U.S.-France Cooperative Research: Variational Data Assimi- lation Using Optimal Control Methods

点击翻译按钮获取中文摘要

This award will support collaborative research between Drs. I.M. Navon and James O'Brien, Florida State University, and two French researchers: Professor F.X. LeDimet, Department of Mathematics, Blaise-Pascal University, Clermont-Ferrand, and Professor Olivier Talagrand, Laboratory for Dynamic Meteoro- logy, Ecole Normale Superieure, Paris. The objective of the project is to investigate optimal control variational methods for data assimilation in meteorology and oceanography with the aim of improving accuracy and predictability of atmospheric and oceanic models. In global forecasting models, errors arise in the numerical forecast mainly due to errors in the initial conditions and deficiencies in the dynamics and physics of the models. Potential improvements in numerical weather prediction (NWP) in the near future are expected to occur as a result of either better physical parameterizations or from data improvement in the accuracy of global observing systems. The use of adjoint models as an application of the theory of optimal control for data assimilation has seen a vigorous development in the last few years. The US and French investigators in past collaboration have developed the adjoint of the NMC spectral model. In this project, they will focus on related theoretical questions in the area of optimization of algorithms, efficiency and scaling of complex problems. In addition, they will study the inclusion in the adjoint NWP of variational data assimilation of physical atmospheric processors, such as radiation and moisture ther- modynamics. This project will benefit from the complementary expertise of the investigators in optimal control theory and meteorology. The results of this research will be important for enhancing the medium-range numerical weather forecasting model's predictability and for taking advantage of various satellite data which will shortly become available.