Development of a tangent linear model (version 1.0) for the High-Order Method Modeling Environment dynamical core

• Main Authors: S. Kim, B.-J. Jung, Y. Jo
• Format: Article
• Language: English
• Published: Copernicus Publications 2014-06-01
• Series: Geoscientific Model Development
• Online Access: http://www.geosci-model-dev.net/7/1175/2014/gmd-7-1175-2014.pdf

We describe development and validation of a tangent linear model for the High-Order Method Modeling Environment, the default dynamical core in the Community Atmosphere Model and the Community Earth System Model that solves a primitive hydrostatic equation using a spectral element method. A tangent linear model is primarily intended to approximate the evolution of perturbations generated by a nonlinear model, provides a computationally efficient way to calculate a nonlinear model trajectory for a short time range, and serves as an intermediate step to write and test adjoint models, as the forward model in the incremental approach to four-dimensional variational data assimilation, and as a tool for stability analysis. Each module in the tangent linear model (version 1.0) is linearized by hands-on derivations, and is validated by the Taylor–Lagrange formula. The linearity checks confirm all modules correctly developed, and the field results of the tangent linear modules converge to the difference field of two nonlinear modules as the magnitude of the initial perturbation is sequentially reduced. Also, experiments for stable integration of the tangent linear model (version 1.0) show that the linear model is also suitable with an extended time step size compared to the time step of the nonlinear model without reducing spatial resolution, or increasing further computational cost. Although the scope of the current implementation leaves room for a set of natural extensions, the results and diagnostic tools presented here should provide guidance for further development of the next generation of the tangent linear model, the corresponding adjoint model, and four-dimensional variational data assimilation, with respect to resolution changes and improvements in linearized physics and dynamics.