Jianhao Zhang, Ph.D.

Shipping Emissions & IMO2020 Climate Impact
[Machine Learning ]

Ship tracks, bright, linear clouds trailing ocean-going ships, are great examples or manifestations of cloud adjustment in response to inadvertent aerosol perturbations from shipping emission. Their radiative characteristics in contrast to the surrounding unperturbed scenes offers invaluable insigts into understanding aerosol-cloud-radiation interactions. In this work, I trained ensembles of neural networks to examine how reduced pollution from shipping affects cloud properties and climate, revealing that while the IMO2020 regulation have significant environmental and public health benefits, they also produce a global warming effect of ~0.1 W m^-2, equivalent to about 3.5 years of CO₂ warming. We concluded even this substantial, abtrupt radiative forcing is challenging to detect, underscoring a key challenge for observing real-world cloud adjustments against cloud natural variability. This study raises concerns about accelerated warming associated with future emission regulations and the detectability of Marine Cloud Brightening (MCB) as a geoengineering proposal.

Related publications: Zhang et al. (2025) Commun. Earth & Environ. [media coverage]

Visible ship tracks and marine clouds as seen from the International Space Station. Linear, bright cloud fields created by sulfur-rich ship exhausts surrounded by less bright clouds. Astronaut photograph ISS059-E-36734. Image credit: Earth Science and Remote Sensing Unit, NASA Johnson Space Center, and NASA Earth Observatory.

(left) Effective radiative forcing estimates of the IMO2020 event from literature are compared to this study which reports the forcing due to changes in the SW cloud radiative effect from the 3 stratocumulus decks. (right) Detectability for changes in cloud radiative effect due to IMO2020. Perfect detectability lies at the upper-left corner (0,1) of the diagram, and a random detectability lies on the 1-to-1 line. Detectability increases from bottom-right to upper-left.

Cloud Albedo Susceptibility & Marine Cloud Brightening
[satellite observations ]

Characterizing and quantifying aerosol-cloud interactions are fundamental to improving future climate projections and assessing the viability of MCB. In this work, I developed a satellite-based, bottom-up framework for deriving cloud albedo susceptibility that controls for co-varying meteorological factors. This method revealed three distinct susceptibility regimes, their meteorological drivers, and their spatiotemporal variability across marine low-cloud regions. The findings stress the importance to consider the co-variability between cloud susceptibility and background aerosol condition in informing strategies for MCB.

Related publications: Zhang & Feingold (2023) ACP, Zhang et al. (2022) ACP

Cloud albedo susceptibility (colored filled circles) in LWP–N_d space for the NE Pacific stratocumulus deck. Size of the filled circles indicates the relative frequency of occurrence. Three susceptibility regimes are identified: (i) precipitating–brightening (light green; positive susceptibility states with effective radii greater than 12µm), (ii) entrainment–darkening (brown; negative susceptibil- ity states and right-hand side of the EEF isoline), and (iii) Twomey– brightening (dark green; non-precipitating states with positive sus- ceptibilities) regimes.

Monthly mean albedo susceptibility, overlaid with monthly mean low-cloud frequency of occurrence and cloud droplet number concentration (N_d).

Temporal Evolution in Cloud Water Adjustment
[LES modeling ]

Characterizing the magnitute and evolution of cloud water adjustment to aerosol perturbations is crucial to understand aerosol effects on clouds (i.e., brightening or darkening). Using a large ensemble of (300+) diurnal large-eddy simulations (LES) of marine stratocumulus and a novel conditional Monte Carlo sampling technique, I demonstrated that solar heating as a key factor buffering cloud responses after sunrise, underscoring the importance of timing for aerosol seeding in MCB scenarios. The conditional Monte Carlo sampling approach has the potential to be applied to also observational datasets and GCM Perturbed Parameter Ensembles (PPEs).

Related publications: Zhang et al. (2024) ACP, Chen, Zhang, et al. (2024) ACP

Diurnal cycle of the N_d–LWP regression slope, separated by colors representing different values at sunrise (large dots). Mean values of the 50 times, 25-member conditional Monte Carlo subsampling approach are shown.

Diurnal cycle of the N_d–SW_reflected regression slope, separated by colors representing aerosol perturbations ("seeding") at different times before sunrise.

Biomass Burning Aerosol Effects on Clouds and Climate
[ground-based observations ]

Eastern subtropical oceans are home to warm and bright low-level stratocumulus cloud decks. Among them, the southeast Atlantic basin presents an unique opportunity to study smoke-cloud-radiation interactions where seasonally emitted biomass burning plumes from the southern African continent are being advected over the oceanic region via the southern African Easterly Jet. In this work, I investigated how the presence of biomass burning smoke affects marine boundary layer clouds, particularly in the remote southeast Atlantic region. I demonstrated that the presence of sunlight-absorbing aerosols in the boundary layer suppress cloud water and coverage by modifying the boundary layer's thermodynamic structure and coupling state. The seaonsal variation in the vertical distribution of smoke contributes to the amplification of the seasonal cycle in low-cloud fraction, with significant implications for regional climate dynamics and radiative effects. These observationally based studies pose a challenging target for both process modeling and GCMs on their representation of aerosols, clouds, and their interactions with atmospheric circulations from diurnal to regional scales.

Related publications: Zhang & Zuidema (2021) ACP, Zhang & Zuidema (2019) ACP

Schematic of the diurnal cycle for days with“less” (blue) and “more” (red) smoke.

Amplified seasonal cycle in cloud fraction, cloud type, and all-sky albedo.

Bridging Observations and Earth System Models
[Neural Networks ]

Climate models, built to help us understand the internal variability of our climate system and its sensitivity to external forcings, are extensively used to project future climates and to examine the risks and viability of climate intervention proposals. Yet, estimates of cloud feedbacks (CF) and the effective radiative forcing due to aerosol-cloud interactions (ERFaci) still vary significantly among climate models in the latest model intercomparison project (CMIP6). In collaboration with Andrew Gettelman (PNNL), I extended the application of a Meteorology-Aerosol-Cloud Neural Network framework (also used in my other work) to GCM outputs. This application reveals striking discrepancies in cloud sensitivities to aerosol and meteorological conditions between models and observations, offering a promising path to benchmark and improve the representation of cloud feedback and aerosol-cloud interactions processes in Earth system models.

Related publications: Zhang, Gettelman, et al. in prep.

Low cloud susceptibility to N_d perturbations in observations and GCMs, derived from the Meteorology-Aerosol-Cloud Neural Network framework.

Inferring Cloud Processes from Satellite Snapshots
[satellite & airborne ]

The idea of inferring time-evolving process from static snapshots derives from the characteristics of an ergodicity system which depends the variabiliry of the boundary conditions of the system and the spatiotemporal scale of the process(es) one is interested in. Taking advantage of the characteristics of a marine cold-air outbreak, i.e. large-scale meteorological conditions are evolving much more slowly than the rate of cloud evolution, enabling a space-time exchange, I show examples of traces in LWP-N_d space that contain information about the effect of underlying boundary layer and microphysical processes on the directionality of the trace. This work showcase the potential of space-time exchange in process inference.

Related publications: Feingold, Glassmeier, Zhang, and Hoffman (2025) ACPD, Zhang et al. to-be-submitted

The concept of time-space exchange for process inference, illustrated with a cold-air outbreak case over the east coast of U.S. during the NASA-ACTIVATE field campaign. A diagram identifying how individual processes drive the system in LWP-N_d space: 1. activation, 2. condensation, 3. collision-coalescence, 4. precipitation, 5. entrainment (5.1: homogeneous, 5.2: inhomogeneous) is shown on the right.

Physical Science Reseach for Marine Cloud Brightening
[Overview]

As evidence for severe impacts of global warming continues to accumulate, the possibility of averting the worst effects of climate change through climate intervention (CI) is a topic of increasing scrutiny and debate. The common goal of these CI proposals is to cool the climate by deliberately modifying the solar (shortwave) or terrestrial (longwave) radiation pathways in the Earth’s atmosphere, aiming to reduce the worst impacts of rising global temperatures, in addition to reducing greenhouse gas (GHG) emissions. Our papers layout existing physical science knowledge and gaps on MCB and discuss research needed to assess the viability and risks of MCB.

Related publications: Feingold et al. (2024) Science Advances, Zhang & Feingold (2024) invited newsletter by GPC-APS

Marine cloud brightening proposals using ship-­based generators. In optimal conditions, many of these haze droplets would be lofted into the cloud by updrafts, where they would modify cloud microphysics processes, such as increasing droplet number concentrations, suppressing rain formation, and extending the coverage and lifetime of the clouds. On the left are shown details of the key aerosol, cloud, dynamics, and radiation processes in the marine boundary layer that are the foundation of MCB in shallow liquid clouds. The strong coupling between these processes presents interesting challenges and opportunities for understanding the outcomes of seawater haze injections into these clouds.

An integrated view of MCB research. Laboratory facilities such as cloud chambers together with observations at a range of scales will help improve the representation of physical processes in models. Parcel models, large eddy models, and cloud-resolving models will inform the global model activities to improve the reliability of regional climate responses. The earth view image is courtesy of the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT). Figure is courtesy of Chelsea Thompson, NOAA/CSL.