When conduct return period analysis on most of extreme events.
For example in flood levels, the difference between the levels of 10-yr and 50 yr floods is significantly higher comparing to values of 100yr-150yr.
What is statistically sound reason for this ?
Your first answer is great and how I have always thought. But as we must face, we are often extrapolating a flood peak curve well beyond the data. It has taken much of my career to reach this, but the statistics of extending say 25 or 50 years of peak flow data to 100 years or beyond has some problems. I am not sure exactly what a statistician would say. But in my career, what is the likelihood of one person having multiple storms in the 50 - 100 year range, and one 400-500 year, and one 1000 year storm. I think our data is generally limited in duration, and the occassional peak outlier might need to be evaluated in another way. We should recognize even with a great curve, extending data well beyond its boundary has limits to be clarified. Sometimes we forget this when we see the smooth curve, and assume it deserves to be extended.
I like the first answer, and I'd like to state it in a different way. Look at floods as differing from the normal level of a river---i.e., how high is the water compared to zero (normal water level). We assume that the more extreme the flood, the rarer it is, i.e., water level is an indicator of rarity. Therefore a 10-year flood will be not-too high, and a 50-year flood (40 years later on average) will be higher but rarer. But a 10-year flood is not really uncommon, nor very high, so a 50 year flood will be significantly rarer but doesn't have to be extremely higher than a 10-year flood. A 100-year flood will be much higher, and a 150-year flood much, much higher and rarer. Both the 100 year and 150-year floods encompass the 50-year floods, in that they are as high as the 50-year flood and actually they surpass it.. As you go further in time, or further from zero water height, the rarity becomes non-linearly more extreme (you can graph the results in the first answer). So a 50 year difference between 100 and 150-year floods will be more extreme in terms of rarity (and water height) than the difference between a 10-year and 50-year flood.
I've always wondered about naming these flood events as a 100-year flood, 500-year flood,, etc. In the US there are simply not records going back 500 years and not even 150 years in many places. Geologists and the Corps of Engineers can look for flood scars on the landscape and estimate the time they occurred but I don't know how well they actually do this. I recall river flooding in my lifetime that was declared to be a 500-year flood, which seems slightly unlikely given that a human doesn't live for 500 years and one is tempted to ask 'how likely is this to occur in my lifetime?' Also given the propensity for old records to be broken at every possible excuse to declare them broken, and that these floods don't occur at exactly every 500 years, makes one wonder how often within 500 years a 500-year flood might actually occur, or indeed how often could it occur within a human lifetime. A probability density function or similar procedure might answer this. But again since we are basically speculating about past flood frequencies, the logic arises that if something occurs twice in, say, 120 years, then is it not more likely to be a once-in-a-century event than a once-in-a-500-year event? As the second answer here says, we shouldn't extend the prediction curve outwards too far---and especially since our past data doesn't reliably go back very far either.
As a side story, in Pittsburgh and the Ohio River valley there was huge flooding in the 1930s. You can see high-water markers that were put on buildings which are still standing. It is staggering that the water level could have been that high---there must have been simultaneous record snowmelt and the remains of a rare hurricane giving drenching rains for days to make that happen. There were literally cubic miles of water above the normal water level in those valleys and it's fast-moving too. And they say there are geologic indicators of even much higher water than that in relatively recent but pre-historic times in the Ohio valley. Rare events do eventually happen!
- Badges
- Science topic