Accounting for the layering of snow and �rn On the link between density and grain size variability

M.W. Hörhold*, S. Linow**, J. Freitag*** IUP, Bremen University, ** Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research

measurements parametrization (Linow et al., 2012)

parametrization (this study)

result

density residual density

residual grain size

annual meantemperature,accumulation rate, ∆T

temperatureand grain size pro�le

meangrain size

absolute grain size

B36 - density

20 40 60 80 100depth (m)

dens

ity (g

/cm

B36 - grain size

0 5 10depth (m)

grai

n ra

dius

(mm

)

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

meanvariabilityrandomAMSR-E ascending / SSM/IAMSR-E descending

220

200

180

160

140

120

100

220

200

180

160

140

120

100

0 2 4 6 8 10 12depth (m)

resi

dual

spe

ci�c

sur

face

(1/c

m)

0

5

10

15

-5

-10

resi

dual

spe

ci�c

sur

face

(1/c

m)

0

5

10

15

-5

-10

-15

-20-0.2 -0.1 0.0 0.1

residual density (g/cm3)

Figure 3: Measured (red dots, CT) and modelled (blue line, from high-resolution density data) residual speci�c surface over depth

frequency (GHz)bi

as (K

)

meanrandomvariability

6 10 18 19 36 37

frequency (GHz)

6 10 18 19 36 37

■ strong linear relationship between residual (detrended) speci�c surface and residual density of the analyzed �rn cores

Parametrization of grain size variability

■ density variability in polar �rn is connected to grain size variability via the snow metamorphism process

■ we use density and grain size from CT measurements from 5 antarctic sites to parametrize grain size variability as a function of density variability

■ this allows us to reconstruct grain size using

1. modeled mean grain size based on annual mean temperature and accumulation rate 2. grain size variability derived from density measurements

Introduction

■ microwave (MW) interaction with dry polar �rn is in�uenced by the variability of �rn density and grain size

■ due to the integration of MW measurements over firn depths of several meters, the effect of layering can be significant

■ in the retrieval of geophysical parameters from MW data in the polar regions, the variability due to layering is o�en unconsidered or treated as a stochastic process

■ in this study, we examine the connection between density and grain size variability to improve the representation of firn layering and examine the impact on the modeled MW signal

Sensitivity of the MW signal to variability

■ we examine the influence of microstructure variability on the MW signal under di�erent assumptions:

1. mean pro�le 2. mean pro�le + random noise 3. parametrization of variability

■ we use microwave data from AMSR-E and SSM/I to analyse the in�uence of layering on the microwave signal

■ models: MEMLS and DMRT/ML

■ bias = [Σ(TB,modeled -TB,satellite)2 /n]0.5 depends on the profile type, for DMRT-ML also on frequency

The B36 test site

■ B36 is a �rn core drilled to ~80 m depth at Kohnen Station, Antarctica

■ the mean annual temperature is -44.6˚C, the accumulation rate is 0.065 m w.e./year

Results (B36)

■ MEMLS:

Conclusions

■ we can show that grain size variability is coupled to density variability, and are able to reconstruct grain size from measured density pro�les

■ first sensitivity studies with MW models show improved results when a realistic variability is used■ future studies will extend this analysis

■ DMRT-ML:

■ reconstructed specific surface variability based on density measurements

Figure 1: Measured grain size (Computer Tomography) and measured high-resolution density (Gamma-absorption)

Figure 2: Measured residual speci�c surface vs. measured residual density (Computer Tomography) of 5 antarctic sites with regression line, correlation coe�cient = 0.886

Figure 5: Measured brightness temperatures of di�erent channels compared to model runs (MEMLS)

Figure 6: Measured brightness temperatures of di�erent channels compared to model runs (DMRT-ML)

Figure 7: Comparison of the bias between modelled and measured brightness temperatures for di�erent �rnpro�le types