OpenACC and OpenMP target offload achieve competitive performance on the graphics processing unit (GPU), but OpenACC performs better for Parameterization of Unified Microphysics Across Scales

GPU can outperform CPU in a practical Community Atmosphere Model simulation and using multiple GPUs per node is beneficial, especially at a high resolution

The data transfer between CPU and GPU and the data allocation on GPU dominate the overall GPU performance

The editors thank the 2022 peer reviewers

The History Matching method is explained in detail then used for tuning a toy coupled model: the Lorenz 96 model

The importance of several design choices is demonstrated, especially when considering forced experiments such as Atmospheric Model Intercomparison Protocol and Ocean Model Intercomparison Project

We argue that this tuning method is semi-automatic & highlight the importance of human expertise when considering it for real coupled models

A tool to trace the process contributions to surface precipitation is developed

The tool confirms dominance of warm processes in convective rain and cold processes in stratiform rain

The tool reveals that the autoconversion formulation produces an excessive amount of rain drops that the microphysics scheme clips

Bin and Lagrangian microphysics schemes with various levels of complexity are used to handle aerosol-cloud interactions in a cloud chamber

Simulations using haze-capable bin and Lagrangian schemes capture the observed haze mode in the chamber

Activation and deactivation rates are overestimated when using a CCN-based bin scheme compared with haze-capable schemes

The impact of coastal indentation on circulation within an idealized square basin is studied systematically for various numerical choices

Solutions are insensitive to coastal indentation with rot-div formulation of the diffusive tensor and sufficiently fine spatial resolution

Indented coastlines become slippery when a true mirror boundary condition is imposed on the flow for all numerical formulations

A rapid infrared gaseous spectral transmittance scheme based on the correlated K-distribution method and the optimal method is developed

The optimal iteration method can significantly improve the accuracy of heating rates simulated with the correlated K-distribution model

Our infrared gaseous spectral transmittance scheme maintains high accuracy across various carbon dioxide concentration scenarios

Model scheme complexity brings about a bouquet of problems, from hindering interpretability to hiding unforced modeler choices

Using sensitivity analysis on an emulated perturbed parameter ensemble we characterize the two cloud microphysics schemes of ECHAM-HAM

We find potential for process simplifications as well as instances where model sensitivities deviate from a priori physical understanding

A matrix version of the Dynamical Orthogonal (DO) field method is presented

DO is used to determine transition probabilities in multi-stable stochastic dynamical systems

The method is applied to study transition behavior of ocean western boundary currents

A physics-incorporated deep learning (DL) framework for parameterization of atmospheric radiative transfer is proposed

The DL model structures with global receptive fields are more suitable for the radiative transfer problem

Cloud resolving simulations for several months using the ICOsahedral Non-hydrostatic model Large-Eddy model in the Arctic are evaluated

High agreement of model and observations on wind flow, integrated water vapor, and light to medium precipitation

More clouds are simulated than observed but at the same time the model clouds lack liquid water

A combination of statistical methods is used to fit cloud size distributions from two simulated convective cloud ensembles

The merging of individual cloud cores is found to control the emergence of power-law cloud size distributions

Community Land Model Version 5 (CLM5) streamflow predictions use default parameters are vulnerable to poor performance especially in the arid Southwest and Central U.S

The effects of CLM5 hydrologic parameter uncertainty are significantly higher for drought versus flood predictions in humid regions

The dominant hydrologic parameters in CLM5 vary significantly across regions, diagnostic error metrics, and temporal scales

Numerical limiters and nonlinear artificial diffusion can inject unintended small-scale features into the solution of the model equations

Improved treatment of limiters and artificial diffusion can reduce noise and improve the quality of solutions and numerical convergence

Convergence testing is a valuable tool for detecting noise in the model solution, as well as unintended discontinuities and grid dependence

Influence of circulation changes on cloud feedbacks is substantial in some cloud-resolving models

Component of cloud feedback associated with circulation changes is coupled to ascent area

Intermodel spread in response of ascent area linked to non-radiative diabatic heating

A globally enhanced ecosystem C stock (ECS) by land greening has been detected since the 1980s from 40 global models

The sensitivity of gross primary productivity to leaf area index (LAI) is the largest contributor to the modeled greening effect on ECS uncertainty

This study estimated the effect of global greening on ECS with the emergent constraint method

