|Paper title||Using cloud-resolving model simulations to assess solar and thermal radiative flux estimation for the EarthCARE BBR instrument|
|Form of presentation||Poster|
The Earth Cloud, Aerosols and Radiation Explorer (EarthCARE) is a joint mission of the European Space Agency (ESA) and the Japan Aerospace Exploration Agency (JAXA). The mission objectives are to improve the understanding of the cloud-aerosol-radiation interactions by acquiring vertical profiles of clouds and aerosols simultaneously with radiance and flux observations for their better representation in numerical atmospheric models
The operational EarthCARE L2 product on top-of-atmosphere (TOA) radiative fluxes is based on a radiance-to-flux conversion algorithm fed mainly by unfiltered broad-band radiances from the BBR instrument, and auxiliary data from EarthCARE L2 cloud products and modelled geophysical databases. The conversion algorithm models the angular distribution of the reflected solar radiation and thermal radiation emitted by the Earth-Atmosphere system, and returns flux estimates to be used for the radiative closure assessment of the Mission.
Different methods are employed for the solar and thermal BBR flux retrieval models. Models for SW radiances are created for different scene types and constructed from Clouds and the Earth’s Radiant Energy System (CERES) data using a feed-forward back-propagation artificial neural network (ANN) technique. LW models are based on correlations between BBR radiance field anisotropy and the spectral information provided by the narrow-band radiances of the imager instrument on-board. Both retrieval algorithms exploit the multi-viewing capability of the BBR (forward, nadir and backward observations of the same target) co-registering radiances and providing flux estimates for every view and checking their integrity before being combined into the optimal flux of the observed target. The reference height where the three BBR measurements are co-registered corresponds to the height where most reflection or emission takes place and depends on the spectral regime. LW observations are co-registered at the cloud top height, but the most radiatively significant height level on SW radiances is very dependent on the cloud. This reference height is instead selected by minimizing the flux differences between nadir, fore and aft fluxes.
The study presented here shows an evaluation of the BBR radiance-to-flux conversion algorithms using scenes from the Environment Canada and Climate Change’s Global Environmental Multiscale (GEM) model. The EarthCARE L2 team has simulated three EarthCARE frames (1/8 of orbit) running a radiative transfer code optimized for the EarthCARE instrument models over the GEM scenes. The test scenes resulting include synthetic L1 EarthCARE data that have been used by the different L2 teams to test and develop L2 products for testing the end-to-end-processor chaining. The test scenes collect data for a ~6200 x 150 km swath, with 1 km along-track sampling, of a simulated EarthCARE orbit. The “Halifax” scene corresponds to an orbit crossing the Atlantic Ocean and Canada in December 7, 2015. This case includes Sun just below the horizon over Greenland, cold air over Labrador, a cold-front near Halifax, dense overcast south of Halifax, and scattered shallow convection south of Bermuda. The “Baja” scene corresponds to an orbit crossing Canada and USA in April 2, 2015. This case includes clear and cold conditions at the northern extremity, scattered cloud through the Canadian Prairies, overcast over the Rocky Mountains, clear through Utah, and cirrus in Arizona and Mexico. The “Hawaii” scene corresponds to an orbit going through the Pacific Ocean and passing near the Hawaiian Islands in June 23rd 2014.
The BBR solar and thermal flux retrieval algorithms were successfully employed to retrieve radiative fluxes over the test scenes. The approach followed to evaluate the flux retrieval algorithms includes both testing the model performance with L2 products directly derived from the geophysical properties included in the GEM simulations (ideal case, no dependence on L2 retrievals analysed) and testing the model performance with L2 products derived from the EarthCARE L2 cloud and radiance processors (operational case, dependence on L2 Lidar and Imager cloud algorithms and L2 radiance unfiltering algorithm analysed). These two exercises allow evaluating discrepancies between retrieved and simulated fluxes, and assessing the sensitivity of the flux retrieval models to uncertainties in the cloud and radiance retrievals over a huge variety of realistic samples in three different scenes.