Statistical Relationship Status? It's Complicated: Geohazards with Alex McKenzie-Johnson

0:00 Rhett On this episode of Pipeline Things, we talk about statistics, and we talk about how statistics never lie, why shark attacks are directly correlated to the amount of ice cream consumed in the United States, and also... what all of that has to do with landslide and hydrotechnical events in the PHMSA geohazard database. It is another great discussion with a return guest, the Dirt Merchant. Thanks for joining us, and we think you'll enjoy it. 00:44 Rhett All right, welcome to this episode of Pipeline Things. It is good to be back as we are in the PPIM series, where we're going through papers that we identified at PPIM 2026. Always a great conference and a great time in Houston, and we're going to be discussing with the findings of the authors. 01:01 Rhett Today's author is, again, another returning guest that I think the audience is going to love. But before we get into that, we're going to be talking a lot about statistics and data, Chris. And before the show, we were talking about some of our favorite statistics. 01:17 Chris I got one I like. So, we tend to not eat a lot of ice cream in the colder months because somewhere we found that there's a correlation with ice cream sales and shark attacks. So we also like to go to the beach in the winter months because people tend to order less ice cream in the winter. So therefore, if there's a correlation between ice cream sales and shark attacks, ice cream sales are down in the winter. Therefore, shark attacks should be down in the winter. So we're planning an October beach vacation. We're trying to make sure we don't get bit by sharks. 01:51 Rhett I mean, I just got back from the Bahamas in December, so that makes sense. Yeah. Chris Did you guys get bit by sharks? Rhett No. Chris Did anybody else? 01:54 Rhett I swam with sharks, with Caribbean reef sharks, when I got my scuba certification. And I can attest to the validity of that. I swam within feet of them, and they did not attack me. And there was no ice cream on the dive boat. I am at least... 37 % positive that had there been ice cream 02:22 Chris 37 % not 33 % not a third Rhett Look let's be quite clear I am 37 % certain with a 20 % confidence interval that if there had been ice cream on the boat I may have been attacked by a shark more likely you would have it's possible it's absolutely yeah Chris So yeah if you guys don't believe it check it out the correlation between ice cream sales and shark attacks do correlate yeah 02:40 Rhett And there's some other really good correlations stay out of warm water Did you know that the popularity of the first name Gerard correlates with air pollution at Anchorage? Gerard? Gerard. G -E -R -A -R -D. Almost one to one. Look how good that correlation is there. Oh, God, that is good. Yeah, look, I got another one for you. This one's even better, and this one might have real meaning for you. The popularity of the first name Sarah with an H correlates with Google searches for learn Spanish. 03:08 Chris Sarah, are you learning Spanish? Rhett Both have trended down since 2004. So I don't know what we take. Chris What happened to Sarah in Spanish in 2004? Rhett I don't know. But since that time, the use of the name Sarah has trended down, as has the number of Google searches for learning Spanish. Yeah. I mean, but really, there are amazing correlation statistics out there that bear absolutely no resemblance to anything at all. Right. But what's fun anytime you're looking at data. is I feel strongly you have to ask the right questions, right? To blindly look at a chart and derive the notion that the name Gerard correlates with air pollution and Anchorage, or to come to the conclusion that I was somehow safer diving in December with no ice cream on the boat than I would be diving in, by the way, when I return diving with my kids. In June oh so you better you better check the season's question don't let Morgan know this because I didn't think what if the kids eat ice cream before they die This is real problems. I'm not going to tell my kids about this statistic. 04:19 Chris Makes you wonder if popsicles count or is it just ice cream? Rhett But you got to ask the right questions of the data. And that's what today's paper is all about. 04:26 Chris And not all trends make sense. Rhett Yep. So without further ado, I'm sure our guest has been listening intently in that conversation. Probably has really strong opinions about both the name Gerard and the name Sarah. Our guest is the returning dirt merchant. So Alex, I want to say welcome back to the show. It's good to see you again. How have you been? 04:44 Alex It's good to be here. I've been well. Looking forward to the conversation. Rhett Yeah. So, you know, I don't think that the audience needs to reintroduce you per se, but I do think there have been some changes since the last time we had you on the show. You care to talk about those? Drum roll. Oh, yeah. Drum roll. Drum roll. 05:03 Alex So since I was last on the show, I went independent, formed my own company, TriCoast Geoservices, exclusively focused on pipeline geo hazards and helping operators to identify hazards and come up with plans to manage them and all that kind of good stuff. 05:14 Rhett Well, congratulations. Welcome to the team of small business owners. Alex Well, thank you. Thank you. Rhett Fantastic. So the reason we're having you on is that you are the first author, the primary author on a paper. I was a co –author on the paper and I was happy to review it, but look. to the audience and to, to Alex, Alex did the lion's share of the work on this paper and really did a great job. The title of the paper, can you give us the title? I actually, I was looking at it right hand and I didn't have it off hand. This is what happens when you live ask the title. Look, there's Chris. 05:49 Alex Yep. Yep. So the audience bears with us a deep dive into the PHMSA significant incident database and what it tells us about trends in pipeline geohazards. 