Geography asks three questions: |
| Rook, aka Missionary Poker Image: Wikipedia |
The young dealer insisted that this highly unlikely outcome had occurred naturally. After all, every combination is equally likely, and since every player had an equal hand, there had been no motive for the "crime" of setting up a deal this way. No motive, perhaps, but enough opportunity to leave the other players skeptical. Skeptical indeed.
This came to mind -- actually the mind of my wife and favorite librarian -- as we contemplated the "thousand-year flood" that devastated the beautiful little town of Ellicott City, Maryland while we were visiting family nearby over the weekend. Off-duty guardsman Eddison Hermond perished trying to assist others in the storm.
The video below was taken just 24 hours after we had driven through the area, admiring how well the downtown had recovered from 2016 flooding. More harrowing, it was filmed just as my mother-in-law was returning from a family gathering with us to her home just a mile from these scenes.
The video below was taken just 24 hours after we had driven through the area, admiring how well the downtown had recovered from 2016 flooding. More harrowing, it was filmed just as my mother-in-law was returning from a family gathering with us to her home just a mile from these scenes.
This was the second such flood in two years, perhaps even more destructive than the deadly 2016 flood that put the entire downtown area out of business for nearly a year. I will not detail that flood here, as I have covered it in some detail in other Ellicott City blog posts, especially Flood Flash, Flood Peak, and Houston.
 |
| I had driven past this place in the sunshine the day before. Image: Kim Hairston, Baltimore Sun |
Rather, I want to focus on the use of recurrence intervals, in this case 1,000 years. Even NPR used this expression uncritically in reporting on the flood's aftermath. A 100-year flood was more commonly cited in the past, perhaps because it is the interval most important to the delineation of flood plains for insurance purposes. Whatever the specific interval, the term lends itself to misinterpretation. The fact that we now routinely hear of 500- and 1,000-year events should be a hint that something is seriously awry.
To understand how a recurrence interval can be misinterpreted, it is useful to understand how it is defined. A flood level with a recurrence interval of n years has a 1/n probability of being met or exceeded in any given year. The "expected" interval between such events is n years, but this does not mean it will be exceeded every nth year. Consider a simple card example -- the probability of drawing a green card in Rook is 1/4, so the expected interval is one green card every 4th draw. Of course they may be drawn consecutively or with a longer interval between them. So the first way in which recurrence intervals may be misunderstood is that they do not define intervals in the sense of regular timing.
This basic fallacy is rather easy to understand, and most people know that low-frequency, high-interval floods could occur within a short span of time, and this is what seems to have happened in Ellicott City. But just as highly unusual coincidences can make us suspicious of card sharps, so too should highly improbably flood events lead us to question the calculations.
Page One of any discussion of probability (whether in floods, industrial quality control, or card games) states some basic assumptions that are often forgotten by the time the probabilities are calculated. Those assumptions are that the events in question are individual observations drawn from a population of discrete, random, independent events. If there is any kind of relationship among successive observations, or if there is any kind of secular (long-term) trend in the population being observed, then all calculations of probability are meaningless.
To calculate flood recurrence intervals, a record of gage readings may be taken from a single stream over a period of just 30 years to define a curve of probabilities that will estimate recurrence intervals of up to 100 years within acceptable error tolerances (typically 0.05, or 5 percent). This is valid ONLY if the factors affecting flood levels remain constant. The main factors are climate and land use, and this is the problem worldwide, and especially in suburban areas. Climate is changing everywhere, and land use is changing in most places. So terms like 100-year and 1,000-year describe how rare today's floods would have been in the past. But without understanding why the past is no longer a guide, we will continue to put lives and property in danger if we use these terms for decision-making.
Further Reading
Much has been and will be written about these events, so I would like to cut through the noise to recommend two articles recommended by two friends who were undergraduate students of geography with me -- studying hydrology just a few miles from this disastrous flood in between many long walks along nearby streams. The first is Ellicott City flood was no 'act of God,' an opinion piece by Baltimore Sun journalist Dan Rodricks. He laments the role of politics in deliberate misunderstanding of the science that explains these events and the resultant dismissal of common-sense remedies such as rain taxes. He cites some of the papers in which hydrologists called for the elimination of the language of recurrence intervals.
The second is Does it make sense to rebuild Main Street Ellicott City? by blogger and architect Klaus Philipsen. His focus is on the infrastructure in the immediate area of Main Street itself, where some structural features do seem to amplify the damage done by flooding. He compares the options Ellicott City faces with steps taken by Frederick, another of our favorite downtowns that appears to have found some increase in resilience through civil engineering. All would agree, I think, that dramatic intervention in upstream land uses is also needed, with particular attention to the reduction of impervious surfaces.
