C152 Optimization of GOSAT Atmospheric VMR Profile Retrieval : Solving a Non-Linear Inverse Problem with Empirical Orthogonal Functions

概要

論文の詳細を見る
The EOF representation proposed allows reducing significantly the number of parameters, and speeds-up Jacobian calculation, while preserving accuracy of maximum a posteriori (MAP) retrieval. Using EOF derived from clear sky conditions for CO_2 profile proved to be accurate when atmospheric particles such as clouds and aerosols are included in the forward model. For GOSAT CO_2 profile retrieval, less than 6 EOF are required to represent accurately more than 20 atmospheric layers. Computation time grows linearly with the number of layers, instead of current quadratic grow. Therefore, the cost of increasing the number of layers is smaller when considering EOF approach. The computation time of simulations was improved by a factor 2.5. Further improvement will occur for a larger number of layers and when retrieving more gas species.
社団法人日本気象学会の論文
2008-04-30