SEA ICE RADAR ALTIMETER SIGNATURE MODELLING EXPERIMENTS

CONTACT:RASMUS TONBOE (1)SØREN ANDESEN (1)

LEIF TOUDAL PEDERSEN (2)

(1) Danish Meteorological InstituteLyngbyvej 100

Dk-2100 Copenhagen Ørtt@dmi.dk

(2) Ørsted●Technical University of DenmarkBuilding 348

2800 Kgs. Lyngbyltp@oersted.dtu.dk

Initial snow/ice profile Meteorological input

Thermodynamicand mass model

Backscatter model

Simulated leading edge of the wave-form for two measured first-year ice profiles with thin 7cm (dotted line) and thick 36cm (solidline) snow cover. The ice thickness in the two cases is the same(1.5m). In buoyant equilibrium, the thick snow surface is 20cmabove the thin snow surface. The thin snow profile and its ½ –power time has therefore been delayed by 68x10 -11s (the time it

takes light to travel 20cm). The ½ – power time for the thin(169x10 -11s) and thick (252x10 -11s) profile is marked with thevertical dashed and solid lines respectively. The thin snow icesurface free-board is 13cm and the thick snow ice surface 4cm.The thin snow penetration depth is 10.5cm and the thick snow

profile 40.0cm.

Simulated ½ – power time [1x10 -11 s] as a functionof snow density and snow depth. The snow densityand the snow depth of the multiyear ice profile isvaried. The radar is at a fixed height above the

snow surface. The contour interval time (5x10 -11s)is equivalent to 1.5cm in free space.

The simulated backscatter coefficient [dB](dotted line), liquid water content of the uppersnow layer [%](dash-dotted line) and penetration depth [cm](dashed line) using a

modified meteorological record from the GreenIce camp in the Lincoln Sea May 2004and the multiyear ice profile in Appendix I. The air temperature [ºC] is marked by thesolid line and precipitation events on day 134, 135 and 136 of 5kg/m 2 are marked byvertical pins. The original meteorological record had persistent cold (about -10 ºC)

conditions during the entire 10 day period and both backscatter and snow/ice parametersvariability was small. In order to study the effect of different meteorological conditionsthe three precipitation events were added and 10 ºC and 20w/m2 were added to the air

temperature and incoming long-wave radiation respectively after day 137.

(continued) The range + ½ – power time [1x10 -11s] (solid line) and the upper snow layerdensity [kg/m3] (dotted line) is simulated using a modified meteorological record from

the GreenIce camp in the Lincoln Sea May 2004 and the multiyear ice profile

BACKSCATTER MODELLING PRINCIPLES AND TERMSAbstract: Scattering at the large horizontal dielectric contrasts of the snow/sea ice system i.e. the snow and ice surface dominate the altimeter total backscatter coefficient. To investigate the temporal and seasonal altimeter backscatter signature variability a forward model is coupled to a thermodynamic and mass model for the snow and sea-ice temperature profile, accumulation, growth, melt and metamorphosis. The thermodynamic and mass model uses an initial snow and ice profile and meteorological data input. In situ data collected during the 2003 CRYOVEX joint Danish-German-ESA campaign and the 2004 GreenIce ice camp are used. Our results show that the snow cover is indeed important for the microwave signature of sea ice as well as its buoyant balance.

BACKSCATTER MODELLING RESULTS

First-year ice in Fram Strait 2003 Multiyear ice in Lincoln Sea 2004

CONCLUSIONS1) Different snow cover i.e. snow depth, layering and density has a significant impact on theradar altimeter leading edge.

2) In particular the snow depth increase the ½ – power time. The snow depth suppress the icefree-board and elevate the snow surface. The simulated buoyant balance and the simulated radarice elevation measurement counter-balance to some extent (extension of the radar range andelevation of the snow free board). The simulated measurement of ice thickness is accuratewithin +/- 1m under the different snow conditions.

3) Backscatter parameters are fairly stable during cold conditions (air temperature < 0 ºC).However melt and liquid water in the snow change backscatter parameters dramatically. In ourexample the range + ½ - power time is reduced by 100x10 -11s at the onset of melt, equivalent to30cm apparent elevation difference (increase).