05:59 Rhett Right. And that paper says it all. You went into the PHMSA significant database with an idea to learn what you could about geohazards in the United States. And I'm going to say in particular, the trending of the geohazards in the United States. So tell us what started you on this journey. 06:14 Alex Yeah, so I think most of us who've been involved in this space can really attest the last decade or so there's been a lot more interest around geo hazards for pipelines, right? So whether you look at like conference presentations, operators who have in -house geo hazard programs, spend it on geo hazards, guidance document, there's definitely been a lot more interest around the subject. And so as someone who's been practicing in the space for a long time and kind of interested in... out what are the reasons why we're seeing so much more interest. I think there's a lot of theories about why that might be, but one of the things I was interested in is what does the data actually tell us? Are geo hazards happening more often? Are we seeing more incidents? Are we seeing more incidents in certain types, certain places? And we have the PHMSA of Significant Incident Database available to us. So for those of you who don't know what that database is, there's a regulatory requirement for certain kinds of incidents. that is reported to PHMSA. And these are usually incidents involving some kind of injury, some kind of property damage, or a significant release of a product. And then when those incidents happen for PHMSA -regulated pipelines, they have to then report it to PHMSA. All of those reports then go into this gigantic database. So it forms this record that goes back all the way to 1970 for the original records. And then... relatively unbiased right because you're having all these individual incidents are being reported by the operators themselves the data is publicly available so that anybody can download and look at it it's not behind a paywall or some other restriction um and so the thought was this data would be a good source to try to go and parse out some of these trends and see what do we actually see within these incidents that are recorded. 08:08 Rhett It was relatively straightforward to dive into that database and parse out the information you wanted, right? 08:14 Alex It is absolutely not. Chris I will say this, it has gotten a lot better though. Yeah what do you mean we got to give PHMSA a little bit of credit right so before there wasn't a lot of interactive capabilities associated with the portal it was just you know you would download the excel or commuter limited file and then you have to sort through it yeah at least now i think there is some gooey component to it and some trends that you can see so it's a little bit it is a little bit easier to like we were talking about earlier like they have trending built into it right and then once you filter to a degree you can then export the data and do what you want which is a little bit neat 08:53 Alex Yeah, that's absolutely true. I think PHMSA has recognized the importance of trying to make the data more transparent and easier to parse. In this case, we decided to go back to 1985 because for a couple of reasons. One, it presents 40 years of data from 1985 to 2024, so it's kind of a good round number. But it's also when narratives started to be included with the... reported significant incidents. So prior to that time period, incidents were recorded, but there was no description of what happened. The complication comes about as one is a self-reported database. And so as the operators are filling the data out, they fill it out. Some fill out very complete, very... very good records it's very easy to tell what happened and others it's very ambiguous in terms of what's actually happening the other part is the uh you know Chris you mentioned the comment delimited files yeah some of these when you download them have over 600 separate columns and so trying to go through the data parse it out is kind of complicated um it can make the results a little bit ambiguous depending on what you're looking at 09:56 Alex So one of the things that that I did is I went I reviewed these records and tried to find incidents that were attributed to natural forces, and then reviewed the narratives to try to then assign a type of cause to it. And again, sometimes it's pretty clear and sometimes it's pretty ambiguous. So just for example, sometimes it might say, hey, a hillside gave way and it bent and ruptured the pipeline. So, okay, that's a landslide. That's pretty clear. Sometimes it might say the pipeline failed in the Red River or the Mississippi River. And then we replaced it by HDD, but no real cause is attributed to it. There's no real description of what actually happened. So you can look at it and say, okay, did it happen because of the river or is it just coincidental that it happened at that location? Yeah. 10:44 Rhett & Chris So after you comb through all 40 years of data, got to make it sound big, 40 years of data, good thing it didn't take him 40 years to go through it. Felt like it. During that same time span, the number of searches for I want to learn Spanish has gone down. Has gone down. That is true. Let's not forget that. 2004. 11:10 Rhett What was your first step? So you obviously, you found all of them that you believed were attributed to geohazards. But did you, I mean, you just took that database and started slicing it? Or did you have even questions within the portion of the database that you identified as potentially? associated with geohazards. 