Walter W. Piegorsch is the Director of Statistical Research & Education at the University of Arizona’s BIO5 Institute. He is also a Professor of Mathematics, a Professor of Public Health, a Member and former Chair of the University’s Graduate Interdisciplinary Program (GIDP) in Statistics. Dr. Piegorsch’s research focuses on data science and informatics for environmental hazards and risk assessment.
Episode Description
When planning for potential disasters, we often focus on hurricanes that might ravage coastal areas or tornados and droughts that strike rural parts of the Midwest. But researchers are also working to uncover the vulnerabilities faced by urban areas and that’s the focus of this episode of Stats and Stories
+Full Transcript
Rosemary Pennington: Planning for potential disasters we often focus on hurricanes that might ravage coastal areas or droughts and fires that might strike rural places, but researchers are also working to uncover the vulnerabilities faced by urban areas, that's the focus of this episode of Stats and Stories where we explore the statistics behind the stories and the stories behind the statistics I'm Rosemary Pennington Stats and Stories is a production of Miami University's Department of Statistics and media journalism and film, as well as the American Statistical Association. Joining me were our panelists, John Bailer Chair of Miami statistics department, and Richard Campbell Professor Emeritus of media journalism and film. Our guest today is Walter Piegorsch, Piegorsch is the director of statistical research and edge gal, Walter, is it bio five Institute. Piegorsch is the director of statistical research and education at the University of Arizona's BIO5 Institute. He's also a professor of mathematics, a professor of Public Health, and a member and former chair of the University's Graduate interdisciplinary program in statistics Piegorsch His research focuses on data science and informatics for environmental hazards and risk assessment. He recently co-authored a significant article about the threats facing cities today, Walter. Thank you so much for being here.
Piegorsch: Oh, you're welcome Rosemary I'm looking forward to this.
Pennington: For those who haven't read your article yet, could you describe one of the big takeaways, as far as sort of city vulnerabilities goes,
Piegorsch: Sure. I think the biggest takeaway that we had our message, maybe our narrative was that place matters, and cities. These are large urban areas, we use the term city very loosely here. But place matters. If a city's vulnerability to any kind of hazard, artificial nature really changes depending upon a number of important factors that each city has and each city's factors are almost always very, very different.
Bailer: So what am I, you know, I want to take you back in time a little bit to when you started on this project, what was it that inspired you to start working on this.
Piegorsch: The Great question, John. So my collaborator we were sitting at a university professors were sitting at a university committee that we got told to go to. I just sat down and turned to her and I said hi I'm walking she said Hi I'm Susan, what do you do Susan I do geographic risk assessment. And I said, Oh, I do statistical risk assessment, and I'll stop here, the rest is history. So one day we were sitting in a conference room of sorts, and we were working on the data that underlie these 132 cities that are part of our data set. And it just kind of dawned on me that the kind of risk assessment that I do could be applied to those data, and that would be pretty novel, and it's called a benchmark, risk assessment, John, in fact, you and I published on that recently. So it's something that I just looked at it and went, Holy cow, you could take this and it's completely different than what the people do in benchmark risk assessment, they're open toxicology cancer risk assessment, but pick it up, drop it back down and all of a sudden you're looking at terrorism vulnerability. Vulnerability in cities. And just the light bulb went off.
Campbell: So, to go back to the significance article that Rosemary mentioned so say I'm a journalist, which I used to be a long, long time ago, and I'm looking at your significance article I'm going to write a newspaper story about it and my lead would be something like, data scientists and mathematicians predicted the January 6 terrorist attack. That would be that. What would be wrong with that or is that, does that get it.
Piegorsch: So you're talking about the insurrection at the Capitol on January 6
Campbell: Yes, yeah.
Piegorsch: That's, uh, you know, I never thought of it quite that way. Honestly, maybe we did because it isn't Washington DC, listed number one, number one on our list, yeah 77% probability is that right, am I gonna have to look at the number of myself, to be honest but Washington came in pretty high and it came in pretty high for a number of reasons. I would not have said that if someone asked me that on January 5 Yes, but at the same time, come to think of it, boy we did, didn't we, pattern our backs. Yes. That was the first thing I thought of when I saw that list and I said oh my gosh, talk about risk and vulnerability and vulnerability they got this right. So, you're right.
