EC-09. Model development for E3SM at high resolutions

Abstract
Inadequate model resolution in vertical and horizontal directions is a major cause of the underestimation of low-level shallow cloud amount in global circulation models, a well-known problem in climate studies. We have found that remarkable improvements in the representation of shallow clouds in U.S. DOE’s Energy Exascale Earth System Model (E3SM) can be achieved by using eight-times or higher vertical resolutions in the boundary layer. To achieve an efficient vertical resolution enhancement, an adaptive vertical grid method, combined with the Framework for Improvement by Vertical Enhancement (FIVE), is under development for E3SM. With this method, only the most relevant processes are computed at high vertical resolution while the rest of the model runs at the standard resolution. To further save computational costs, high-resolution vertical grids will only be turned on when and where necessary. In the first part of this presentation, we will show some results from the application of neural networks for decision making on the meteorological conditions that are likely to benefit most from high-resolution vertical grids with respect to low cloud formation. The E3SM will be adopting a 3-km horizontal resolution in the near future. Testing the adaptive-FIVE and re-tuning model parameters will be necessary before the climate simulation. However, these tasks are computationally too expensive to accomplish in full three-dimensional simulations given current computer performance. We propose a vertical-meridional two-dimensional pole-to-pole simulation framework to address this issue. By neglecting the zonal direction, a large number of test simulations at high resolution become easily affordable. In the second part of this presentation, we will introduce this two-dimensional modeling framework and demonstrate its usefulness.