I am beyond thrilled to finally get to share some news that I am rather excited about - I have had the privilege of coauthoring an article accepted in Science, and it made the cover! The article, "Hydraulic jumps above supercell storms", authored by Dr. Morgan O'Neill (Stanford), Dr. Leigh Orf (SSEC/CIMSS), Dr. Gerald Heymsfield (NASA Goddard), and myself (UW-Mad/CIMSS), is to be published on Friday, September 10th, 2021. I feel like this is kind of a big deal and was worth dusting off the blog again. The article describes simulations of Above Anvil Cirrus Plumes (AACPs) over a supercell thunderstorm, and the mechanisms responsible for their formation. The simulations, compared to observations from previous NASA campaigns with the ER-2 aircraft, conclude that in this case hydraulic jumps are responsible. This would be the first confirmed instance of hydraulic jumps occurring without the presence of solid topography, and also shows that these AACPs are responsible for injecting large amounts of water vapor into the lower stratosphere. This water vapor injection is shown to be significantly larger than previous research suggests, where it can have a climate impact through the destruction of ozone.
Some press release sources discussing the article are linked below. If this is as far as you read -- thank you for stopping by! Otherwise, I'll have more below on exactly what it was I contributed to the study and how it was done. First, I seriously cannot begin to express how crazy this all is to me. I honestly went into this year incredibly disappointed in the lack of work I accomplished in 2020... only for the fruits of a side project and amazing collaborators to come fourth from it. I don't have any real conclusions to draw from that other than to keep persisting even when you don't see where things are going, and always say yes to collaborating with good people. Thanks Morgan and Leigh for letting me contribute to this project.
The cool thing about this paper is that we each were able to bring crucial components to the analysis with our different backgrounds and datasets -- and to anyone who knows me, it should come as no surprise that my contributions were in conducting the Lagrangian trajectory analysis.
Since the beginning of my PhD, we have been working to build a pipeline in which we can do the following:
- Run high resolution simulations while saving data to disk at the model time step for analysis
- Save data efficiently using lossy data compression and domain subsetting
- Save only the bare minimum of base variables and calculate everything else offline
- Conduct Lagrangian trajectory analysis without having to re-run the entire simulation
While we've been gradually building and refining this pipeline over time, this publication is the first peer-reviewed research to make use of this whole pipeline from start to finish. This was done using our modified version of CM1 with the LOFS I/O driver, the LOFS-Read API, and the Lagrangian Offline Flow Trajectories (LOFT) package I've been building. LOFT was built specifically for conducting the trajectory integrations on the highly parallelizable GPU architecture.
With these tools we were able to run simulations with 50m uniform/isotropic grid spacing, save data at the model time step of 1/3s, and run tens of thousands of forward and backward trajectories through the updraft and through the hydraulic jump without additional simulations. One of the figures displaying the trajectories is below.
Using the parcel trajectories, we were able to quantify the amount of mixing taking place in the hydraulic jump, as well as tracing the water vapor content of parcels that enter the stratosphere. Various volumes of parcels were seeded to either enter the jet streak of the hydraulic jump, or to be at an altitude above the jump to display the displacement caused by the jump. A parcel descending into the jump accelerated from 35 m/s to 108 m/s in 33s, and another parcel reached a storm relative maximum wind speed of 119 m/s. Some parcels warmed by as much as 100 K of potential temperature.
I'm currently finishing up a paper I hope to have submitted soon that details everything about the offline trajectory process and quantifies the error incurred through lossy compression for the gain in disk space. While I don't want to share too much while I'm still writing it, I will show you this figure of how the size of a 100 m resolution simulation changes based on the compression accuracy parameter used. Needless to say, you can get a powerful reduction in file size for minimal loss in information... but more on that once that paper is finished. The kind of analysis that resulted in this Science article would not be possible without these tools.
Once again, super stoked and humbled by getting to be a part of this. Dr. O'Neill did some incredible work and deserves all of the credit due. I'm also thrilled to have something to show for all of this coding I've been doing to get this to work correctly. It feels pretty good.