A variational cloud retrieval scheme combining radar, lidar and radiometer observations

It is widely recognised that the synergy of remote-sensing instruments should offer the most accurate retrievals of clouds, which are essential if we are to understand their important role in the radiation budget. However, despite the availability of extensive multi-instrument datasets from the ground (e.g. the US ARM and European Cloudnet sites) and shortly from space (CloudSat, Calipso and MODIS as part of the A-train), most such algorithms combine only two instruments (e.g. radar+lidar or radar+radiometer). In this talk a "unified" retrieval scheme is described that uses a variational approach to combine cloud radar, backscatter lidar and both infrared and microwave radiometers, to retrieve the properties of liquid, ice and mixed-phase clouds. The rigorous treatment of observational errors and careful use of additional constraints enables the retrieval to blend smoothly in the vertical between regions where different instruments are sensitive. For example, a deep ice cloud viewed from above by CloudSat and Calipso would typically consist of a top region detected only by the lidar, a central region where both lidar and radar can be used to infer particle size, and a base region detected only by radar. The MODIS infrared channels contain particle size information but are weighted towards the cloud top region. The unified scheme retrieves an optimum profile of cloud variables that best fits the observations. The efficiency of the method is facilitated by a new ultra-fast lidar multiple-scattering forward model. The scheme has been tested on ground-based observations from the Cloudnet project, and is shortly to be applied to global observations from the A-train of satellites.