A New Method to Filter Out Radio-Frequency Interference (RFI) from SMOS Level 1C Data for Sea Ice Applications - Towards an Interdisciplinary Approach in Earth System Science - page 84

2 Data Description

The SMOS satellite observes the earth surface by taking so called snapshots. One
snapshot is acquired every 1.3 s and extends more than 1,000 km along and cross
track, so that the obtained snapshots overlap to a large extent. This overlap means
that every location is observed at several incidence angles during one single

over ight. The SMOS Level 1C data product is organized swath-wise but gridded

into the Snyder equal area grid (Sahr et al. 2003 ) with a grid spacing of 15 km. As
common in microwave remote sensing, the data are expressed as brightness tem-
peratures (TB) which is de

fl

ned as the temperature a black body would have to
have, if it emits this amount of radiation at a given frequency. The brightness
temperature is essentially the product of the emissivity, a material constant, and the
physical temperature.
Since mid 2010 SMOS is in full polarization mode, i.e., it obtains all four
components of the Stokes vector. However, it needs several snapshots to create the
whole Stokes vector for a single geographical point since the instrument switches
between the polarisations each snapshot. These data still have to be converted from
the satellite reference (X, Y) frame into the earth reference frame (horizontal, ver-
tical) or often just (H, V) (Martin-Neira et al. 2002 ). The corresponding horizontal
and vertical polarized brightness temperatures are denoted as TBh and TBv
respectively. After this transformation the data are ready to serve as, e.g., input for
retrievals of surface parameters like sea ice thickness.
One major problem with SMOS is the Radio Frequency Interference (RFI)
contamination of the data especially in the early phase of SMOS (Oliva et al. 2012 ).
It originates from anthropogenic L-band transmitters from the earth surface mostly
from radar or communication devices. Even though mostly located on land, some
RFI can also be traced back to ships. In addition, since SMOS uses a synthetic
aperture antenna with a Fourier transform based reconstruction of one snapshot,
strong RFI may extend over the whole snapshot producing ringing effects. In the
current L1C v5.05 data product RFI is flagged sparsely and ringing effects are not
addressed.

3 Filtering Procedure

For current snow- and sea ice thickness retrieval algorithms using SMOS data, a
rather simple RFI cient. Whenever a single observation in one snapshot

exceeds 300 K, the whole snapshot is discarded. This leads to a considerably large
data loss of potentially unaffected data, especially when the RFI is not strong
enough to cause ringing effects in the whole snapshot. This loss of data is not an
issue on daily averages and averages over large incidence angle ranges as used by
Huntemann et al. ( 2014 ), Kaleschke et al. ( 2012 ), Tian-Kunze et al. ( 2014 ) and
Maa

filter is suf

ß et al. ( 2013 ).