Noah's Hark!
August 3, 2019 11:01 AM Subscribe
What exactly is a 100 year flood? An explanation of the statistics behind the term, courtesy of the USGS.
What exactly is a 100 year flood?
Due to climate change, it's the kind of flood that happens about once every ten years.
posted by justsomebodythatyouusedtoknow at 12:10 PM on August 3, 2019 [14 favorites]
Due to climate change, it's the kind of flood that happens about once every ten years.
posted by justsomebodythatyouusedtoknow at 12:10 PM on August 3, 2019 [14 favorites]
I appreciated that they highlighted that the 1% flood level can change over time: just getting better information, urbanization changing the inputs, or flood control measures reducing the occurrence.
“As the below charts show, changes in local land use, new river impoundments, changes in the amount of impervious surfaces, and long-term climate patterns can affect at what point a "100-year flood" is designated.”
posted by Huffy Puffy at 12:15 PM on August 3, 2019
“As the below charts show, changes in local land use, new river impoundments, changes in the amount of impervious surfaces, and long-term climate patterns can affect at what point a "100-year flood" is designated.”
posted by Huffy Puffy at 12:15 PM on August 3, 2019
Nice quote from BeesWing's linked video, which really clear and nice -- it's a channel called Practical Engineering:
You don't call snake eyes the 36-roll dice throw, and you don't call a full house the 700-deal poker hand.posted by LobsterMitten at 12:16 PM on August 3, 2019 [3 favorites]
Skip to 5:38 in that video, to see what bullshit "C" is. By "C", engineers are drowning you. "C" is the reason that engineering isn't adapting to climate change. Anything that is "environmental" gets flattened into "C".
I would make the joke here, about how "assume the cow is a sphere", but it's not very funny, we have massive industrial facilities built inside of rivers (Exxon Beaumont in Harvey) because of this bullshit.
I've got some spreadsheets on how to communicate return periods and flood risk to the public, because it's something we do all the time in Louisiana. When local government can't read a percent exceedance chart, people die.
I've also seen some decent resources from Ag Extension offices, like this one from UFL: How Likely Is a 100-Year Rainfall Event During the Next Ten Years?1
Young Gu Her, William Lusher, and Kati Migliaccio2
What is insane is that the idea that a "25-yr" storm is a "storm that happens once every 25 years" is currently legal precedent in the USA, based on TP-40, a set of statistics dating back to the 1950's, that we can't seem to update, even though National Weather Service is trying, with its Atlas 14 project.
Atlas 14 is pretty good, in that it gives you actual inches of water from rain, given a location, duration, and return period, and also gives you a confidence interval--you know, like science do.
Too bad engineers won't use it.
For the USA, where housing exists not to house people, but as a form of bank account, it's important to know that
"the 100-yr flood has a 26% chance of happening over a thirty-year mortgage."
Here's a ULL engineer trying to unfuck this in a lecture to the public after the 2016 climate rains.
But the Engineers have hopelessly abused the mathematics--the system we have is almost worthless, because the engineering resists the reality of rivers, wetlands, climate and floodplains. It's infuriating.
The Statisticians and Mathematicians among us would have us say the "1%" storm--as that is mathematically correct. The "100-yr" storm is really the storm that has a 1% chance of happening every year (in a given location and duration, based on rainfall stats from 1950).
But the issue with that is that the communicative number decreases as the amount and power of the water increases, so journalists and communications professionals revert to the old system. A journalist is always going to say "This 1000-yr storm" over "This 0.02% storm".
The Dutch love to scam dumb americans with their rhetoric about the Netherlands being built to a "10,000-year" standard--as if their levees and walls are that much higher or stronger than the USA's "100-year" or "500-year" standards.
They are not! It's just easier to build to the "10,000-yr" storm in the Netherlands, because they don't get the cyclones the USA gets. The trick is that the Netherlands doesn't really get storms (yet, i guess, let's not ignore climate change), only ocean flooding.
So, our challenge is to convert all this into real amounts of water, and use the statistics to communicate the actual probability over time of worst case scenarios (as opposed to the best estimates, which are always going to be lower, less amounts of water).
