One of the biggest challenges facing a governor when a dangerous hurricane is barrelling towards their state is to convince people to evacuate. This happened in North Carolina last month as the state grappled with Hurricane Florence.
Flooding threatened resident Dawn Davis’ house in the midst of the storm, according to a story in Time magazine by Maya Rhodan. Sandbags placed along a local river had failed to keep the water at bay. Still, Davis wasn’t ready to go. “I guess if it gets a little bit more higher, we’ll all end up leaving,” she said. “I’m just trying to put it off.”
This challenge can be harder when predicted rainfall, storm surges, or wind damage end up being less drastic than expected, making people are less likely to heed future warnings. At the same time, predictions that a storm will be less forceful than it is can be equally harmful to residents.
To better understand and predict how such storms will track and how damaging they may be, you need a better understanding of all of the ingredients involved, says Catrina Nowakowski, a graduate student at the University of Rhode Island’s Graduate School of Oceanography (GSO).
Nowakowski, along with GSO Ph.D. candidate Kevin Rosa, spoke at the Bay Informed Series in September about the importance of numerical models in preparing communities for potentially devastating storm events.
She likened building models to baking cookies, comparing the mixing bowl as the framework in which all of the ingredients are added.
“All of the steps that go into making cookies are similar to the same steps of putting models together that are responsible for us getting our weekly weather forecasts,” she explained.
But instead of flour, eggs, and butter, hurricanes require heat, moisture, and wind. After that, a variety of variables, such as wind speed, water density, and geography are plugged into the models to predict what kind of storm it will be.
“We build each [variable] as one puzzle piece and then put them all together so we can ask different questions with our models,” said Nowakowski, explaining that every model is built to answer a specific question, and includes the appropriate physics to answer that question. One model could look at how a storm may track, while another may look at how strong its wind speed is or how much rain it will produce.
“Certain models are better at predicting certain things…it’s still up to human [interpretation]. We use [models] to help make our decisions and forecasts.”
Storm surge models, said Rosa, are used to predict which areas on land will be inundated at various levels and are essentially the result of combining hurricane and ocean models.
“This is important because the main cause of death from hurricanes doesn’t have anything to do with wind. It’s drowning, both from the storm surge and river flow,” he said.
Another type of model that Rosa discussed is the Regional Ocean Modeling System (ROMS), which he says is popular around the globe because of its flexibility to evaluate a variety of variables such as sea ice and biogeochemistry. It’s used in Narragansett Bay by researchers, like himself, who are looking at circulation patterns or the dynamics of phytoplankton.
“You get to create whatever scenario you want even if it hasn’t happened yet so you can help people prepare, but you can also learn a lot about a system by running the model and changing the knobs,” he added.
One of these scenarios includes potential oil or chemical spills in Providence resulting from storm damages. Hurricanes Katrina and Rita combined, Rosa said, dumped millions of barrels of oil into the surrounding neighborhoods––more than the historic Exxon Valdez oil spill in 1989.
To get a sense of what such a spill might look like in Providence, Rosa used ROMS to simulate conditions of the bay after a hypothetical hurricane. The model provided a 50-hour snapshot and showed a significant southward progression of a spill well beyond Greenwich Bay. To determine which factors drove that progression, Rosa started by looking at one variable––water density.
A 3D model applying a cross-section of temperature, salinity, and density was initially used to account for the physics of an estuary––where there’s a gradient of lighter, less dense freshwater that meets heavier, denser saltwater. When those elements were removed, the model showed a spill in Providence Harbor would be relatively contained to that area.
He explained that this one variable highlighted the need to consider gradient densities of an estuary. It is important, he said, for understanding the flow of a spill and potential effects.
“[An estuary] creates a two-layer flow. Without this, a spill in Providence wouldn’t make it’s way as far south from Providence,” said Rosa. “Density has effects.”
The quality of a model is dependent on the type of question being asked. Is it a specific or broad-based question and is all the appropriate information being accounted for?
“You want to make sure you include enough [information] to answer your question but not so much that it takes hours to run your model,” she said, explaining that adding too much information could overwhelm the model and not provide a clear answer. “Everyone is working together in the grand scheme to build these models, and it takes large teams, lots of resources, and different types of understanding to put them together.”
The scientific community, she said, is continuously refining these models with the ultimate goal to build faster and higher resolution models that have the most accurate predictions whether it’s for hurricanes or oil spills.
The Bay Informed Discussion Series is sponsored by Rhode Island Sea Grant in partnership with GSO. This series is held every third Thursday of the month at 7 p.m. at the URI Graduate School of Oceanography Bay campus in Narragansett. These events are designed for the community to get involved and learn more about research at GSO.
[divider style=”solid” color=”#eeeeee” width=”1px”]
– Meredith Haas and Monica Allard-Cox | Rhode Island Sea Grant