Introducing a Deep-Atmosphere Version of E3SM’s Dynamical Core to Reduce Climate Model Biases
Recent work has created controversy around the validity of a pair of simplifying assumptions that are often made in representing (a) the strength of the gravitational force and (b) the form of the Coriolis force in the dynamical cores of weather and climate models. They are called the Shallow-Atmosphere Approximation (SA) and Traditional Approximation (TA) which together form the basis of the dynamical core design of the Department of Energy’s (DOE) Energy Exascale Earth System Model version 2 (E3SMv2). E3SMv2’s shallow-atmosphere (SA & TA) dynamical core is called the “High-Order Methods Modeling Environment” (HOMME) which is accompanied by a new software design in DOE’s latest “Simple Cloud-Resolving E3SM Atmosphere Model” (SCREAM). The HOMME dynamical core supports non-hydrostatic and hydrostatic configurations as well as Fortran-based (HOMME) and C++/Kokkos-based software architectures for SCREAM. Recent increases in computational capacity have now allowed high-resolution (25 km) and even cloud-permitting (3 km) assessments of the climate system with the E3SM/HOMME and SCREAM configurations. At these high resolutions the validity of the shallow-atmosphere configuration is especially questionable, as diabatic effects become better represented. In fact, there are indications that the neglect of the cosine-based Coriolis accelerations in the momentum budget can cause biases on the order of 10% for diabatically driven tropical and planetary-scale flows. In addition, the neglect of the gravitational variations needs to be questioned as models raise the position of the model lid (60-65 km for E3SM). These biases are systematic, and while they might be less relevant for short weather simulations, they accumulate over time in long-term climate simulations. It is therefore hypothesized that long-standing “stubborn” climate model biases like the position and strength of the Intertropical Convergence Zone (ITCZ) in E3SM are connected to E3SM’s shallow-atmosphere design, and a deep-atmosphere extension of HOMME is proposed to test this hypothesis. The expectation is that the deep-atmosphere configuration of E3SM has the potential to reduce climate model biases, especially in the tropics.
The research objectives are:
- A deep-atmosphere equation set will be designed and implemented for HOMME. Close attention will be paid to HOMME’s Hamiltonian design and conservation properties.
- A hierarchy of HOMME/E3SM configurations and test cases of increasing complexity will be used to analyze the biases caused by HOMME’s shallow-atmosphere and traditional approximations. The fidelity of the HOMME deep-atmosphere variant will be evaluated.
- Various deep-atmosphere HOMME/E3SMv2 configurations with prescribed sea surface temperatures and sea-ice forcing (representing an aqua-planet and so-called AMIP mode) will be compared to their SA + TA counterparts. This will remove systematic sources of error from the full-complexity HOMME-E3SM configuration and will help reveal whether the E3SM ITCZ and other tropical biases can be, at least partially, attributed to the TA and SA approximations.
Deep-atmosphere modeling is an emerging frontier for the dynamical core community. This project is innovative, computationally advanced, and will benefit the long-term development of E3SM with HOMME/SCREAM. It will provide a paramount step for future ultra-high cloud-permitting climate model assessments at DOE. The project is built upon a close collaboration between the University of Michigan and DOE’s key dynamical core developers at Sandia National Laboratories. The deep-atmosphere extensions of the HOMME prototype will be transferable to other code bases and will thereby provide a benefit for future versions of both E3SM and SCREAM.