This is what I want from a news story:
"Hurricane Barry is going to drop 14 inches of rain in 48 hours onto the Mosaic Phosphate Waste Pile. The Waste pile, which enables the manufacture of most of the petrochemical fertilizer in the United States, is only engineered to handle three inches of rain in 48 hours. the Pile is collapsing, and we totally were able to predict why. but the company was greedy and too thrifty, and the Farm Bureau lobby too powerful, so it was built to substandard, and now radioactive, acid water will likely flow into Blind River."
If someone knows how to unfuck EPA, so that their engineers learn to require that facilities are built to the the worst case 4% storm --(highest part of the confidence interval, given a return period, duration, and location), i'm all ears.
Just promise me, promise me, please, that i will never have to hear "no one could have predicted this flood / this levee failure / this petrochemical spill of radioactive acid waste" ever, ever again.
posted by eustatic at 12:30 PM on August 3, 2019 [32 favorites]
I would make the joke here, about how "assume the cow is a sphere", but it's not very funny, we have massive industrial facilities built inside of rivers (Exxon Beaumont in Harvey) because of this bullshit.
I've got some spreadsheets on how to communicate return periods and flood risk to the public, because it's something we do all the time in Louisiana. When local government can't read a percent exceedance chart, people die.
I've also seen some decent resources from Ag Extension offices, like this one from UFL: How Likely Is a 100-Year Rainfall Event During the Next Ten Years?1
Young Gu Her, William Lusher, and Kati Migliaccio2
What is insane is that the idea that a "25-yr" storm is a "storm that happens once every 25 years" is currently legal precedent in the USA, based on TP-40, a set of statistics dating back to the 1950's, that we can't seem to update, even though National Weather Service is trying, with its Atlas 14 project.
Atlas 14 is pretty good, in that it gives you actual inches of water from rain, given a location, duration, and return period, and also gives you a confidence interval--you know, like science do.
Too bad engineers won't use it.
For the USA, where housing exists not to house people, but as a form of bank account, it's important to know that
"the 100-yr flood has a 26% chance of happening over a thirty-year mortgage."
Here's a ULL engineer trying to unfuck this in a lecture to the public after the 2016 climate rains.
But the Engineers have hopelessly abused the mathematics--the system we have is almost worthless, because the engineering resists the reality of rivers, wetlands, climate and floodplains. It's infuriating.
The Statisticians and Mathematicians among us would have us say the "1%" storm--as that is mathematically correct. The "100-yr" storm is really the storm that has a 1% chance of happening every year (in a given location and duration, based on rainfall stats from 1950).
But the issue with that is that the communicative number decreases as the amount and power of the water increases, so journalists and communications professionals revert to the old system. A journalist is always going to say "This 1000-yr storm" over "This 0.02% storm".
The Dutch love to scam dumb americans with their rhetoric about the Netherlands being built to a "10,000-year" standard--as if their levees and walls are that much higher or stronger than the USA's "100-year" or "500-year" standards.
They are not! It's just easier to build to the "10,000-yr" storm in the Netherlands, because they don't get the cyclones the USA gets. The trick is that the Netherlands doesn't really get storms (yet, i guess, let's not ignore climate change), only ocean flooding.
So, our challenge is to convert all this into real amounts of water, and use the statistics to communicate the actual probability over time of worst case scenarios (as opposed to the best estimates, which are always going to be lower, less amounts of water).
This is what I want from a news story:
"Hurricane Barry is going to drop 14 inches of rain in 48 hours onto the Mosaic Phosphate Waste Pile. The Waste pile, which enables the manufacture of most of the petrochemical fertilizer in the United States, is only engineered to handle three inches of rain in 48 hours. the Pile is collapsing, and we totally were able to predict why. but the company was greedy and too thrifty, and the Farm Bureau lobby too powerful, so it was built to substandard, and now radioactive, acid water will likely flow into Blind River."
If someone knows how to unfuck EPA, so that their engineers learn to require that facilities are built to the the worst case 4% storm --(highest part of the confidence interval, given a return period, duration, and location), i'm all ears.
