What a curve number actually stands for
The runoff curve number is not a percentage and it is not a coefficient you multiply rainfall by. It is a shorthand for one physical quantity: how much water the ground can still absorb once a storm is under way. That quantity is called the potential maximum retention, written S, and the relationship is S = 1000 / CN − 10, with S in inches.
Run a few numbers through it and the scale stops looking arbitrary. CN 98, which is what a roof or a sealed driveway gets, gives S = 0.20 inches — a fifth of an inch of storage and then everything else runs off. CN 61, a lawn in good condition on a well-drained loam, gives S = 6.39 inches, which is over half a foot of storage capacity sitting in the soil profile. CN 30 gives S = 23.3 inches, essentially a sponge. The curve number compresses a soil type and a ground cover and a management condition into one number on a scale where high means hard.
The runoff itself then comes from Q = (P − Ia)² / (P − Ia + S), where P is the storm rainfall and Ia is the initial abstraction — the water lost to interception on leaves, filling of surface depressions and the first wetting of the ground before any of it starts moving. Ia is conventionally taken as 0.2 S.
The threshold nobody expects the first time
Because Ia has to be satisfied before the equation produces anything, there is a rainfall depth below which the answer is exactly zero. For that lawn at CN 61, Ia is 0.2 x 6.39 = 1.28 inches. A one-inch storm on it yields no runoff whatsoever. A 1.5-inch storm yields 0.0074 inches. A two-inch storm yields 0.073 inches. The curve is almost flat and then it is not.
The roof beside it has Ia = 0.04 inches, so it starts producing runoff within the first few minutes of almost any storm, and by two inches of rain it has yielded 1.77 inches. Same weather, same square footage, twenty-four times the water. This is the single most useful thing the method teaches a homeowner, and it is why 1,800 square feet of roof can be a bigger drainage problem than the 6,200 square feet of grass around it.
| Rainfall P | Q at CN 98 (roof) | Q at CN 89 (gravel) | Q at CN 61 (lawn) |
|---|---|---|---|
| 0.5" | 0.32" | 0.04" | 0 |
| 1.0" | 0.79" | 0.28" | 0 |
| 2.0" | 1.77" | 1.03" | 0.073" |
| 4.0" | 3.77" | 2.82" | 0.81" |
| 6.0" | 5.76" | 4.74" | 2.01" |
Read down the last column rather than across. The lawn contributes essentially nothing until the storm gets large, then it begins contributing seriously. That is why a drainage arrangement that has worked for six years can fail in one afternoon: the storm crossed the point where the pervious two-thirds of the property joined in.
Why the calculator shows you two answers
There are two defensible ways to handle a site made of several different surfaces. You can area-weight the curve numbers into one composite CN and run the equation once, or you can run the equation once per surface and add the volumes. Both are standard. They do not agree.
They disagree because the runoff equation is curved. Averaging the inputs to a curved function is not the same as averaging its outputs, and here the direction of the error is predictable: blending a small high-CN area into a large low-CN area drags the composite down into the flat part of the curve, and the hard surface loses most of the runoff it would actually have produced. The gap is largest exactly where homeowners live — small storms, small impervious fraction. On the default numbers in the form, with a two-inch storm, the composite route comes out meaningfully lower than the sum of the parts.
Neither answer is wrong as arithmetic. What matters is which one the person reviewing your drawing expects, and that is worth a question rather than an assumption. Where impervious cover is a small share of a mostly-vegetated site, running the surfaces separately is the more conservative and generally the more honest choice.
The initial abstraction ratio is not a settled number
The 0.2 in Ia = 0.2 S was fitted to a limited data set decades ago and has been argued about ever since. Re-analysis of large rainfall-runoff data sets has repeatedly suggested a value nearer 0.05 fits observations better, which produces noticeably more runoff from small and medium storms and rather less difference at large ones. Some manuals have adopted it, most have not, and the curve number tables themselves were calibrated with 0.2 baked in — so changing the ratio without also adjusting the CN is not a clean substitution.
The field is there so you can see how much your answer depends on it, and so you can match whatever your jurisdiction asks for. If you are not being reviewed by anyone, leave it at 0.2 and treat the result as a bracket.
Where this method stops being appropriate
The curve number method estimates a runoff volume for a storm of roughly one day, on a watershed measured in acres or square miles, using soil hydrologic groups mapped for agriculture. Applied to a suburban lot and a twenty-minute thunderstorm it is well outside its calibration, and it says nothing at all about when the water arrives or how fast.
For rate rather than volume you need a different tool — the rational method peak flow calculator is the usual one at this scale, and it needs a rainfall intensity read at the site's time of concentration. If the runoff is going into something rather than through it, the volume from this page feeds the rain garden, dry well and detention basin pages directly.
Questions people ask
Where do I get the curve number for my yard?
From the NRCS tables, which cross a land cover description against a hydrologic soil group — A through D, from sand at one end to clay at the other. Your local drainage manual almost certainly reprints the relevant rows, and the soil group for your property can be looked up in the USDA Web Soil Survey. The two halves matter equally: "lawn in good condition" spans CN 39 on group A soil to CN 80 on group D, which is nearly the whole range from sponge to pavement. Guessing the cover and ignoring the soil is the most common way people arrive at a confident wrong number.
Why did my lawn produce zero runoff for a one-inch storm?
Because one inch is below the initial abstraction for that curve number, and the method is built so that nothing runs off until the initial abstraction is satisfied. At CN 61 the potential retention S is 6.39 inches and Ia is 0.2 of that, so 1.28 inches of rain has to fall before the equation produces anything at all. Physically this represents interception, depression storage and the first wetting of dry soil. The honest caveat is that a saturated lawn in the third day of rain does shed water in a storm that size, and the standard curve number describes an average antecedent moisture condition rather than a wet one.
Is the curve number the same as a runoff coefficient?
No, though the calculator shows you the coefficient it works out to. A rational-method C is a ratio of peak runoff rate to rainfall intensity and it is treated as fixed for a surface. A curve number stands for a storage depth in inches, and the fraction of rainfall that runs off changes with the size of the storm — the same lawn sheds nothing at one inch and a third of the rain at six. If you need a single equivalent C to hand somebody, take the ratio of runoff depth to rainfall depth for the specific design storm you care about, and say which storm it came from.
Should I use 0.2 or 0.05 for the initial abstraction ratio?
Use whatever the authority reviewing the project expects, and if nobody is reviewing it, use 0.2. The published curve number tables were calibrated with 0.2 in place, so switching to 0.05 without a corresponding adjustment to the curve numbers themselves mixes two calibrations. The argument for 0.05 comes from re-analysis of large observed data sets and it is a serious argument, but it is a change to the whole method rather than a knob to turn. The field is on this page so you can see the sensitivity, not so you can shop for an answer.
Does this tell me how big a pipe or a swale needs to be?
No. This is a volume for the whole storm, and pipes and channels are sized on a flow rate in cubic feet per second, which depends on how quickly the water arrives as well as how much of it there is. Peak flow at this scale usually comes from the rational method, which is on the culvert and swale flow page, and it needs a rainfall intensity read at a duration equal to the time of concentration. The volume from this page is the input for things that store water rather than move it: a rain garden, a dry well, a detention basin, a cistern.