Is anyone using the NOAA Atlas 14 Temporal Distributions for design? If so, is there anything that has been constructed based upon this design to verify performance? I'm struggling getting comfortable with the design approach being recommended by a client where we don't have any data (gauge or public input noting high water marks) to confirm the modeling. The modeling results are significantly lower than current design standards would indicate. To up the flow we are using a 6-hr median 1st quartile distribution assuming 100% impervious. I know current approaches are "conservative", but a 10-year peak flow with this new approach is indicating a 1-yr to 2-yr peak flow rate when comparing to the current design method. Other methods analyzed for comparison were Huff 1st Quartile with a 1-hr duration, SCS Type II (Current Std), NOAA Atlas 14 IDF, Frequency Modified Pilgrim Cordery (FMPC) Method. The FMPC was from another community and was used for comparison only. Any thoughts and discussion would be greatly appreciated.

For many years on this list Bill James would have responded to your question with his standard admonishment that one should never use a design storm but instead assess performance from historic events. Have you considered making that comparison? Using SWMM's Events feature, it's now fairly easy to simulate selected events from a long-term record.

We are using a second quartile median storm for an SSO study in the southeast. In that case, we know that SSOs primarily occur in response to moderate intensity winter storms. We (Hazen is the engineer; my firm is program manager) compared simulated results from the synthetic event with a suite of historical storms, and assessed SSO volume, number of overflow locations, and peak overflow rates to ensure the NOAA storm produced reasonable results. In general though, I'd agree with your hesitation to use one of the median shapes for a design event where critical locations in the system may have both short and long times of concentration, as the NOAA hyetographs are not sharply peaked. You might consider picking among the 10th percentile shapes, which are much steeper than the median shapes. Keep in mind too that while the SCS shapes have a long history, even NRCS is in the process of deprecating them, e.g. http://www.wcc.nrcs.usda.gov/ftpref/wntsc/H&H/rainDist/NOAA_Atlas_14_MW_SE_rainfall_%20state_county_documentation_2015.pdf . There's a nice presentation online that you may have seen about this issue: http://www.floods.org/Files/Conf2017_ppts/B3_Wilson.pdf

I presume you're working in the Midwest where it's not unusual for the peak hour of a heavy storm to account for half or more of the event's 6-hour total. I suspect that the median NOAA events are considerably more attenuated than that. As an example of historic events, here are a few events in Kansas City with roughly equivalent average recurrence intervals at 1 and 6 hours: 5/26/16 (5-y), 9/16/01 (10-y), and 9/10/15 (15-y). Characteristics of these events and other events can be viewed by selecting Local Statistics at http://www.dynsystem.com/netstorm, and choosing Kansas City from the map.

I would've accepted that criticism. You are correct this is in the Midwest (Kansas City). I ran a simulation for the July and August 2017 flood events based upon rainfall data near our project site. The July event showed more of an impact to our project area. Although the overall Blue River Watershed experienced a higher peak from the August event due to the nature of the event coming from the S/SE instead of the typical W or NW for the area.

I actually have the presentation you linked sitting on my desk for reference and actually sat through a similar presentation before this one at the MfSMA conference. I know KC is different than STL, but I used the FMPC data from MSD and ran it through our watershed for the different duration events. I haven't included the Greene County distribution. They have recently adopted an FMPC distribution as well. Everything I'm seeing indicates the Temporal Distribution as correlating to a 1-yr to 2-yr event per current design standards as noted. It is my understanding the KC metro was approached to do the same analysis, but declined.

I may pull the events you have noted and run those and see what the results look like.

If you use Huff distribution - one should perform a critical duration analysis - that is DONT just look at the system 1-hour duration. As Atlas 14 distributions were built in a similar yet different manner, you should look at the responses across the range of all probabilities.

I think Circular 173 covers Huff in for design methods. I might have my reference incorrect.

The Huff 1hr was the critical duration. The watershed is approx. 240 acres, which falls in line with a guidance table I have. When you state all range of probabilities are you saying to do the 90% through 10% for the 1st quartile or all quartiles? I'll have to review, but are the quartiles similar to Huff for the durations 6hr and less use the 1st quartile, 6to12 use 2nd quartile, 12to24 use 3rd quartile, and greater than 24 use 4th quartile?

If Huff 1-hour 1st Quartile is your critical duration - I wouldn't entertain NOAA is your critical Huff is early and short.

The NOAA temporals are fixed duration - 6, 12, 24 hours etc. They don't scale, or NOAA does not recommend scaling them down.

NOAA critical duration analysis would look at the 4 Quartiles - and at least run 10/90 (early/late) for each quartile per duration. It could be as many as 40 events per duration.

Do you have a document that explains the critical duration analysis for use of the NOAA Temporal Distributions?

Since this is a topic near and dear to me as well as a major part of the research I am working on, I have added a paper that we published on the variability of flood response with storm temporal patterns as well as the impact of climate change. Since the paper is based on NOAA temporal distributions for the midwest, I thought it might be appropriate and maybe useful. HESS is open access, hence the link should work even if you do not have access to academic journals. The paper supports the ideas discussed in the presentation as well.

I agree with Mitch on trending towards the 10th percentile shapes and moving away from the SCS distributions. I cannot stress enough the need to use an ensemble of shapes too, including initial conditions, or running continuous simulations.

https://www.hydrol-earth-syst-sci.net/22/2041/2018

Another perspective to this discussion is that the rainfall distributions considered have different time intervals. Compared to 1-hr Huff, SCS etc. which uses a hyetograph time interval of 5-6 min, NOAA temporal distributions has a 30-min interval. This generally corresponds to a large peak rainfall intensity for HUFF/SCS distributions compared to NOAA temporal distributions. SWMM results for peak runoff are sensitive to the peak rainfall intensity and I think that is probably a main factor for Ron's observation of lower peak flow rate when NOAA temporal distributions are used.

Not sure I like how this discussion bifurcated on GMAIL - but Suresh has a nice paper on the research he did: https://www.hydrol-earth-syst-sci.net/22/2041/2018

The general process of a critical duration analysis is to find the proper critical values (Flow, Depth, Volume) for the design problem - bigger pipe, how deep does it flood, and how much storage occurs. In many cases - these may be very different - the pipe might require one duration, but the flooding volume a different duration. Any design decision should be made appropriately. There are a number of state DOT requirements on design require a review of pre- and post-development critical durations. Let me see what I can dig up for you.