Cloud retrievals from satellite data using optimal estimation: evaluation and application to ATSR

Abstract. Clouds play an important role in balancing the Earth's radiation budget. Hence, it is vital that cloud climatologies are produced that quantify cloud macro and micro physical parameters and the associated uncertainty. In this paper, we present an algorithm ORAC (Oxford-RAL retrieval of Aerosol and Cloud) which is based on fitting a physically consistent cloud model to satellite observations simultaneously from the visible to the mid-infrared, thereby ensuring that the resulting cloud properties provide both a good representation of the short-wave and long-wave radiative effects of the observed cloud. The advantages of the optimal estimation method are that it enables rigorous error propagation and the inclusion of all measurements and any a priori information and associated errors in a rigorous mathematical framework. The algorithm provides a measure of the consistency between retrieval representation of cloud and satellite radiances. The cloud parameters retrieved are the cloud top pressure, cloud optical depth, cloud effective radius, cloud fraction and cloud phase. The algorithm can be applied to most visible/infrared satellite instruments. In this paper, we demonstrate the applicability to the Along-Track Scanning Radiometers ATSR-2 and AATSR. Examples of applying the algorithm to ATSR-2 flight data are presented and the sensitivity of the retrievals assessed, in particular the algorithm is evaluated for a number of simulated single-layer and multi-layer conditions. The algorithm was found to perform well for single-layer cloud except when the cloud was very thin; i.e., less than 1 optical depths. For the multi-layer cloud, the algorithm was robust except when the upper ice cloud layer is less than five optical depths. In these cases the retrieved cloud top pressure and cloud effective radius become a weighted average of the 2 layers. The sum of optical depth of multi-layer cloud is retrieved well until the cloud becomes thick, greater than 50 optical depths, where the cloud begins to saturate. The cost proved a good indicator of multi-layer scenarios. Both the retrieval cost and the error need to be considered together in order to evaluate the quality of the retrieval. This algorithm in the configuration described here has been applied to both ATSR-2 and AATSR visible and infrared measurements in the context of the GRAPE (Global Retrieval and cloud Product Evaluation) project to produce a 14 yr consistent record for climate research.