11:25 Alex Yeah so I mean if you kind of go through the kind of nitty -gritty of it there's over 16,000 incidents recorded in the database during that that time span I want to sort out a couple things one is I'm interested in onshore pipelines so I sorted out all the offshore incidents um sorted out all the facilities incidents. So kind of excluded those from the analysis. Then the first cut is you go and you look for all of the incidents that are recorded as natural forces. So that's one of the categories that PHMSA includes in their reporting. You go through and review those. But even in there, you can tell there's certain errors, right? Like, for instance, some of the incidents might describe a corrosion -related... failure within the product flowed into a river and the operator for whatever reason had recorded that under natural forces. So kind of knowing that you said, okay, there's probably things on the opposite side where someone's recorded something that's a different incident type that probably actually is a geohazard. So I went through and did keyword searches and those kinds of things until I came up with the final list that I believed had some relationship to a geohazard. 12:42 Rhett Cool. All right. So, and that's when we get into the data. So I'm going to tell the audience that, you know, if you really want to understand the paper. I think you have to look at the plots that are in the paper because that's where the data is. And so this episode is going to be a little bit unique because we, if you're on the YouTube channel, we're actually going to show you on the screen the plots that we're individually tackling. So we're going to jump in. So let's start. Like I'm looking at plot one here, right? And plot one is you're all geohazard incidents by time, right? And so it's this real nice bar chart. And to give the audience some idea. of the number of or the types of incidents that are geohazard related. They include hydrotechnical, debris flow, landslide. Those are the things we would commonly think of. Flotation, not so much. Wind, lightning, subsidence, frost heave, and tornado. Now I'm curious, how does tornado land in the category of geohazard? 13:36 Alex So in this case, what I was trying to do is be kind of that expansive definition of natural forces that PHMSA uses. And so you can see, as you alluded to, we got lightning and wind and tornadoes. So quite often, those would not be included in a geo hazard program. But there's kind of whether or not side force category that we oftentimes talk about. 13:55 Chris So to be clear, in chart one, these are subcategories in the PHMSA database? Or are these your interpretations? 14:04 Alex They are. It's a mix. There's some of those subcategories and some of the interpretations. So you're going through the narrative, you're reviewing what happened, assigning a type to it. 14:12 Rhett Cool. Got it. And so do you take anything? What do you think this chart tells us? Do you take anything from this chart? 14:19 Alex So I think a couple things are really relevant. One is if we're talking about geohazards, unequivocally, we see this almost any way we slice the data, landslides and hydrotechnical hazards are the dominant type over and over again. So it's about 70 % of all incidents are the landslides of hydrotechnical. And if you can find it to ruptures, which is not shown on this chart, it's about 80 % of all incidents are landslides of hydrotechnical. 14:44 Rhett So that's good. So that means in our mind, I think when we think about geohazards, the first thing that comes to a lot of people's minds is landslides, at least the PHMSA incident database. tends to bear that out. 14:50 Rhett Okay, that's cool. Yeah. 14:53 Alex The other thing I think is this chart, and if you're on the YouTube channel, you can see it, is clustered by five-year clustering. So going through and writing this paper, my first experiment is doing it on a yearly basis. It becomes really noisy to show 40 years of data, especially with all these different hazard types. So kind of settled on the idea of let's cluster it by five -year intervals. But you can see there's a lot of... variability in it right so you can see the time period 1990 so 1994 has a big high spike uh the time period 2000 to 2004 is relatively low and it kind of goes up and down from that time span but you know what if you look at this kind of superficially it looks like geohazards are kind of decreasing over time so if you kind of casually look at it and say hey we've got a lot of incidents in 1990 to 1994 and But fast forward to 2020 to 2024, it looks like we're doing pretty good, right? And so, but I think we start to parse it out and we're going to talk about it some more. There's a lot more to that story than just that kind of superficial appearance. 15:54 Rhett And you kind of already hit on the first topic, audience, that I think becomes a challenge because this, I told Alex when I was reviewing this paper. I had just finished reading a book, Fooled by Randomness, that talks about trying to read into noise. And I was like, oh man, when you're dealing sometimes with these numbers, it's difficult. I think you can, you raise a really good point there, right? And that figure one, you can look and be like, man, that one big peak that happens in 1990, 1994 makes it look a certain way, but you had a huge portion of earthquakes that happened from Northridge, which you mentioned in there. If you take those out suddenly, then it... doesn't give this this look that it's decreasing right which again goes to that nature of trying to judge too much from truly random events sometimes is a challenge so you did something else after this right and I want to 16:39 Chris Hold on we missed something hold on hold on again we'll see if this makes any sense like what would you say in general as a human we're like oh like you're more likely to hit struck by lightning but then we think about it's like what are really the odds of us getting struck by lightning like you would think that's a pretty rare event right In theory? Yeah, yeah. But this data doesn't say that. On average, there's one lightning incident per year. Per group of five years? Yeah. No, per year, right? So from 20 to 24, there was five. From 15 to 19, there's six. From 2010 to 2014, there's five. From 2005 to 2006, on average, there's one year. 17:16 Rhett Going a different direction on this podcast. We're not going to talk about landslides, Alex. We need to know your thoughts about lightning strikes on pipelines. 