Search

Global reanalyzes like ERA5 accurately capture atmospheric processes at spatial scales of O(10) $\mathcal{O}(10)$ km or larger. By downscaling ERA5 with large‐eddy simulation (LES), LES can provide details about processes at spatio‐temporal scales down to meters and seconds. Here, we present an open‐source Python package named the "Large‐eddy simulation and Single‐column model—Large‐Scale Dynamics," or (LS)2D in short, designed to simplify the downscaling of ERA5 with doubly‐periodic LES. A validation with observations, for several sensitivity experiments consisting of month‐long LESs over Cabauw (the Netherlands), demonstrates both its usefulness and limitations. The day‐to‐day variability in the weather is well captured by (LS)2D and LES, but the setup under‐performs in conditions with broken or near overcast clouds. As a novel application of this modeling system, we used (LS)2D to study surface solar irradiance variability, as this quantity directly links land‐surface processes, turbulent transport, and clouds, to radiation. At a horizontal resolution of 25 m, the setup reproduces satisfactorily the solar irradiance variability down to a timescale of seconds. This demonstrates that the coupled LES‐ERA5 setup is a useful tool that can provide details on the physics of turbulence and clouds, but can only improve on its host reanalysis when applied to meteorological suitable conditions. Plain Language Summary: Modern global weather models are accurate in predicting atmospheric processes at scales of around 10 km or larger, but are less good at predicting smaller scale processes, like for example, the interaction between solar radiation, individual clouds, and the resulting clouds shadows that are cast onto the land surface. High spatio‐temporal resolution research models are able to capture these smaller scale processes, but require a coupling to a weather model to account for the day‐to‐day variability in our weather. In this paper, we present a framework to couple large to small scale models, and demonstrate both the benefits and challenges of using this coupled model setup. The coupled setup excels in capturing the aforementioned high frequency interactions between small clouds and surface solar radiation. However, the chaotic nature of broken to overcast clouds is proven difficult to represent. The coupled model setup is published as open‐source code, and is therefore freely available to the research community. Key Points: We developed an open‐source Python package named (LS)2D, designed to downscale the ERA5 reanalysis with turbulence and cloud‐resolving large‐eddy simulation (LESs)One month long experiments with (LS)2D and MicroHH over the Netherlands demonstrate both the skill and limitations of the coupled setupCapturing high‐frequency interactions between clouds and surface solar irradiance requires high resolution (O $\mathcal{O}$(10) m) LES [ABSTRACT FROM AUTHOR]

We present simulations of cumulus convection over land with shortwave radiation computed by a Monte Carlo ray tracer coupled to the flow solver. Ray tracing delivers very realistic in‐cloud heating rates and global horizontal irradiance fields. The necessary performance of the ray tracer has been enabled by the raw power of graphics processing unit computing and from techniques for accelerating data lookup and ray tracer convergence. We used a case study over a grassland in the Netherlands to compare simulations with a coupled ray tracer to those with a conventional two‐stream solver, and to test ray tracer convergence. We demonstrate that the simulated cloud evolution is insensitive to the convergence of the ray tracing across a wide range of samples per pixel per spectral quadrature point. Furthermore, simulations with a coupled ray tracer produce surface irradiance patterns that resemble observations and that strongly feed back to the evolution of clouds via locally enhanced surface heat fluxes. Plain Language Summary: Clouds absorb and reflect solar radiation and create spatial patterns at the land surface of cloud shadows interspersed with sunny regions. These patterns are currently largely simplified in most weather models because radiation computations cost a lot of computer power. We have developed a fast and realistic numerical radiation model that runs on a modern computer graphics card. By doing so, we can do simulations that produce radiation patterns that closely resemble reality and study how these patterns affect clouds. This enables us to improve our understanding of the complex interactions between clouds, solar radiation, and the Earth's surface. Key Points: We present graphics processing unit‐accelerated cloud‐resolving simulations with a coupled ray tracerSimulated cloud statistics are insensitive to ray tracer convergence across a wide range of sample sizesStrong coupling between surface irradiance patterns and cloud evolution underlines need for coupled 3D radiative transfer [ABSTRACT FROM AUTHOR]