Improving the effective temperature estimation over sea ice using low frequency microwave radiometer data and Arctic buoys
In the frame of the OSI SAF Visiting Scientist Program, Lise Kilic from the "Observatoire de Paris" (LERMA) has worked at the Danish Meteorological Institute (DMI) on Improving the effective temperature estimation over sea ice using low frequency microwave radiometer data and Arctic buoys.
This work took place in 2017 and was supervised by Rasmus Tonboe (DMI).
Objectives and framework of the study :
Low microwave frequencies are less sensitive to the water vapor in the atmosphere than infrared frequencies. Therefore microwave observations are possible under cloudy conditions. Low microwave frequencies have also an higher penetration depth so they can give information from the ice layer below the snow layer.
In this study, we use observations of a passive microwave satellite instrument: the Advanced Microwave Scanning Radiometer (AMSR-2) in collocation with in-situ measurements from Arctic buoys: the Ice Mass Balance (IMB) buoys.
The objective was to improve the estimation of the sea ice effective temperature (Teff) which is part of the sea ice emissivity (OSI-404) product and which could potentially be used for improving the physical correction of sea ice concentration (OSI-401-b). The Teff is strongly correlated with the snow-ice interface temperature. Empirical relationships (regression models) for the snow-ice interface temperature and the snow depth as a function of the AMSR brightness temperatures were derived to estimate the Teff.
The Teff is the integrated emitting layer temperature of the snow and ice. Teff is related to the upper profile physical snow and ice temperature by the snow and ice microwave emission and scattering processes [Tonboe et al., 2011]. These relationships between Teff and the physical temperature profile are complicated especially at higher microwave frequencies (>10 GHz) where scattering from snow grains and voids in the sea ice plays an important role. Also layering and vertical structure in the snowpack is affecting the microwave emission processes and especially the polarisation difference of the brightness temperature (Tb) at horizontal polarisation [Tonboe, 2010]. Because of these complicated relationships between the physical temperature, Teff and the brightness temperature, an empirical algorithm based on regression analysis was pursued using the extensive sea ice buoy and in situ database developed in the ESA sea ice CCI project (http://www.seaice.dk/ecv2/rrdb-v1.1/).
At 6 GHz there is a very high linear correlation between the snow-ice interface temperature and the Tb and Teff [Tonboe et al., 2011]. This is because the microwave penetration though the snow layer and because the emission is relatively unaffected by scattering from snow grains. Higher frequencies have more shallow penetration than at 6 GHz due to increasing scattering and absorption as a function of frequency and preliminary results show that it is possible to derive the vertical temperature profile using a suite of frequencies and simple regression models [Grönfeldt, 2015]. Even though these relationships get more noisy at higher microwave frequencies (>10 GHz) there is still a very high correlation between Teff at microwave channels between 6 and 50 GHz [Tonboe et al., 2011]. This is exploited in the current operational set-up for deriving Teff in OSI-404-a (yet unpublished but updated ATBD).
In the final report (below), a description of the dataset is given in section 2, the method used to estimate the snow-ice interface temperature, the snow depth and Teff is presented in section 3, and the results are shown and discussed in section 4.
Report conclusion : Equations for the snow-ice interface temperature (TSI), the Teff and the snow depth in terms of brightness temperatures were estimated using RRDP data. The TSI and the snow depth are needed to estimate the Teff. Multi-linear regressions have been used. The parametric equation to estimate the TSI gives an error of 2.11 K with the 10.65 GHz channel and an error 2.22 K with the 6.9 GHz channel. The multi-linear model to estimate the snow depth from 6.9, 18.7 and 36.5 GHz channels, gives an error of 5 cm. Comparisons between the observations and the simulations from our equations to retrieve the TSI and the snow depth are shown in the figure at the top of this page. The linear regression equations to estimate the Teff at different microwave frequencies as function of the TSI are presented in the report. This work gives the keys to derive the Teff from the microwave TBs. These new Teff estimations can be used for correcting brightness temperature over sea ice in order to compute sea ice concentration.
Benefits for the SAF : The effective temperatures will be used for regional noise reduction in the OSI SAF sea ice concentration products. Further, the effective temperature is part of the OSI-404 emissivity product. During CDOP 3 (2017-2022), the emissivity product will be extended to cover other (MWI) frequencies. During the effective temperature algorithm development, Lise also developed an algorithm for snow thickness on sea ice and for the snow – ice interface physical temperature.