Multiscale modeling of a diurnal cycle of real‐world conditions is presented for the first time, validated using data from the CWEX‐13 field experiment. Dynamical downscaling from synoptic‐scale down to resolved three‐dimensional eddies in the atmospheric boundary layer (ABL) was performed, spanning 4 orders of magnitude in horizontal grid resolution: from 111 km down to 8.2 m (30 m) in stable (convective) conditions. Computationally efficient mesoscale‐to‐microscale transition was made possible by the generalized cell perturbation method with time‐varying parameters derived from mesoscale forcing conditions, which substantially reduced the fetch to achieve fully developed turbulence. In addition, careful design of the simulations was made to inhibit the presence of under‐resolved convection at convection‐resolving mesoscale resolution and to ensure proper turbulence representation in stably‐stratified conditions. Comparison to in situ wind‐profiling lidar and near‐surface sonic anemometer measurements demonstrated the ability to reproduce the ABL structure throughout the entire diurnal cycle with a high degree of fidelity. The multiscale simulations exhibit realistic atmospheric features such as convective rolls and global intermittency. Also, the diurnal evolution of turbulence was accurately simulated, with probability density functions of resolved turbulent velocity fluctuations nearly identical to the lidar measurements. Explicit representation of turbulence in the stably‐stratified ABL was found to provide the right balance with larger scales, resulting in the development of intra‐hour variability as observed by the wind lidar; this variability was not captured by the mesoscale model. Moreover, multiscale simulations improved mean ABL characteristics such as horizontal velocity, vertical wind shear, and turbulence.

Realistic multiscale simulations involve coupling of mesoscale and large-eddy simulation (LES) models, thus requiring efficient generation of turbulence in nested LES domains. Herein, we extend our previous work on the cell perturbation (CP) method to nonneutral atmospheric boundary layers (ABLs). A modified Richardson number scaling is proposed to determine the amplitude of the potential temperature perturbations in stable ABLs, with [ ] −1.0 overall providing optimum turbulence transition to a fully developed state (fetch reduced by a factor of 4–5, compared to the unperturbed cases). In the absence of perturbations, turbulence onset is triggered by a Kelvin–Helmholtz instability, typically occurring in the vicinity of the low-level jet maximum. It is found that a turbulent length scale [ ] can be used to more accurately estimate the optimum [ ], where q is the turbulence kinetic energy, and N is the Brunt–Väisälä frequency. In convective ABLs, a perturbation amplitude based on mixed layer temperature variance scaling is proposed: [ ]. For that criterion to be optimum, the ratio [ ], where [ ] is the wind speed at the top of the capping inversion, and [ ] is the convective velocity scale, needs to be incorporated: [ ]. This allows us to account for the competing roles of the surface thermal instability and the mean flow advection. For [ ] 10, the development fetch is reduced by a factor of [ ]6, while when [ ] 3, the use of the CP method does not provide a significant advantage in the ability to generate turbulence, provided a smooth mesoscale inflow.

This paper introduces a new large‐eddy simulation model, FastEddy®, purpose built for leveraging the accelerated and more power‐efficient computing capacity of graphics processing units (GPUs) toward adopting microscale turbulence‐resolving atmospheric boundary layer simulations into future numerical weather prediction activities. Here a basis for future endeavors with the FastEddy® model is provided by describing the model dry dynamics formulation and investigating several validation scenarios that establish a baseline of model predictive skill for canonical neutral, convective, and stable boundary layer regimes, along with boundary layer flow over heterogeneous terrain. The current FastEddy® GPU performance and efficiency gains versus similarly formulated, state‐of‐the‐art CPU‐based models is determined through scaling tests as 1 GPU to 256 CPU cores. At this ratio of GPUs to CPU cores, FastEddy® achieves 6 times faster prediction rate than commensurate CPU models under equivalent power consumption. Alternatively, FastEddy® uses 8 times less power at this ratio under equivalent CPU/GPU prediction rate. The accelerated performance and efficiency gains of the FastEddy® model permit more broad application of large‐eddy simulation to emerging atmospheric boundary layer research topics through substantial reduction of computational resource requirements and increase in model prediction rate.

