With apologies to those of you who understandably do not like the design storm concept, I have a question on Design Storm Duration - to go with the hydrograph method. In most typical WR projects I worked on, a critical storm duration was found by testing a number of candidate durations, starting with a rounded-upward time of concentration of the total catchment and ending with some integer multiple of 24 hours (the rationale at the time was an easy access to such rainfall data). For urban drainage of the cities we simply used to use a 6-hour storm where almost no concentrated storage was involved and a 24-hour storm for otherwise.

I'd like to know what's the rule & rationale for selection of a design storm duration for medium to large city ( 6' 000 - 60' 000 ha ~ 1' 5000 - 150' 000 acres)? Working on a capacity-enhancement assignment of an urban subbasin (~ 25 sq km ; Tc ~ 1.25 hrs) I just realized that the parameter is too important to be overlooked and that a 'conservative' selection here can be too expensive & thus must have a solid basis. Any set rules, guidelines, ideas and/or suggestions are highly appreciated.

With the size of the drainage basins you are describing, it is more than likely that you would have more than one critical duration. This is especially true if you are looking at solving more than one problem at different points in the system. You will need to identify the critical locations within your watershed and run multiple durations and determine how much the elevation and flow vary at these critical locations. Based on this comparison, you can determine which storm(s) is the critical duration(s). If you are adding/reducing storage or increasing conveyance in the system as part of your solution to the problem, this may end up changing your critical durations downstream of the solution for your with project conditions.

You are right that this approach can become as expensive and as time consuming as using continuous simulation. As a person who uses both continuous simulation and design storms based on client and regulatory requirements, you will have to decide which approach will give you the better answer and will be accepted in the regulatory framework you are working within.

Sometimes a design storm is a better approach if you do not have a historic rainfall series with a long enough period, has few large events, or the data is collected only at a daily time interval. If you have an adequate historic rainfall series nearby, this may be the better approach. There are too many variables to identify the better approach.

I do agree that the general route to take is to search for a critical duration/s, but having no detention there seems to be no reasonable clear cut on where to stop increasing the duration; it keeps producing bigger and higher flows!

There's no set rules on design storm selection for the semi-aridic environment of the project area; neither is available a continuous suitable rain record. The client just wishes to make it 'safe' against a 50-year storm by conveyance improvement only! A previous simulation of a 1-hour 50-yr storm was the design basis of the improvement works & a relief conduit been planned for construction. The large open channel with the deficient capacity is a branch of the city's main drainage system & has a free outfall with ample capacity. I'm here to check to see if the system can safely & economically handle the 50-year flow!

The question is, what's an appropriate storm duration for such a case? Though obvious, I've tested the 2-,3-,4-, 6- and 24-hour 50-year storms, only to see that a successively larger volumes & peaks are produced at the diversion point to the relief works and beyond. I know the general climate history (and the type & specs of the project) favours a shorter duration away from a 24-hour storm, but how short a shorter duration should be?

Among others, some say : 'As a minimum, the rainfall duration should be twice the time of travel through the drainage area to allow development of a full hydrograph' (p. 563, ASCE,1996; Hydrology Handbook ) -- what's an upper limit which is both not wasteful and is conservative where no obvious critical duration is found? Many just throw an easy ultra conservative choice of a 24-hour storm but it hardly can be justified when no detention storage is involved. How about considering the whole main drainage system concentration time (~ 4.0-4.5 hours) and raising the project's storm duration to the one compatible with the larger enclosing system?

My past experience on a small (25 acre) urban, very impervious, drainage area was that the peaks were almost the same no matter the duration, and that they were driven by the peak 5-minute rainfall in the middle of the design storm. I got essentially the same results for peak discharge from the 24-hour event and the 5-minute event.

Based on that single data point (!) I would anticipate that there is a duration beyond which the peak flows remain the same. At some point, the very low depth tails on either side of the middle of the design storm would no longer contribute anything substantial to the peak part of the hydrograph.

This applies only to peak flow rate, not to volume, which obviously should increase as the duration increases.

Apologies if you've tried this and it hasn't worked.

I've found a similar situation - the local regulator has prescribed 4- and 24-hour (Chicago) storm events for design. My peak flows are similar, irrespective of the event duration - at least for urbanized catchments. Therefore, given my models are relatively small, I just end up using the 24-hour storm.

That's the advantage of a design storm that's "nested" so that each bracketed time interval contains the depth of a storm of the corresponding duration (the peak 10 minutes has the depth of a 10 minute storm, and that plus the adjacent 20 minutes has the depth of a 30 minutes storm, etc. It's the approach the Corps of Engineers used in HEC-1 watershed simulations and the reason that SCS distributions work in urban areas. You get the runoff that applies to projects sensitive to peak rate (storm drains in the urban core or large parking lots), the rate that occurs further down the watershed where TC becomes significant, and also the volume that becomes critical when outflow is restricted such as in detention.

In effect, if your drainage area TC is sensitive to a 4 hour storm, the nested 24 hour storm gives you the response to both.

The procedures typically employed will result in bigger and bigger peak flow values as the duration of the design storm increases for large watersheds. But in semi-arid regions such as the one I live in (Denver, Colorado) the large intensity storms have a limited footprint on the ground. Unfortunately, there are no reliable area correction recommendations for semi-arid areas and we found the ones recommended by NOAA not to work well in our region either. Radar images clearly show that the areal extent of storms in larger watersheds such as yours do not have heaviest intensities over the entire area. As a result, you have an unsolvable problem at this time that would give you reliable 50-year flows.

We have adopted a standard design storm series in the Denver area based on intense studies of hydrology in the mid 1970s. All of our design storms are front loaded intensity types and range from 2-hour to 6-hour durations depending on watershed area. In addition, the temporal distributions of all storms were calibrated against long term simulations of runoff from smaller catchments using models that were calibrated for them using rainfall/runoff data obtained over 8-year and longer periods.

You may access the resulting recommended rainfall distributions and durations by downloading the Volume 1 of the Urban Storm Drainage Criteria Manual (USDCM) from www.udfcd.org under Downloads and then Criteria. In addition, the site has spreadsheets that help calculate design storms for the Denver area by merely entering 1-hour precipitation depth. Hope these will help you make some choices in the protocols you adopt.

Bill, for a small highly impervious lot, that's indeed what to be expected. However, a larger mixed-response system is somehow lazier & does not behaves so straight, though as Gene pointed out it should finally get there.

Gene, I'm using a 'nested' (or 'balanced') pattern locally built but find it too expensive to jump from a shorter duration event, say a 3- or a 4-hr storm, to a 24-hour or longer one just because it can handle many remotely possible cases conservatively.

I believe it goes quite exactly the way Ben bothered to explain: for some reasonable durations flow keeps on increasing. In fact, Mike's been lucky enough to 'soon' get to an equilibrium state. Ben, thanks for your valuable comments plainly stated; I'll follow your lead and try to make the best of it. I'm somewhat aware of your previous works and actually did consult your table of duration versus drainage area first --- which for this case points to a 2-hour storm (for an area of about 10 sq miles). I can't clearly say why but I feel a 2-hour duration is not a critical enough duration for the case perhaps because of some fragmented info I got on the local rainfall regime & storms' spatio-temporal distribution; in addition, I just can't neglect the fact that it's a branch (a subbasin) of a larger system and that I would have moved in your D.A.-Duration table towards a higher drainage area (e.g., to a >20 sq miles) to select a 6-hour storm if I'd been to derive a 50-year flood for a somewhat larger area covering the current project domain.