A potential Feature Mask algorithm for the EarthCARE lidar

The Earth Clouds, Aerosol and Radiation Explorer (EarthCARE) is a combined ESA/JAXA mission to be flow in 2013. EarthCARE will study the spatial (3D) distribution of clouds and aerosols and their impact on the Earth's radiative balance. To do this, the EarthCARE platform will carry a combination of active and passive sensors. The aims of the EarthCARE mission will be pursued exploiting various synergies from combining two or more of the instruments (lidar+radar, lidar+msi or lidar+radar+msi). Next to the standard synergetic algorithms (level L2b), there will be single instrument algorithms (level L2a) which either produce high level input for the L2b algorithms (e.g. the Lidar Target classification and Radar target classification can be combined to a Lidar+Radar target classification), or produce retrievals in the case of an instrument malfunction.

In this work, a potential FeatureMask algorithm for the EarthCARE high spectral resolution lidar is discussed which was developed within the ESA sponsored CASPER study. A feature mask identifies 'significant return' in the lidar signal. Significant in this sense means that the lidar return is higher than some threshold level, which in turn is based on the noise of the signal. It does not specify the nature of the feature. In order to be able to derive reliable extinction and backscatter profiles, as well as a target classification, which specifies the nature of the feature (ice cloud, liquid cloud or aerosol layer etc.); an accurate feature mask is essential. The feature mask is considered a standard output needed for the correct determination of both the L2a and L2b extinction and target mask products and L2a aerosol retrievals.

The algorithm described here is intended to find the feature mask based on the correlation of the data without relying on a number of hard coded or input dependent thresholds. As the signal strength of aerosol or very optically thin ice clouds on the single shot grid can be comparable to the noise levels it was chosen to rely on image reconstruction techniques and not on signal to noise ratios and thresholds. The main reason why an image reconstruction technique can be so effective for the EarthCARE lidar data is that in principle the Mie signals contain only particle backscatter, background noise and noise due to the Mie-Rayleigh cross-talk. It also ensures the derivation of a feature mask on the single shot resolution instead of directly going to a lower horizontal resolution of 1km. This enables both the use of variable masks, e.g. use only those profiles which are sure to have no clouds to derive the mean aerosol signals and calculation of feature fractions which can result in a better determination of higher order L2a and L2b products.

The algorithm and results for a number of different scenes including ice clouds, liquid clouds and aerosol layers will be presented.