A better understanding and prediction of turbulence dissipation rate ε in the atmospheric boundary layer (ABL) is important for many applications. Herein, sonic anemometer data from the Experimental Planetary boundary layer Instrumentation Assessment (XPIA) field campaign (March–May 2015) are used to derive energy dissipation rate (EDR; =) within the first 300 m above the ground employing second-order structure functions. Turbulence dissipation rate is found to be strongly driven by the diurnal evolution of the ABL, presenting a distinct statistical behavior between daytime and nighttime conditions that follows log–Weibull and lognormal distributions, respectively. In addition, the vertical structure of EDR is characterized by a decrease with height above the surface, with the largest gradients occurring within the surface layer (z < 50 m). Convection-permitting mesoscale simulations were carried out with all of the 1.5-order turbulent kinetic energy (TKE) closure planetary boundary layer (PBL) schemes available in the Weather Research and Forecasting (WRF) Model. Overall, the three PBL schemes capture the observed diurnal evolution of EDR as well as the statistical behavior and vertical structure. However, the Mellor–Yamada-type schemes underestimate the large EDR levels during the bulk of daytime conditions, with the quasi-normal scale elimination (QNSE) scheme providing the best agreement with observations. During stably stratified nighttime conditions, Mellor–Yamada–Janjić (MYJ) and QNSE tend to exhibit an artificial “clipping” to their background TKE levels. A reduction in the model constant in the dissipation term for the Mellor–Yamada–Nakanishi–Niino (MYNN) scheme did not have a noticeable impact on EDR estimates. In contrast, application of a postprocessing statistical remapping technique reduced the systematic negative bias in the MYNN results by 75%.

The level‐set method is typically used to track and propagate the fire perimeter in wildland fire models. Herein, a high‐order level‐set method using fifth‐order WENO scheme for the discretization of spatial derivatives and third‐order explicit Runge‐Kutta temporal integration is implemented within the Weather Research and Forecasting model wildland fire physics package, WRF‐Fire. The algorithm includes solution of an additional partial differential equation for level‐set reinitialization. The accuracy of the fire‐front shape and rate of spread in uncoupled simulations is systematically analyzed. It is demonstrated that the common implementation used by level‐set‐based wildfire models yields to rate‐of‐spread errors in the range 10–35% for typical grid sizes (Δ = 12.5–100 m) and considerably underestimates fire area. Moreover, the amplitude of fire‐front gradients in the presence of explicitly resolved turbulence features is systematically underestimated. In contrast, the new WRF‐Fire algorithm results in rate‐of‐spread errors that are lower than 1% and that become nearly grid independent. Also, the underestimation of fire area at the sharp transition between the fire front and the lateral flanks is found to be reduced by a factor of ≈7. A hybrid‐order level‐set method with locally reduced artificial viscosity is proposed, which substantially alleviates the computational cost associated with high‐order discretizations while preserving accuracy. Simulations of the Last Chance wildfire demonstrate additional benefits of high‐order accurate level‐set algorithms when dealing with complex fuel heterogeneities, enabling propagation across narrow fuel gaps and more accurate fire backing over the lee side of no fuel clusters.

Multiscale atmospheric simulations can be computationally prohibitive, as they require large domains and fine spatiotemporal resolutions. Grid‐nesting can alleviate this by bridging mesoscales and microscales, but one turbulence scheme must run at resolutions within a range of scales known as the terra incognita (TI). TI grid‐cell sizes can violate both mesoscale and microscale subgrid‐scale parametrization assumptions, resulting in unrealistic flow structures. Herein we assess the impact of unrealistic lateral boundary conditions from parent mesoscale simulations at TI resolutions on nested large eddy simulations (LES), to determine whether parent domains bias the nested LES. We present a series of idealized nested mesoscale‐to‐LES runs of a dry convective boundary layer (CBL) with different parent resolutions in the TI. We compare the nested LES with a stand‐alone LES with periodic boundary conditions. The nested LES domains develop ∼20% smaller convective structures, while potential temperature profiles are nearly identical for both the mesoscales and LES simulations. The horizontal wind speed and surface wind shear in the nested simulations closely resemble the reference LES. Heat fluxes are overestimated by up to ∼0.01 K m s⁻¹ in the top half of the PBL for all nested simulations. Overestimates of turbulent kinetic energy (TKE) and Reynolds stress in the nested domains are proportional to the parent domain's grid‐cell size, and are almost eliminated for the simulation with the finest parent grid‐cell size. Based on these results, we recommend that LES of the CBL be forced by mesoscale simulations with the finest practical resolution.