Table of Contents
Introduction
Almost a decade ago, one of my students investigated the influence of terrain on thunderstorms crossing the Hudson River Valley. It was a unique approach to the problem of how terrain impacts storms, examining correlations between radar parameters and elevation. After his project was completed, I expanded the dataset and played around with it, but lacked a rigorous statistical methodology to make sense of the data.
For a long time, the project sat on my back burner, awaiting the right statistical approach and a framework to explain the results. Now, thanks to a colleague with experience in time series modeling and some new conceptual insight, this longstanding project is finally reaching completion. While the paper puts forth a new way of understanding internal structure in convective storms, this post provides an introduction to some of the concepts in the paper using a different context: cloud growth.
One Day in the Driveway
I typically work on my research in my home office. I step away from my desk every once in a while to get some fresh air and to chew on my latest idea. One afternoon at the end of June 2025, I noticed some clouds. They were changing rapidly, and this robust evolution brought to mind the core framework of my paper. You can flip through these cloud photos (taken over the course of 20 minutes) in the widget below.
In the first photo, it’s hard to tell whether this is one cloud or two. The greatest growth rate appears to be occurring in the cloud to the right, given the lumpiness on top. It looks to have a decently strong, vertical updraft. Meanwhile, the cloud at upper left is evaporating.
In the second photo, the cloud at right looks to be strongly sheared, breaking apart. It looks to be dissipating as drier air from the surroundings evaporates its cloud droplets. The cloud at upper left now appears to be the center of activity. Connecting the two, the updraft appears to be weak, slanted, and discontinuous.
In the third photo, the cloud on the right has rebounded, while the cloud at upper left has partially evaporated. The updraft again appears to be vertical, but perhaps a bit weaker than in the first photo.
And finally, in the fourth photo, the cloud appears to be sheared again, with the most droplet growth in the upper left cloud, while the cloud at right has mostly evaporated.
But here’s the thing: there was minimal wind shear. The winds at ground level were calm, and even up to more than 5 km (3.1 miles), winds were quite light. The lack of cloud movement observed throughout the time frame of the above photos suggests minimal influence from lower tropospheric wind conditions. The data (from Birmingham, Alabama) are limited because the much closer Peachtree City sounding was unavailable, but they seem to be representative of the conditions here in Columbus. Due to the lack of wind shear, the rapid changes in cloud behavior were almost certainly internally-driven (although the process appears to involve quite a bit of mixing with the drier surrounding air).
In the sounding above, wind speeds throughout the depth of the cloud layer that I had observed were generally less than 10 knots. Upper tropospheric winds were roughly 30 knots, indicating weakly forced conditions. And with a moist environment, it might be surprising that the clouds are evaporating so readily.
So what is going on with these clouds? They have a sheared appearance, but wind shear is minimal. They environment is very moist, but the clouds are evaporating quickly. And how does this relate to the paper?
A Taste of the Framework
As meteorologists, we know that cloud growth is built on the phase changes of water. And we know that as hydrometeors form through condensation, heat energy is released to the environment. The new framework is built upon these phase changes of water – net hydrometeor production – and how they build structure in a storm as reflectivity centroids shift and thermal gradients grow in response to droplet and ice crystal growth.
The cloud sequence above illustrates “side-feeding growth”. The lower cloud is seeding the slanted growth toward the upper cloud. Rather than the strictly vertical motion that upper-level divergence over low-level convergence implies, horizontal momentum from low-level convergence is carried slantwise by rising air, lifting the condensing parcels as a tilted formation. Considering the clouds together as a single system, the dynamical processes shaping cloud growth and structure are largely internal. Rather than being externally imposed by shear, the tilt is built by parcel-scale cloud droplet growth.
The direction of the tilt can be explained by small-scale processes such as entrainment, localized turbulence, and hydrometeor loading. These internal mechanisms act to favor one flank of the cloud over the other, guiding upward growth into the side where relative humidity is slightly greater or inhibition is slightly weaker. At the same time, evaporation along the opposite flank accentuates the asymmetry. Thus, we see ragged clouds above the right flank of the lower cloud. The oscillations in growth and evaporation between the right and upper left clouds reflect similar patterns often seen in my thunderstorm dataset. Perhaps even more surprisingly, the time scale of these oscillations (10-20 minutes) matches, as well.
These processes illustrate the new framework’s principle of self-organization: the growth and evaporation of cloud droplets dictate the evolution of the clouds’ structure. The cloud sequence discussed here illustrates one thread of a broader story that the upcoming paper will explore in depth. The fundamental self-organizing mechanism explained here scales up to convective storms and perhaps even mesoscale convective systems. In addition to expanding upon this basic framework, the paper further incorporates considerations of storm motion and terrain impacts on convection.
Look for the paper during the second half of 2026/