17:21 Alex It is actually kind of interesting. It was actually one of the biggest surprises of doing this work is how many lightning-related incidents there are. What's interesting is you read the narratives over and over again. You're having incidents where lightning is striking a surface facility. It's traveling into the pipe, but then when it's... you know, exiting the pipe, it's leaving a pinhole, a little pinhole behind. And so you get a leak that associated with that lightning strike. 17:45 Chris Yeah. I was saying this is interesting because these are significant incidents, right? Which means they were either $50,000 worth of damage according to like 1980 something dollars, which is over a hundred thousand in today's dollars or fatality or hospitalization. And or some volume of loss of product. So the lightning is doing something. 18:05 Chris Yeah. I mean, I guess the good news in the lightning ones, they seem to be resulting in leaks, not full blown ruptures. Rhett So I still think we need to get in the lightning integrity management. It's so cool to tell my family. It's true story. So you did a really good job next to breaking down the incidents by state. Right. And I think this is cool because it tells me that North Carolina and South Carolina are on the ball. Because they have absolutely zero geohazard incidents. And so if you're a pipeline, at least for the perspective of geohazards, North Carolina and South Carolina are the place to be, right? 18:40 Alex So I think that kind of goes back to the misleading statistics that you brought up before. So I think when you look at this distribution, there's two things that really show up. One is places like California. West Virginia kind of geologically hazardous areas most people kind of know that except it have a lot of incidents and that's kind of makes sense uh but then you look at Texas and Oklahoma and Louisiana and they have a very uh significant number of incidents in those and there's really a strong correlation where you have pipelines you have potential geo hazard exposure so those you know those far east coast uh Atlantic coast states uh in comparison places like Texas don't have a lot of miles of pipeline and i think that's more of what we're seeing in the data 19:26 Rhett yeah so it's really cool you had a you had another plot in here that showed the the congestion of pipelines if you were to take figure five and then almost overlay it and look at it through the light your your more lit up areas would be where the congested pipelines are which makes sense right and we obviously know you have a high congestion of pipelines in Texas and that goes through the midwest uh, of the United States, if you will. Um, but I think this also does speak to the fact, you know, I think that for instance, you, you have nine incidents right there in Oklahoma. That is Oklahoma, right? My geography is not very good. There's nine. I mean, you're not, and we don't think about Oklahoma as a geologically, let's say prone region, right? Like it's not one that comes up a whole lot. You have four in Arkansas. Right. I don't hear whenever we go to conferences, people are like, man, you know, those darn Arkansas landslides. Right. So can you comment on that? 20:21 Chris Those are hydrotechnical incidents. Right. So you're looking at number nine. So nine for Oklahoma was hydrotechnical, 18 in total. Rhett You're right. Thank you. Chris I got you, bro. Maybe I should look at the data correctly. But 18 comprehensively, nine hydrotechnical. Yeah. This, when we, when we look at this together, what I want to, what I was getting at with Alex, thank you for bringing that is this is a common theme, right? That geohazard incidents. Don't just target the place that comes to everybody's mind first. 20:45 Chris I thought tornado, that was one of them. I thought you were going to say tornadoes, but it's not. It's hydrotechnical for nine. 20:52 Alex Well, I think that's a great point, Chris. I think that a couple things, and maybe we're kind of skipping ahead a little bit, but if we look at the central U.S., most of these hydrotechnical incidents are happening in the central U.S. because we have some combinations of a lot of miles of pipeline. We have intense storm systems. I think anyone who's lived in Texas or Oklahoma is familiar with the thunderstorms. Tropical storms or the remnants of those storms. We have river systems that tend to flood very dramatically. And so you can think of a lot of the rivers around Houston, which, you know, during, right now it's been pretty dry. You can probably walk across a lot of those rivers. And during a big flood, they might be two miles wide and 30 or 40 feet deep. I think the combination of those things really predisposes those areas to having a lot of incidents related to the hydrotechnical. But I think we'll kind of get into it too with the landslides. We still see landslide incidents in places like Texas, Mississippi, Louisiana. I think one of the biggest myths that's out there is that geohazards only affect certain parts of the country. And people are very predisposed to think, okay, we only have geohazards in places like West Virginia. California and there's some truth to that those areas in a given geographic area are going to be more likely to have geo hazards but you still find a lot of those same forces at play in other locations around the U.S. 22:18 Rhett Yeah and so you did a good job with that right and I apologize like you really have to compare figure figure one and figure four if you're going to take a look at the landslide events and the hydrotechnical events. But my point was still valid, even if I completely misread the plot, that the hydrotechnical and what we lump in as geohazard events don't have favorite states that they congregate towards, right? That they really are dispersed, with the exception of North and South Carolina, throughout the United States. 