We explore a high-level programming model for porting numerical weather prediction (NWP) model codes to graphics processing units (GPUs). We present a Python rewrite with the domain-specific library GT4Py (GridTools for Python) of two renowned cloud microphysics schemes and the associated tangent-linear and adjoint algorithms. We find excellent portability, competitive GPU performance, robust execution on diverse computing architectures, and enhanced code maintainability and user productivity.

The ultimate objective of this model evaluation is to improve boundary layer flow representation over complex terrain. The numerical model is tested against observations retrieved during the Perdigão 2017 field campaign (moderate complex terrain). We observed that the inclusion of a forest parameterization in the numerical model significantly improves the representation of the wind field in the atmospheric boundary layer.

Fog forecasting is one of the biggest problems for numerical weather prediction. By comparing many models used for fog forecasting with others used for fog research, we hoped to help guide forecast improvements. We show some key processes that, if improved, will help improve fog forecasting, such as how water is deposited on the ground. We also showed that research models were not themselves a suitable baseline for comparison, and we discuss what future observations are required to improve them.

Weather and climate models need to represent the water cycle on land in order to provide accurate estimates of moisture and energy exchange between the land and the atmosphere. Infiltration of water into the soil is often modeled with an equation describing water transport in porous media. Here, we point out some challenges arising in the numerical solution of this equation and show the consequences for the representation of the water cycle in modern weather and climate models.