Time of Concentration and Design Storms

The design storm lasts just long enough for the whole watershed to contribute.

Before this Lesson

A culvert under a county road drains 40 acres. The rainfall records list three equally rare storms: a 10-minute burst at 4.80 in./hr, a half-hour storm at 2.67 in./hr, and a one-hour soaker at 1.60 in./hr. Which one do you size the culvert for? And how rare is rare enough, when a 100-year flood has only about a 63 percent chance of showing up within 100 years?

The worst storm isn't the hardest rain

Two facts pull against each other. The first is that short storms rain harder. Every storm has a worst few minutes. A 10-minute average catches just that burst, while a one-hour average dilutes it with everything around it.

The second is that the outlet needs time. As you saw in The Rational Method, the flow keeps growing until water from the farthest point arrives, at the time of concentration, tct_c. Stop the rain sooner, and the near water has drained away before the far water gets there. Only part of the area is ever contributing at once.

So try all three storms on the culvert's site, which has C=0.50C = 0.50 and tc=30t_c = 30 min. For the short storm, take the simplest picture, where the contributing area grows steadily with time. A 10-minute storm then has a third of the site contributing at its peak. Only the contributing area counts in Q=CIAQ = CIA, so that storm gives (0.50)(4.80)(13.3)=32(0.50)(4.80)(13.3) = 32 cfs.

Three equally rare storms on the 40-acre site
StormI (in./hr)Area contributingPeak, Q (cfs)
10 min4.8013.3 acres32
30 min2.6740 acres53
60 min1.6040 acres32

Run every other duration the same way, and the whole trade-off appears.

Intensity, I (in./hr) 0 2 4 6 8 Share of the area contributing 0 50% 100% Peak at the outlet, Q (cfs) 0 20 40 60 0 15 30 45 60 75 Storm duration (min) tc = 30 min 10 min: 32 cfs 30 min: 53 cfs 60 min: 32 cfs
Why the peak tops out at the time of concentration
The bottom curve is the top two curves multiplied together, times C=0.50C = 0.50 and A=40A = 40 acres.

Below tct_c, a longer storm gains more from added area than it loses in intensity. Past tct_c, there's no more area to gain, and the rain only gets softer. The peak sits right at the kink, so the design storm lasts tct_c, and the II in Q=CIAQ = CIA is the intensity for that duration.

Time of concentration: the last drop home

Mark a raindrop on every spot of the site and start a stopwatch when the rain starts. Each drop runs downhill along its own route to the outlet, and tct_c is the moment the last one arrives. That's what "the travel time from the hydraulically most distant point" means: most distant in minutes, not in feet.

One drop's trip usually has legs: a slow sheet across the ground, called overland flow, and then quicker flow once it reaches a ditch, gutter, or pipe. That drop travels the legs one after another, so their times add. Different routes don't add, because those drops are all running at the same time. Only the slowest route counts.

outlet B Paved gutter: 18 min A Overland, grass: 20 min Ditch: 10 min Path A: 20 + 10 = 30 min Path B: 18 min
Two routes across the 40-acre site
Schematic, not to scale.

Path B starts farther away, but a paved gutter is far faster than a sheet of water pushing through grass. Path A sets tc=30t_c = 30 min. Adding all three legs gives 48 min, which is nobody's trip.

The Handbook prints no time-of-concentration formula, so a problem has to give you each leg's time, or its length and velocity. Then the time is length over velocity, and it's worth watching for seconds that need to become minutes.

Reading the design intensity

Rainfall records come as intensity-duration-frequency data, or IDF for short. The duration is how long the storm lasts. The frequency is how rare it is, stated as a return period TT in years, which you'll unpack later in this Lesson. The intensity is the average rate over that duration. Every location has its own IDF data, and the Handbook prints none, so a problem supplies it. To use it, go in with tct_c as the duration and the stated return period, and read II.

