In the event of an accidental or intentional hazardous material release in the atmosphere, researchers often run physics-based atmospheric transport and dispersion models to predict the extent and variation of the contaminant spread. These predictions are imperfect due to propagated uncertainty from atmospheric model physics (or parameterizations) and weather data initial conditions. Ensembles of simulations can be used to estimate uncertainty, but running large ensembles is often very time consuming and resource intensive, even using large supercomputers. In this paper, we present a machine-learning-based method which can be used to quickly emulate spatial deposition patterns from a multi-physics ensemble of dispersion simulations. We use a hybrid linear and logistic regression method that can predict deposition in more than 100,000 grid cells with as few as fifty training examples. Logistic regression provides probabilistic predictions of the presence or absence of hazardous materials, while linear regression predicts the quantity of hazardous materials. The coefficients of the linear regressions also open avenues of exploration regarding interpretability—the presented model can be used to find which physics schemes are most important over different spatial areas. A single regression prediction is on the order of 10,000 times faster than running a weather and dispersion simulation. However, considering the number of weather and dispersion simulations needed to train the regressions, the speed-up achieved when considering the whole ensemble is about 24 times. Ultimately, this work will allow atmospheric researchers to produce potential contamination scenarios with uncertainty estimates faster than previously possible, aiding public servants and first responders.

Industrial accidents, chemical spills, and structural fires can release large amounts of harmful materials that disperse into urban atmospheres and impact populated areas. Computer models are typically used to predict the transport of toxic plumes by solving fluid dynamical equations. However, these models can be computationally expensive due to the need for many grid cells to simulate turbulent flow and resolve individual buildings and streets. In emergency response situations, alternative methods are needed that can run quickly and adequately capture important spatiotemporal features. Here, we present a novel deep learning model called ST-GasNet inspired by the mathematical equations that govern the behavior of plumes as they disperse through the atmosphere. ST-GasNet learns the spatiotemporal dependencies from a limited set of temporal sequences of ground-level toxic urban plumes generated by a high-resolution large eddy simulation model. On independent sequences, ST-GasNet accurately predicts the late-time spatiotemporal evolution, given the early-time behavior as an input, even when a building splits a large plume into smaller plumes. By incorporating large-scale wind boundary condition information, ST-GasNet achieves a prediction accuracy of at least 90% on test data for the entire prediction period.

In this paper, two different computationally inexpensive methods for nowcasting/data filling spatially varying meteorological variables (wind velocity components, specific humidity, and virtual potential temperature) covering scales ranging from 100 m to 5 km in regions marked by complex terrain are compared. Multivariable linear regression and artificial neural networks are used to predict micrometeorological variables at eight locations using the measurements from three nearby weather stations. The models are trained using data gathered from a system of eleven low-cost automated weather stations that were deployed in the Cadarache Valley of southeastern France from December 2016 to June 2017. The models are tested on two held-out periods of measurements of thermally-driven flow and synoptically forced flow. It is found that the models have statistically significant performance differences for the wind components during the synoptically driven flow period (p = 6.6 × 10⁻³ and p = 2.0 × 10⁻² for U and V, respectively), but perform the same otherwise. These methods can be used to spatially fill gaps in micrometeorological datasets. Recommended future work should include statistically interpreting the predictive models and testing their capabilities on meteorological datasets from different locations.

Cosmogenic beryllium-10 and beryllium-7, and the ratio of the two (10Be/7Be), are powerful atmospheric tracers of stratosphere–troposphere exchange (STE) processes; however, measurements are sparse for altitudes well above the tropopause. We present a novel high-altitude balloon campaign aimed to measure these isotopes in the mid-stratosphere called Beryllium Isotopes for Resolving Dynamics in the Stratosphere (BIRDIES). BIRDIES targeted gravity waves produced by tropopause-overshooting convection to study their propagation and impact on STE dynamics, including the production of turbulence in the stratosphere. Two custom-designed payloads called FiSH and GASP were flown at altitudes approaching 30 km to measure in situ turbulence and beryllium isotopes (on aerosols), respectively. These were flown on nine high-altitude balloon flights over Kansas, USA, in summer 2022. The atmospheric samples were augmented with a ground-based rainfall collection targeting isotopic signatures of deep convection overshooting. Our GASP samples yielded mostly negligible amounts of both 10Be and 7Be collected in the mid-stratosphere but led to design improvements to increase aerosol capture in low-pressure environments. Observations from FiSH and the precipitation collection were more fruitful. FiSH showed the presence of turbulent velocity, temperature, and acoustic fluctuations in the stratosphere, including length scales in the infra-sonic range and inertial subrange that indicated times of elevated turbulence. The precipitation collection, and subsequent statistical analysis, showed that large spatial datasets of 10Be/7Be can be measured in individual rainfall events with minimum terrestrial contamination. While the spatial patterns in rainfall suggested some evidence for overshooting convection, inter-event temporal variability was clearly observed and predicted with good agreement using the 3D chemical transport model GEOS-CHEM.

Accurately representing the formation of precipitation due to coalescence of cloud droplets remains a major challenge for atmospheric models. This study introduces and compares three data‐driven approaches for modeling the time evolution of droplet size distributions within a reduced‐order latent space: a polynomial‐based Sparse Identification of Nonlinear Dynamics (SINDy) framework, numerical integration of a neural‐network‐predicted time derivative, and direct prediction of state at the next timestep via neural network. All three methods are coupled and simultaneously trained with an autoencoder that discovers the latent space variables. We train with high‐fidelity data from large eddy simulations equipped with the superdroplet method, which captures the stochastic nature of the coalescence process. Our results show that while all three approaches can successfully emulate the nonlinear behavior of coalescence, the simplest, a polynomial‐based SINDy representation, generalizes most robustly to out‐of‐sample cloud regimes, achieving strong performance and physical consistency. We further quantify model uncertainty using conformal prediction intervals, thereby providing rigorous estimates of predictive confidence across the modeling pipeline. These findings highlight that simple time evolution representations, when coupled with the correct prognostic variables, are a promising approach to advancing data‐driven microphysics parameterizations toward operational atmospheric modeling.