Numerical models for large-scale water age/quality simulations are absent in communities of hydrology and Earth Surface Processes

A parallel framework for accelerating Lagrangian particle tracking to continental-scale on distributed, multi-Graphics Processing Unit platforms is established

The parallelized particle tracking model, EcoSLIM, is a promising tool to accelerate our understanding of the terrestrial water cycle

A simple coupling of black and organic carbon emissions from wildfires to cloud microphysics

Environmental atmospheric moisture has a more significant influence than surface moisture release on the pyroconvective cloud formation

In our case study, the formation of pyroconvective clouds increased the surface winds and fire spread rate by 8% and 25%, respectively

Subgrid models of all interaction classes between eddies, meanfield and topography are calculated from reference global ocean simulations

Inter-eddy interactions are modeled both stochastically and deterministically, with deterministic models used for the other interactions

Coarse simulations using these subgrid coefficients without tuning are in very good agreement with those of the high-resolution simulation

The representation of turbulence and convection in the convective gray zone is investigated at global 5 km resolution

In the examined context a reduced mass-flux convection scheme is beneficial in 5 km-resolution global model forecasts

The assessment is more conclusive with regard to convective organization and tropical variability than extratropical predictability

We developed a deep learning method that significantly improves the accuracy and resolution of gridded sea surface height anomalies

This data-driven method takes advantage of combining sea surface temperature and altimetry observations

Inferred surface geostrophic currents are quantitatively and qualitatively more realistic than those from existing sea surface height maps

The stochastic unified convection scheme (UNICON) is extended to deep convection by parameterizing the impact of mesoscale organized flow

The free parameters of stochastic UNICON are optimized using single-column model simulations over the ocean

The biases presented in the simulations of original UNICON are reduced by stochastic UNICON for tropical deep convection

Machine-learned (ML) corrective tendencies and radiative fluxes for a coarse-grid global model are trained using a year-long global storm-resolving simulation

The ML corrections reduce systematic errors of land surface temperature and precipitation in multiyear coarse-model simulations

The ML corrections nevertheless induce systematic biases in the simulated Hadley circulation

The study determines how turbulent kinetic energy (TKE) advection influences tropical cyclone simulations subject to storm motion and ocean coupling

The inclusion of TKE advection changes the storm-relative location of maximum TKE and the vertical distribution of TKE in each quadrant

The inclusion of TKE advection produces a more realistic distribution of TKE and should be used in closure schemes when possible

We present an interactive representation of the stratospheric sulfur cycle in GFDL ESM4 that replaced prescribed aerosol optical properties

We simulated years 1989–2014 with a focus on the Mt. Pinatubo eruption and evaluated against observations

Simulated stratospheric sulfate burden and optical depth are sensitive to injection height, emission amount, and aerosol size

A stochastic reaction-diffusion model of the tropics can simulate aggregated and random convective states

The model produces a similar sensitivity to domain size and resolution as full-physics models

A dimensionless parameter called the aggregation number combines model and experiment configuration parameters to predict aggregation onset

In a library of weather research and forecasting model simulations, we compare methods for estimating 3D gravity wave drag force that are un- and under-resolved by general circulation models

For drag associated with vertical fluxes, different methods agree on time- and zonal-mean but not on instantaneous spatiotemporal patterns

Drag associated with horizontal fluxes is significant but is very sensitive to the estimation methodology

Expanding on other kinematic studies of clustering buoyant tracers, we include simple temporal structure in the carrying velocity fields

We utilize global measures of clustering, as well as a cluster tracking algorithm, to compare the process quantitatively between models

Our results show the sensitivity of clustering to distortions of time structure, which can inform future studies in dynamical contexts

Tropical cyclone (TC) large-eddy simulations were conducted as benchmarks to provide statistical information about subgrid cloud at convection-permitting resolution

Subgrid convective clouds were mainly located in the lower troposphere, and exhibited shallow vertical extents of less than 4 km

Local subgrid clouds differed in mass-flux profile shape and magnitude at various cloud regimes in the TC convection system

This study introduces a simultaneous multiscale data assimilation method using spatial scale-dependent and variable-dependent localization (VDL)

The scale-dependent method concurrently updates storm and its ambient environments compared to a traditional single scale method

Further applying VDL is beneficial in the analysis and forecast of a tornadic supercell