0 10 20 30 40 50 60 0 1 2 3 4 5 6 7 8 Storm duration, t (min) Intensity, I (in./hr) 100-year 25-year 10-year 2.67 tc = 30 min
IDF curves for a made-up location
The culvert's site reads the 25-year curve at its tct_c of 30 min.
The same data as a table: intensity in in./hr
Duration (min)10-year25-year100-year
104.004.806.40
153.334.005.33
202.863.434.57
252.503.004.00
302.222.673.56
451.672.002.67
601.331.602.13

A problem may also hand you the data as an equation. Some problems write it with a lowercase ii, for the same intensity:

I=at+bI = \frac{a}{t + b}

These curves have b=15b = 15, with aa = 100, 120, and 160 for the 10-, 25-, and 100-year storms. Here tt is in minutes, because that's how aa and bb were fit. Enter a half hour as 0.5 and you get 120/15.5=7.7120/15.5 = 7.7 in./hr, harder than the 10-minute burst, which is a good sign that something's off.

Worked Example From travel times to a peak
1 of 3

A 20-acre site has C=0.60C = 0.60. Water from its most distant point flows overland for 15 min, then along 1,800 ft of channel at 3 ft/sec. Use the table above to find the 25-year peak runoff.

Faster water, bigger peak

That chain lets you judge a change without computing anything. Whatever gets water down the slowest route sooner shortens tct_c. A shorter tct_c means a shorter design storm, which sits farther left on the IDF curve, where the curve is higher. With CC and AA unchanged, the peak rises in step with II.

There's something a little perverse about this. Improve a ditch, and the flood at its outlet gets worse. No more rain has fallen. It just all shows up together.

Checkpoint

A site's area and CC stay the same. Which of these changes raise its design peak?

A.
Replacing a winding grass swale with a straight concrete channel
B.
Rerouting the main ditch the long way around the site
C.
Adding inlets so that sheet flow reaches a pipe after a shorter run
D.
Replacing a smooth concrete channel with a rough rock one
E.
Steepening the grade of the main ditch

Development hits the peak twice

Now let a 20-acre meadow become a subdivision. Roofs and streets raise CC from 0.25 to 0.75. Gutters and pipes replace the slow trip across the grass, so tct_c drops from 45 min to 15 min. On the 25-year storm in the table, that moves II from 2.00 to 4.00 in./hr.

Before: meadow After: subdivision slow overland flow gutters and pipes outlet C = 0.25 tc = 45 min I = 2.00 in./hr Q = 10 cfs outlet C = 0.75 tc = 15 min I = 4.00 in./hr Q = 60 cfs
The same 20 acres, before and after development
Site plans are schematic, not to scale.

CC tripled and II doubled, so the peak grows sixfold, from 10 cfs to 60 cfs. Scaling by CC alone predicts 30 cfs, which is half the real answer. That shortcut only works when a problem says the intensity holds steady. Otherwise, find each peak with its own CC and its own II, and subtract. Those two peaks return in Hydrographs and Detention, which holds the second down to the first. The Handbook has nothing on development, so this is just Q=CIAQ = CIA used twice.

Worked Example The increase in peak runoff
1 of 3

A 15-acre site has C=0.30C = 0.30 and tc=30t_c = 30 min today. After development, C=0.70C = 0.70 and tc=15t_c = 15 min. The 10-year intensity is I=100/(t+15)I = 100/(t + 15) in./hr, with tt in minutes. Find the increase in the 10-year peak.

A return period is a probability in disguise

Now the culvert's other question: how rare is rare enough? The "frequency" in IDF is the return period TT, and it's really an annual probability. A 100-year flood is the flood with a 1 percent chance of being equaled or exceeded in any given year:

p=1Tp = \frac{1}{T}

"100-year" is an unfortunate name. It sounds like a schedule: one flood per century, and then the clock starts over. Instead, picture a 100-sided die, rolled once a year. Roll a 100, and the flood comes. Over thousands of years, those rolls average out to one per hundred years, and that's all TT means. The die has no memory, so a flood last year does nothing to this year's roll. The flood can come two years running.

