Framework for an Earth System Reanalysis – EARTH^(RA)

点击翻译按钮获取中文摘要

The EARTH^(RA) project will develop and apply the first framework for a dynamically consistent climate reanalysis in which mathematically rigorous, adjoint-based climate data assimilation and parameter optimization will become an integral part of coupled Earth System Models (ESM). The new EARTH^(RA) framework will advance climate modeling by overcoming the long-standing problems of model biases through parameter estimation hitherto hampering progress in climate predictions, thereby enabling the development of new climate change information through improved climate predictions needed to support decision-making.The EARTH^(RA) work program will have two major elements: (1) establishment of a novel adjoint modeling and parameter estimation framework for fully coupled ESM, required to combine non-linear ESMs with in situ and satellite data of the ocean, atmosphere and cryosphere, pioneering a fully coupled Earth system reanalysis, advancing assimilation techniques to deal with non-linearities and the chaotic nature of the coupled climate system, and enabling the use of mathematically-rigorous data assimilation approaches by advancing modern automatic differentiation tools; (2) performance of a first pilot observations-constrained Earth system reanalysis through the application of control theory to fully-coupled ESMs over a year to a decade. The framework will be used in this project to (1) objectively and transparently bring full Earth System Models (ESM) into consistency with climate observations through dynamically consistent data assimilation and parameter optimization (Earth system reanalysis), (2) study adjoint climate sensitivities and feedbacks thereby improving our understanding of the climate system and its adjustment processes. The framework will firmly be based on a new adjoint earth system model whose principle can be adopted subsequently by any modeling center. Through this novel framework, observations-informed model improvement will become a firm part of model improvement in a way that is dual/complementary to classical forward model developments.