Soon, Steve Guimond and his students will begin exploring a new angle of his hurricane research. They want to better understand the fundamental physics that drives hurricanes. Specifically, they want to know how small disturbances in a hurricane’s wind flow, similar to a strong gust on a windy day, may affect its overall structure and intensity. The findings could have implications for hurricane forecasting.
A new, $682,000 National Science Foundation (NSF) grant in collaboration with the New Jersey Institute of Technology will fund the team’s work, which primarily involves developing, running, and analyzing complex numerical models on supercomputers. However, Guimond might never have received the grant if he hadn’t received a UMBC Strategic Award for Research Transitions (START) first.
In 2018, the NSF rejected a related proposal from Guimond, who is an associate research professor with UMBC’s physics department and the Joint Center for Earth Systems Technology, a partnership with NASA. The proposal outlined a broad research program using two different methods to learn more about hurricanes: remote sensing and numerical modeling. “The START funding helped us evaluate those two sides of the project,” he says.
The START program funded initial research that was carried out by Guimond, Devin Protzko ’20, physics and mathematics, and Badrul Hasan, a Ph.D. student in mechanical engineering. They gathered preliminary data showing that the modeling path had significant potential, so they wrote a fresh NSF proposal with a narrower focus. It won approval. The small START grant, Guimond says, “was key to helping us identify the most fruitful path for follow-on funding from NSF.”
Turbulence in focus
The NSF proposal will support work to understand the role of turbulence in how or whether hurricanes intensify. Guimond’s team will also look at the value of using numerical hurricane models with very high resolution. High resolution is important because a hurricane’s “gusts” appear and disappear very quickly. They also generally take up very little physical space. Because they’re so ephemeral, you can’t simulate them precisely or accurately with current models. And if you can’t simulate them, you can’t figure out their role in overall hurricane development. High resolution models would make those simulations possible, which can feed back into improved forecasts in the future.
The fact that Guimond has the chance to do this work exemplifies START’s goal. The program, funded by UMBC’s Office of the Vice President for Research, offers a maximum of $25,000 to UMBC faculty who wish to pursue new avenues of research. The hope is that the funds will put them in a stronger position for much larger external grants from places like NSF—exactly what happened for Guimond.
A first step
A new cohort of UMBC faculty will receive START funding this summer. Lira Yoon, associate professor of psychology, hopes to follow a similar path to Guimond. Yoon will collect initial data on how Asian Americans regulate their emotions in response to overt racism, and how or whether the strategies they practice affect their well-being. This work is particularly relevant given the sharp rise in anti-Asian discrimination during the pandemic. Yoon is also working on a project to help long-term care workers better manage their stress, another group impacted heavily by the pandemic.
“I’m eager to start a new line of research examining the effects of racial discrimination on psychological well-being, particularly in Asian Americans,” Yoon shares. “Although the ultimate solution to the problems resulting from racial discrimination lies at the policy and systems levels, understanding mechanisms operating at an individual level could help mitigate the adverse effects of discrimination.”
“This project will be the first step in that direction,” she says, “and it will provide preliminary data to secure external funding for future larger-scale research.”
Maryam Rahnemoonfar, information systems, is another member of the new START cohort. Her team is developing algorithms for use on drones flying over affected areas after natural disasters. The goal is for the drones to relay important information about conditions on the ground in real time. Her team has created a unique dataset, called FloodNet, that can train the algorithms to recognize disaster impacts—for example, roads that are flooded or blocked by debris.
With a previous AI for Earth grant from Microsoft, more than 20 students at all levels in Rahnemoonfar’s research group spent more than a thousand hours creating the FloodNet dataset. Input from first responders and the Federal Emergency Management Administration (FEMA) guided their work, plus mentoring from faculty and more advanced students. Team members labeled each pixel in images to teach the artificial intelligence (AI) system what a flooded road or damaged building looks like.
“It was a very difficult and challenging task, but we are the first in the world to prepare this sophisticated dataset,” Rahnemoonfar says.
Already, German authorities have requested (and been granted) access to the FloodNet set of labeled images to help them respond to recent catastrophic flooding in the country. The FloodNet dataset is the first and only dataset of its kind that can be used for training AI systems in the wake of natural disasters.
Saving time, saving lives
With the new START funding, Rahnemoonfar’s team hopes to improve the algorithm that evaluates how much damage buildings have sustained, on a scale from none to total destruction. Rahnemoonfar also hopes to build an interface where a person can ask the AI system a verbal question and get a useful answer. Her team is developing this ability for data already collected, “but when a new hurricane happens, we hope that while a drone is flying you can ask and get answers to these questions in real time,” she says.
To save time—of the essence during disaster response—Rahnemoonfar’s graduate students are working on adding a layer to the algorithms that would allow the AI to learn on its own, with a minimal training dataset.
“Now that we have developed this AI system, for any new hurricane that happens, we don’t need to label data again,” she says. “With the algorithm that we are developing, with very few labeled images we can get insights for any new hurricane.”
A ripple effect
The START program has a ripple effect beyond UMBC faculty. Many of the awardees involve students in their work, enhancing the students’ UMBC experience and helping set them up for success later on. Students in Guimond’s group, for example, get first-hand experience with NASA scientists and facilities.
Also, by helping researchers hone their projects and, as a result, future proposals, START increases their chance of success with applying for larger grants. Like Guimond, Rahnemoonfar is already looking to use the START support to lead to bigger research awards.
“The START program enables UMBC faculty members (and with them, UMBC students) to move in new directions with their scholarly work,” shares Don Engel, associate vice president for research. “We are proud of the success past recipients have found in turning their proposed ideas into longer-term initiatives.”
Banner image: Hurricane Matthew bears down on Haiti in 2016. Image by NASA Goddard Space Flight Center.