Risk over a design life

A culvert with a design life of nn years gets nn rolls of that die. What's the chance of at least one exceedance? "At least one" can happen in a pile of ways, but "none" happens just one way: every year escapes. Each year escapes with probability 1−1/T1 - 1/T, and the years are independent, so the chance they all escape is that number multiplied nn times. The risk RR is everything else:

R=1−(1−1T)nR = 1 - \left(1 - \frac{1}{T}\right)^n

The tempting answer is n/Tn/T. It works when nn is small next to TT, and then it fails, for the same reason adding failed in the Probability Lesson: a stretch with three floods gets counted three times.

0 50 100 150 200 0% 20% 40% 60% 80% 100% Design life, n (years) Chance of at least one exceedance n/T says 100% n/T n/T 1 − (1 − 1/T)n 1 − (1 − 1/T)n n = 100: 63.4% n = 100: 63.4% T = 100 years T = 100 years
Risk of at least one exceedance of a 100-year flood

For exactly one exceedance, use the binomial from the Handbook's Binomial Distribution section, with p=1/Tp = 1/T and x=1x = 1. There are C(n,1)=nC(n, 1) = n choices for which year it is, so:

Pn(1)=n(1T)(1−1T)n−1P_n(1) = n \left(\frac{1}{T}\right)\left(1 - \frac{1}{T}\right)^{n-1}

The Handbook prints nothing about return periods, and its Cumulative Binomial Probabilities table has no column below p=0.10p = 0.10, so anything rarer than a 10-year flood takes the formula.

Worked Example A 50-year flood in 20 years
1 of 3

A structure has a 20-year design life. Find the probability that the 50-year flood is equaled or exceeded at least once, and exactly once, during that life.

The 100-year flood in 100 years

Put T=100T = 100 and n=100n = 100 into the risk formula: 1−0.99100=0.6341 - 0.99^{100} = 0.634. So a 100-year flood has about a 63 percent chance of arriving within 100 years, not a certainty. Any long return period gives about the same answer when n=Tn = T. Run the binomial for each count, and here's how the possible centuries divide up.

No flood: 37% One flood: 37% Two floods: 18% back to back Three or more: 8% 0 25 50 75 100 Year
Four ways a century can go, with a 100-year flood
The flood years shown are examples. Only the counts and chances are calculated.
Checkpoint

A 50-year flood struck a town last year. Which statements are true?

A.
The same flood has a 2 percent chance of occurring this year
B.
The chance this year is below 2 percent, because it just happened
C.
It could occur in each of the next two years
D.
It can't occur again for about 50 years
E.
It is certain to occur again within the next 50 years
F.
It is more likely than not to occur at least once in the next 50 years

Choosing the return period

Often the risk is what's given. An owner names a design life and the risk they'll accept, and you find the return period that delivers it. Undo the risk formula one step at a time:

(1−1T)n=1−R⇒T=11−(1−R)1/n\left(1 - \frac{1}{T}\right)^n = 1 - R \quad\Rightarrow\quad T = \frac{1}{1 - (1 - R)^{1/n}}

Your calculator's power function handles the 1/n1/n exponent. Keep several digits of that root, because everything hangs on its small difference from 1, and round TT up, since a smaller TT means more risk. When the options are a list of return periods, you can instead put each one into the risk formula and see which ones pass.

Worked Example A culvert with a 10 percent limit
1 of 3

A culvert has a 25-year design life, and the owner accepts at most a 10 percent risk that its design flood is exceeded during that life. What return period does it need?

Recap

  • Short storms rain harder, but the outlet's peak needs the whole area contributing, so the design storm lasts tct_c. Any shorter and part of the area isn't contributing. Any longer and the intensity is lower.
  • tct_c is the longest travel time along one route: its legs add, and separate routes don't. Read II from IDF data at that duration and the stated return period. A shorter, smoother path lowers tct_c and raises II and the peak.
  • Development raises CC and shortens tct_c, so find each peak with its own CC and II. The peak rises by more than the ratio of the CC values.
  • A TT-year flood has a 1/T1/T chance in any year, with no schedule and no memory. It can come in consecutive years, and a 100-year flood has only about a 63 percent chance within 100 years.
  • Over nn years, the risk is R=1−(1−1/T)nR = 1 - (1 - 1/T)^n, not n/Tn/T, and exactly once is the binomial with p=1/Tp = 1/T. To hold a risk, solve for TT and round up.