22:50 Alex Yeah, but I guess the one big exception, that would be seismic events, right? I think California is the most seismically active. state in the lower 48. So there's a reason we don't see a lot of seismic incidents outside of California. But if we're talking about flooding landslides, they are not just confined to one area of the country. I think the other thing that's kind of important to keep in mind is we have this tendency to talk about geo hazards of weather and outside forces as if it's one causal type and it's really a bunch of different things that are very different i think the easiest way to think about that you know earthquakes and flooding really have no correlation between each other maybe there's kind of a looser correlation between flooding and landslides or landslides 23:34 Rhett what if the earthquake damages the dam and the dam gives way and then you get flooding Alex Well I mean I guess that could happen but i mean i think really again you know if you're uh you know an integrity person it'd be kind of like lumping external corrosion internal corrosion and stress corrosion cracking into a single category because i'll have corrosion in the and the name and so i think that's where to your point about muddying the data, we kind of have to separate out these causal types to really see what it says. 24:04 Rhett All right. So we are actually going to take a break. When we get back, audience, I want you to hang on because we're going to get into the meat of the conversation, which is asking the question from the data, are we actually doing better? Does the incident database show that we're learning geohazard or hydrotechnical incidents are actually decreasing with time? Hang on as we get into that question when we get right back. 24:34 Rhett All right. Welcome back, audiences. We are on with the dirt merchant, Alex McKenzie Johnson, in I think his fourth episode now, talking about his deep dive into the PHMSA incident database, which he published at PPIM this year. So, Alex, where I want to go now, I thought was probably what the audience is more interested in. And I thought was the more fun part, because it led to lots of good conversations between you and I, which is after you took all this data out of the database and you sliced it, and we didn't really cover it. But you actually came up with high likelihood and moderate likelihood and then low likelihood incidents. You separated out the high and moderate likelihood incidents. And then you further separated it by geohazard events. And you separated it by landslides and by hydrotechnical events. And then you grouped those together. And you tried to draw some conclusions about time or patterns. And so I'm just going to take the audience. We're going to go through those step by step and talk through them. And first we're going to start with is actually figure seven or chart seven, right? So it's all geohazard incidents by time for all regions. And as you alluded to, you clustered it in five–year groups. And the chart, not only is it clustered by five-year groups, it also clusters the incidents by west, central, and east regions of the United States. What was the first thing that you took from this plot whenever you were looking at it in the paper? 26:03 Alex Yeah, so I think one of the big takeaways is there are regional differences. And when we talk about the regions, kind of think of the West as being like Pacific and Mountain time zones, Central being like Central time zone, and then East being kind of Eastern time zone. Not exactly, line up with time zones, pretty close. And so if we look at like the West, we see what looks like this big...drop in the number of incidents so we go back to the 1990s to 1994 time frame we have like 19 incidents and it goes all the way down to three and the 2020 to 2024 time frame but it's you know we kind of talked about this muddying of data it's really biased by the fact that 1994 we have the Northridge earthquake And that causes almost all of the seismic incidents that are recorded in the database. 26:54 Rhett Which are lumped into geohazard incidents on the west side. 26:56 Alex All the other stuff. And so if we take out those, this kind of seismic event, which we don't have any other seismic events that really show up in the data, it really changes that perception of the west. You know, we see that the central region also appears to kind of be declining in the number of incidents. And I think there is some... validity to that and you know I think we're going to get into that but then we see the eastern region we see an increasing number number of incidents during this time from really starting around the 2005 to 2010 uh time we're seeing more and more incidents are happening in the east I think there is some something really going on there which we'll talk about the landslides coming up 27:38 Rhett Cool it's one of the this is gets to like one of the first challenges I have with looking at data right so we just focus on the western side of things the red bars in the chart It's interesting to me because they oscillate through your five -year periods. You've got 19, 10, 19, 11, 18. So it's almost like on this five-year interval, you really have, let's say, 14 incidents plus or minus five, which means at any given five-year period, you could have as few as 10 or as many as 19 and be completely normal. But if you try and draw a trend between any 10 -year span, like for instance, if we... If somebody erroneously took the data from, for instance, 2015 on, you'd say, man, we've gotten so much better. We've cut our number of geohazard incidents in half in the western half of the United States. When in reality, that's just been par for the course for the last 40 years, as you're showing here. 28:28 Alex Oh, I totally agree. And I think that's why it's so impactful to look at 40 years of data versus focusing on the last five years of data, the last 10 years of data. We see a lot of temporal variability. that happens right exactly like you're saying there's the significant variations that are happening the other thing is again looking at this chart seven is we're bumping all these geohazard types into a single category and so it muddies that those trends in a way that's