Indeed these model experiments indicates that the snow cover and the upper ice characteristicssignificantly influence the backscatter signature. In order to proceed making sea ice free-boardand ice thickness measurements with radar altimeter a description of the snow cover isnecessary i.e. snow thickness, density, water content, grain-size. However, reliable hemisphericsea ice snow mapping algorithms using satellite data does not exist today. Future effort toachieve this is urgent if the prophesy of Rothrock should not become reality: ‘... to estimate theice surface ht [free-board] and then multiply by ... 10 to obtain thickness h introducesunsatisfactory errors.’ [Rothrock, 1986, p. 563].

BACKSCATTER SIGNATURE RESPONSE TO METEOROLOGY AND STATE OF THE SNOW/ICE PROFILE USING A COUPLED BACKSCATTER AND THERMODYNAMIC MASS MODEL

The snow load vs. a simulation of measured ice thicknessusing ice ½ – power + range time and a water reference

for different snow loads (product of snow density andthickness i.e. the snow water equivalent). In thesimulation it is assumed that the ice free board is

measured by the ½ – power time + snow surface range tothe radar, this range measurement is compared to a

'perfect' (constant) range measurement over water. Thefree board is then multiplied by 10 to obtain the

simulated ice thickness measurement.

Snow and ice profiles collected at the 2003 CRYOVEX joint Danish-German-ESA campaign (two first-year ice profiles from Fram Strait) and the 2004 GreenIce ice camp (typical multiyear ice profile from the Lincoln Sea) are used as examples to illustrate the significance of the snow cover for the altimeter backscatter signature.

The permittivity of two first-year ice profiles used in the simulation of the backscatter leading edge (right).

No specific correction is applied for antenna gain or pulse modulation. Themodel is comparable in principle to Ridley & Partington [1988]. Surfacescattering is here synonymous with scattering at horizontal interfaces betweenlayers. Volume scattering is synonymous with scattering from e.g. snow grainswithin the layers. Surface scattering is computed using the sea ice forwardmodel by Ulander & Carlström [1991] where the backscatter coefficient is afunction of the fractional area of flat patches (about 3% for sea ice, [Fetterer etal., 1992]) and the Fresnel reflection coefficient.

The range measured by the altimeter is the sum of 'range' and '½ - power' time. The '1/2-power elevation' is not a physical interface in the snow/ice profile, but rather a function of snow depth and other snow properties.

REFERENCESFetterer, F. M., M. R. Drinkwater, K. C. Jezek, S. W. C. Laxon, R. G. Onstott, & L. M. H. Ulander, In: F. D.

Carsey, Ed.,. Microwave Remote Sensing of Sea Ice, Geophysical Monograph 68 (pp. 111-135). Washington DC: American Geophysical Union, 1992.

Ridley, J. K. & K. C. Partington, Int. J. Rem.Sens.9(4), 601-624, 1988.Rothrock, D. A., In: N. Untersteiner (Ed.) The Geophysics of Sea Ice (pp. 551-575). NATO ASI series, Series

B: Physics Vol. 146. Plenum Press, 1986.Ulander, L. M. H., & A. Carlström, Proc. IGARSS'91, 1215-1218, 1991.

The meteorological record from the GreenIce camp in the Lincoln Sea 2004 is used in the coupled thermodynamic mass and backscatter model. The thermodynamic mass model ensures a realistic description of the snow/ ice profile under different meteorological forcing. The initial snow and multiyear ice profile consist of 25cm snow and 350cm ice.

Ice and Remote Sensing Division ● Danish Meteorological Institute Ørsted ● Technical University of Denmark

Polar bear weight

Free-board weight

Ice buoyancy

output output

density

thickness

Sea ice conditions even within a CRYOSAT resolution cell are often diverse. Large myltiyear ice floes have undulating topography with refrozen melt-ponds and hummocks caused by differential melt in summer, uneven snow cover distribution and pressure ridges. These features all add the backscattered signal and complicates its analysis. However, the melt-pond altimeter backscatter is much higher than backscatter from ridges and hummocks. Therefore, if the model describes backscatter from the melt-ponds adequately it is possible to explain the most important backscatter processes. Our modelling experiments can be seen as an attempt to model multiyear ice melt-pond altimeter backscatter. The snow cover plays a vital role for the melt-pond signature (photo: National Ice Center).