Object-Oriented Data Assimilation System Design and Experience

Modern data assimilation systems are complex software systems. It not only has complicated in handling various observation instruments, error characteristics but require frequent system upgrade and improvement as new observation system deployed for operations. Maintenance and upgrade usually require resources both human and computer time. The most challenging issue for modern data assimilation systems is that these systems are not well designed and implemented. This makes the maintenance and upgrade as well its quality assurance even more difficult. For instance, when one to introduce a new observation dataset, he/she must goes through almost every routines in the system for the upgrade. It is inefficient and error prone process. Object-oriented design and programming have been widely used in computer industry and other fields. In this presentation, we discuss some design issues in a variational cloud data assimilation for Gridpoint Statistical Interpolation (GSI) using ensemble covariance and share some experience of using Fortran 2008 for object-oriented programming.