To assess the ability of Earth system model (ESM) predictions, we developed a tool called ESMAC Diags to understand how aerosols, clouds, and aerosol–cloud interactions are represented in ESMs. This paper describes its version 2 functionality. We compared the model predictions with measurements taken by planes, ships, satellites, and ground instruments over four regions across the world. Results show that this new tool can help identify model problems and guide future development of ESMs.

A primary sink of air pollutants is dry deposition. Dry deposition estimates differ across the models used to simulate atmospheric chemistry. Here, we introduce an effort to examine dry deposition schemes from atmospheric chemistry models. We provide our approach’s rationale, document the schemes, and describe datasets used to drive and evaluate the schemes. We also launch the analysis of results by evaluating against observations and identifying the processes leading to model–model differences.

PyFLEXTRKR is a flexible atmospheric feature tracking framework with specific capabilities to track convective clouds from a variety of observations and model simulations. The package has a collection of multi-object identification algorithms and has been optimized for large datasets. This paper describes the algorithms and demonstrates applications for tracking deep convective cells and mesoscale convective systems from observations and model simulations at a wide range of scales.

Atmospheric aerosols play a critical role in Earth's climate, but it is too computationally expensive to directly model their interaction with radiation in climate simulations. This work develops a new neural-network-based parameterization of aerosol optical properties for use in the Energy Exascale Earth System Model that is much more accurate than the current one; it also introduces a unique model optimization method that involves randomly generating neural network architectures.

We developed an Earth system model (ESM) diagnostics package to compare various types of aerosol properties simulated in ESMs with aircraft, ship, and surface measurements from six field campaigns across spatial scales. The diagnostics package is coded and organized to be flexible and modular for future extension to other field campaign datasets and adapted to higher-resolution model simulations. Future releases will include comprehensive cloud and aerosol–cloud interaction diagnostics.

Radars can suffer from missing or poor-quality data regions for several reasons: beam blockage, instrument failure, and near-ground blind zones, etc. Here, we demonstrate how deep convolutional neural networks can be used for filling in radar-missing data regions and that they can significantly outperform conventional approaches in terms of realism and accuracy.

In addition to detecting clouds, vertically pointing cloud radars detect individual insects passing over head. If these insects are not identified and removed from raw observations, then radar-derived cloud properties will be contaminated. This work identifies clouds in radar observations due to their continuous and smooth structure in time, height, and velocity. Cloud masks are produced that identify cloud vertical structure that are free of insect contamination.

The activation of aerosol into cloud droplets is an important but uncertain process in the Earth system. The physical and chemical interactions that govern this process are too computationally expensive to explicitly resolve in modern Earth system models. Here, we demonstrate how hybrid machine learning approaches can provide a potential path forward, enabling the representation of the more detailed physics and chemistry at a reduced computational cost while still retaining physical information.