Research
Aerosols, clouds, radiation, and climate — using NASA satellite and airborne observations, climate model output, radiative-transfer and cloud-resolving modeling, and physics-informed machine learning.
Research Interests
1. Cloud Microphysics & Precipitation in Convective Storms
How aerosols shape the warm-rain and ice-crystal pathways of precipitation in mixed-phase and deep convective clouds, and which microphysical processes dominate ice amplification in convective anvils (Waman et al., 2022; Gupta et al., 2023). Current work extends this to Saharan dust and tropical Atlantic convection using NASA airborne field-campaign and satellite observations together with cloud-resolving modeling.
2. Global Dust Cycle & Its 21st-Century Decline
Observation-constrained reconstruction of the global dust cycle, showing that global dust optical depth declined by −10 ± 8 % during 2003–2023 (Gupta et al., Sci. Adv., 2026), quantification of the longwave radiative effect of dust missing from climate models (Kok, Gupta et al., Nat. Commun., 2026), and ongoing diagnosis of the environmental drivers of the decline.
3. Tropical Cirrus Radiative Heating
Whether radiative heating stabilizes or destabilizes tropical cirrus depends on where within the cloud it is deposited. Using multi-year spaceborne radar–lidar observations and radiative-transfer calculations, I examine how ice loading and the day–night cycle shape this vertical heating structure and what it implies for cirrus evolution and for proposed cirrus cloud thinning (Gupta & Bennartz, submitted to PNAS).
4. Remote Sensing of Aerosol & Cloud Properties
Satellite- and ground-based retrievals of aerosol and cloud properties — aerosol optical depth, size, and absorption; liquid and ice water paths and effective radius — to narrow uncertainties in Earth's radiation budget.
5. Volcanic Eruptions: Umbrella Clouds, the Stratosphere & Climate
Machine-learning and physics-informed methods for detecting and tracking umbrella clouds in geostationary satellite imagery and for describing how they grow and are transported during simple and multipulse explosive eruptions (Gupta & Bennartz, 2022; in preparation), together with quantification of how volcanic injections of water vapor and SO₂ perturb stratospheric composition, radiative forcing, and surface temperature — most recently for the 2022 Hunga eruption (Gupta et al., 2025).
6. Arctic Mixed-Phase Clouds & Ice Formation
The role of ice-nucleating particles and ice-formation processes in the liquid–ice balance of high-latitude mixed-phase clouds and their influence on the polar surface energy budget.
7. Machine Learning for Atmospheric Science
Physics-informed deep learning and Bayesian methods for satellite retrievals, observation-consistent aerosol datasets, and the representation of clouds and convection in climate models, learned from superparameterized simulations.
Current Research
As Staff Scientist at Vanderbilt University, I work on the radiative structure of ice clouds — including the diurnal variability of tropical cirrus — in support of NASA's PolSIR (Polarized Submillimeter Ice-cloud Radiometer) mission, which will deliver global, all-weather measurements of ice clouds to constrain their impact on Earth's radiation budget.
In tropical cirrus, I am using multi-year spaceborne radar–lidar observations with radiative-transfer calculations to characterize how ice loading controls the vertical distribution of radiative heating within the cloud, with implications for cirrus evolution and for proposed cirrus cloud thinning. This work is submitted to PNAS.
I have reconstructed the 21st-century global dust cycle by combining ground-based and satellite observations, three aerosol reanalyses, and a large-ensemble inverse modeling framework. We show that global dust aerosol optical depth declined by −10 ± 8 % between 2003 and 2023 — the strongest evidence to date that the planet has become less dusty in the early 21st century. This work is published in Science Advances (2026).
Building on the dust-decline result, I am now diagnosing what is driving the decline by combining the inverse-model ensemble with reanalysis meteorology, surface-station wind observations, and independent dust-emission modeling to identify the environmental mechanisms behind regional emission trends and assess how they may shape dust emissions in a future climate.
As a co-author on an observationally constrained assessment of the longwave direct radiative forcing of desert dust led by Jasper Kok (UCLA), I contributed the dust absorption calculations underlying the constraint. The study finds a present-day clear-sky dust longwave DRE of +0.25 ± 0.06 W m⁻² — nearly double current climate model estimates — and is published in Nature Communications (2026).
A related effort examines how Saharan dust modulates deep convection and precipitation over the tropical Atlantic, where dust, African easterly waves, and developing tropical cyclones coincide. Combining NASA airborne field-campaign and satellite observations with cloud-resolving simulations, this work seeks to determine which warm-rain, ice-phase, and radiative pathways dust activates in convective clouds, building on my earlier process-tagging studies of precipitation formation (Gupta et al., 2023).
I also continue developing physics-informed, scene-adaptive machine-learning models for volcanic umbrella clouds using Himawari-8/9 imagery, including work on how umbrella clouds grow and are transported during multipulse eruptions such as the 2022 Hunga sequence, with application to other recent explosive eruptions.
Across these projects, physics-informed machine learning is a core tool rather than an add-on: I am developing deep-learning and Bayesian methods for satellite retrievals of cloud and aerosol properties, building observation-consistent mineral-dust datasets designed for training and independently testing AI models, and working on data-driven representations of convection and cloud processes learned from superparameterized simulations as a computationally affordable alternative to conventional parameterizations in climate models.
A new line of research uses NASA's PREFIRE far-infrared and infrared spectral radiances to quantify spectrally resolved longwave cloud radiative effects across both polar regions, providing observationally constrained benchmarks for evaluating reanalyses and climate models directly in measurement space.
I am also investigating the impact of mineral-dust ice-nucleating particles on Arctic mixed-phase clouds, examining how INP parameterizations and ice-formation processes interact to control the liquid–ice partitioning that governs the high-latitude surface energy budget, with implications for sea-ice and permafrost projections.
News Coverage
Our work on the Hunga Tonga eruption and its unexpected cooling effect has been widely covered: