The team’s central finding is that the rate at which an aerosol plume spreads sideways through the marine atmosphere, while genuinely variable, barely matters for the total cooling that the Twomey effect can deliver. That result surprised the researchers, because plume spreading rates turned out to differ dramatically between the two meteorological regimes they examined. In their simulations of the northeast Pacific, the second day of a two-day air-mass trajectory, characterized by deeper boundary layers, stronger turbulence, and cleaner background air, produced plume spreading rates that exceeded the first day’s rates by more than 2 kilometers per hour at certain times. Yet when the researchers swapped the turbulence fields between the two regimes while holding everything else fixed, the resulting change in global mean Twomey forcing was smaller than 1 percent.

To reach that conclusion, the team built a modeling framework with two interlocking components. The first was a library of 54 two-day large-eddy simulations, or LES, of northeast Pacific stratocumulus clouds, run at 100-meter horizontal grid spacing over domains 51.2 kilometers on a side. These simulations were initialized along realistic Lagrangian trajectories derived from ERA5 reanalysis data, spanning the meteorological and aerosol conditions of summers from 2018 to 2021. Because LES explicitly resolves much of the energy-containing turbulence in the boundary layer, quantities like horizontal velocity variance and turbulence dissipation rate, which are usually parameterized and uncertain in coarser models, could be read directly from the simulated flow field. From the full library, the researchers selected 17 cases with median cloud fractions above 50 percent, which showed the best agreement with satellite observations from CERES and represent the most plausible candidate environments for cloud brightening.

The second component was a Langevin particle model, a statistical technique that describes the motion of individual aerosol parcels as a combination of a deterministic drift toward the mean wind and a random walk whose intensity is set by the resolved turbulence. Instead of solving the full fluid equations on a grid, which would have been computationally prohibitive across hundreds of injection scenarios, the particle model stochastically simulates thousands of superparticles released from moving ship-like sources on a 200-by-200-kilometer domain. Each superparticle represents a vast multiplicity of real aerosol particles, injected at a rate of 10 to the 16th particles per second, and the model tracks their positions, ages, and eventual removal through exponential decay timescales ranging from a quarter of a day to two days. The approach had already been validated against explicit LES of individual ship tracks in earlier work by the same group.

The clever step was merging the two components. At 20-minute intervals, the two-dimensional aerosol perturbation fields from the particle model were overlaid onto the two-dimensional reflectance fields from the LES, which preserve the detailed cloud morphology of each simulated scene. A modified version of Twomey’s albedo susceptibility equation then converts the local ratio of perturbed to background aerosol concentration into a change in reflectance, which is multiplied by the incoming solar radiation to yield the instantaneous radiative forcing in each grid column. Averaging over the domain and time, and scaling by an assumed sprayed area of about 3 percent of the global ocean surface, produces a global mean Twomey forcing. The resulting magnitudes, roughly 1.8 watts per square meter for a mid-range deployment injecting about 8 teragrams of salt per year, sit in a plausible range compared with earlier heuristic estimates, though somewhat larger because the sprayers in this study are concentrated in the cloudiest regions.

When the researchers probed why the forcing was so insensitive to spreading rate, a compensating web of effects emerged. In a sensitivity analysis, deeper boundary layers and lower cloud fractions on the second day of the trajectories each acted to reduce Twomey forcing by 10 to 20 percent, while mean winds had a negligible influence. But the cleaner background aerosol environment on day two boosted the forcing by nearly 50 percent, offsetting the losses. The physics behind this compensation is the nonlinearity at the heart of the Twomey effect: cloud albedo is most susceptible to added droplets when the background air is clean, so the same injection produces more brightening in pristine conditions. Faster spreading also reshapes the distribution of aerosol concentrations, reducing the frequency of the smallest concentrations and increasing the frequency of extreme ones, but these distributional shifts largely cancel out in the domain average.

The study also delivered practical guidance for anyone designing a cloud brightening deployment. Perhaps most strikingly, the forcing proved almost entirely insensitive to sprayer velocity across the range from stationary to 10 meters per second, as long as the ambient mean winds were realistic. Only in artificial no-wind experiments did slow-moving sprayers lose efficiency dramatically, because the plumes then pooled into saturated concentrations where additional aerosol yields diminishing returns. This suggests that stationary or slow-moving platforms could slash the energy and mechanical demands of a deployment without sacrificing radiative efficacy. The team also found that extending daily injections from 6 to 12 hours increases forcing by roughly 50 percent, and that starting injections a few hours before sunrise, rather than after, enhances forcing by 10 to 25 percent, consistent with earlier modeling by Jenkins and colleagues.

Where spreading rate does matter is in the assumptions embedded in global climate models. Most GCM studies of marine cloud brightening use uniform subgrid aerosol concentrations, an implicit assumption that plumes spread infinitely fast within a model grid box. The researchers tested the consequences using idealized simulations with a single fixed source and unity cloud fraction, designed to maximize the importance of spreading. Comparing the fastest and slowest spreading cases in their library, which differed by daily-average rates of 3.18 versus 0.2 kilometers per hour, they found that ignoring the natural range of spreading could bias forcing by 5 to 30 percent at low sprayer densities. But the infinite-spreading assumption produces far larger errors: at ship densities below 2.5 times 10 to the minus 4 ships per square kilometer, global mean Twomey forcing could be overestimated by 10 to 200 percent, depending on the assumed aerosol lifetime, and even high-end deployments could see errors above 10 percent when aerosol lifetimes fall below one day.

Indeed, aerosol lifetime emerged as the leading-order control on brightening potential, overshadowing every meteorological variable the team tested. Quadrupling the number of sprayers only doubled the forcing at a two-day lifetime, a signature of saturation, whereas changing the lifetime reshaped the forcing far more substantially. Particles older than one day contributed between 25 and 65 percent of the total forcing depending on the decay assumption, and the contribution of three-day-old particles swung from a trivial 2 percent at a one-day lifetime to 41 percent at a four-day lifetime. Because lifetime is governed by poorly constrained processes such as collision-coalescence scavenging in precipitating clouds, which can remove aerosols within hours, versus the two to three days suggested by non-precipitating LES, narrowing this uncertainty is arguably more urgent than refining plume dynamics.

The authors are careful to note the limits of their framework. By coupling aerosol fields offline to the LES reflectance fields, they assumed that all injected aerosol activates into droplets, an upper-bound choice, and that cloud adjustments, changes in liquid water path and cloud fraction triggered by the aerosol itself, are negligible compared with the Twomey effect. Observational studies increasingly support the latter assumption, but recent work suggests that in clean, precipitating regimes, cloud adjustments could enhance brightening, meaning the true picture may be more favorable than these estimates. The team also derived a promising statistical relationship, showing that the product of boundary-layer turbulent kinetic energy and the Lagrangian relaxation timescale explains roughly 70 percent of the variance in spreading rate once the first ten hours after injection are excluded, offering a ready-made parameterization for global models that carry these quantities. For now, the message is twofold: nature’s variability in plume spreading is unlikely to derail cloud brightening, but the convenient fiction of instantaneous spreading in climate models may be quietly inflating its promised cooling, and fixing that could be as simple as assuming a finite spread of about 1.5 kilometers per hour.

Subject of Research: Sensitivity of the Twomey cloud-albedo effect to aerosol plume spreading rates in marine stratocumulus and implications for Marine Cloud Brightening

Article Title: Estimating Twomey forcing sensitivity to aerosol plume spreading rates

Article References: McMichael, L. A., Erfani, E., Wood, R., & von Salzen, K. (2026). Estimating Twomey forcing sensitivity to aerosol plume spreading rates. Atmospheric Chemistry and Physics, 26(19), 14111-14132. https://doi.org/10.5194/acp-26-14111-2026

Image Credits: AI Generated

DOI: 10.5194/acp-26-14111-2026

Keywords: Twomey effect, Marine Cloud Brightening, aerosol plume spreading, marine stratocumulus, large-eddy simulation, Langevin particle model, cloud albedo, aerosol lifetime, boundary layer turbulence, climate models, radiative forcing, northeast Pacific