um meaningful and it also kind of impacts what is the the implication for the pipeline uh operators 29:07 Rhett Well then you know what let me address your concern or you address the concern for the audience and we'll get into figure eight which is now just hydrotechnical incidents by region or by time. So I have one question I'll ask in your parsing of the PHMSA database, right? And you did the same thing in this chart for the audience. It shows five-year groupings of hydrotechnical incidents with West, Central, and East. Hydrotechnical, how did you identify it as a hydrotechnical since it's self -reporting? Did you just look for the keywords river? Did they have to check hydrotechnical? How did the incident get noted as a hydrotechnical for you? 29:36 Alex Yeah, so it certainly helps support it if they report it as a subcategory of natural forces damage called heavy rains and floods. But even in there, looking for things of the narrative like pipeline was exposed during the flood and ruptured, or pipeline was just oscillating before it failed, or something else that indicated that the river itself, that flow of water, caused the...failure. So what I'm kind of trying to do, separate out hydrotechnical from landslides, if the failure is because a hillside gave way, even if it's during a heavy storm, I'm saying that's a landslide. But the mechanism that broke the pipeline is the sliding soil versus hydrotechnical. It's from the flowing water that causes the failure. 30:28 Rhett Gotcha. All right. So what do you take from this plot? This might be our first argument on the show. Let's go. 30:29 Alex So, I mean, to me, it looks like, so again, kind of going back to the early 90s or the mid -90s, we have a lot of hydrotechnical incidents. Rhett Particularly in the west part of the United States, for sure, right? Alex The central part. Red bars of central U .S. 30:50 Alex So we have 31 incidents, 1990 to 1994. We have 16 from 1995 to 1999. more or less it seems we're getting fewer and fewer incidents over time with the real low point being this uh last four-year period 2020 to 2024. now I think there's some real validity to what you're saying there's even a five -year time span may not be enough to really parse out long -term trends but I do think there is something going on here that's real um which I think we're you know we're constructing your pipelines when replacing pipelines we're increasingly doing it by horizontal directional drilling which reduces the vulnerability the likelihood the pipelines are going to be exposed got to be damaged during flood events and so I think that's partly what we're seeing is that a lot of those really vulnerable pipelines have already been exposed and damaged or been replaced within the data that being said you know the the data back to 1990 -1994 reflects two really big flood events, one in Arizona and one here locally in the Houston area. So kind of to your point, if you took out those big events, the data may not look quite as dramatic. 32:05 Rhett So if I took a plot and I just covered the right-hand side of it like this, it's harder to draw the conclusion. In the last 25 years, basically. If you looked at the last 25 years, it's harder to draw the conclusion that it is decreasing, with the exception of the last five years does look really good. Now, I'm not suggesting that. I mean, again, maybe this was like a seminal event here that has kind of shaped things. Maybe we are indeed getting better. I think my point is, overall, it would show a trend in the last 40 years. But if we looked at only the last 25 years, there's potential you might make, at least in the central United States, that maybe we're not getting as good as we could. But maybe there's room for improvement. 32:43 Alex I mean, unequivocally, right? I mean, we're in the integrity business. All of us are trying to work towards zero incidents. And the fact that we're still having these types of incidents, it means there's room for... 32:56 Rhett But you're feeling on this, and I did like, because I really like that you went for a cause, that we should expect river crossings and hydrotechnical incidents to get better by virtue of, you know, it's really funny. I had a meeting with a client today that was telling me in the 1970s, he's like, I don't know what happened in the Mississippi River, but if you look like almost all the Mississippi River just traversed in the 1970s because of all these floods that happened, moved to hydrotechnical. HDD crossings in the 1970s and in the Mississippi. It makes me think of that here. That is a valid reason why we should expect to see at hydrotechnical or river crossings fewer incidents because HDDs eliminate that threat. 33:39 Alex Yeah, and we can see it another way. Because of length limitations for conferences, I couldn't include this in there, but if you look at the ages, the installation age, pipelines are damaged. for these hydrotechnical incidents with one exception they're all installed before 2000. they're all older pipelines when conventional trenching was more the method to um to lay pipelines in some cases they're just laying the pipelines directly on the river bottoms um so we see a lot of older pipelines being affected we don't see newer pipelines being affected which to me tells me that the HDDs are um likely reducing that that risk 34:16 Rhett It's also interesting because it looks like the east coast is relatively holding constant I mean, they don't have very many incidents, but they don't really pick up until the 80s and you only get a few. Chris Hydro -technically. Hydro -technical. Rhett We should clarify with the audience. Thank you, Chris. We are talking about hydro -technical incidents right now. 34:31 Alex Yeah, I think that's a function of both pipeline mileage and then also just geography. The rivers in the central U .S. tend to be much more dynamic, much more prone to these major flood events and channel migration and all those kinds of things. 