Operations Research in Storm: An Interview with Dr. Clint Dawson

Shreya Gupta

Interviewee : Dr. Clint Dawson

Interviewer :  Shreya Gupta

Can you briefly point out the series of events undertaken to simulate the results when a storm approaches and how this information is passed on for evacuation measures?

Every six hours the National Hurricane Center releases a forecast based on nine different weather models. They use the results of these models to come up with what is called the “consensus forecast.” This consensus forecast is released at 4am, 10am, 4pm and 10pm. So every six hours you are getting an updated forecast of the hurricane track, intensity, wind speed and size.

We take that information, process it, and run a storm surge model based on this forecast. So, first this data gets translated into a mathematical model of wind and wind field. This is just a very simple vortex-shaped wind field because we don’t have a complete description of the hurricane and the event. Post this translation of data, we run our storm surge model out several days depending on the length of the forecast. This information is post-processed, and the results are posted on the Coastal Emergency Risks Assessment website, cera.cct.lsu.edu. Anybody in the world can look at the results on this website.

For Harvey, in fact for all hurricanes this year, the simulations were done here at the Texas Advanced Computing Center (TACC) in The University of Texas at Austin. So during Harvey, which was the first major storm this year, I helped get the forecast system for Texas in place. This is a big collaboration between many universities, and making sure that we have a forecast system for Texas was our responsibility. So when Harvey happened — which by the way happened very quickly! On Wednesday morning it was barely a tropical storm, by Thursday it was a full blown hurricane and by Friday it made landfall — we did not have a lot of time to get our act together, but we did and got things running. We started picking up the forecasts and producing results.

I was also producing results on my local machine for the Texas State Operations Center which is run through the Division of Emergency Management, a part of the Texas state government. They are responsible for working with all the government, county and city officials, and whoever is involved in deciding where and when to evacuate. They also have to figure out where to stage the resources and supplies, like trucks, first responders, etc., so that they can go in immediately after. They also have to work with the Department of Transportation to figure out which roads to use. So our storm surge model helps in these decisions. Because we model the flooding, we can tell which roads might be covered or clear, or which bridges may need to be raised because they are too close to the sea level. In other words, which evacuation routes to use and which routes to use for the first responders. This is the very first level of information that they are interested in.

After the storm is over, we work with people to go in and collect the data we use to physically model the storm as it really happened. This helps us analyze how well our model performed in terms of having the right physics in the concerned areas, identifying things we may have missed, etc.

Which other universities are collaborating in this project?

The collaboration is divided into the Gulf Coast and East Coast. The Gulf Coast comprises of Texas and Louisiana State University (LSU). LSU does storms that are going to impact Louisiana, Mississippi and Alabama. We don’t work so much with Florida since they have their own system. We also work with the University of North Carolina. They do most of the East Coast.

How long does it take to run the simulation?

About an hour.

Okay, that’s quite fast. So that should allow you to cover multiple scenarios since that’s not very long?

Yes, we try to.

You mentioned that you get the weather forecast data from consensus weather report released every six hours by the National Hurricane Center. Is there any other kind of data that you need and utilize for this analysis?

Yes, we need to have a model ready with all the data for Texas, for example the bathymetry data and land cover data. This data already has to exist because you can’t obtain and incorporate it on the fly. This data has to be ready to go pre-storm.

How well do the simulation results confirm to reality?

Our simulations did well because people went in right after the storm to collect all the data. Our simulation results were found to be pretty close, and they were a lot better than what was released by the federal government.

And how do you define pretty close?

Pretty close is subjective, but I would say we were within about half a meter to a meter of the surge in most cases. For example, we would predict say five meters of surge whereas the federal government predicted ten meters.

So given that there is an error interval of being within half a meter to a meter of the surge, what are the risks associated with the error, even if it’s just by half a meter?

