You might have seen the images of the Great Pacific Garbage Patch or read about how microbeads are ruining aquatic ecosystems. Whatever the blight, the message is clear: when plastics mix into bodies of water, trouble is on the horizon. Now, researchers from Pennsylvania State University at Behrend and the Rochester Institute of Technology (RIT) have leveraged the Comet system at the San Diego Supercomputer Center (SDSC) to examine the distribution of plastics in one of the Great Lakes: Lake Erie.
From 2012 to 2014, Sherri Mason – a chemistry professor at Penn State-Behrend – set out on the waters of the Great Lakes with a team of undergraduate students to collect water samples from the lakes’ surfaces. “We used three different vessels to collect our samples, taking advantage of every opportunity afforded to us,” Mason said in an interview with SDSC’s Kimberly Mann Bruch. “We would hop aboard, collect our samples, then take them back to the lab for analysis. Such fieldwork is important to better understand how plastic is not only impacting our lake, but freshwater lakes around the world.”
Once at the lab, the team measured the levels of plastics in the samples, finding a generally high level of plastic pollution (which the researchers attribute to nearby population centers and the position of the lakes), mostly consisting of sub-millimeter plastic particles.
The researchers didn’t want to stop there, though – they wanted to understand how much plastic was in Lake Erie, and how it was distributed. To do that, they used a Lagrangian transport model that simulated the three-dimensional movement of plastic particles, accounting for plastics being moved, sinking and mixing into the waters and sediments of the lake.
Running this model called for much greater computing power than the universities’ labs had on hand, so the researchers turned to the Comet system at SDSC via a time allocation from the Extreme Science and Engineering Discovery Environment (XSEDE) program. Comet comprises 1,944 Intel Haswell-based nodes and 72 Nvidia-based GPU nodes, delivering 2.76 peak petaflops in total.
“Comet allowed us to run three years of hydrodynamic simulations of Lake Erie that drive the transport of plastic particles within Lake Erie,” said Matthew Hoffman, a professor of mathematics at RIT and a co-author on both of the papers that emerged from the research. “We don’t have the computing infrastructure at my university to run these simulations efficiently, so having access to Comet enabled us to run state-of-the-art code that is not possible on our local desktops and laptops.”
The researchers calibrated their transport model based on the types and quantities of plastic they had found in their water samples, then ran the simulations on Comet. The supercomputer-powered model estimated that there are likely 381 ± 204 metric tons of plastic in the waters of Lake Erie – somewhere around the weight of 50 to 150 mid-sized elephants. Even though this staggering number is orders of magnitude higher than previous sampling-based estimates, the researchers wrote, it’s still just a fraction of estimated plastic input – leaving future research to continue closing the gap in the “plastic mass balance” by figuring out exactly how much plastic is entering the system and exactly where it’s going.
To read the reporting from SDSC’s Kimberly Mann Bruch, click here.