Pennington: So how are you defining a vulnerability and how is that different from risk.
Piegorsch: Oh that's a, it's a great question. It's also very technical. There's multiple definitions of what scientists will call risk, there're multiple definitions of what scientists would call vulnerability risk for us, boiled down is essentially a probability of some adverse event vulnerability is a little more complicated and much more multi dimensional. It's the, it's a city or two locations of features that make it vulnerable to some sort of adverse event. And what we did in this is the fifth of six papers that we've written on this over 14 years, what we've done has evolved, the, the, the imaginations and then technology to really kind of give us a better understanding of where vulnerability is and then how it relates to things like risk.
Bailer: I thought it was really, really cool that you're, you're taking these ideas of, you know this, this physical hazard vulnerability built environment vulnerability and social vulnerability, and that you're integrating these into a, into a measure that ultimately is is kind of your, your signposts for whether you worry or not in a community, you know that's going to be driving your, your kind of prediction of what you think's going to happen. Could you talk about each of those in turn and give an example of, of social vulnerability or built environment vulnerability and physical hazard vulnerability for us.
Piegorsch: Sure, let me start with the social which is why we call it SoV. That was the earliest metric that was developed by my colleague Susan cutter, and it's, it's, it's, it's a measure it summarizes socio economic and demographic characteristics that that influence a community's vulnerability to well anything but we focusing on natural or artificial hazards. What's a good example you know my favorite example of the SoV metric is the percentage of single parent households. And that's not a favorite because it's, it's fun, it's actually unfortunate statistic, but think about it if you have a large number of single parent house, if you've got a major event that occurs terrorist attack, hurricane, flood, and you have a single parent that single parent is almost certainly not working. If kids are at home somehow and maybe you're lucky enough to have some sort of supervisor, but oftentimes you don't. And if that parent gets taken out in some unfortunate fashion by that hazard. It's typically a woman so she's going to be in the hospital in a coma. Who's taking care of the kids, who's getting them out of there, if this flood is going to wipe out that street that city block. So that's kind of, there're many factors of social vulnerability, but as I say, that's one of my favorites, not because it's pleasant but because it's a really good marker that gets all of us thinking about what the story is behind these kinds of numbers.
Bailer: So can you follow up with like some of the ones that sort of jumped out for you for the built environment, or for physical hazard. Sure, so
Piegorsch: For a built environment. What's another good one? Oh, here's one I like a lot, it's, this is kind of subtle actually the easy things to say are bridges and tunnels. How fast can you get people out of a hazard event has this event in your city, but one of my favorites on this one as well, is what if you are doing the right things with your first responders. They have the same electronics, do they have the same communications, do they know what dial to turn on the radio or transmission device that the ambulance driver can tell the hospital who's coming in where and the fire department can tell the sheriff. Oh wait, we need this road open, and some cities actually have these kinds of things integrated others don't, and they're a little behind the curve on a bit, but they're again if you had this integration if you've got some sort of important hazardous event that occurred in your location. If you've got that integration in that system, a lot of things go a lot faster. A lot of people don't lose their lives or don't get hurt badly. And so that's an important thing, that's why I am getting one of my favorites to have a built environment. So,
Campbell: I'm picking off the infrastructure bills that politicians are considering they've been considering this for 10 years and we haven't seen much. Do you have some frustration about your work getting transmitted to guys that actually make the decisions about whether to improve the build environment I mean, I think, I mean that's what I'm thinking what I'm thinking of the, the risk to Washington on January 6 Then you know are they paying attention then our, I guess, you know, thinking of this infrastructure bill coming up work that you do, seems like should should be in right in front of politicians right now are making decisions about this stuff.