34:41 Chris It also makes me wonder, and again, this is an anecdote here, right? But when I look at... With the chart where you had the US, I think it's a figure one where you had, you know, the number of incidents geographically and we look at the hydrotechnical, it's really concentrated in the third coast, right? So Texas, Oklahoma, Arkansas, Louisiana, Mississippi, and it just, it kind of makes sense, right? If we think of like extreme storm events, right? um and so the correlation there with hydrotechnical it just begins to make sense whereas like the further north you go hydrotechnical really starts dropping off as integrity people i think what we're talking about is that just makes sense right because if from an integrity perspective it's you do all your water water, water water crossing surveys and then you start doing P&M so it's it's not necessarily integrity assessments and looking for response times rather hey we've identified we have crossings what's our P&M And so it's an easier managed threat the more north you go, whereas the more south you are towards that third coast, it becomes a little bit more of a time–independent threat. 35:48 Alex Yeah, exactly. We have big, dynamic, sudden storms. You can think of how hard it is for forecasters here to predict the actual amount of precipitation that's going to fall within a given storm. We have a lot of thunderstorms. We have fronts that come through. We have tropical storms, all those things. Combined with that, we have a lot of very relatively flat topography, and that makes it much easier for those rivers to migrate, change their position, all those kinds of things. 36:18 Chris Whereas, again, if we look at figure four, we're seeing the landslides being far west coast, California, or in the northeast, which, again, continues to make sense. P&M is, what would you say there, pretty well understood now in gaining momentum, though. So it's also a little bit more time dependent, right? Because you're able to monitor it. It's less, I mean, it's still somewhat time independent, but I would say it's still more of a time dependent threat. How would you manage it? 36:39 Alex Probably more time dependent. It's probably more, you know, than hydrotechnical for sure, because usually you have some kind of advanced indication. You have an ability to evaluate it, to monitor it, to respond to it. 37:00 Rhett I mean, it leads really nicely in our next question because we just talked about hydrotechnical. The next chart, which is figure nine, divides landslide incidents along the same regions by five-year intervals. And this plot, I think, is unfortunately much muddier. So what conclusions did you try and draw from this plot? Or did you recommend that the audience should draw from this plot? 37:24 Alex So I think there's a lot going on here. And I think what's interesting, again, is we kind of dive into the data. 37:30 Rhett We're getting better, though, right? Cross the board. That's the conclusion you draw about landslides. plot shows that landslides are decreasing, right? 37:37 Alex I would say there's a complicated story here. And I think it might be too soon to tell in some ways, right? So I think if we look at this plot, there's a couple of things that really show up. One is we have this big spike in the West, 1995, 1999. And that correlates if you're like myself, I'm from the Pacific Northwest, a lot of very, very wet winters back to back. And many failures that happened during that time. So kind of the one I know best is Northwest Pipeline, which is now part of Williams. They had back -to –back landslide failures the same day, 100 miles apart. 38:17 Rhett How many years ago did you say this was? How many years ago? 1997. 1997, so 30 years ago. Yep. But aren't we right now going through another cycle in the West? This is the second year in a row where they've had torrential downpours and these atmospheric rivers being described over California. And it was only a few weeks ago we had an incident in the news related to a landslide in California, right? So is this like a, I don't want to use words climate change. I'm not trying to create a buzzword. But climate cycle type thing where you could be going through again? 38:47 Alex Very possibly. But I think in the case of these incidents in the 90s, Northwest Pipeline Institute was probably the first real geo hazard program. And they haven't had a, to my knowledge, an incident since 1999. So it's a real kind of success story. They go from having many incidents in a relatively short amount of time to having none for over 25 years at this point. So I think there's some things with some of the operators in those Western states are, they're evaluating the threat, they're doing something to manage it, and they're addressing it. Now, we see this kind of real low point around the year 2000 to 2004, and then we start to see this uptick again. 39:31 Rhett Those are the golden years of landslide management pipeline integrity. 39:35 Alex And what's interesting to me and what really kind of stood out is, you know, we go from having most of the incidents happening in the west to most of the incidents happening in the east. So what gives? Well, I think there's a couple things going on here. So I think there's one thing is we do start to have some wet. winters in some of these areas like Ohio, Kentucky, West Virginia, and a lot of these slides in those areas are kind of slow moving. They're pretty subtle. They're kind of hard to identify. And some of these older pipelines start to fail. But the other thing I think what happens is the shale gas revolution. So we have the Marcellus, the Utica Shell. You had some new pipelines built in that region. A lot of new pipelines are being built. 40:17 Rhett So did you hear that? He just blamed new pipelines and old pipelines in the same breath. 