Forward operator developments - errors and biases in representativity

Forward operator developments - errors and biases in representativity
Date Published
WP-1000 Report for ESA contract 1-5576/07/NL/CB: Project QuARL - Quantitative Assessment of the Operational Value of Space-Borne Radar and Lidar Measurements of Cloud and Aerosol Profiles
Document Number
2009:WP-1000 contract 1-5576/07/NL/CB
S. Di Michele
Event Series/Collection
ESA Contract Report
Abstract A radar-and-lidar reflectivity forward model is required to enable the verification and data assimilation work planned for this project using CloudSat and CALIPSO data. At ECMWF, a radar forward operator exists, explicitily designed for assimilation purposes. It has been used for ground-based 14 and 35 GHz radar observations and it has been adapted for the CloudSat radar frequency of 94 GHz in this study. Also, a recent pre-existing radar-and-lidar forward operator, called CFMIP (Cloud Feedback Model Intercomparison Project) Observation Simulator Package (COSP), has been implemented. This operator is being used only for verification purposes since the assimilation system requires stringent specifications of computationally efficiency. The first part of this report compares the two forward models, shows their sensitivities to different microphysical and sub-grid variability assumptions and describes the changes to the ECMWF forward model for use in this project. Since CloudSat and CALIPSO observations have high vertical and horizontal resolution, but they lack of spatial coverage, it is important to address the issue of representativity errors. These errors are likely to form a large part of the total observation error and their magnitude varies for different wheather regimes. The second part of this report presents a statistical approach for computing a flow dependent estimate for the representativity error. The proposed method derives a quasi-empirical relationship between the error and a statistical measure ("score") which can be computed from satellite measurements. The robustness of this method is demonstrated for observations which have larger horizontal coverages and therefore allow a direct verification.