A meter can certainly make a difference. If we are down to the level of a particular house or street that is going to get flooded, then whether you are off by a meter is very important. But at least in the early stages of the hurricane, they want to get an idea of the range. The codes are never going to be exact in any situation because we don’t have the exact wind data to start with. For example, one thing that happened in Rockport and Port Aransas area, from what I understand, was that a beach was eroded, and that caused a lot more surge to come in than anticipated. And it came in earlier than anticipated as well. And this was unanticipated because our code does not account for beach erosion, the reason being that it’s just hard to know exactly where and how much erosion is going to happen. Erosion is sort of a binary thing, either the beach stays in place and does its job, or you could have a complete failure. For example, an entire section of the beach could get washed away, and that would create a new channel for all this water to get in and cause even more erosion. So you can have this sort of catastrophic failure. This happens a lot in hurricanes, and it’s hard to predict exactly where and when it will happen.

I know you already mentioned not having exact wind data. What are the other bottlenecks you face?

Yes that’s one bottleneck we face, and we could probably do better if we used, say, a very highly accurate weather model, if it existed. The existing models are getting better at predicting and modeling hurricanes, but they are still not perfect.

We are sort of limited by what we think the coast looks like at a particular time. For example, this is 2017, and we actually built our model of the coast probably six or seven years ago. There has been a lot of development and changes since then, but we just don’t have the money and resources to go back and redo everything every year. It’s just too costly. So we just have to go off of older data, and sometimes very old data.

How much does it cost to redo data collection?

A lot, millions!

So, your current work is current state of the art and cutting edge. What are the future directions for your research?

Well, one of the things is that we don’t model rainfall. Rainfall is an important component of a hurricane. There are other codes that do that, but they are not integrated with ours. We also don’t model sediment erosion, as I mentioned earlier, so at present we are not doing a good job of modeling how the coast changes due to a storm. We are not modeling any sort of 3D effects. We are just doing 2D.

And what do you mean by 3D and 2D effects?

For example, to model any sort of sediment, you would need to do scouring. You would need to have a vertical component of velocity. If you have scouring, then you have velocity coming in, you get entrainment of sediment, and then it gets deposited again. We can’t account for that right now because it’s very computationally expensive. Running a 3D model is about ten times more expensive than running a 2D model.

That’s interesting. Thanks for sharing all this information about your work with me. Is there anything about your work that I missed and you want to highlight?

Well another thing that we are concerned about is climate change and sea level rise. As we go into the future, we don’t know, for example, how different the coast is going to look in ten, twenty, thirty or forty years. Even conservative projections look pretty drastic. But sea level rise varies in different places, so you don’t know how much it’s going to be even in, say, Texas versus Louisiana versus other places.

We also don’t know exactly how future hurricanes are going to look and whether they are going to be stronger. This year was kind of eye opening in that respect in that we had so many strong hurricanes, and they maintained their strength for a long time, which does not usually happen. That was probably due to global warming since the ocean was very warm over very large distances.

Okay, and how did global warming cause the hurricane to maintain its strength for a longer period of time?

Because the hurricane gets it energy out of the ocean. So if the water is warm, it gets more energy and can sustain itself for longer. If you look at Harvey, it just blew up from a tropical storm to suddenly this Category 4 and then this huge rain bomb! And from what I understand, because I just got back from AGU (American Geophysical Union), the thing about Harvey that was unusual was that even after making landfall, it kept its rotation. So it sat over the coast, and it never broke apart. That’s why you had these rain bands that were coming over and over and over again, and that’s why the Houston area got so much rain. So Harvey was kind of unusual in that it just sat there and spun and never broke apart. But I understand that we don’t get anything like what they get in India or like the Pacific storms. For example, in Taiwan they get tens of meters of rain. Sometimes it could be ten and twenty meters of rainfall. That’s astonishing! And we got fifty inches, which is about a meter and a half.

Wow, yes! I guess the impact of the storm also varies from region to region depending on the infrastructure.

Yes, exactly.