Trees
As I was writing this piece -- over the course of several days -- I learned of a project being undertaken in Chicago that is relevant, even if the circumstances are not exactly comparable.
From the Yale Climate Connections project, I learned of Blacks in Green (BIG) a community-scale project that is fostering climate justice and building climate resilience. As the Yale team and BIG founder Naomi Davis understand, resilience is not a zero-sum game. Treating the land and water better is also good for equity and economies.
Lagniappe
In the midst of the many calamities and dislocations of the first half of 2020, extreme flooding in Michigan was the direct result of climate change and a refusal to enforce dam-safety laws. As I describe in Dam Shame, the damage was compounded by terrible inattention to other environmental regulations.
To understand how a recurrence interval can be misinterpreted, it is useful to understand how it is defined. A flood level with a recurrence interval of n years has a 1/n probability of being met or exceeded in any given year. The "expected" interval between such events is n years, but this does not mean it will be exceeded every nth year. Consider a simple card example -- the probability of drawing a green card in Rook is 1/4, so the expected interval is one green card every 4th draw. Of course they may be drawn consecutively or with a longer interval between them. So the first way in which recurrence intervals may be misunderstood is that they do not define intervals in the sense of regular timing.
This basic fallacy is rather easy to understand, and most people know that low-frequency, high-interval floods could occur within a short span of time, and this is what seems to have happened in Ellicott City. But just as highly unusual coincidences can make us suspicious of card sharps, so too should highly improbably flood events lead us to question the calculations.
Page One of any discussion of probability (whether in floods, industrial quality control, or card games) states some basic assumptions that are often forgotten by the time the probabilities are calculated. Those assumptions are that the events in question are individual observations drawn from a population of discrete, random, independent events. If there is any kind of relationship among successive observations, or if there is any kind of secular (long-term) trend in the population being observed, then all calculations of probability are meaningless.
To calculate flood recurrence intervals, a record of gage readings may be taken from a single stream over a period of just 30 years to define a curve of probabilities that will estimate recurrence intervals of up to 100 years within acceptable error tolerances (typically 0.05, or 5 percent). This is valid ONLY if the factors affecting flood levels remain constant. The main factors are climate and land use, and this is the problem worldwide, and especially in suburban areas. Climate is changing everywhere, and land use is changing in most places. So terms like 100-year and 1,000-year describe how rare today's floods would have been in the past. But without understanding why the past is no longer a guide, we will continue to put lives and property in danger if we use these terms for decision-making.
Further Reading
Much has been and will be written about these events, so I would like to cut through the noise to recommend two articles recommended by two friends who were undergraduate students of geography with me -- studying hydrology just a few miles from this disastrous flood in between many long walks along nearby streams. The first is Ellicott City flood was no 'act of God,' an opinion piece by Baltimore Sun journalist Dan Rodricks. He laments the role of politics in deliberate misunderstanding of the science that explains these events and the resultant dismissal of common-sense remedies such as rain taxes. He cites some of the papers in which hydrologists called for the elimination of the language of recurrence intervals.
The second is Does it make sense to rebuild Main Street Ellicott City? by blogger and architect Klaus Philipsen. His focus is on the infrastructure in the immediate area of Main Street itself, where some structural features do seem to amplify the damage done by flooding. He compares the options Ellicott City faces with steps taken by Frederick, another of our favorite downtowns that appears to have found some increase in resilience through civil engineering. All would agree, I think, that dramatic intervention in upstream land uses is also needed, with particular attention to the reduction of impervious surfaces.
Trees
As I was writing this piece -- over the course of several days -- I learned of a project being undertaken in Chicago that is relevant, even if the circumstances are not exactly comparable.
 |
| Resilient Chicago community -- Blacks in Green |
Lagniappe
In the midst of the many calamities and dislocations of the first half of 2020, extreme flooding in Michigan was the direct result of climate change and a refusal to enforce dam-safety laws. As I describe in Dam Shame, the damage was compounded by terrible inattention to other environmental regulations.
2023 Follow-up
July 2023 flooding in Vermont provided a reminder that terms like "100-year flood" are no longer meaningful. NPR did a story about this, but wasted the opportunity to provide a fully useful explanation. The story focused on this as a problem of risk communication, which it is, but only hinted at the much more serious issues described above. Probability models should not be used only with discrete observations that are not connected to each other by trend lines. And probability models of one population should not be used to describe other populations. Because of land-use and climate change, flood and temperature data from different centuries might as well be from different continents.
No comments:
Post a Comment
Thanks for your comment and your interest in my blog. I will approve your comment as soon as possible. I had to activate comment moderation because of commercial spam; I welcome debate of any ideas I present, but this will not be a platform for dubious commercial messages.