Creating Recurrence Intervals from Historical Daily Rainfall Data

[cab454]

I have long suspected that the Atlas 14 data for my community is potentially overestimated. Our City's stormwater ordinance requires 2-, 10-, & 100-yr, 24-hour detention.

I have taken 122 years worth of rainfall data from a NOAA recognized gauge in our community which provides 24-hour rainfall depth. I ran a daily study for all 122 years of data and found the following:

100-yr
-expected: 0 - 1
-occurred: 0

50-yr
-expected: 0 - 2
-occurred: 1

25-yr
-expected: 2 - 4
-occurred: 1

10-yr
-expected: 7 - 11
-occurred: 5

5-yr
-expected: 18 - 22
-occurred: 7

2-yr
-expected: 52 - 58
-occurred: 21

This seems to tell me that most of the 24-hour storm depths provided by Atlas 14 are higher than what we actually experience. I would like to use the 122 year data set to create my own recurrence intervals for a 24-hour storm.

I know there are quite a few limitations with a rain gauge of this type: user error, user consistency, missing data, assumes linear distribution or rainfall, doesn't not account for storms that bridge the "24-hour tracking period", etc., but I want to see what the recurrence intervals would be if I were to assume these errors did not exist.

After scouring the internet, I am looking for some guidance on how one would determine what the recurrence intervals are based on a daily data set.

 

[CarlB]

A fairly simple approach is presented in this video:

https://www.youtube.com/watch?v=2s3siU9jf6g

Since you are interested in "x-year" storms, I'd think you would use just the largest 24-hr storm from each year, so working with 122 data points.

 

[cab454]

That actually seemed to work rather well, although I wonder if there is a way to modify the equation to make N = days rather than years. It seems like calculating the recurrence intervals off of the granularity of daily data would far exceed in accuracy those based on yearly data.

 

[CarlB]

Sure you could use days rather than years, but with that length of record, result should be not that much different.
So a 2-year storm would be a (365 x 2)-day storm.

 

[FacEngrPE]

Look at the first several sections of the hard copy NOAA atlas 14 located at this link. The derivation philosophy is explained. 122 years of data is not enough to develop precipitation return from strictly statistical analysis, even when ignoring some weather cycles are much longer than 100 years.

Using the NOAA atlas 14 is easy to justify as it is cited in International Plumbing Code and, Minimum Design Loads and Associated Criteria for Buildings and Other Structures (ASCE/SEI 7). If you want to use a different approach to developing rainfall returns, be prepared to justify your approach to the AHJ.

In my region (east coast) we have observed some rainfall events having returns more frequent than predicted in the Atlas 14.
NOAA atlas 14 supplementary information tab has this graphic, and other information relevant to understanding the derivation of NOAA Atlas 14 data.