Piegorsch: Yeah, and so my frustrated well personally kind of has been around long enough to know that, translating even any kind of good science into a call it a political or decision making process takes time. Yes. So, I like to think that there's things that we do that impact. Move forward. My favorite example of this is one of the early works had Boise, Idaho, identified as a high vulnerability city, for various reasons, I could go into July, but it was also kind of sticking out there one because there's not a whole lot of big cities, Boise. And two, it just turned out there weren't a whole lot of high vulnerability locations, west of the Mississippi. Oh, so that got picked up by the news media. And I got moving on it and I got calls from newspapers and radios and, you know, local TV and everything. But then one day I got a call from the Boise office of the FBI.
Pennington: Oh my god.
Piegorsch: I have to tell you, it changes the entire tenor of your day. When you pick up the phone. Hi, this is the FBI, they were doing their job and it did very very well, I must say, they wanted to know on What's this new thing we're hearing about Boise what's going on here. Oh, you know…
Bailer: Now you’ve got to tell us about Boise now well, you can't leave us hanging here, pal. Why why is, why is Boise higher up on the vulnerability,
Piegorsch: I actually have the I pulled out the numbers in dissipating. You asked about the three components we have the social component, we call it Sophie we have the built environment component we call it Betty, we had this physical hazard component has V O, to answer your last question, John, quick example of the physical environment is, how often do you see hurricanes come through. How often do you have to deal with, with floods, even if you're not in a hurricane prone city tornadoes in the in the in Tornado Alley, those factors will drive up a city's hazard vulnerability, and in particular for Boise, they experienced two very significant events in the span of our study, a major flood which really hammered I'm pretty good, and a wildfire which really cut them up pretty good. And created large losses of property and some even some casualties. And that's the kind of thing that drove up that drives up our hazard vulnerability index. So they're just way high up there on just basically two very subtle kinds of vulnerabilities, but boy is these about. I should look this up. I forgot to do that 15 miles I think away from a very large dam. And if you wanted to take out this is the terrorism analysis we did if you wanted to basically level Boise, take out that dam. Now, the Boise FBI kind of knew that. Oh yeah sure we understand and good for them and I'm sure they're doing great job of the security of that they're also bigger than them I don't take it out, it's gonna be. This ain't no firecracker you're gonna use. And then the wildfires as well they had all you need is one crazy little thunderstorm in the middle of summer hits with one lightning bolt, and that can take out a big chunk of town if you, if you have that high vulnerability to wildfires. That's going to be a major factor, and we're seeing more and more that with more of our cities as climate change is affecting them.
Pennington: You're listening to Stats and Stories and today we're talking with Walter Piegorsch about urban vulnerability. I wonder if there are if there is data, or there are findings from this work that you've been doing that really did sort of catch you, you know off your guard or that you found particularly compelling or surprising.
Piegorsch: All of it. To be honest. We had no preconceptions when we were doing this, like I said it was a light bulb event that said hey you know we can take this statistical method over from from this cancer biology method and translate it, very efficient, effective effectively to this terrorism issue and then we've since then done flood analysis and we've got further work we're doing on other kinds of hazards as well. It really does plug and play very very quickly. Once you get once you fine tune it to be appropriate. so no preconceived notions which is I think the big step was the maps that we produce then you if you've seen the article you've seen these maps, because they really do visualize, far more than I expected but not in the sense that I'm in any sense upset about or disappointed about. They really do visualize the kind of hazard scape that we're seeing very are very proud of, I've got to say I realize how painfully ironic it is for us to be talking about color coded cartographic maps on an audio pod.
Bailer: Well we use colorful language.
Piegorsch: Sometimes you listeners out there, you got to get the, get your recent copy of Significance
Bailer: Buy one for your family.