40:18 Alex And we can see it again. If we look at that installation, which unfortunately I didn't have room to include in the paper, is we get kind of a typical curve we'd expect. Older pipelines seem to be more vulnerable. And telling you to get to these new pipelines. and then you start to have a bunch of new pipelines start to fail from from landslides once a new pipelines pipelines installed in 2010 or more recently if we look at that normalize it by pipeline mileage actually appear to be more susceptible than these these old pipelines so i think what's happening is we have a combination of natural events and then with new construction a lot of it's happening in landslide prone areas where they're disturbing the the soil serving the rock and then you're having failures happening relatively quickly after that that construction happens 41:09 Rhett So what's it going to take to abate that that trend I think is the question that i'd want to ask right I mean it's because you're pointing to the fact that at least in the eastern united states if you were to look at this data just eastern part of the United States it would suggest that the trend is increasing I mean maybe we turned a corner since 2020 to 24 but i think that that's risky to draw from the data when you consider the noise so what do we need to do better 41:32 Alex Well, I think that's a good point. It kind of goes back to the original genesis of the question. Are geo hazards actually increasing? And I think if you're an operator in the central U.S., eastern U.S., it's reasonable to say we're seeing more landslide incidents than we've seen in the past. And these incidents are happening and they're pretty dramatic. They're getting attention. So I think some of that increased interest is actually following incidents. In terms of... do we do about it how are we doing about it you know again you think about most shoe hazard programs and you guys know this are really new right I mean most of those programs started in 2020 or more recently and it's probably just not going to show up on the data whether or not those programs are having the desired effect of reducing long -term you know the effectiveness of things like lidar bending strength and things that you and I work in from day to day it'll be 42:21 Rhett It'll be very interesting to see the effectiveness of that if you redo this paper and you know 15 years 42:29 Alex Yep so I mean hopefully that's what we see right hopefully we see is that this kind of at this point starts to uh see reduced reduced risks and reduced a number of incidents I think the other thing is we have to look at uh you know construction practices how are companies designing new construction in areas that are prone to things like landslides are they taking measures to account for that um in that design 42:53 Chris I'll take a little bit of a different take on this I would say putting my integrity hat on one of the ways that PHMSA communicates what they're doing with this data is they call it performance measures right so if you were to go to the PHMSA website and you're trying to look for where these significant incidents are I think at the top of the page and obviously I'm not looking at I'm going a little bit off of memory here I think it says performance measures and I think they say that like for example the liquid guys They developed these performance measures with API, AOPL, for example, Pipeline Safety Trust for the natural gas performance measures. They worked with Pipeline Safety Trust, API, INGA, NAPSR. And this data set sits in that, right? And so when I think of those words, that's what grabbed onto me is performance measures, right? So in your integrity management program, you're supposed to define performance measures that you're monitoring. So a bit of a rhetorical question here for the audience is it's, I wonder, hopefully now this is made light to you if you hadn't thought about it this way, right? Is it for those operators that have stood up a geohazards program, you know, are they using the PHMSA database? to kind of benchmark their performance measures right so if they're taking a monitor plus P&M approach strategy to geohazards versus your traditional integrity approach of assessment response and then P&M I'm curious if as we see people they can give us some feedback and see how are they using these PHMSA related performance measures in their integrity program? 44:26 Rhett Well I'm wondering what the score would be on this chart right here um maybe PHMSA don't look at it so uh the story is and I think it's really Alex you did a good job i want to thank you for putting the paper together audience I'd encourage you to take a look at it the story is you know from a hydrotechnical perspective I want to summarize I think you could make the argument that trends might be decreasing right but there's causes for that or reasons for that that makes sense the prevalence of HDDs I would even argue um i think that the hydrotechnical hazards might be even slightly more mature than landslides, but that might be something you and I could debate over a beer. The story, though, is not so clear, unfortunately, with landslides, right? The prevalence of pipelines in the eastern United States, maybe more wet seasons in the Appalachian region, make it much more difficult to draw the conclusion that we're decreasing. If anything, it might tend to say that we're increasing, which... tends to support some of that anecdotal evidence we have of, hey, there's more awareness or there's more interest, more buzz around geohazards, certainly in the last decade that we've been at conferences. 45:42 Rhett Alex, I want to say thanks for joining us on the podcast. It's always great to have you back again. Again, congratulations on TriCoast. Look forward to working with you in the future. All the best. To our audience. Again, thanks for joining us. Remember, be careful about sharks and ice cream and watch out for those sharks in the summer. When you see ice cream near a beach, you better watch out for sharks. Thanks for joining us. Be ready here first. 46:03 This episode was executively produced by Sarah Etier. Thanks to our guest, the dirt merchant, Alex McKenzie Johnson. And for his paper, for reference, a deep dive into the PHMSA Significant Incident Database and what it tells us about trends in pipeline geohazards, PIPIM 2026. Thank you.