vectors n (Flin et al., 2005)
• for each normal vector, compute angles between n and the x, y, z axes
• compute the proportion of normals corresponding to a particular angle
• draw the resulting histogram in a polar plot
• estimate the ratio h_max_z / h_max_horizontal
ice
α x
Introduction Methods Results & Discussion Conclusions
x z
n
Computation of Physical Properties - Homogenisation
zz zy
zx
yz yy
yx
xz xy
xx
k k
k
k k
k
k k
k
k
effSolving a specific boundary problem arising from the homogenisation process, on a 3D image, by Geodict software.
L
Snow layer
Macroscopic scale Microscopic scale
T
α
αα
k
F
T
k
effF
ice) air, ( α
l
Auriault et al. 2009
Periodic homogenisation
k
eff(micros., k
air, k
ice)
Example for the effective thermal conductivity k
effIntroduction Methods Results & Discussion Conclusions
REV
Specific Surface Area (SSA)
• SSA decreases during metamorphism
• a fast SSA decrease corresponds to a high TG
Introduction Methods Results & Discussion Conclusions
10 12 14 16 18 20 22 24 26 28 30
0 100 200 300 400 500 600 700 800 900 1000 1100 1200
Time (hours) S S A ( m
2k g
-1)
This study Flin et al., 2004
Schneebeli et al., 2004
Isothermal conditions
~ -2°C
100 K m
-1~ -7°C
43 K m
-1~ -4°C
Histograms of Mean Curvature (C)
Introduction Methods Results & Discussion Conclusions
After 0 hours
~2 mm
Histograms of Mean Curvature (C)
Introduction Methods Results & Discussion Conclusions
After 73 hours
~2 mm
Histograms of Mean Curvature (C)
Introduction Methods Results & Discussion Conclusions
After 144 hours
~2 mm
Histograms of Mean Curvature (C)
Introduction Methods Results & Discussion Conclusions
After 217 hours
~2 mm
Histograms of Mean Curvature (C)
Introduction Methods Results & Discussion Conclusions
After 313 hours
~2 mm
Histograms of Mean Curvature (C)
Introduction Methods Results & Discussion Conclusions
After 409 hours
~2 mm
Histograms of Mean Curvature (C)
Introduction Methods Results & Discussion Conclusions
After 500 hours
~2 mm
Histograms of Mean Curvature (C)
Conclusions:
+ the peak is moving to 0 grains are growing + the peak is narrowing grains are facetting
+ the negative part of the peak increases concave shapes are created
Introduction Methods Results & Discussion Conclusions
After 500 hours
~2 mm
Histograms of Gaussian Curvature (G)
Introduction Methods Results & Discussion Conclusions
After 0 hours
~2 mm
Histograms of Gaussian Curvature (G)
Introduction Methods Results & Discussion Conclusions
After 73 hours
~2 mm
Histograms of Gaussian Curvature (G)
Introduction Methods Results & Discussion Conclusions
After 144 hours
~2 mm
Histograms of Gaussian Curvature (G)
Introduction Methods Results & Discussion Conclusions
After 217 hours
~2 mm
Histograms of Gaussian Curvature (G)
Introduction Methods Results & Discussion Conclusions
After 313 hours
~2 mm
Histograms of Gaussian Curvature (G)
Introduction Methods Results & Discussion Conclusions
After 409 hours
~2 mm
Histograms of Gaussian Curvature (G)
Introduction Methods Results & Discussion Conclusions
After 500 hours
~2 mm
Histograms of Gaussian Curvature (G)
Introduction Methods Results & Discussion Conclusions
After 500 hours Conclusions:
+ the 0-centered peak is increasing and narrowing grains are facetting + the proportion of convex shapes is decreasing grains are growing + the proportion of concave shapes is decreasing necks are growing
~2 mm
Intrinsic Permeability from Numerical Computations
• Permeability increases with time
• Larger increase in z direction
Introduction Methods Results & Discussion Conclusions
Effective Thermal Conductivity from Num. Computations
Introduction Methods Results & Discussion Conclusions
• Thermal conductivity increases with time
• Larger increase in z direction Density = 300 kg m
-3In overall
Effective Thermal Conductivity from Num. Computations
Introduction Methods Results & Discussion Conclusions
• Thermal conductivity decreases with time
• Larger decrease in horizontal direction Density = 300 kg m
-3At the beginning
Stage 1 Stage 2
Effective Thermal Conductivity from Num. Computations
Introduction Methods Results & Discussion Conclusions
• Thermal conductivity decreases with time
• Larger decrease in horizontal direction Density = 300 kg m
-3At the beginning
Schneebeli et al. 2004
Stage 1 Stage 2
Anisotropy
Relationship between Microstructure and Properties
• All anisotropic coefficients increase with time
highlights the link between microstructural and physical properties
Introduction Methods Results & Discussion Conclusions
Conclusions
• We presented some numerical tools to quantitatively monitor the geometrical and physical properties of snow under TG metamorphism
• Observations are qualitatively consistent with previous experiments described in the litterature
• Anisotropy of K and k
effare linked to snow microstructural anisotropy Perspectives
• Use this dataset as guidelines or validation tools for microstructural metamorphism models
Conclusion
Introduction Methods Results & Discussion Conclusions
Acknowledgements:
H. Arakawa, F. Domine for datasets and discussions ESRF ID19 beamline, 3SR lab for tomographic acquisitions
French National Research Agency (DigitalSnow ANR-11-BS02-009) for funding
General Conclusion
Both approaches are particularly helpful for better snowpack modeling One property for many snow samples:
Obtain robust relationships between complex physical
properties (K, keff) and more basic properties (density, SSA)
Many properties for a particular sample undergoing
metamorphism: better
understanding of the physical
processed involved
3D Snow Imaging by X-ray Microtomography
1200 pixels
12 00 p ix el s
Energy 18-20 keV
1 pixel
=
4.92 to 10 microns
10 mm Room temperature experiments:
A cold cell is needed
REV
Precipitation Particles Facetted Crystals
REV
• Smallest volume from which a variable representative of the whole can be estimated
• Computation of K on volume at least equal to the REV
• REV estimated by calculating K on increasing volumes
Representative Elementary Volume (REV)
Curvature of a line in a 2D space
R R’
Osculating circles
C = 1/R
Gaussian curvature :
1 2
1 1
.
G R R [L
-2
]
1 2