Piegorsch: So on a really practical level, if I'm advising my, my son or daughter. And they, they tell me they want to move to Norland Should I, should I, what would you what would you advise your, your children if you have children I, because New Orleans is like number two on the terrorism list and I think number one on flooding, 99.8 I think or 98.0 Probability of flood damage. I've been in New Orleans, and it was this before Katrina, and it's so beautiful and lovely, yes, I want to say positive things about it first. But it's hazard escape is just, wow. And that's it's got social vulnerabilities that actually were exposed during Katrina. It's got built environment vulnerabilities because it's sitting on a river, and a big river, and it's forever that floods and gets hit by Hurricane so it's got physical hazard vulnerability, it's just it's just, it's hitting on all cylinders, but these that are on cylinders right. I don't want to do that. So, I don't want to be flippant and say don't move to New Orleans, because it's a beautiful city, But there's a lot of vulnerability in that town. It's, among other things, it's when we did our flood analysis it's. Keep in mind, New Orleans is below sea level. Yeah, yeah, and sea levels are only going up these days. So, I don't see our vulnerability analysis, helping them out any in the immediate future.
Campbell: Let's hope we get that big infrastructure bill this year though, a lot of money into New Orleans because it is a good city,
Piegorsch: in fact that was one of our messages has been one of our messages all along. Yeah, and there are places where you know you don't get a pump that much into the, the economy, I'll say the. Let's take the built environment as a good example. You don't have to do too much. They've got it all figured out, they're doing a great job. There's other places that are well sitting below sea level, maybe we've got to really understand flooding in that area, far better than we do right now. So, yes, I would really want to. What's another good example I would want someone to say in Norfolk, pretty high on our list as well for multiple reasons and multiple outputs. Inputs, to also be beefing up things like it's anti terrorism is it's it's built environment, how do you get out of that Norfolk is actually Norfolk and a couple associated cities located right nearby. How do you get out of there because they're sitting on a I guess a river, and it's a naval base so of course, motion coming up boat traffic going down. It can you build that, can you create that built environment that really is much more protective of the location, and they may be doing and I haven't followed up with what's going on in Norfolk, to be honest but these are the kinds of things we thought were we felt our we feel are our one of our biggest messages.
Bailer: So you know, if you've had a really, a lot of work invested in defining these vulnerabilities, which is a key part of the story, but there's also some interesting questions about trying to define what is a terrorist event. Yeah, you know, and then all of a sudden I thought, well okay, you've got some quantification here, and it also makes me think that you only. You don't know about the censored observations. You know the ones that are the kind of the near misses that might have been in place that maybe if there was if you had had some access to some other information if that would have happened that would have told you more, but that's, you have to deal with the observed data that you have. So, how did you operationally define this kind of terrorist event for the models that you invent, you developed.
Piegorsch: Yeah, we actually seeded that to an existing database called the Global GTD global I think it's global terrorism database. And then connected that up with other databases that just identified a numbers of events in these 132 metropolitan areas. And in fact this is my one of the geographers working on this, they just basically did all that, that data compilation and curation. And then we kind of fine tuned it a little bit and said you know, terrorist events are just one thing. But casualties from a terrorist event that's that kind of raises the bar of it. Let's use let's identify that let's target that is the important thing So did a city that had a terrorist event report casualties or deaths. Okay, that's basically what we did based on these, on these existing databases.
Campbell: I was a little surprised this is a follow up question. You know I, I was when I was looking at your list of these top 10 us metropolitan areas. I was, I was kind of thinking. I was expecting to see New York City as showing up higher in the list or possibly Atlanta, or possibly LA. So there were sort of, you know, the cities that I was, I think that some of them may have shown up a little bit later than the top, you know, they didn't make the top 10 But I was just I kind of my, my bias my A priority, intuition, had them higher on the scale.
Piegorsch: Yeah, and like I said we didn't have any preconceived notions ourselves so a lot of this was kind of interesting and surprising we thought that the knowledge discovery here was just off the charts. Yeah, for us, and what wisdom we can derive from that knowledge discovery of course is the next step. But sure, New York is actually pretty high, it just didn't make the top 10 on the, on terrorism, it did make the top 10 on flooding is that right, yeah. And it's New York Newark, I should add it, we these these are, these are conglomerations of multiple locations in a given city or urban area in the United States. In fact, it sounds like 132 cities but I think it's over 600 counties into these constructed areas. So you, but did I answer your question there John is so why not la Why not Atlanta. Yeah, they have certain levels of vulnerability but other places had more and that's kind of Rose them up on the list.