Just promise me, promise me, please, that i will never have to hear "no one could have predicted this flood / this levee failure / this petrochemical spill of radioactive acid waste" ever, ever again.
posted by eustatic at 12:30 PM on August 3, 2019 [32 favorites]
I'm not the intended audience for this piece, but I'd love to read something that digs deep into the statistical methods the USGS uses to infer what should happen once every 100 years in places where there are only one or two centuries of records. Even ignoring the fact that climate change has made "normal" a moving target, there has to be a lot of inherent uncertainty in predicting the extreme tails of a distribution when you only have a few data points from that part of the distribution.
posted by aws17576 at 2:03 PM on August 3, 2019
posted by aws17576 at 2:03 PM on August 3, 2019
One of the problems with "100 year flood/drought/fire" etc. besides the moving target of climate change and the way people don't understand that number is that many input factors are long term cyclic and can exasperate each other and result in strong bi modal averages. Similar to the way solar and lunar tides interact.
So if there is an irregular warming (wet)/cooling (dry) cycle in an ocean with an average period of 19 years and a independent irregular warming (wet)/cooling (dry) cycle with an average period of 17 years actual rain fall in an area where the two factors overlap can vary with a much greater variance than either cycle alone would cause. And those peaks (wet) / valleys (dry) will align and reinforce with each other once over the course of ~323 years. But in a bracket around the 323 year peak you will have a significant period of dry/wet that is far outside the average norms where peaks are cancelled out by valleys.
So in our example you might start off with 50 years where the area gets about 10" of rain annually with a low of 5 inches. Then things kind of tend towards wetter but nothing radically so year to year for the next hundred years but whose cumulative effect means years 150-200 sees the area get 30 inches of rain on average with peaks of 45". and then back down for the next hundred years, give or take, until the cycle repeats.
If by the luck of the draw your "average" rainfall figure is derived from years 25-125 you are going to have a bad time, flood wise, between years 125-175.
Or if you've built infrastructure (IE:cities) to take advantage of the wet 50 things are going to get ugly when the trend reverts to the much dryer mean.
For example this sort of multi cycle min/max interaction seems to describe water in the southwest US and we've done all our "planning" with data from the wet quarter of the cycle. If so the next couple hundred years are going to much dryer than even the short historical year to year variation would project.
These cycles are everywhere in natural systems. The route the Mississippi takes to the gulf moves left to right and back again over the course of hundreds of years. The Cascadia Earthquake zone whose weird cyclic nature means we are either over due for a major earthquake effecting much of the North American west coast or are in the "break" part of the cycle and won't have "The One" for hundreds of years. Sunspots.
posted by Mitheral at 6:28 PM on August 4, 2019
So if there is an irregular warming (wet)/cooling (dry) cycle in an ocean with an average period of 19 years and a independent irregular warming (wet)/cooling (dry) cycle with an average period of 17 years actual rain fall in an area where the two factors overlap can vary with a much greater variance than either cycle alone would cause. And those peaks (wet) / valleys (dry) will align and reinforce with each other once over the course of ~323 years. But in a bracket around the 323 year peak you will have a significant period of dry/wet that is far outside the average norms where peaks are cancelled out by valleys.
So in our example you might start off with 50 years where the area gets about 10" of rain annually with a low of 5 inches. Then things kind of tend towards wetter but nothing radically so year to year for the next hundred years but whose cumulative effect means years 150-200 sees the area get 30 inches of rain on average with peaks of 45". and then back down for the next hundred years, give or take, until the cycle repeats.
If by the luck of the draw your "average" rainfall figure is derived from years 25-125 you are going to have a bad time, flood wise, between years 125-175.
Or if you've built infrastructure (IE:cities) to take advantage of the wet 50 things are going to get ugly when the trend reverts to the much dryer mean.
For example this sort of multi cycle min/max interaction seems to describe water in the southwest US and we've done all our "planning" with data from the wet quarter of the cycle. If so the next couple hundred years are going to much dryer than even the short historical year to year variation would project.
These cycles are everywhere in natural systems. The route the Mississippi takes to the gulf moves left to right and back again over the course of hundreds of years. The Cascadia Earthquake zone whose weird cyclic nature means we are either over due for a major earthquake effecting much of the North American west coast or are in the "break" part of the cycle and won't have "The One" for hundreds of years. Sunspots.
posted by Mitheral at 6:28 PM on August 4, 2019
« Older "Good stop! I really liked that halt, guys." | The Birth of the Semicolon Newer »
This thread has been archived and is closed to new comments
posted by Bee'sWing at 11:30 AM on August 3, 2019 [4 favorites]