Bailer: So my follow up my follow up to that job would be. Should we move south. We're in a red low we're in the red zone for flooding in the orange zone for terrorism So
Campbell: John, you know, you didn't mention tornadoes either. So you know, Oxford, I think we're, I'm surprised Where's Oxford, Ohio on this list, Walter, you know, certainly this is, this has got to be a major,
Piegorsch: I would have bet a priority that when you folks saw the maps, your eyes immediately were drawn to South Western Ohio. That's right, back when I give a talk on this, it's the statistics yeah the students or the or the audiences. Yeah that's nice, that's nice. Then I'll pop one of these maps up, and everybody's eyes open, even the ones that were sleeping in the back. Yeah. You can see them all. I have learned and watched this. You can see them all look somewhere their heads tilt their eyes open, and they're not looking at the same spot because you're looking maybe where they are right now maybe where their parents live maybe where their grandmothers retired to or whatever the place that's important. Right, yeah, that's important to them. Yeah. And so, for the record. Let's see, I can't tell you, Oxford Exactly. Well, cuz you're kind of small, and Hamilton your county seat, yeah we're close to Cincinnati and Dayton so the closest Hamilton closer said is the County County that Hamilton is the seat of is included in our definition of the Cincinnati Metro. And that probably doesn't surprise you folks because there's pieces of Indiana that.
Bailer: Northern Kentucky. Yeah, Cincinnati area so if you are in there, which, being kind of coalesced in with, with the rest of the CNC, the larger sense image found area.
Pennington: Now, I do like Richard's been trying to get you to turn into a lobbyist, a life coach.
Campbell: So far, you already do risk assessments.
Bailer: So you know what you've built these, these, these really interesting predictive models. So the natural question is when, anytime we're doing modeling, how do you know the models work. You know what's what, what did you do I know you did some of it. You did some stuff with with out of sample prediction, but can you talk a little bit about kind of checking and investigating whether or not the models are doing what you claim.
Piegorsch: Yes, and that's a good thing any good statisticians should be thinking about is just handing out a number is not enough. You got to look and see what that number means in fact, I just take a little tangent here for you folks in any class statistics class like statistics class, I teach, I, I try to always put this one line forward there's one thing that I hope the students will get, and that's that no user, no, no producer no reader of any statistic should ever be satisfied with a point estimate, never be satisfied with the point estimate, always insist we the only to producing or discovering a some quantification of the uncertainty. And it's kind of funny as a student wasn't my student, but it was, it was doing his PhD defense and he actually probably remembered what I was saying in my, in the classes I took he said yes and I'm not satisfied with the point estimate.
Piegorsch: But we did this yeah it's in the significance article, John, we actually just asked, okay, what do we predict a city's going to be. And we had data from 1970 to 2004, and we looked at, okay, what did we predict and what happened in that in that 35 year period, but then we were able to go back into that same database I mentioned earlier, and see what's happened in since 2005 It ended in 2018 for the data we had, and then ask, okay, what would we have predicted for those same cities, and what happened since then. And we saw pretty good predictive responses Predictive analytics always kind of happens your test, your, your training data will kind of give you high numbers you don't want to over train, but then go in, going against the test data set and see how it does and it did pretty well. Certainly above what we would call the coin flip, a metric which is just if it's 5050 You might as well flip a coin, you don't need statistics for this.
Pennington: Well that's all the time we have for this episode Stats and Stories well thank you so much for being here.
Piegorsch: Oh, it's been my pleasure. Thanks.
Bailer: Thanks. Well,
Pennington: Stats and Stories is a partnership between Miami University’s Departments of Statistics, and Media, Journalism and Film, and the American Statistical Association. You can follow us on Twitter, Apple podcasts, or other places you can find podcasts. If you’d like to share your thoughts on the program send your email to statsandstories@miamioh.edu or check us out at statsandstories.net, and be sure to listen for future editions of Stats and Stories, where we discuss the statistics behind the stories and the stories behind the statistics.